Improve accessibility of microphone, camera, and screen share buttons

Taking Valere's suggestion of giving them the 'switch' role. Also, the aria-label attributes were redundant (having tooltips already gives the buttons aria-labelledby).

.github/PULL_REQUEST_TEMPLATE.md

@@ -0,0 +1,39 @@
+<!-- Thanks for submitting a PR! Please ensure the following requirements are met in order for us to review your PR -->
+
+## Content
+
+<!-- Describe shortly what has been changed -->
+
+## Motivation and context
+
+<!-- Provide link to the corresponding issue if applicable or explain the context -->
+
+## Screenshots / GIFs
+
+<!--
+
+You can use a table like this to show screenshots comparison.
+Uncomment this markdown table below and edit the last line `|||`:
+|copy screenshot of before here|copy screenshot of after here|
+
+|Before|After|
+|-|-|
+|||
+ -->
+
+## Tests
+
+<!-- Explain how you tested your development -->
+
+- Step 1
+- Step 2
+- Step ...
+-
+
+## Checklist
+
+- [ ] I have read through [CONTRIBUTING.md](https://github.com/element-hq/element-call/blob/livekit/CONTRIBUTING.md).
+- [ ] Pull request includes screenshots or videos if containing UI changes
+- [ ] Tests written for new code (and old code if feasible).
+- [ ] Linter and other CI checks pass.
+- [ ] I have licensed the changes to Element by completing the [Contributor License Agreement (CLA)](https://cla-assistant.io/element-hq/element-call)

.github/workflows/blocked.yaml

@@ -1,7 +1,16 @@
 name: Prevent blocked
 on:
+  # zizmor: ignore[dangerous-triggers]
+  # Reason: This workflow does not checkout code or use secrets.
+  # It only reads labels to set a failure status on the PR.
   pull_request_target:
     types: [opened, labeled, unlabeled, synchronize]
+
+permissions:
+  pull-requests: read
+  # Required to fail the check on the PR
+  statuses: write
+
 jobs:
   prevent-blocked:
     name: Prevent blocked

.github/workflows/build-and-publish-docker.yaml

@@ -20,10 +20,13 @@ jobs:
     runs-on: ubuntu-latest
     permissions:
       contents: write # required to upload release asset
-      packages: write
+      packages: write # needed for publishing packages to GHCR
+      id-token: write # needed for login into tailscale with GitHub OIDC Token
     steps:
       - name: Check it out
         uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+        with:
+          persist-credentials: false
 
       - name: 📥 Download artifact
         uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v4
@@ -40,12 +43,50 @@ jobs:
           username: ${{ github.actor }}
           password: ${{ secrets.GITHUB_TOKEN }}
 
+      - name: Connect to Tailscale
+        uses: tailscale/github-action@306e68a486fd2350f2bfc3b19fcd143891a4a2d8 # v4
+        if: github.event_name != 'pull_request'
+        with:
+          oauth-client-id: ${{ secrets.TS_OAUTH_CLIENT_ID }}
+          audience: ${{ secrets.TS_AUDIENCE }}
+          tags: tag:github-actions
+
+      - name: Compute vault jwt role name
+        id: vault-jwt-role
+        if: github.event_name != 'pull_request'
+        run: |
+          echo "role_name=github_service_management_$( echo "${{ github.repository }}" | sed -r 's|[/-]|_|g')" | tee -a "$GITHUB_OUTPUT"
+
+      - name: Get team registry token
+        id: import-secrets
+        uses: hashicorp/vault-action@4c06c5ccf5c0761b6029f56cfb1dcf5565918a3b # v3
+        if: github.event_name != 'pull_request'
+        with:
+          url: https://vault.infra.ci.i.element.dev
+          role: ${{ steps.vault-jwt-role.outputs.role_name }}
+          path: service-management/github-actions
+          jwtGithubAudience: https://vault.infra.ci.i.element.dev
+          method: jwt
+          secrets: |
+            services/voip-repositories/secret/data/oci.element.io username | OCI_USERNAME ;
+            services/voip-repositories/secret/data/oci.element.io password | OCI_PASSWORD ;
+
+      - name: Login to oci.element.io Registry
+        uses: docker/login-action@c94ce9fb468520275223c153574b00df6fe4bcc9 # v3
+        if: github.event_name != 'pull_request'
+        with:
+          registry: oci-push.vpn.infra.element.io
+          username: ${{ steps.import-secrets.outputs.OCI_USERNAME }}
+          password: ${{ steps.import-secrets.outputs.OCI_PASSWORD }}
+
       - name: Extract metadata (tags, labels) for Docker
         id: meta
         uses: docker/metadata-action@c299e40c65443455700f0fdfc63efafe5b349051 # v5.10.0
         with:
-          images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
-          tags: ${{ inputs.docker_tags}}
+          images: |
+            ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
+            oci-push.vpn.infra.element.io/element-call
+          tags: ${{ inputs.docker_tags }}
           labels: |
             org.opencontainers.image.licenses=AGPL-3.0-only OR LicenseRef-Element-Commercial
 

.github/workflows/build-element-call.yaml

@@ -33,6 +33,8 @@ jobs:
     steps:
       - name: Checkout code
         uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+        with:
+          persist-credentials: false
       - name: Enable Corepack
         run: corepack enable
       - name: Yarn cache
@@ -43,7 +45,7 @@ jobs:
       - name: Install dependencies
         run: "yarn install --immutable"
       - name: Build Element Call
-        run: ${{ format('yarn run build:{0}:{1}', inputs.package, inputs.build_mode) }}
+        run: yarn run build:"$PACKAGE":"$BUILD_MODE"
         env:
           SENTRY_ORG: ${{ secrets.SENTRY_ORG }}
           SENTRY_PROJECT: ${{ secrets.SENTRY_PROJECT }}
@@ -52,6 +54,8 @@ jobs:
           CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}
           VITE_APP_VERSION: ${{ inputs.vite_app_version }}
           NODE_OPTIONS: "--max-old-space-size=4096"
+          PACKAGE: ${{ inputs.package }}
+          BUILD_MODE: ${{ inputs.build_mode }}
       - name: Upload Artifact
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4
         with:

.github/workflows/build.yaml

@@ -49,7 +49,9 @@ jobs:
     permissions:
       contents: write
       packages: write
+      id-token: write
     uses: ./.github/workflows/build-and-publish-docker.yaml
+    secrets: inherit
     with:
       artifact_run_id: ${{ github.run_id }}
       docker_tags: |

.github/workflows/changelog-label.yml

@@ -1,8 +1,16 @@
 name: PR changelog label
 
 on:
+  # zizmor: ignore[dangerous-triggers]
+  # This is safe because we do not use actions/checkout or execute untrusted code.
+  # Using pull_request_target is necessary to allow status writes for PRs from forks.
   pull_request_target:
     types: [labeled, unlabeled, opened]
+
+permissions:
+  pull-requests: read
+  statuses: write
+
 jobs:
   pr-changelog-label:
     runs-on: ubuntu-latest

.github/workflows/deploy-to-netlify.yaml

@@ -63,11 +63,15 @@ jobs:
 
       - name: Add config file
         run: |
-          if [ "${{ inputs.package }}" = "full" ]; then
-            curl -s "https://raw.githubusercontent.com/${{ inputs.pr_head_full_name }}/${{ inputs.pr_head_ref }}/config/config_netlify_preview.json" > webapp/config.json
+          if [ "${INPUTS_PACKAGE}" = "full" ]; then
+            curl -s "https://raw.githubusercontent.com/${INPUTS_PR_HEAD_FULL_NAME}/${INPUTS_PR_HEAD_REF}/config/config_netlify_preview.json" > webapp/config.json
           else
-            curl -s "https://raw.githubusercontent.com/${{ inputs.pr_head_full_name }}/${{ inputs.pr_head_ref }}/config/config_netlify_preview_sdk.json" > webapp/config.json
+            curl -s "https://raw.githubusercontent.com/${INPUTS_PR_HEAD_FULL_NAME}/${INPUTS_PR_HEAD_REF}/config/config_netlify_preview_sdk.json" > webapp/config.json
           fi
+        env:
+          INPUTS_PACKAGE: ${{ inputs.package }}
+          INPUTS_PR_HEAD_FULL_NAME: ${{ inputs.pr_head_full_name }}
+          INPUTS_PR_HEAD_REF: ${{ inputs.pr_head_ref }}
       - name: ☁️ Deploy to Netlify
         id: netlify
         uses: nwtgck/actions-netlify@4cbaf4c08f1a7bfa537d6113472ef4424e4eb654 # v3.0

.github/workflows/lint.yaml

@@ -8,6 +8,8 @@ jobs:
     steps:
       - name: Checkout code
         uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+        with:
+          persist-credentials: false
       - name: Enable Corepack
         run: corepack enable
       - name: Yarn cache

.github/workflows/pr-deploy.yaml

@@ -1,5 +1,7 @@
 name: Deploy previews for PRs
 on:
+  # zizmor: ignore[dangerous-triggers]
+  # Reason: This is now restricted to internal PRs only using the 'if' condition below.
   workflow_run:
     workflows: ["Build"]
     types:
@@ -7,7 +9,14 @@ on:
 
 jobs:
   prdetails:
-    if: ${{ github.event.workflow_run.conclusion == 'success' && github.event.workflow_run.event == 'pull_request' }}
+    # Logic:
+    # 1. Build must be successful
+    # 2. Event must be a pull_request
+    # 3. Head repository must be the SAME as the base repository (No Forks!)
+    if: >
+      github.event.workflow_run.conclusion == 'success' && 
+      github.event.workflow_run.event == 'pull_request' &&
+      github.event.workflow_run.head_repository.full_name == github.repository
     runs-on: ubuntu-latest
     outputs:
       pr_number: ${{ steps.prdetails.outputs.pr_id }}
@@ -60,7 +69,9 @@ jobs:
     permissions:
       contents: write
       packages: write
+      id-token: write
     uses: ./.github/workflows/build-and-publish-docker.yaml
+    secrets: inherit
     with:
       artifact_run_id: ${{ github.event.workflow_run.id || github.run_id }}
       docker_tags: |

.github/workflows/publish-embedded-packages.yaml

@@ -22,8 +22,18 @@ jobs:
       TAG: ${{ steps.tag.outputs.TAG }}
     steps:
       - name: Calculate VERSION
-        # We should only use the hard coded test value for a dry run
-        run: echo "VERSION=${{ github.event_name == 'release' && github.event.release.tag_name || 'v0.0.0-pre.0' }}" >> "$GITHUB_ENV"
+        # Safely store dynamic values in environment variables
+        # to prevent shell injection (template-injection)
+        run: |
+          # The logic is executed within the shell using the env variables
+          if [ "$EVENT_NAME" = "release" ]; then
+            echo "VERSION=$RELEASE_TAG" >> "$GITHUB_ENV"
+          else
+            echo "VERSION=v0.0.0-pre.0" >> "$GITHUB_ENV"
+          fi
+        env:
+          RELEASE_TAG: ${{ github.event.release.tag_name }}
+          EVENT_NAME: ${{ github.event_name }}
       - id: dry_run
         name: Set DRY_RUN
         # We perform a dry run for all events except releases.
@@ -71,7 +81,9 @@ jobs:
       contents: write # required to upload release asset
     steps:
       - name: Determine filename
-        run: echo "FILENAME_PREFIX=element-call-embedded-${{ needs.versioning.outputs.UNPREFIXED_VERSION }}" >> "$GITHUB_ENV"
+        run: echo "FILENAME_PREFIX=element-call-embedded-${NEEDS_VERSIONING_OUTPUTS_UNPREFIXED_VERSION}" >> "$GITHUB_ENV"
+        env:
+          NEEDS_VERSIONING_OUTPUTS_UNPREFIXED_VERSION: ${{ needs.versioning.outputs.UNPREFIXED_VERSION }}
       - name: 📥 Download built element-call artifact
         uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v4
         with:
@@ -80,9 +92,9 @@ jobs:
           name: build-output-embedded
           path: ${{ env.FILENAME_PREFIX}}
       - name: Create Tarball
-        run: tar --numeric-owner -cvzf ${{ env.FILENAME_PREFIX }}.tar.gz ${{ env.FILENAME_PREFIX }}
+        run: tar --numeric-owner -cvzf ${FILENAME_PREFIX}.tar.gz ${FILENAME_PREFIX}
       - name: Create Checksum
-        run: find ${{ env.FILENAME_PREFIX }} -type f -print0 | sort -z | xargs -0 sha256sum | tee ${{ env.FILENAME_PREFIX }}.sha256
+        run: find ${FILENAME_PREFIX} -type f -print0 | sort -z | xargs -0 sha256sum | tee ${FILENAME_PREFIX}.sha256
       - name: Upload
         if: ${{ needs.versioning.outputs.DRY_RUN == 'false' }}
         uses: softprops/action-gh-release@a06a81a03ee405af7f2048a818ed3f03bbf83c7b # v2
@@ -104,6 +116,8 @@ jobs:
     steps:
       - name: Checkout
         uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+        with:
+          persist-credentials: false
 
       - name: 📥 Download built element-call artifact
         uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v4
@@ -123,13 +137,16 @@ jobs:
       - name: Publish npm
         working-directory: embedded/web
         run: |
-          npm version ${{ needs.versioning.outputs.PREFIXED_VERSION }} --no-git-tag-version
+          npm version ${NEEDS_VERSIONING_OUTPUTS_PREFIXED_VERSION} --no-git-tag-version
           echo "ARTIFACT_VERSION=$(jq '.version' --raw-output package.json)" >> "$GITHUB_ENV"
-          npm publish --provenance --access public --tag ${{ needs.versioning.outputs.TAG }} ${{ needs.versioning.outputs.DRY_RUN == 'true' && '--dry-run' || '' }}
+          npm publish --provenance --access public --tag ${NEEDS_VERSIONING_OUTPUTS_TAG} ${{ needs.versioning.outputs.DRY_RUN == 'true' && '--dry-run' || '' }}
+        env:
+          NEEDS_VERSIONING_OUTPUTS_PREFIXED_VERSION: ${{ needs.versioning.outputs.PREFIXED_VERSION }}
+          NEEDS_VERSIONING_OUTPUTS_TAG: ${{ needs.versioning.outputs.TAG }}
 
       - id: artifact_version
         name: Output artifact version
-        run: echo "ARTIFACT_VERSION=${{env.ARTIFACT_VERSION}}" >> "$GITHUB_OUTPUT"
+        run: echo "ARTIFACT_VERSION=${ARTIFACT_VERSION}" >> "$GITHUB_OUTPUT"
 
   publish_android:
     needs: [build_element_call, versioning]
@@ -143,6 +160,8 @@ jobs:
     steps:
       - name: Checkout
         uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+        with:
+          persist-credentials: false
 
       - name: 📥 Download built element-call artifact
         uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v4
@@ -161,16 +180,19 @@ jobs:
       - name: Get artifact version
         # Anything that is not a final release will be tagged as a snapshot
         run: |
-          if [[ "${{ needs.versioning.outputs.TAG }}" == "latest" ]]; then
-            echo "ARTIFACT_VERSION=${{ needs.versioning.outputs.UNPREFIXED_VERSION }}" >> "$GITHUB_ENV"
-          elif [[ "${{ needs.versioning.outputs.TAG }}" == "rc" ]]; then
-            echo "ARTIFACT_VERSION=${{ needs.versioning.outputs.UNPREFIXED_VERSION }}" >> "$GITHUB_ENV"
+          if [[ "${NEEDS_VERSIONING_OUTPUTS_TAG}" == "latest" ]]; then
+            echo "ARTIFACT_VERSION=${NEEDS_VERSIONING_OUTPUTS_UNPREFIXED_VERSION}" >> "$GITHUB_ENV"
+          elif [[ "${NEEDS_VERSIONING_OUTPUTS_TAG}" == "rc" ]]; then
+            echo "ARTIFACT_VERSION=${NEEDS_VERSIONING_OUTPUTS_UNPREFIXED_VERSION}" >> "$GITHUB_ENV"
           else
-            echo "ARTIFACT_VERSION=${{ needs.versioning.outputs.UNPREFIXED_VERSION }}-SNAPSHOT" >> "$GITHUB_ENV"
+            echo "ARTIFACT_VERSION=${NEEDS_VERSIONING_OUTPUTS_UNPREFIXED_VERSION}-SNAPSHOT" >> "$GITHUB_ENV"
           fi
+        env:
+          NEEDS_VERSIONING_OUTPUTS_TAG: ${{ needs.versioning.outputs.TAG }}
+          NEEDS_VERSIONING_OUTPUTS_UNPREFIXED_VERSION: ${{ needs.versioning.outputs.UNPREFIXED_VERSION }}
 
       - name: Set version string
-        run: sed -i "s/0.0.0/${{ env.ARTIFACT_VERSION }}/g" embedded/android/lib/src/main/kotlin/io/element/android/call/embedded/Version.kt
+        run: sed -i "s/0.0.0/${ARTIFACT_VERSION}/g" embedded/android/lib/src/main/kotlin/io/element/android/call/embedded/Version.kt
 
       - name: Publish AAR
         working-directory: embedded/android
@@ -184,7 +206,7 @@ jobs:
 
       - id: artifact_version
         name: Output artifact version
-        run: echo "ARTIFACT_VERSION=${{env.ARTIFACT_VERSION}}" >> "$GITHUB_OUTPUT"
+        run: echo "ARTIFACT_VERSION=${ARTIFACT_VERSION}" >> "$GITHUB_OUTPUT"
 
   publish_ios:
     needs: [build_element_call, versioning]
@@ -200,6 +222,7 @@ jobs:
         uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
         with:
           path: element-call
+          persist-credentials: false
 
       - name: 📥 Download built element-call artifact
         uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v4
@@ -215,15 +238,18 @@ jobs:
           repository: element-hq/element-call-swift
           path: element-call-swift
           token: ${{ secrets.SWIFT_RELEASE_TOKEN }}
+          persist-credentials: false
 
       - name: Copy files
         run: rsync -a --delete --exclude .git element-call/embedded/ios/ element-call-swift
 
       - name: Get artifact version
-        run: echo "ARTIFACT_VERSION=${{ needs.versioning.outputs.UNPREFIXED_VERSION }}" >> "$GITHUB_ENV"
+        run: echo "ARTIFACT_VERSION=${NEEDS_VERSIONING_OUTPUTS_UNPREFIXED_VERSION}" >> "$GITHUB_ENV"
+        env:
+          NEEDS_VERSIONING_OUTPUTS_UNPREFIXED_VERSION: ${{ needs.versioning.outputs.UNPREFIXED_VERSION }}
 
       - name: Set version string
-        run: sed -i "s/0.0.0/${{ env.ARTIFACT_VERSION }}/g" element-call-swift/Sources/EmbeddedElementCall/EmbeddedElementCall.swift
+        run: sed -i "s/0.0.0/${ARTIFACT_VERSION}/g" element-call-swift/Sources/EmbeddedElementCall/EmbeddedElementCall.swift
 
       - name: Test build
         working-directory: element-call-swift
@@ -235,17 +261,22 @@ jobs:
           git config --global user.email "ci@element.io"
           git config --global user.name "Element CI"
           git add -A
-          git commit -am "Release ${{ needs.versioning.outputs.PREFIXED_VERSION }}"
-          git tag -a ${{ env.ARTIFACT_VERSION }} -m "${{ github.event.release.html_url }}"
+          git commit -am "Release ${NEEDS_VERSIONING_OUTPUTS_PREFIXED_VERSION}"
+          git tag -a ${ARTIFACT_VERSION} -m "${GITHUB_EVENT_RELEASE_HTML_URL}"
+        env:
+          NEEDS_VERSIONING_OUTPUTS_PREFIXED_VERSION: ${{ needs.versioning.outputs.PREFIXED_VERSION }}
+          GITHUB_EVENT_RELEASE_HTML_URL: ${{ github.event.release.html_url }}
 
       - name: Push
         working-directory: element-call-swift
         run: |
-          git push --tags ${{ needs.versioning.outputs.DRY_RUN == 'true' && '--dry-run' || '' }}
+          git push "https://x-access-token:${SWIFT_RELEASE_TOKEN}@github.com/element-hq/element-call-swift.git" --tags ${{ needs.versioning.outputs.DRY_RUN == 'true' && '--dry-run' || '' }}
+        env:
+          SWIFT_RELEASE_TOKEN: ${{ secrets.SWIFT_RELEASE_TOKEN }}
 
       - id: artifact_version
         name: Output artifact version
-        run: echo "ARTIFACT_VERSION=${{env.ARTIFACT_VERSION}}" >> "$GITHUB_OUTPUT"
+        run: echo "ARTIFACT_VERSION=${ARTIFACT_VERSION}" >> "$GITHUB_OUTPUT"
 
   release_notes:
     needs: [versioning, publish_npm, publish_android, publish_ios]
@@ -257,9 +288,13 @@ jobs:
     steps:
       - name: Log versions
         run: |
-          echo "NPM: ${{ needs.publish_npm.outputs.ARTIFACT_VERSION }}"
-          echo "Android: ${{ needs.publish_android.outputs.ARTIFACT_VERSION }}"
-          echo "iOS: ${{ needs.publish_ios.outputs.ARTIFACT_VERSION }}"
+          echo "NPM: ${NEEDS_PUBLISH_NPM_OUTPUTS_ARTIFACT_VERSION}"
+          echo "Android: ${NEEDS_PUBLISH_ANDROID_OUTPUTS_ARTIFACT_VERSION}"
+          echo "iOS: ${NEEDS_PUBLISH_IOS_OUTPUTS_ARTIFACT_VERSION}"
+        env:
+          NEEDS_PUBLISH_NPM_OUTPUTS_ARTIFACT_VERSION: ${{ needs.publish_npm.outputs.ARTIFACT_VERSION }}
+          NEEDS_PUBLISH_ANDROID_OUTPUTS_ARTIFACT_VERSION: ${{ needs.publish_android.outputs.ARTIFACT_VERSION }}
+          NEEDS_PUBLISH_IOS_OUTPUTS_ARTIFACT_VERSION: ${{ needs.publish_ios.outputs.ARTIFACT_VERSION }}
       - name: Add release notes
         if: ${{ needs.versioning.outputs.DRY_RUN == 'false' }}
         uses: softprops/action-gh-release@a06a81a03ee405af7f2048a818ed3f03bbf83c7b # v2

.github/workflows/publish.yaml

@@ -38,9 +38,9 @@ jobs:
           name: build-output-full
           path: ${{ env.FILENAME_PREFIX }}
       - name: Create Tarball
-        run: tar --numeric-owner --transform "s/dist/${{ env.FILENAME_PREFIX }}/" -cvzf ${{ env.FILENAME_PREFIX }}.tar.gz ${{ env.FILENAME_PREFIX }}
+        run: tar --numeric-owner --transform "s/dist/${FILENAME_PREFIX}/" -cvzf ${FILENAME_PREFIX}.tar.gz ${FILENAME_PREFIX}
       - name: Create Checksum
-        run: find ${{ env.FILENAME_PREFIX }} -type f -print0 | sort -z | xargs -0 sha256sum | tee ${{ env.FILENAME_PREFIX }}.sha256
+        run: find ${FILENAME_PREFIX} -type f -print0 | sort -z | xargs -0 sha256sum | tee ${FILENAME_PREFIX}.sha256
       - name: Upload
         uses: softprops/action-gh-release@a06a81a03ee405af7f2048a818ed3f03bbf83c7b # v2
         with:
@@ -55,7 +55,9 @@ jobs:
     permissions:
       contents: write
       packages: write
+      id-token: write
     uses: ./.github/workflows/build-and-publish-docker.yaml
+    secrets: inherit
     with:
       artifact_run_id: ${{ github.event.workflow_run.id || github.run_id }}
       docker_tags: |

.github/workflows/test.yaml

@@ -10,6 +10,8 @@ jobs:
     steps:
       - name: Checkout code
         uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+        with:
+          persist-credentials: false
       - name: Enable Corepack
         run: corepack enable
       - name: Yarn cache
@@ -34,6 +36,8 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+        with:
+          persist-credentials: false
       - name: Enable Corepack
         run: corepack enable
       - uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020 # v4

.github/workflows/translations-download.yaml

@@ -14,6 +14,8 @@ jobs:
     steps:
       - name: Checkout the code
         uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+        with:
+          persist-credentials: false
 
       - name: Enable Corepack
         run: corepack enable

.github/workflows/translations-upload.yaml

@@ -15,6 +15,8 @@ jobs:
     steps:
       - name: Checkout the code
         uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+        with:
+          persist-credentials: false
 
       - name: Upload
         uses: localazy/upload@27e6b5c0fddf4551596b42226b1c24124335d24a # v1

.github/workflows/zizmor.yml

@@ -0,0 +1,23 @@
+name: GitHub Actions Security Analysis with zizmor 🌈
+
+on:
+  push:
+    branches: ["livekit", "full-mesh"]
+  pull_request: {}
+
+permissions: {}
+
+jobs:
+  zizmor:
+    name: Run zizmor 🌈
+    runs-on: ubuntu-latest
+    permissions:
+      security-events: write # Required for upload-sarif (used by zizmor-action) to upload SARIF files.
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          persist-credentials: false
+
+      - name: Run zizmor 🌈
+        uses: zizmorcore/zizmor-action@71321a20a9ded102f6e9ce5718a2fcec2c4f70d8 # v0.5.2

dev-backend-docker-compose.yml

@@ -47,7 +47,7 @@ services:
       - ecbackend
 
   livekit:
-    image: livekit/livekit-server:v1.9.4
+    image: livekit/livekit-server:v1.9.11
     pull_policy: always
     hostname: livekit-sfu
     command: --dev --config /etc/livekit.yaml
@@ -67,7 +67,7 @@ services:
       - ecbackend
 
   livekit-1:
-    image: livekit/livekit-server:v1.9.4
+    image: livekit/livekit-server:v1.9.11
     pull_policy: always
     hostname: livekit-sfu-1
     command: --dev --config /etc/livekit.yaml
@@ -88,7 +88,7 @@ services:
 
   synapse:
     hostname: homeserver
-    image: ghcr.io/element-hq/synapse:pr-18968-dcb7678281bc02d4551043a6338fe5b7e6aa47ce
+    image: ghcr.io/element-hq/synapse:latest
     pull_policy: always
     environment:
       - SYNAPSE_CONFIG_PATH=/data/cfg/homeserver.yaml
@@ -106,7 +106,7 @@ services:
 
   synapse-1:
     hostname: homeserver-1
-    image: ghcr.io/element-hq/synapse:pr-18968-dcb7678281bc02d4551043a6338fe5b7e6aa47ce
+    image: ghcr.io/element-hq/synapse:latest
     pull_policy: always
     environment:
       - SYNAPSE_CONFIG_PATH=/data/cfg/homeserver.yaml

embedded/android/gradle/libs.versions.toml

@@ -2,11 +2,11 @@
 # https://docs.gradle.org/current/userguide/platforms.html#sub::toml-dependencies-format
 
 [versions]
-android_gradle_plugin = "8.13.1"
+android_gradle_plugin = "8.13.2"
 
 [libraries]
 android_gradle_plugin = { module = "com.android.tools.build:gradle", version.ref = "android_gradle_plugin" }
 
 [plugins]
 android_library = { id = "com.android.library", version.ref = "android_gradle_plugin" }
-maven_publish = { id = "com.vanniktech.maven.publish", version = "0.35.0" }
+maven_publish = { id = "com.vanniktech.maven.publish", version = "0.36.0" }

embedded/android/gradle/wrapper/gradle-wrapper.properties

@@ -1,6 +1,6 @@
 distributionBase=GRADLE_USER_HOME
 distributionPath=wrapper/dists
-distributionUrl=https\://services.gradle.org/distributions/gradle-8.14.3-all.zip
+distributionUrl=https\://services.gradle.org/distributions/gradle-8.14.4-all.zip
 networkTimeout=10000
 validateDistributionUrl=true
 zipStoreBase=GRADLE_USER_HOME

knip.ts

@@ -34,6 +34,12 @@ export default {
     // then Knip will flag it as a false positive
     // https://github.com/webpro-nl/knip/issues/766
     "@vector-im/compound-web",
+    // Yarn plugins are allowed to depend on packages provided by the Yarn
+    // runtime. These shouldn't be listed in package.json, because plugins
+    // should work before Yarn even installs dependencies for the first time.
+    // https://yarnpkg.com/advanced/plugin-tutorial#what-does-a-plugin-look-like
+    "@yarnpkg/core",
+    "@yarnpkg/parsers",
     "matrix-widget-api",
   ],
   ignoreExportsUsedInFile: true,

locales/en/app.json

@@ -250,11 +250,11 @@
   "video_tile": {
     "always_show": "Always show",
     "camera_starting": "Video loading...",
-    "change_fit_contain": "Fit to frame",
     "collapse": "Collapse",
     "expand": "Expand",
     "mute_for_me": "Mute for me",
     "muted_for_me": "Muted for me",
+    "screen_share_volume": "Screen share volume",
     "volume": "Volume",
     "waiting_for_media": "Waiting for media..."
   }

package.json

@@ -43,12 +43,12 @@
     "@codecov/vite-plugin": "^1.3.0",
     "@fontsource/inconsolata": "^5.1.0",
     "@fontsource/inter": "^5.1.0",
-    "@formatjs/intl-durationformat": "^0.9.0",
+    "@formatjs/intl-durationformat": "^0.10.0",
     "@formatjs/intl-segmenter": "^11.7.3",
     "@livekit/components-core": "^0.12.0",
     "@livekit/components-react": "^2.0.0",
     "@livekit/protocol": "^1.42.2",
-    "@livekit/track-processors": "^0.6.0 || ^0.7.1",
+    "@livekit/track-processors": "^0.7.1",
     "@mediapipe/tasks-vision": "^0.10.18",
     "@playwright/test": "^1.57.0",
     "@radix-ui/react-dialog": "^1.0.4",
@@ -79,7 +79,7 @@
     "@vector-im/compound-design-tokens": "^6.0.0",
     "@vector-im/compound-web": "^8.0.0",
     "@vitejs/plugin-react": "^4.0.1",
-    "@vitest/coverage-v8": "^3.0.0",
+    "@vitest/coverage-v8": "^4.0.18",
     "babel-plugin-transform-vite-meta-env": "^1.0.3",
     "classnames": "^2.3.1",
     "copy-to-clipboard": "^3.3.3",
@@ -101,7 +101,7 @@
     "i18next-browser-languagedetector": "^8.0.0",
     "i18next-parser": "^9.1.0",
     "jsdom": "^26.0.0",
-    "knip": "^5.27.2",
+    "knip": "^5.86.0",
     "livekit-client": "^2.13.0",
     "lodash-es": "^4.17.21",
     "loglevel": "^1.9.1",
@@ -118,7 +118,7 @@
     "qrcode": "^1.5.4",
     "react": "19",
     "react-dom": "19",
-    "react-i18next": "^16.0.0 <16.1.0",
+    "react-i18next": "^16.0.0 <16.6.0",
     "react-router-dom": "^7.0.0",
     "react-use-measure": "^2.1.1",
     "rxjs": "^7.8.1",
@@ -128,17 +128,22 @@
     "unique-names-generator": "^4.6.0",
     "uuid": "^13.0.0",
     "vaul": "^1.0.0",
-    "vite": "^7.0.0",
+    "vite": "^7.3.0",
     "vite-plugin-generate-file": "^0.3.0",
     "vite-plugin-html": "^3.2.2",
     "vite-plugin-node-stdlib-browser": "^0.2.1",
     "vite-plugin-svgr": "^4.0.0",
-    "vitest": "^3.0.0",
+    "vitest": "^4.0.18",
     "vitest-axe": "^1.0.0-pre.3"
   },
   "resolutions": {
     "@livekit/components-core/rxjs": "^7.8.1",
-    "@livekit/track-processors/@mediapipe/tasks-vision": "^0.10.18"
+    "@livekit/track-processors/@mediapipe/tasks-vision": "^0.10.18",
+    "minimatch": "^10.2.3",
+    "tar": "^7.5.11",
+    "glob": "^10.5.0",
+    "qs": "^6.14.1",
+    "js-yaml": "^4.1.1"
   },
   "packageManager": "yarn@4.7.0"
 }

playwright/create-call.spec.ts

@@ -22,8 +22,8 @@ test("Start a new call then leave and show the feedback screen", async ({
   await expect(page.getByTestId("lobby_joinCall")).toBeVisible();
 
   // Check the button toolbar
-  // await expect(page.getByRole('button', { name: 'Mute microphone' })).toBeVisible();
-  // await expect(page.getByRole('button', { name: 'Stop video' })).toBeVisible();
+  // await expect(page.getByRole('switch', { name: 'Mute microphone' })).toBeVisible();
+  // await expect(page.getByRole('switch', { name: 'Stop video' })).toBeVisible();
   await expect(page.getByRole("button", { name: "Settings" })).toBeVisible();
   await expect(page.getByRole("button", { name: "End call" })).toBeVisible();
 

playwright/reconnect.spec.ts

@@ -49,12 +49,12 @@ test("can only interact with header and footer while reconnecting", async ({
   ).toBeVisible();
 
   // Tab order should jump directly from header to footer, skipping media tiles
-  await page.getByRole("button", { name: "Mute microphone" }).focus();
+  await page.getByRole("switch", { name: "Mute microphone" }).focus();
   await expect(
-    page.getByRole("button", { name: "Mute microphone" }),
+    page.getByRole("switch", { name: "Mute microphone" }),
   ).toBeFocused();
   await page.keyboard.press("Tab");
-  await expect(page.getByRole("button", { name: "Stop video" })).toBeFocused();
+  await expect(page.getByRole("switch", { name: "Stop video" })).toBeFocused();
   // Most critically, we should be able to press the hangup button
   await page.getByRole("button", { name: "End call" }).click();
 });

playwright/widget/huddle-call.test.ts

@@ -55,13 +55,10 @@ widgetTest("Create and join a group call", async ({ addUser, browserName }) => {
     const frame = user.page
       .locator('iframe[title="Element Call"]')
       .contentFrame();
-
     // No lobby, should start with video on
-    // The only way to know if it is muted or not is to look at the data-kind attribute..
-    const videoButton = frame.getByTestId("incall_videomute");
-    await expect(videoButton).toBeVisible();
-    // video should be off by default in a voice call
-    await expect(videoButton).toHaveAttribute("aria-label", /^Stop video$/);
+    await expect(
+      frame.getByRole("switch", { name: "Stop video", checked: true }),
+    ).toBeVisible();
   }
 
   // We should see 5 video tiles everywhere now
@@ -101,13 +98,15 @@ widgetTest("Create and join a group call", async ({ addUser, browserName }) => {
   const florianFrame = florian.page
     .locator('iframe[title="Element Call"]')
     .contentFrame();
-  const florianMuteButton = florianFrame.getByTestId("incall_videomute");
-  await florianMuteButton.click();
+  const florianVideoButton = florianFrame.getByRole("switch", {
+    name: /video/,
+  });
+  await expect(florianVideoButton).toHaveAccessibleName("Stop video");
+  await expect(florianVideoButton).toBeChecked();
+  await florianVideoButton.click();
   // Now the button should indicate we can start video
-  await expect(florianMuteButton).toHaveAttribute(
-    "aria-label",
-    /^Start video$/,
-  );
+  await expect(florianVideoButton).toHaveAccessibleName("Start video");
+  await expect(florianVideoButton).not.toBeChecked();
 
   // wait a bit for the state to propagate
   await valere.page.waitForTimeout(3000);

playwright/widget/pip-call-button-interaction.test.ts

@@ -0,0 +1,73 @@
+/*
+Copyright 2026 Element Creations Ltd.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import { expect, test } from "@playwright/test";
+
+import { widgetTest } from "../fixtures/widget-user.ts";
+import { HOST1, TestHelpers } from "./test-helpers.ts";
+
+widgetTest("Footer interaction in PiP", async ({ addUser, browserName }) => {
+  test.skip(
+    browserName === "firefox",
+    "The is test is not working on firefox CI environment. No mic/audio device inputs so cam/mic are disabled",
+  );
+
+  test.slow();
+
+  const valere = await addUser("Valere", HOST1);
+
+  const callRoom = "CallRoom";
+  await TestHelpers.createRoom("CallRoom", valere.page);
+
+  await TestHelpers.createRoom("OtherRoom", valere.page);
+
+  await TestHelpers.switchToRoomNamed(valere.page, callRoom);
+
+  // Start the call as Valere
+  await TestHelpers.startCallInCurrentRoom(valere.page, false);
+  await expect(
+    valere.page.locator('iframe[title="Element Call"]'),
+  ).toBeVisible();
+
+  await TestHelpers.joinCallFromLobby(valere.page);
+  // wait a bit so that the PIP has rendered
+  await valere.page.waitForTimeout(600);
+
+  // Switch to the other room, the call should go to PIP
+  await TestHelpers.switchToRoomNamed(valere.page, "OtherRoom");
+
+  // We should see the PIP overlay
+  const iFrame = valere.page
+    .locator('iframe[title="Element Call"]')
+    .contentFrame();
+
+  {
+    // Check for a bug where the video had the wrong fit in PIP
+    const audioBtn = iFrame.getByRole("switch", { name: /microphone/ });
+    const videoBtn = iFrame.getByRole("switch", { name: /video/ });
+    await expect(
+      iFrame.getByRole("button", { name: "End call" }),
+    ).toBeVisible();
+    await expect(audioBtn).toBeVisible();
+    await expect(videoBtn).toBeVisible();
+    await expect(audioBtn).toHaveAccessibleName("Mute microphone");
+    await expect(audioBtn).toBeChecked();
+    await expect(videoBtn).toHaveAccessibleName("Stop video");
+    await expect(videoBtn).toBeChecked();
+
+    await videoBtn.click();
+    await audioBtn.click();
+
+    // stop hovering on any of the buttons
+    await iFrame.getByTestId("videoTile").hover();
+
+    await expect(audioBtn).toHaveAccessibleName("Unmute microphone");
+    await expect(audioBtn).toBeChecked();
+    await expect(videoBtn).toHaveAccessibleName("Start video");
+    await expect(videoBtn).not.toBeChecked();
+  }
+});

playwright/widget/pip-call.test.ts

@@ -0,0 +1,72 @@
+/*
+Copyright 2026 Element Creations Ltd.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import { expect, test } from "@playwright/test";
+
+import { widgetTest } from "../fixtures/widget-user.ts";
+import { HOST1, TestHelpers } from "./test-helpers.ts";
+
+widgetTest("Put call in PIP", async ({ addUser, browserName }) => {
+  test.skip(
+    browserName === "firefox",
+    "The is test is not working on firefox CI environment. No mic/audio device inputs so cam/mic are disabled",
+  );
+
+  test.slow();
+
+  const valere = await addUser("Valere", HOST1);
+  const timo = await addUser("Timo", HOST1);
+
+  const callRoom = "TeamRoom";
+  await TestHelpers.createRoom(callRoom, valere.page, [timo.mxId]);
+
+  await TestHelpers.createRoom("DoubleTask", valere.page);
+
+  await TestHelpers.acceptRoomInvite(callRoom, timo.page);
+
+  await TestHelpers.switchToRoomNamed(valere.page, callRoom);
+
+  // Start the call as Valere
+  await TestHelpers.startCallInCurrentRoom(valere.page, false);
+  await expect(
+    valere.page.locator('iframe[title="Element Call"]'),
+  ).toBeVisible();
+
+  await TestHelpers.joinCallFromLobby(valere.page);
+
+  await TestHelpers.joinCallInCurrentRoom(timo.page);
+
+  const frame = timo.page
+    .locator('iframe[title="Element Call"]')
+    .contentFrame();
+
+  // check that the video is on
+  await expect(
+    frame.getByRole("switch", { name: "Stop video", checked: true }),
+  ).toBeVisible();
+
+  // Switch to the other room, the call should go to PIP
+  await TestHelpers.switchToRoomNamed(valere.page, "DoubleTask");
+
+  // We should see the PIP overlay
+  await expect(valere.page.getByTestId("widget-pip-container")).toBeVisible();
+
+  {
+    // wait a bit so that the PIP has rendered the video
+    await valere.page.waitForTimeout(600);
+
+    // Check for a bug where the video had the wrong fit in PIP
+    const frame = valere.page
+      .locator('iframe[title="Element Call"]')
+      .contentFrame();
+    const videoElements = await frame.locator("video").all();
+    expect(videoElements.length).toBe(1);
+
+    const pipVideo = videoElements[0];
+    await expect(pipVideo).toHaveCSS("object-fit", "cover");
+  }
+});

playwright/widget/test-helpers.ts

@@ -276,4 +276,16 @@ export class TestHelpers {
       });
     }
   }
+
+  /**
+   * Switches to a room in the room list by its name.
+   * @param page - The EW page
+   * @param roomName - The name of the room to switch to
+   */
+  public static async switchToRoomNamed(
+    page: Page,
+    roomName: string,
+  ): Promise<void> {
+    await page.getByRole("option", { name: `Open room ${roomName}` }).click();
+  }
 }

playwright/widget/voice-call-dm.spec.ts

@@ -51,34 +51,36 @@ widgetTest(
       .contentFrame();
 
     // ASSERT the button states for whistler (the callee)
-    {
-      // The only way to know if it is muted or not is to look at the data-kind attribute..
-      const videoButton = whistlerFrame.getByTestId("incall_videomute");
-      // video should be off by default in a voice call
-      await expect(videoButton).toHaveAttribute("aria-label", /^Start video$/);
-
-      const audioButton = whistlerFrame.getByTestId("incall_mute");
-      // audio should be on for the voice call
-      await expect(audioButton).toHaveAttribute(
-        "aria-label",
-        /^Mute microphone$/,
-      );
-    }
+    // video should be off by default in a voice call
+    await expect(
+      whistlerFrame.getByRole("switch", {
+        name: "Start video",
+        checked: false,
+      }),
+    ).toBeVisible();
+    // audio should be on for the voice call
+    await expect(
+      whistlerFrame.getByRole("switch", {
+        name: "Mute microphone",
+        checked: true,
+      }),
+    ).toBeVisible();
 
     // ASSERT the button states for brools (the caller)
-    {
-      // The only way to know if it is muted or not is to look at the data-kind attribute..
-      const videoButton = brooksFrame.getByTestId("incall_videomute");
-      // video should be off by default in a voice call
-      await expect(videoButton).toHaveAttribute("aria-label", /^Start video$/);
-
-      const audioButton = brooksFrame.getByTestId("incall_mute");
-      // audio should be on for the voice call
-      await expect(audioButton).toHaveAttribute(
-        "aria-label",
-        /^Mute microphone$/,
-      );
-    }
+    // video should be off by default in a voice call
+    await expect(
+      whistlerFrame.getByRole("switch", {
+        name: "Start video",
+        checked: false,
+      }),
+    ).toBeVisible();
+    // audio should be on for the voice call
+    await expect(
+      whistlerFrame.getByRole("switch", {
+        name: "Mute microphone",
+        checked: true,
+      }),
+    ).toBeVisible();
 
     // In order to confirm that the call is disconnected we will check that the message composer is shown again.
     // So first we need to confirm that it is hidden when in the call.
@@ -90,10 +92,7 @@ widgetTest(
     ).not.toBeVisible();
 
     // ASSERT hanging up on one side ends the call for both
-    {
-      const hangupButton = brooksFrame.getByTestId("incall_leave");
-      await hangupButton.click();
-    }
+    await brooksFrame.getByRole("button", { name: "End call" }).click();
 
     // The widget should be closed on both sides and the timeline should be back on screen
     await expect(
@@ -142,34 +141,30 @@ widgetTest(
       .contentFrame();
 
     // ASSERT the button states for whistler (the callee)
-    {
-      // The only way to know if it is muted or not is to look at the data-kind attribute..
-      const videoButton = whistlerFrame.getByTestId("incall_videomute");
-      // video should be on by default in a voice call
-      await expect(videoButton).toHaveAttribute("aria-label", /^Stop video$/);
-
-      const audioButton = whistlerFrame.getByTestId("incall_mute");
-      // audio should be on for the voice call
-      await expect(audioButton).toHaveAttribute(
-        "aria-label",
-        /^Mute microphone$/,
-      );
-    }
+    // video should be off by default in a video call
+    await expect(
+      whistlerFrame.getByRole("switch", { name: "Stop video", checked: true }),
+    ).toBeVisible();
+    // audio should be on too
+    await expect(
+      whistlerFrame.getByRole("switch", {
+        name: "Mute microphone",
+        checked: true,
+      }),
+    ).toBeVisible();
 
     // ASSERT the button states for brools (the caller)
-    {
-      // The only way to know if it is muted or not is to look at the data-kind attribute..
-      const videoButton = brooksFrame.getByTestId("incall_videomute");
-      // video should be on by default in a voice call
-      await expect(videoButton).toHaveAttribute("aria-label", /^Stop video$/);
-
-      const audioButton = brooksFrame.getByTestId("incall_mute");
-      // audio should be on for the voice call
-      await expect(audioButton).toHaveAttribute(
-        "aria-label",
-        /^Mute microphone$/,
-      );
-    }
+    // video should be off by default in a video call
+    await expect(
+      whistlerFrame.getByRole("switch", { name: "Stop video", checked: true }),
+    ).toBeVisible();
+    // audio should be on too
+    await expect(
+      whistlerFrame.getByRole("switch", {
+        name: "Mute microphone",
+        checked: true,
+      }),
+    ).toBeVisible();
 
     // In order to confirm that the call is disconnected we will check that the message composer is shown again.
     // So first we need to confirm that it is hidden when in the call.
@@ -181,10 +176,7 @@ widgetTest(
     ).not.toBeVisible();
 
     // ASSERT hanging up on one side ends the call for both
-    {
-      const hangupButton = brooksFrame.getByTestId("incall_leave");
-      await hangupButton.click();
-    }
+    await brooksFrame.getByRole("button", { name: "End call" }).click();
 
     // The widget should be closed on both sides and the timeline should be back on screen
     await expect(

src/button/Button.tsx

@@ -22,23 +22,25 @@ import {
 import styles from "./Button.module.css";
 
 interface MicButtonProps extends ComponentPropsWithoutRef<"button"> {
-  muted: boolean;
+  enabled: boolean;
+  size?: "sm" | "lg";
 }
 
-export const MicButton: FC<MicButtonProps> = ({ muted, ...props }) => {
+export const MicButton: FC<MicButtonProps> = ({ enabled, ...props }) => {
   const { t } = useTranslation();
-  const Icon = muted ? MicOffSolidIcon : MicOnSolidIcon;
-  const label = muted
-    ? t("unmute_microphone_button_label")
-    : t("mute_microphone_button_label");
+  const Icon = enabled ? MicOnSolidIcon : MicOffSolidIcon;
+  const label = enabled
+    ? t("mute_microphone_button_label")
+    : t("unmute_microphone_button_label");
 
   return (
     <Tooltip label={label}>
       <CpdButton
         iconOnly
-        aria-label={label}
         Icon={Icon}
-        kind={muted ? "primary" : "secondary"}
+        kind={enabled ? "primary" : "secondary"}
+        role="switch"
+        aria-checked={enabled}
         {...props}
       />
     </Tooltip>
@@ -46,23 +48,25 @@ export const MicButton: FC<MicButtonProps> = ({ muted, ...props }) => {
 };
 
 interface VideoButtonProps extends ComponentPropsWithoutRef<"button"> {
-  muted: boolean;
+  enabled: boolean;
+  size?: "sm" | "lg";
 }
 
-export const VideoButton: FC<VideoButtonProps> = ({ muted, ...props }) => {
+export const VideoButton: FC<VideoButtonProps> = ({ enabled, ...props }) => {
   const { t } = useTranslation();
-  const Icon = muted ? VideoCallOffSolidIcon : VideoCallSolidIcon;
-  const label = muted
-    ? t("start_video_button_label")
-    : t("stop_video_button_label");
+  const Icon = enabled ? VideoCallSolidIcon : VideoCallOffSolidIcon;
+  const label = enabled
+    ? t("stop_video_button_label")
+    : t("start_video_button_label");
 
   return (
     <Tooltip label={label}>
       <CpdButton
         iconOnly
-        aria-label={label}
         Icon={Icon}
-        kind={muted ? "primary" : "secondary"}
+        kind={enabled ? "primary" : "secondary"}
+        role="switch"
+        aria-checked={enabled}
         {...props}
       />
     </Tooltip>
@@ -71,6 +75,7 @@ export const VideoButton: FC<VideoButtonProps> = ({ muted, ...props }) => {
 
 interface ShareScreenButtonProps extends ComponentPropsWithoutRef<"button"> {
   enabled: boolean;
+  size: "sm" | "lg";
 }
 
 export const ShareScreenButton: FC<ShareScreenButtonProps> = ({
@@ -88,13 +93,19 @@ export const ShareScreenButton: FC<ShareScreenButtonProps> = ({
         iconOnly
         Icon={ShareScreenSolidIcon}
         kind={enabled ? "primary" : "secondary"}
+        role="switch"
+        aria-checked={enabled}
         {...props}
       />
     </Tooltip>
   );
 };
 
-export const EndCallButton: FC<ComponentPropsWithoutRef<"button">> = ({
+interface EndCallButtonProps extends ComponentPropsWithoutRef<"button"> {
+  size?: "sm" | "lg";
+}
+
+export const EndCallButton: FC<EndCallButtonProps> = ({
   className,
   ...props
 }) => {
@@ -105,7 +116,6 @@ export const EndCallButton: FC<ComponentPropsWithoutRef<"button">> = ({
       <CpdButton
         className={classNames(className, styles.endCall)}
         iconOnly
-        aria-label={t("hangup_button_label")}
         Icon={EndCallIcon}
         destructive
         {...props}
@@ -114,9 +124,10 @@ export const EndCallButton: FC<ComponentPropsWithoutRef<"button">> = ({
   );
 };
 
-export const SettingsButton: FC<ComponentPropsWithoutRef<"button">> = (
-  props,
-) => {
+interface SettingsButtonProps extends ComponentPropsWithoutRef<"button"> {
+  size?: "sm" | "lg";
+}
+export const SettingsButton: FC<SettingsButtonProps> = (props) => {
   const { t } = useTranslation();
 
   return (

src/button/ReactionToggleButton.tsx

@@ -166,6 +166,7 @@ export function ReactionPopupMenu({
 interface ReactionToggleButtonProps extends ComponentPropsWithoutRef<"button"> {
   identifier: string;
   vm: CallViewModel;
+  size?: "sm" | "lg";
 }
 
 export function ReactionToggleButton({

src/grid/TileWrapper.tsx

@@ -27,7 +27,13 @@ interface Props<M, R extends HTMLElement> {
       state: Parameters<Handler<"drag", EventTypes["drag"]>>[0],
     ) => void
   > | null;
+  /**
+   * The width this tile will have once its animations have settled.
+   */
   targetWidth: number;
+  /**
+   * The width this tile will have once its animations have settled.
+   */
   targetHeight: number;
   model: M;
   Tile: ComponentType<TileProps<M, R>>;

src/livekit/MatrixAudioRenderer.test.tsx

@@ -30,7 +30,13 @@ import {
 } from "../utils/test";
 import { initializeWidget } from "../widget";
 initializeWidget();
-export const TestAudioContextConstructor = vi.fn(() => testAudioContext);
+export const TestAudioContextConstructor = vi.fn(
+  class {
+    public constructor() {
+      return testAudioContext;
+    }
+  },
+);
 
 const MediaDevicesProvider = MediaDevicesContext.MediaDevicesContext.Provider;
 

src/room/InCallView.module.css

@@ -65,6 +65,7 @@ Please see LICENSE in the repository root for full details.
 .footer.overlay.hidden {
   display: grid;
   opacity: 0;
+  pointer-events: none;
 }
 
 .footer.overlay:has(:focus-visible) {
@@ -107,22 +108,9 @@ Please see LICENSE in the repository root for full details.
   }
 }
 
-@media (max-width: 370px) {
-  .shareScreen {
-    display: none;
-  }
-
-  @media (max-height: 400px) {
-    .footer {
-      display: none;
-    }
-  }
-}
-
-@media (max-width: 320px) {
-  .invite,
-  .raiseHand {
-    display: none;
+@media (max-height: 800px) {
+  .footer {
+    padding-block: var(--cpd-space-8x);
   }
 }
 
@@ -132,9 +120,27 @@ Please see LICENSE in the repository root for full details.
   }
 }
 
-@media (max-height: 800px) {
-  .footer {
-    padding-block: var(--cpd-space-8x);
+@media (max-width: 370px) {
+  .shareScreen {
+    display: none;
+  }
+
+  /* PIP custom css */
+  @media (max-height: 400px) {
+    .shareScreen {
+      display: flex;
+    }
+    .footer {
+      padding-block-start: var(--cpd-space-3x);
+      padding-block-end: var(--cpd-space-2x);
+    }
+  }
+}
+
+@media (max-width: 320px) {
+  .invite,
+  .raiseHand {
+    display: none;
   }
 }
 

src/room/InCallView.tsx

@@ -9,8 +9,8 @@ import { IconButton, Text, Tooltip } from "@vector-im/compound-web";
 import { type MatrixClient, type Room as MatrixRoom } from "matrix-js-sdk";
 import {
   type FC,
-  type PointerEvent,
-  type TouchEvent,
+  type MouseEvent as ReactMouseEvent,
+  type PointerEvent as ReactPointerEvent,
   useCallback,
   useEffect,
   useMemo,
@@ -110,8 +110,6 @@ import { ObservableScope } from "../state/ObservableScope.ts";
 
 const logger = rootLogger.getChild("[InCallView]");
 
-const maxTapDurationMs = 400;
-
 export interface ActiveCallProps extends Omit<
   InCallViewProps,
   "vm" | "livekitRoom" | "connState"
@@ -334,40 +332,20 @@ export const InCallView: FC<InCallViewProps> = ({
     ) : null;
   }, [ringOverlay]);
 
-  // Ideally we could detect taps by listening for click events and checking
-  // that the pointerType of the event is "touch", but this isn't yet supported
-  // in Safari: https://developer.mozilla.org/en-US/docs/Web/API/Element/click_event#browser_compatibility
-  // Instead we have to watch for sufficiently fast touch events.
-  const touchStart = useRef<number | null>(null);
-  const onTouchStart = useCallback(() => (touchStart.current = Date.now()), []);
-  const onTouchEnd = useCallback(() => {
-    const start = touchStart.current;
-    if (start !== null && Date.now() - start <= maxTapDurationMs)
-      vm.tapScreen();
-    touchStart.current = null;
-  }, [vm]);
-  const onTouchCancel = useCallback(() => (touchStart.current = null), []);
-
-  // We also need to tell the footer controls to prevent touch events from
-  // bubbling up, or else the footer will be dismissed before a click/change
-  // event can be registered on the control
-  const onControlsTouchEnd = useCallback(
-    (e: TouchEvent) => {
-      // Somehow applying pointer-events: none to the controls when the footer
-      // is hidden is not enough to stop clicks from happening as the footer
-      // becomes visible, so we check manually whether the footer is shown
-      if (showFooter) {
-        e.stopPropagation();
-        vm.tapControls();
-      } else {
-        e.preventDefault();
-      }
+  const onViewClick = useCallback(
+    (e: ReactMouseEvent) => {
+      if (
+        (e.nativeEvent as PointerEvent).pointerType === "touch" &&
+        // If an interactive element was tapped, don't count this as a tap on the screen
+        (e.target as Element).closest?.("button, input") === null
+      )
+        vm.tapScreen();
     },
-    [vm, showFooter],
+    [vm],
   );
 
   const onPointerMove = useCallback(
-    (e: PointerEvent) => {
+    (e: ReactPointerEvent) => {
       if (e.pointerType === "mouse") vm.hoverScreen();
     },
     [vm],
@@ -606,8 +584,8 @@ export const InCallView: FC<InCallViewProps> = ({
           vm={layout.spotlight}
           expanded
           onToggleExpanded={null}
-          targetWidth={gridBounds.height}
-          targetHeight={gridBounds.width}
+          targetWidth={gridBounds.width}
+          targetHeight={gridBounds.height}
           showIndicators={false}
           focusable={!contentObscured}
           aria-hidden={contentObscured}
@@ -662,20 +640,21 @@ export const InCallView: FC<InCallViewProps> = ({
 
   const buttons: JSX.Element[] = [];
 
+  const buttonSize = layout.type === "pip" ? "sm" : "lg";
   buttons.push(
     <MicButton
+      size={buttonSize}
       key="audio"
-      muted={!audioEnabled}
+      enabled={audioEnabled}
       onClick={toggleAudio ?? undefined}
-      onTouchEnd={onControlsTouchEnd}
       disabled={toggleAudio === null}
       data-testid="incall_mute"
     />,
     <VideoButton
+      size={buttonSize}
       key="video"
-      muted={!videoEnabled}
+      enabled={videoEnabled}
       onClick={toggleVideo ?? undefined}
-      onTouchEnd={onControlsTouchEnd}
       disabled={toggleVideo === null}
       data-testid="incall_videomute"
     />,
@@ -683,11 +662,11 @@ export const InCallView: FC<InCallViewProps> = ({
   if (vm.toggleScreenSharing !== null) {
     buttons.push(
       <ShareScreenButton
+        size={buttonSize}
         key="share_screen"
         className={styles.shareScreen}
         enabled={sharingScreen}
         onClick={vm.toggleScreenSharing}
-        onTouchEnd={onControlsTouchEnd}
         data-testid="incall_screenshare"
       />,
     );
@@ -695,30 +674,30 @@ export const InCallView: FC<InCallViewProps> = ({
   if (supportsReactions) {
     buttons.push(
       <ReactionToggleButton
+        size={buttonSize}
         vm={vm}
         key="raise_hand"
         className={styles.raiseHand}
         identifier={`${client.getUserId()}:${client.getDeviceId()}`}
-        onTouchEnd={onControlsTouchEnd}
       />,
     );
   }
   if (layout.type !== "pip")
     buttons.push(
       <SettingsButton
+        size={buttonSize}
         key="settings"
         onClick={openSettings}
-        onTouchEnd={onControlsTouchEnd}
       />,
     );
 
   buttons.push(
     <EndCallButton
+      size={buttonSize}
       key="end_call"
       onClick={function (): void {
         vm.hangup();
       }}
-      onTouchEnd={onControlsTouchEnd}
       data-testid="incall_leave"
     />,
   );
@@ -751,7 +730,6 @@ export const InCallView: FC<InCallViewProps> = ({
           className={styles.layout}
           layout={gridMode}
           setLayout={setGridMode}
-          onTouchEnd={onControlsTouchEnd}
         />
       )}
     </div>
@@ -760,12 +738,13 @@ export const InCallView: FC<InCallViewProps> = ({
   const allConnections = useBehavior(vm.allConnections$);
 
   return (
+    // The onClick handler here exists to control the visibility of the footer,
+    // and the footer is also viewable by moving focus into it, so this is fine.
+    // eslint-disable-next-line jsx-a11y/no-static-element-interactions, jsx-a11y/click-events-have-key-events
     <div
       className={styles.inRoom}
       ref={containerRef}
-      onTouchStart={onTouchStart}
-      onTouchEnd={onTouchEnd}
-      onTouchCancel={onTouchCancel}
+      onClick={onViewClick}
       onPointerMove={onPointerMove}
       onPointerOut={onPointerOut}
     >

src/room/LayoutToggle.tsx

@@ -5,7 +5,7 @@ SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
 Please see LICENSE in the repository root for full details.
 */
 
-import { type ChangeEvent, type FC, type TouchEvent, useCallback } from "react";
+import { type ChangeEvent, type FC, useCallback } from "react";
 import { useTranslation } from "react-i18next";
 import { Tooltip } from "@vector-im/compound-web";
 import {
@@ -22,15 +22,9 @@ interface Props {
   layout: Layout;
   setLayout: (layout: Layout) => void;
   className?: string;
-  onTouchEnd?: (e: TouchEvent) => void;
 }
 
-export const LayoutToggle: FC<Props> = ({
-  layout,
-  setLayout,
-  className,
-  onTouchEnd,
-}) => {
+export const LayoutToggle: FC<Props> = ({ layout, setLayout, className }) => {
   const { t } = useTranslation();
 
   const onChange = useCallback(
@@ -47,7 +41,6 @@ export const LayoutToggle: FC<Props> = ({
           value="spotlight"
           checked={layout === "spotlight"}
           onChange={onChange}
-          onTouchEnd={onTouchEnd}
         />
       </Tooltip>
       <SpotlightIcon aria-hidden width={24} height={24} />
@@ -58,7 +51,6 @@ export const LayoutToggle: FC<Props> = ({
           value="grid"
           checked={layout === "grid"}
           onChange={onChange}
-          onTouchEnd={onTouchEnd}
         />
       </Tooltip>
       <GridIcon aria-hidden width={24} height={24} />

src/room/LobbyView.tsx

@@ -230,12 +230,12 @@ export const LobbyView: FC<Props> = ({
           {recentsButtonInFooter && recentsButton}
           <div className={inCallStyles.buttons}>
             <MicButton
-              muted={!audioEnabled}
+              enabled={audioEnabled}
               onClick={toggleAudio ?? undefined}
               disabled={toggleAudio === null}
             />
             <VideoButton
-              muted={!videoEnabled}
+              enabled={videoEnabled}
               onClick={toggleVideo ?? undefined}
               disabled={toggleVideo === null}
             />

src/room/__snapshots__/InCallView.test.tsx.snap

@@ -285,14 +285,14 @@ exports[`InCallView > rendering > renders 1`] = `
         class="buttons"
       >
         <button
+          aria-checked="false"
           aria-disabled="true"
-          aria-label="Unmute microphone"
           aria-labelledby="_r_8_"
           class="_button_13vu4_8 _has-icon_13vu4_60 _icon-only_13vu4_53"
-          data-kind="primary"
+          data-kind="secondary"
           data-size="lg"
           data-testid="incall_mute"
-          role="button"
+          role="switch"
           tabindex="0"
         >
           <svg
@@ -309,14 +309,14 @@ exports[`InCallView > rendering > renders 1`] = `
           </svg>
         </button>
         <button
+          aria-checked="false"
           aria-disabled="true"
-          aria-label="Start video"
           aria-labelledby="_r_d_"
           class="_button_13vu4_8 _has-icon_13vu4_60 _icon-only_13vu4_53"
-          data-kind="primary"
+          data-kind="secondary"
           data-size="lg"
           data-testid="incall_videomute"
-          role="button"
+          role="switch"
           tabindex="0"
         >
           <svg
@@ -354,7 +354,6 @@ exports[`InCallView > rendering > renders 1`] = `
           </svg>
         </button>
         <button
-          aria-label="End call"
           aria-labelledby="_r_n_"
           class="_button_13vu4_8 endCall _has-icon_13vu4_60 _icon-only_13vu4_53 _destructive_13vu4_110"
           data-kind="primary"

src/settings/DeveloperSettingsTab.test.tsx

@@ -5,13 +5,16 @@ SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
 Please see LICENSE in the repository root for full details.
 */
 
-import { describe, expect, it, vi } from "vitest";
-import { render, waitFor } from "@testing-library/react";
-import { type Room as LivekitRoom } from "livekit-client";
+import { afterEach, describe, expect, it, type Mock, vi } from "vitest";
+import { render, waitFor, screen } from "@testing-library/react";
+import userEvent from "@testing-library/user-event";
+import { TooltipProvider } from "@vector-im/compound-web";
 
 import type { MatrixClient } from "matrix-js-sdk";
+import type { Room as LivekitRoom } from "livekit-client";
 import { DeveloperSettingsTab } from "./DeveloperSettingsTab";
-
+import { getSFUConfigWithOpenID } from "../livekit/openIDSFU";
+import { customLivekitUrl as customLivekitUrlSetting } from "./settings";
 // Mock url params hook to avoid environment-dependent snapshot churn.
 vi.mock("../UrlParams", () => ({
   useUrlParams: (): { mocked: boolean; answer: number } => ({
@@ -20,6 +23,14 @@ vi.mock("../UrlParams", () => ({
   }),
 }));
 
+// IMPORTANT: mock the same specifier used by DeveloperSettingsTab
+vi.mock("../livekit/openIDSFU", () => ({
+  getSFUConfigWithOpenID: vi.fn().mockResolvedValue({
+    url: "mock-url",
+    jwt: "mock-jwt",
+  }),
+}));
+
 // Provide a minimal mock of a Livekit Room structure used by the component.
 function createMockLivekitRoom(
   wsUrl: string,
@@ -86,6 +97,7 @@ describe("DeveloperSettingsTab", () => {
     const { container } = render(
       <DeveloperSettingsTab
         client={client}
+        roomId={"#room:example.org"}
         livekitRooms={livekitRooms}
         env={{ MY_MOCK_ENV: 10, ENV: "test" } as unknown as ImportMetaEnv}
       />,
@@ -99,4 +111,141 @@ describe("DeveloperSettingsTab", () => {
 
     expect(container).toMatchSnapshot();
   });
+  describe("custom livekit url", () => {
+    afterEach(() => {
+      customLivekitUrlSetting.setValue(null);
+    });
+    const client = {
+      doesServerSupportUnstableFeature: vi.fn().mockResolvedValue(true),
+      getCrypto: () => ({ getVersion: (): string => "x" }),
+      getUserId: () => "@u:hs",
+      getDeviceId: () => "DEVICE",
+    } as unknown as MatrixClient;
+    it("will not update custom livekit url without roomId", async () => {
+      const user = userEvent.setup();
+
+      render(
+        <TooltipProvider>
+          <DeveloperSettingsTab
+            client={client}
+            env={{} as unknown as ImportMetaEnv}
+          />
+        </TooltipProvider>,
+      );
+
+      const input = screen.getByLabelText("Custom Livekit-url");
+      await user.clear(input);
+      await user.type(input, "wss://example.livekit.invalid");
+
+      const saveButton = screen.getByRole("button", { name: "Save" });
+      await user.click(saveButton);
+      expect(getSFUConfigWithOpenID).not.toHaveBeenCalled();
+
+      expect(customLivekitUrlSetting.getValue()).toBe(null);
+    });
+    it("will not update custom livekit url without text in input", async () => {
+      const user = userEvent.setup();
+
+      render(
+        <TooltipProvider>
+          <DeveloperSettingsTab
+            client={client}
+            roomId="#testRoom"
+            env={{} as unknown as ImportMetaEnv}
+          />
+        </TooltipProvider>,
+      );
+
+      const input = screen.getByLabelText("Custom Livekit-url");
+      await user.clear(input);
+
+      const saveButton = screen.getByRole("button", { name: "Save" });
+      await user.click(saveButton);
+      expect(getSFUConfigWithOpenID).not.toHaveBeenCalled();
+
+      expect(customLivekitUrlSetting.getValue()).toBe(null);
+    });
+    it("will not update custom livekit url when pressing cancel", async () => {
+      const user = userEvent.setup();
+
+      render(
+        <TooltipProvider>
+          <DeveloperSettingsTab
+            client={client}
+            roomId="#testRoom"
+            env={{} as unknown as ImportMetaEnv}
+          />
+        </TooltipProvider>,
+      );
+
+      const input = screen.getByLabelText("Custom Livekit-url");
+      await user.clear(input);
+      await user.type(input, "wss://example.livekit.invalid");
+
+      const cancelButton = screen.getByRole("button", {
+        name: "Reset overwrite",
+      });
+      await user.click(cancelButton);
+      expect(getSFUConfigWithOpenID).not.toHaveBeenCalled();
+
+      expect(customLivekitUrlSetting.getValue()).toBe(null);
+    });
+    it("will update custom livekit url", async () => {
+      const user = userEvent.setup();
+
+      render(
+        <TooltipProvider>
+          <DeveloperSettingsTab
+            client={client}
+            roomId="#testRoom"
+            env={{} as unknown as ImportMetaEnv}
+          />
+        </TooltipProvider>,
+      );
+
+      const input = screen.getByLabelText("Custom Livekit-url");
+      await user.clear(input);
+      await user.type(input, "wss://example.livekit.valid");
+
+      const saveButton = screen.getByRole("button", { name: "Save" });
+      await user.click(saveButton);
+      expect(getSFUConfigWithOpenID).toHaveBeenCalledWith(
+        expect.anything(),
+        expect.anything(),
+        "wss://example.livekit.valid",
+        "#testRoom",
+      );
+
+      expect(customLivekitUrlSetting.getValue()).toBe(
+        "wss://example.livekit.valid",
+      );
+    });
+    it("will show error on invalid url", async () => {
+      const user = userEvent.setup();
+
+      render(
+        <TooltipProvider>
+          <DeveloperSettingsTab
+            client={client}
+            roomId="#testRoom"
+            env={{} as unknown as ImportMetaEnv}
+          />
+        </TooltipProvider>,
+      );
+
+      const input = screen.getByLabelText("Custom Livekit-url");
+      await user.clear(input);
+      await user.type(input, "wss://example.livekit.valid");
+
+      const saveButton = screen.getByRole("button", { name: "Save" });
+      (getSFUConfigWithOpenID as Mock).mockImplementation(() => {
+        throw new Error("Invalid URL");
+      });
+      await user.click(saveButton);
+      expect(
+        screen.getByText("invalid URL (did not update)"),
+      ).toBeInTheDocument();
+      expect(customLivekitUrlSetting.getValue()).toBe(null);
+    });
+  });
 });

src/settings/DeveloperSettingsTab.tsx

@@ -22,6 +22,7 @@ import {
 import { logger } from "matrix-js-sdk/lib/logger";
 import {
   EditInPlace,
+  ErrorMessage,
   Root as Form,
   Heading,
   HelpMessage,
@@ -45,9 +46,11 @@ import {
 } from "./settings";
 import styles from "./DeveloperSettingsTab.module.css";
 import { useUrlParams } from "../UrlParams";
+import { getSFUConfigWithOpenID } from "../livekit/openIDSFU";
 
 interface Props {
   client: MatrixClient;
+  roomId?: string;
   livekitRooms?: {
     room: LivekitRoom;
     url: string;
@@ -60,6 +63,7 @@ interface Props {
 export const DeveloperSettingsTab: FC<Props> = ({
   client,
   livekitRooms,
+  roomId,
   env,
 }) => {
   const { t } = useTranslation();
@@ -97,6 +101,8 @@ export const DeveloperSettingsTab: FC<Props> = ({
     alwaysShowIphoneEarpieceSetting,
   );
 
+  const [customLivekitUrlUpdateError, setCustomLivekitUrlUpdateError] =
+    useState<string | null>(null);
   const [customLivekitUrl, setCustomLivekitUrl] = useSetting(
     customLivekitUrlSetting,
   );
@@ -234,14 +240,36 @@ export const DeveloperSettingsTab: FC<Props> = ({
         savingLabel={t("developer_mode.custom_livekit_url.saving")}
         cancelButtonLabel={t("developer_mode.custom_livekit_url.reset")}
         onSave={useCallback(
-          (e: React.FormEvent<HTMLFormElement>) => {
-            setCustomLivekitUrl(
-              customLivekitUrlTextBuffer === ""
-                ? null
-                : customLivekitUrlTextBuffer,
-            );
+          async (e: React.FormEvent<HTMLFormElement>): Promise<void> => {
+            if (
+              roomId === undefined ||
+              customLivekitUrlTextBuffer === "" ||
+              customLivekitUrlTextBuffer === null
+            ) {
+              setCustomLivekitUrl(null);
+              return;
+            }
+
+            try {
+              const userId = client.getUserId();
+              const deviceId = client.getDeviceId();
+
+              if (userId === null || deviceId === null) {
+                throw new Error("Invalid user or device ID");
+              }
+              await getSFUConfigWithOpenID(
+                client,
+                { userId, deviceId, memberId: "" },
+                customLivekitUrlTextBuffer,
+                roomId,
+              );
+              setCustomLivekitUrlUpdateError(null);
+              setCustomLivekitUrl(customLivekitUrlTextBuffer);
+            } catch {
+              setCustomLivekitUrlUpdateError("invalid URL (did not update)");
+            }
           },
-          [setCustomLivekitUrl, customLivekitUrlTextBuffer],
+          [customLivekitUrlTextBuffer, setCustomLivekitUrl, client, roomId],
         )}
         value={customLivekitUrlTextBuffer ?? ""}
         onChange={useCallback(
@@ -256,7 +284,12 @@ export const DeveloperSettingsTab: FC<Props> = ({
           },
           [setCustomLivekitUrl],
         )}
-      />
+        serverInvalid={customLivekitUrlUpdateError !== null}
+      >
+        {customLivekitUrlUpdateError !== null && (
+          <ErrorMessage>{customLivekitUrlUpdateError}</ErrorMessage>
+        )}
+      </EditInPlace>
       <Heading as="h3" type="body" weight="semibold" size="lg">
         {t("developer_mode.matrixRTCMode.title")}
       </Heading>

src/settings/SettingsModal.tsx

@@ -213,6 +213,7 @@ export const SettingsModal: FC<Props> = ({
         env={import.meta.env}
         client={client}
         livekitRooms={livekitRooms}
+        roomId={roomId}
       />
     ),
   };

src/settings/settings.ts

@@ -34,15 +34,20 @@ export class Setting<T> {
 
     this._value$ = new BehaviorSubject(initialValue);
     this.value$ = this._value$;
+    this._lastUpdateReason$ = new BehaviorSubject<string | null>(null);
+    this.lastUpdateReason$ = this._lastUpdateReason$;
   }
 
   private readonly key: string;
 
   private readonly _value$: BehaviorSubject<T>;
+  private readonly _lastUpdateReason$: BehaviorSubject<string | null>;
   public readonly value$: Behavior<T>;
+  public readonly lastUpdateReason$: Behavior<string | null>;
 
-  public readonly setValue = (value: T): void => {
+  public readonly setValue = (value: T, reason?: string): void => {
     this._value$.next(value);
+    this._lastUpdateReason$.next(reason ?? null);
     localStorage.setItem(this.key, JSON.stringify(value));
   };
   public readonly getValue = (): T => {

src/state/CallViewModel/CallViewModel.ts

@@ -51,14 +51,7 @@ import { v4 as uuidv4 } from "uuid";
 import { type IMembershipManager } from "matrix-js-sdk/lib/matrixrtc/IMembershipManager";
 
 import {
-  LocalUserMediaViewModel,
-  type MediaViewModel,
-  type RemoteUserMediaViewModel,
-  ScreenShareViewModel,
-  type UserMediaViewModel,
-} from "../MediaViewModel";
-import {
-  accumulate,
+  createToggle$,
   filterBehavior,
   generateItem,
   generateItems,
@@ -70,7 +63,7 @@ import {
   playReactionsSound,
   showReactions,
 } from "../../settings/settings";
-import { isFirefox } from "../../Platform";
+import { isFirefox, platform } from "../../Platform";
 import { setPipEnabled$ } from "../../controls";
 import { TileStore } from "../TileStore";
 import { gridLikeLayout } from "../GridLikeLayout";
@@ -92,8 +85,6 @@ import { type MuteStates } from "../MuteStates";
 import { getUrlParams } from "../../UrlParams";
 import { type ProcessorState } from "../../livekit/TrackProcessorContext";
 import { ElementWidgetActions, widget } from "../../widget";
-import { UserMedia } from "../UserMedia.ts";
-import { ScreenShare } from "../ScreenShare.ts";
 import {
   type GridLayoutMedia,
   type Layout,
@@ -144,6 +135,14 @@ import {
 import { Publisher } from "./localMember/Publisher.ts";
 import { type Connection } from "./remoteMembers/Connection.ts";
 import { createLayoutModeSwitch } from "./LayoutSwitch.ts";
+import {
+  createWrappedUserMedia,
+  type MediaItem,
+  type WrappedUserMediaViewModel,
+} from "../media/MediaItem.ts";
+import { type ScreenShareViewModel } from "../media/ScreenShareViewModel.ts";
+import { type UserMediaViewModel } from "../media/UserMediaViewModel.ts";
+import { type MediaViewModel } from "../media/MediaViewModel.ts";
 
 const logger = rootLogger.getChild("[CallViewModel]");
 //TODO
@@ -193,7 +192,6 @@ interface LayoutScanState {
   tiles: TileStore;
 }
 
-type MediaItem = UserMedia | ScreenShare;
 export type LivekitRoomItem = {
   livekitRoom: LivekitRoom;
   participants: string[];
@@ -283,7 +281,6 @@ export interface CallViewModel {
   allConnections$: Behavior<ConnectionManagerData>;
   /** Participants sorted by livekit room so they can be used in the audio rendering */
   livekitRoomItems$: Behavior<LivekitRoomItem[]>;
-  userMedia$: Behavior<UserMedia[]>;
   /** use the layout instead, this is just for the sdk export. */
   matrixLivekitMembers$: Behavior<RemoteMatrixLivekitMember[]>;
   localMatrixLivekitMember$: Behavior<LocalMatrixLivekitMember | null>;
@@ -334,10 +331,6 @@ export interface CallViewModel {
   gridMode$: Behavior<GridMode>;
   setGridMode: (value: GridMode) => void;
 
-  // media view models and layout
-  grid$: Behavior<UserMediaViewModel[]>;
-  spotlight$: Behavior<MediaViewModel[]>;
-  pip$: Behavior<UserMediaViewModel | null>;
   /**
    * The layout of tiles in the call interface.
    */
@@ -721,7 +714,7 @@ export function createCallViewModel$(
   /**
    * List of user media (camera feeds) that we want tiles for.
    */
-  const userMedia$ = scope.behavior<UserMedia[]>(
+  const userMedia$ = scope.behavior<WrappedUserMediaViewModel[]>(
     combineLatest([
       localMatrixLivekitMember$,
       matrixLivekitMembers$,
@@ -767,36 +760,35 @@ export function createCallViewModel$(
             }
           }
         },
-        (scope, _, dup, mediaId, userId, participant, connection$, rtcId) => {
-          const livekitRoom$ = scope.behavior(
-            connection$.pipe(map((c) => c?.livekitRoom)),
-          );
-          const focusUrl$ = scope.behavior(
-            connection$.pipe(map((c) => c?.transport.livekit_service_url)),
-          );
-          const displayName$ = scope.behavior(
-            matrixMemberMetadataStore
-              .createDisplayNameBehavior$(userId)
-              .pipe(map((name) => name ?? userId)),
-          );
-
-          return new UserMedia(
-            scope,
-            `${mediaId}:${dup}`,
+        (scope, _, dup, mediaId, userId, participant, connection$, rtcId) =>
+          createWrappedUserMedia(scope, {
+            id: `${mediaId}:${dup}`,
             userId,
-            rtcId,
+            rtcBackendIdentity: rtcId,
             participant,
-            options.encryptionSystem,
-            livekitRoom$,
-            focusUrl$,
+            encryptionSystem: options.encryptionSystem,
+            livekitRoom$: scope.behavior(
+              connection$.pipe(map((c) => c?.livekitRoom)),
+            ),
+            focusUrl$: scope.behavior(
+              connection$.pipe(map((c) => c?.transport.livekit_service_url)),
+            ),
             mediaDevices,
-            localMembership.reconnecting$,
-            displayName$,
-            matrixMemberMetadataStore.createAvatarUrlBehavior$(userId),
-            handsRaised$.pipe(map((v) => v[mediaId]?.time ?? null)),
-            reactions$.pipe(map((v) => v[mediaId] ?? undefined)),
-          );
-        },
+            pretendToBeDisconnected$: localMembership.reconnecting$,
+            displayName$: scope.behavior(
+              matrixMemberMetadataStore
+                .createDisplayNameBehavior$(userId)
+                .pipe(map((name) => name ?? userId)),
+            ),
+            mxcAvatarUrl$:
+              matrixMemberMetadataStore.createAvatarUrlBehavior$(userId),
+            handRaised$: scope.behavior(
+              handsRaised$.pipe(map((v) => v[mediaId]?.time ?? null)),
+            ),
+            reaction$: scope.behavior(
+              reactions$.pipe(map((v) => v[mediaId] ?? undefined)),
+            ),
+          }),
       ),
     ),
   );
@@ -821,11 +813,9 @@ export function createCallViewModel$(
   /**
    * List of MediaItems that we want to display, that are of type ScreenShare
    */
-  const screenShares$ = scope.behavior<ScreenShare[]>(
+  const screenShares$ = scope.behavior<ScreenShareViewModel[]>(
     mediaItems$.pipe(
-      map((mediaItems) =>
-        mediaItems.filter((m): m is ScreenShare => m instanceof ScreenShare),
-      ),
+      map((mediaItems) => mediaItems.filter((m) => m.type === "screen share")),
     ),
   );
 
@@ -888,39 +878,39 @@ export function createCallViewModel$(
       merge(userHangup$, widgetHangup$).pipe(map(() => "user" as const)),
     ).pipe(scope.share);
 
-  const spotlightSpeaker$ = scope.behavior<UserMediaViewModel | null>(
+  const spotlightSpeaker$ = scope.behavior<UserMediaViewModel | undefined>(
     userMedia$.pipe(
       switchMap((mediaItems) =>
         mediaItems.length === 0
           ? of([])
           : combineLatest(
               mediaItems.map((m) =>
-                m.vm.speaking$.pipe(map((s) => [m, s] as const)),
+                m.speaking$.pipe(map((s) => [m, s] as const)),
               ),
             ),
       ),
-      scan<(readonly [UserMedia, boolean])[], UserMedia | undefined, null>(
-        (prev, mediaItems) => {
-          // Only remote users that are still in the call should be sticky
-          const [stickyMedia, stickySpeaking] =
-            (!prev?.vm.local && mediaItems.find(([m]) => m === prev)) || [];
-          // Decide who to spotlight:
-          // If the previous speaker is still speaking, stick with them rather
-          // than switching eagerly to someone else
-          return stickySpeaking
-            ? stickyMedia!
-            : // Otherwise, select any remote user who is speaking
-              (mediaItems.find(([m, s]) => !m.vm.local && s)?.[0] ??
-                // Otherwise, stick with the person who was last speaking
-                stickyMedia ??
-                // Otherwise, spotlight an arbitrary remote user
-                mediaItems.find(([m]) => !m.vm.local)?.[0] ??
-                // Otherwise, spotlight the local user
-                mediaItems.find(([m]) => m.vm.local)?.[0]);
-        },
-        null,
-      ),
-      map((speaker) => speaker?.vm ?? null),
+      scan<
+        (readonly [UserMediaViewModel, boolean])[],
+        UserMediaViewModel | undefined,
+        undefined
+      >((prev, mediaItems) => {
+        // Only remote users that are still in the call should be sticky
+        const [stickyMedia, stickySpeaking] =
+          (!prev?.local && mediaItems.find(([m]) => m === prev)) || [];
+        // Decide who to spotlight:
+        // If the previous speaker is still speaking, stick with them rather
+        // than switching eagerly to someone else
+        return stickySpeaking
+          ? stickyMedia!
+          : // Otherwise, select any remote user who is speaking
+            (mediaItems.find(([m, s]) => !m.local && s)?.[0] ??
+              // Otherwise, stick with the person who was last speaking
+              stickyMedia ??
+              // Otherwise, spotlight an arbitrary remote user
+              mediaItems.find(([m]) => !m.local)?.[0] ??
+              // Otherwise, spotlight the local user
+              mediaItems.find(([m]) => m.local)?.[0]);
+      }, undefined),
     ),
   );
 
@@ -934,7 +924,7 @@ export function createCallViewModel$(
         return bins.length === 0
           ? of([])
           : combineLatest(bins, (...bins) =>
-              bins.sort(([, bin1], [, bin2]) => bin1 - bin2).map(([m]) => m.vm),
+              bins.sort(([, bin1], [, bin2]) => bin1 - bin2).map(([m]) => m),
             );
       }),
       distinctUntilChanged(shallowEquals),
@@ -944,9 +934,7 @@ export function createCallViewModel$(
   const spotlight$ = scope.behavior<MediaViewModel[]>(
     screenShares$.pipe(
       switchMap((screenShares) => {
-        if (screenShares.length > 0) {
-          return of(screenShares.map((m) => m.vm));
-        }
+        if (screenShares.length > 0) return of(screenShares);
 
         return spotlightSpeaker$.pipe(
           map((speaker) => (speaker ? [speaker] : [])),
@@ -956,7 +944,7 @@ export function createCallViewModel$(
     ),
   );
 
-  const pip$ = scope.behavior<UserMediaViewModel | null>(
+  const pip$ = scope.behavior<UserMediaViewModel | undefined>(
     combineLatest([
       // TODO This also needs epoch logic to dedupe the screenshares and mediaItems emits
       screenShares$,
@@ -968,28 +956,17 @@ export function createCallViewModel$(
           return spotlightSpeaker$;
         }
         if (!spotlight || spotlight.local) {
-          return of(null);
+          return of(undefined);
         }
 
         const localUserMedia = mediaItems.find(
-          (m) => m.vm instanceof LocalUserMediaViewModel,
-        ) as UserMedia | undefined;
-
-        const localUserMediaViewModel = localUserMedia?.vm as
-          | LocalUserMediaViewModel
-          | undefined;
-
-        if (!localUserMediaViewModel) {
-          return of(null);
+          (m) => m.type === "user" && m.local,
+        );
+        if (!localUserMedia) {
+          return of(undefined);
         }
-        return localUserMediaViewModel.alwaysShow$.pipe(
-          map((alwaysShow) => {
-            if (alwaysShow) {
-              return localUserMediaViewModel;
-            }
-
-            return null;
-          }),
+        return localUserMedia.alwaysShow$.pipe(
+          map((alwaysShow) => (alwaysShow ? localUserMedia : undefined)),
         );
       }),
     ),
@@ -998,7 +975,7 @@ export function createCallViewModel$(
   const hasRemoteScreenShares$ = scope.behavior<boolean>(
     spotlight$.pipe(
       map((spotlight) =>
-        spotlight.some((vm) => !vm.local && vm instanceof ScreenShareViewModel),
+        spotlight.some((vm) => vm.type === "screen share" && !vm.local),
       ),
     ),
   );
@@ -1039,8 +1016,10 @@ export function createCallViewModel$(
   );
 
   const spotlightExpandedToggle$ = new Subject<void>();
-  const spotlightExpanded$ = scope.behavior<boolean>(
-    spotlightExpandedToggle$.pipe(accumulate(false, (expanded) => !expanded)),
+  const spotlightExpanded$ = createToggle$(
+    scope,
+    false,
+    spotlightExpandedToggle$,
   );
 
   const { setGridMode, gridMode$ } = createLayoutModeSwitch(
@@ -1053,7 +1032,7 @@ export function createCallViewModel$(
     [grid$, spotlight$],
     (grid, spotlight) => ({
       type: "grid",
-      spotlight: spotlight.some((vm) => vm instanceof ScreenShareViewModel)
+      spotlight: spotlight.some((vm) => vm.type === "screen share")
         ? spotlight
         : undefined,
       grid,
@@ -1085,12 +1064,8 @@ export function createCallViewModel$(
     mediaItems$.pipe(
       map((mediaItems) => {
         if (mediaItems.length !== 2) return null;
-        const local = mediaItems.find((vm) => vm.vm.local)?.vm as
-          | LocalUserMediaViewModel
-          | undefined;
-        const remote = mediaItems.find((vm) => !vm.vm.local)?.vm as
-          | RemoteUserMediaViewModel
-          | undefined;
+        const local = mediaItems.find((vm) => vm.type === "user" && vm.local);
+        const remote = mediaItems.find((vm) => vm.type === "user" && !vm.local);
         // There might not be a remote tile if there are screen shares, or if
         // only the local user is in the call and they're using the duplicate
         // tiles option
@@ -1138,7 +1113,7 @@ export function createCallViewModel$(
                 oneOnOne === null
                   ? combineLatest([grid$, spotlight$], (grid, spotlight) =>
                       grid.length > smallMobileCallThreshold ||
-                      spotlight.some((vm) => vm instanceof ScreenShareViewModel)
+                      spotlight.some((vm) => vm.type === "screen share")
                         ? spotlightPortraitLayoutMedia$
                         : gridLayoutMedia$,
                     ).pipe(switchAll())
@@ -1245,7 +1220,7 @@ export function createCallViewModel$(
             // screen sharing feeds are in the spotlight we still need them.
             return l.spotlight.media$.pipe(
               map((models: MediaViewModel[]) =>
-                models.some((m) => m instanceof ScreenShareViewModel),
+                models.some((m) => m.type === "screen share"),
               ),
             );
           // In expanded spotlight layout, the active speaker is always shown in
@@ -1296,7 +1271,7 @@ export function createCallViewModel$(
       switchMap((mode) => {
         switch (mode) {
           case "pip":
-            return of(false);
+            return of(platform === "desktop" ? true : false);
           case "normal":
           case "narrow":
             return of(true);
@@ -1552,11 +1527,7 @@ export function createCallViewModel$(
     toggleSpotlightExpanded$: toggleSpotlightExpanded$,
     gridMode$: gridMode$,
     setGridMode: setGridMode,
-    grid$: grid$,
-    spotlight$: spotlight$,
-    pip$: pip$,
     layout$: layout$,
-    userMedia$,
     localMatrixLivekitMember$,
     matrixLivekitMembers$: scope.behavior(
       matrixLivekitMembers$.pipe(

src/state/CallViewModel/localMember/Publisher.ts

@@ -30,7 +30,7 @@ import {
   trackProcessorSync,
 } from "../../../livekit/TrackProcessorContext.tsx";
 import { getUrlParams } from "../../../UrlParams.ts";
-import { observeTrackReference$ } from "../../MediaViewModel.ts";
+import { observeTrackReference$ } from "../../observeTrackReference";
 import { type Connection } from "../remoteMembers/Connection.ts";
 import { ObservableScope } from "../../ObservableScope.ts";
 

src/state/MediaViewModel.ts

@@ -1,807 +0,0 @@
-/*
-Copyright 2023, 2024 New Vector Ltd.
-
-SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
-Please see LICENSE in the repository root for full details.
-*/
-
-import {
-  type AudioSource,
-  type VideoSource,
-  type TrackReference,
-  observeParticipantEvents,
-  observeParticipantMedia,
-  roomEventSelector,
-} from "@livekit/components-core";
-import {
-  type LocalParticipant,
-  LocalTrack,
-  LocalVideoTrack,
-  type Participant,
-  ParticipantEvent,
-  type RemoteParticipant,
-  Track,
-  TrackEvent,
-  facingModeFromLocalTrack,
-  type Room as LivekitRoom,
-  RoomEvent as LivekitRoomEvent,
-  RemoteTrack,
-} from "livekit-client";
-import { logger } from "matrix-js-sdk/lib/logger";
-import {
-  BehaviorSubject,
-  type Observable,
-  Subject,
-  combineLatest,
-  filter,
-  fromEvent,
-  interval,
-  map,
-  merge,
-  of,
-  startWith,
-  switchMap,
-  throttleTime,
-  distinctUntilChanged,
-} from "rxjs";
-
-import { alwaysShowSelf } from "../settings/settings";
-import { showConnectionStats } from "../settings/settings";
-import { accumulate } from "../utils/observable";
-import { type EncryptionSystem } from "../e2ee/sharedKeyManagement";
-import { E2eeType } from "../e2ee/e2eeType";
-import { type ReactionOption } from "../reactions";
-import { platform } from "../Platform";
-import { type MediaDevices } from "./MediaDevices";
-import { type Behavior } from "./Behavior";
-import { type ObservableScope } from "./ObservableScope";
-
-export function observeTrackReference$(
-  participant: Participant,
-  source: Track.Source,
-): Observable<TrackReference | undefined> {
-  return observeParticipantMedia(participant).pipe(
-    map(() => participant.getTrackPublication(source)),
-    distinctUntilChanged(),
-    map((publication) => publication && { participant, publication, source }),
-  );
-}
-
-export function observeRtpStreamStats$(
-  participant: Participant,
-  source: Track.Source,
-  type: "inbound-rtp" | "outbound-rtp",
-): Observable<
-  RTCInboundRtpStreamStats | RTCOutboundRtpStreamStats | undefined
-> {
-  return combineLatest([
-    observeTrackReference$(participant, source),
-    interval(1000).pipe(startWith(0)),
-  ]).pipe(
-    switchMap(async ([trackReference]) => {
-      const track = trackReference?.publication?.track;
-      if (
-        !track ||
-        !(track instanceof RemoteTrack || track instanceof LocalTrack)
-      ) {
-        return undefined;
-      }
-      const report = await track.getRTCStatsReport();
-      if (!report) {
-        return undefined;
-      }
-
-      for (const v of report.values()) {
-        if (v.type === type) {
-          return v;
-        }
-      }
-
-      return undefined;
-    }),
-    startWith(undefined),
-  );
-}
-
-export function observeInboundRtpStreamStats$(
-  participant: Participant,
-  source: Track.Source,
-): Observable<RTCInboundRtpStreamStats | undefined> {
-  return observeRtpStreamStats$(participant, source, "inbound-rtp").pipe(
-    map((x) => x as RTCInboundRtpStreamStats | undefined),
-  );
-}
-
-export function observeOutboundRtpStreamStats$(
-  participant: Participant,
-  source: Track.Source,
-): Observable<RTCOutboundRtpStreamStats | undefined> {
-  return observeRtpStreamStats$(participant, source, "outbound-rtp").pipe(
-    map((x) => x as RTCOutboundRtpStreamStats | undefined),
-  );
-}
-
-function observeRemoteTrackReceivingOkay$(
-  participant: Participant,
-  source: Track.Source,
-): Observable<boolean | undefined> {
-  let lastStats: {
-    framesDecoded: number | undefined;
-    framesDropped: number | undefined;
-    framesReceived: number | undefined;
-  } = {
-    framesDecoded: undefined,
-    framesDropped: undefined,
-    framesReceived: undefined,
-  };
-
-  return observeInboundRtpStreamStats$(participant, source).pipe(
-    map((stats) => {
-      if (!stats) return undefined;
-      const { framesDecoded, framesDropped, framesReceived } = stats;
-      return {
-        framesDecoded,
-        framesDropped,
-        framesReceived,
-      };
-    }),
-    filter((newStats) => !!newStats),
-    map((newStats): boolean | undefined => {
-      const oldStats = lastStats;
-      lastStats = newStats;
-      if (
-        typeof newStats.framesReceived === "number" &&
-        typeof oldStats.framesReceived === "number" &&
-        typeof newStats.framesDecoded === "number" &&
-        typeof oldStats.framesDecoded === "number"
-      ) {
-        const framesReceivedDelta =
-          newStats.framesReceived - oldStats.framesReceived;
-        const framesDecodedDelta =
-          newStats.framesDecoded - oldStats.framesDecoded;
-
-        // if we received >0 frames and managed to decode >0 frames then we treat that as success
-
-        if (framesReceivedDelta > 0) {
-          return framesDecodedDelta > 0;
-        }
-      }
-
-      // no change
-      return undefined;
-    }),
-    filter((x) => typeof x === "boolean"),
-    startWith(undefined),
-  );
-}
-
-function encryptionErrorObservable$(
-  room$: Behavior<LivekitRoom | undefined>,
-  participant: Participant,
-  encryptionSystem: EncryptionSystem,
-  criteria: string,
-): Observable<boolean> {
-  return room$.pipe(
-    switchMap((room) => {
-      if (room === undefined) return of(false);
-      return roomEventSelector(room, LivekitRoomEvent.EncryptionError).pipe(
-        map((e) => {
-          const [err] = e;
-          if (encryptionSystem.kind === E2eeType.PER_PARTICIPANT) {
-            return (
-              // Ideally we would pull the participant identity from the field on the error.
-              // However, it gets lost in the serialization process between workers.
-              // So, instead we do a string match
-              (err?.message.includes(participant.identity) &&
-                err?.message.includes(criteria)) ??
-              false
-            );
-          } else if (encryptionSystem.kind === E2eeType.SHARED_KEY) {
-            return !!err?.message.includes(criteria);
-          }
-
-          return false;
-        }),
-      );
-    }),
-    distinctUntilChanged(),
-    throttleTime(1000), // Throttle to avoid spamming the UI
-    startWith(false),
-  );
-}
-
-export enum EncryptionStatus {
-  Connecting,
-  Okay,
-  KeyMissing,
-  KeyInvalid,
-  PasswordInvalid,
-}
-
-abstract class BaseMediaViewModel {
-  /**
-   * The LiveKit video track for this media.
-   */
-  public readonly video$: Behavior<TrackReference | undefined>;
-  /**
-   * Whether there should be a warning that this media is unencrypted.
-   */
-  public readonly unencryptedWarning$: Behavior<boolean>;
-
-  public readonly encryptionStatus$: Behavior<EncryptionStatus>;
-
-  /**
-   * Whether this media corresponds to the local participant.
-   */
-  public abstract readonly local: boolean;
-
-  private observeTrackReference$(
-    source: Track.Source,
-  ): Behavior<TrackReference | undefined> {
-    return this.scope.behavior(
-      this.participant$.pipe(
-        switchMap((p) =>
-          !p ? of(undefined) : observeTrackReference$(p, source),
-        ),
-      ),
-    );
-  }
-
-  public constructor(
-    protected readonly scope: ObservableScope,
-    /**
-     * An opaque identifier for this media.
-     */
-    public readonly id: string,
-    /**
-     * The Matrix user to which this media belongs.
-     */
-    public readonly userId: string,
-    public readonly rtcBackendIdentity: string,
-    // We don't necessarily have a participant if a user connects via MatrixRTC but not (yet) through
-    // livekit.
-    protected readonly participant$: Observable<
-      LocalParticipant | RemoteParticipant | null
-    >,
-
-    encryptionSystem: EncryptionSystem,
-    audioSource: AudioSource,
-    videoSource: VideoSource,
-    protected readonly livekitRoom$: Behavior<LivekitRoom | undefined>,
-    public readonly focusUrl$: Behavior<string | undefined>,
-    public readonly displayName$: Behavior<string>,
-    public readonly mxcAvatarUrl$: Behavior<string | undefined>,
-  ) {
-    const audio$ = this.observeTrackReference$(audioSource);
-    this.video$ = this.observeTrackReference$(videoSource);
-
-    this.unencryptedWarning$ = this.scope.behavior(
-      combineLatest(
-        [audio$, this.video$],
-        (a, v) =>
-          encryptionSystem.kind !== E2eeType.NONE &&
-          (a?.publication.isEncrypted === false ||
-            v?.publication.isEncrypted === false),
-      ),
-    );
-
-    this.encryptionStatus$ = this.scope.behavior(
-      this.participant$.pipe(
-        switchMap((participant): Observable<EncryptionStatus> => {
-          if (!participant) {
-            return of(EncryptionStatus.Connecting);
-          } else if (
-            participant.isLocal ||
-            encryptionSystem.kind === E2eeType.NONE
-          ) {
-            return of(EncryptionStatus.Okay);
-          } else if (encryptionSystem.kind === E2eeType.PER_PARTICIPANT) {
-            return combineLatest([
-              encryptionErrorObservable$(
-                livekitRoom$,
-                participant,
-                encryptionSystem,
-                "MissingKey",
-              ),
-              encryptionErrorObservable$(
-                livekitRoom$,
-                participant,
-                encryptionSystem,
-                "InvalidKey",
-              ),
-              observeRemoteTrackReceivingOkay$(participant, audioSource),
-              observeRemoteTrackReceivingOkay$(participant, videoSource),
-            ]).pipe(
-              map(([keyMissing, keyInvalid, audioOkay, videoOkay]) => {
-                if (keyMissing) return EncryptionStatus.KeyMissing;
-                if (keyInvalid) return EncryptionStatus.KeyInvalid;
-                if (audioOkay || videoOkay) return EncryptionStatus.Okay;
-                return undefined; // no change
-              }),
-              filter((x) => !!x),
-              startWith(EncryptionStatus.Connecting),
-            );
-          } else {
-            return combineLatest([
-              encryptionErrorObservable$(
-                livekitRoom$,
-                participant,
-                encryptionSystem,
-                "InvalidKey",
-              ),
-              observeRemoteTrackReceivingOkay$(participant, audioSource),
-              observeRemoteTrackReceivingOkay$(participant, videoSource),
-            ]).pipe(
-              map(
-                ([keyInvalid, audioOkay, videoOkay]):
-                  | EncryptionStatus
-                  | undefined => {
-                  if (keyInvalid) return EncryptionStatus.PasswordInvalid;
-                  if (audioOkay || videoOkay) return EncryptionStatus.Okay;
-                  return undefined; // no change
-                },
-              ),
-              filter((x) => !!x),
-              startWith(EncryptionStatus.Connecting),
-            );
-          }
-        }),
-      ),
-    );
-  }
-}
-
-/**
- * Some participant's media.
- */
-export type MediaViewModel = UserMediaViewModel | ScreenShareViewModel;
-export type UserMediaViewModel =
-  | LocalUserMediaViewModel
-  | RemoteUserMediaViewModel;
-
-/**
- * Some participant's user media.
- */
-abstract class BaseUserMediaViewModel extends BaseMediaViewModel {
-  private readonly _speaking$ = this.scope.behavior(
-    this.participant$.pipe(
-      switchMap((p) =>
-        p
-          ? observeParticipantEvents(
-              p,
-              ParticipantEvent.IsSpeakingChanged,
-            ).pipe(map((p) => p.isSpeaking))
-          : of(false),
-      ),
-    ),
-  );
-  /**
-   * Whether the participant is speaking.
-   */
-  // Getter backed by a private field so that subclasses can override it
-  public get speaking$(): Behavior<boolean> {
-    return this._speaking$;
-  }
-
-  /**
-   * Whether this participant is sending audio (i.e. is unmuted on their side).
-   */
-  public readonly audioEnabled$: Behavior<boolean>;
-
-  private readonly _videoEnabled$: Behavior<boolean>;
-  /**
-   * Whether this participant is sending video.
-   */
-  // Getter backed by a private field so that subclasses can override it
-  public get videoEnabled$(): Behavior<boolean> {
-    return this._videoEnabled$;
-  }
-
-  private readonly _cropVideo$ = new BehaviorSubject(true);
-  /**
-   * Whether the tile video should be contained inside the tile or be cropped to fit.
-   */
-  public readonly cropVideo$: Behavior<boolean> = this._cropVideo$;
-
-  public constructor(
-    scope: ObservableScope,
-    id: string,
-    userId: string,
-    rtcBackendIdentity: string,
-    participant$: Observable<LocalParticipant | RemoteParticipant | null>,
-    encryptionSystem: EncryptionSystem,
-    livekitRoom$: Behavior<LivekitRoom | undefined>,
-    focusUrl$: Behavior<string | undefined>,
-    displayName$: Behavior<string>,
-    mxcAvatarUrl$: Behavior<string | undefined>,
-    public readonly handRaised$: Behavior<Date | null>,
-    public readonly reaction$: Behavior<ReactionOption | null>,
-  ) {
-    super(
-      scope,
-      id,
-      userId,
-      rtcBackendIdentity,
-      participant$,
-      encryptionSystem,
-      Track.Source.Microphone,
-      Track.Source.Camera,
-      livekitRoom$,
-      focusUrl$,
-      displayName$,
-      mxcAvatarUrl$,
-    );
-
-    const media$ = this.scope.behavior(
-      participant$.pipe(
-        switchMap((p) => (p && observeParticipantMedia(p)) ?? of(undefined)),
-      ),
-    );
-    this.audioEnabled$ = this.scope.behavior(
-      media$.pipe(map((m) => m?.microphoneTrack?.isMuted === false)),
-    );
-    this._videoEnabled$ = this.scope.behavior(
-      media$.pipe(map((m) => m?.cameraTrack?.isMuted === false)),
-    );
-  }
-
-  public toggleFitContain(): void {
-    this._cropVideo$.next(!this._cropVideo$.value);
-  }
-
-  public get local(): boolean {
-    return this instanceof LocalUserMediaViewModel;
-  }
-
-  public abstract get audioStreamStats$(): Observable<
-    RTCInboundRtpStreamStats | RTCOutboundRtpStreamStats | undefined
-  >;
-  public abstract get videoStreamStats$(): Observable<
-    RTCInboundRtpStreamStats | RTCOutboundRtpStreamStats | undefined
-  >;
-}
-
-/**
- * The local participant's user media.
- */
-export class LocalUserMediaViewModel extends BaseUserMediaViewModel {
-  /**
-   * The local video track as an observable that emits whenever the track
-   * changes, the camera is switched, or the track is muted.
-   */
-  private readonly videoTrack$: Observable<LocalVideoTrack | null> =
-    this.video$.pipe(
-      switchMap((v) => {
-        const track = v?.publication.track;
-        if (!(track instanceof LocalVideoTrack)) return of(null);
-        return merge(
-          // Watch for track restarts because they indicate a camera switch.
-          // This event is also emitted when unmuting the track object.
-          fromEvent(track, TrackEvent.Restarted).pipe(
-            startWith(null),
-            map(() => track),
-          ),
-          // When the track object is muted, reset it to null.
-          fromEvent(track, TrackEvent.Muted).pipe(map(() => null)),
-        );
-      }),
-    );
-
-  /**
-   * Whether the video should be mirrored.
-   */
-  public readonly mirror$ = this.scope.behavior(
-    this.videoTrack$.pipe(
-      // Mirror only front-facing cameras (those that face the user)
-      map(
-        (track) =>
-          track !== null &&
-          facingModeFromLocalTrack(track).facingMode === "user",
-      ),
-    ),
-  );
-
-  /**
-   * Whether to show this tile in a highly visible location near the start of
-   * the grid.
-   */
-  public readonly alwaysShow$ = alwaysShowSelf.value$;
-  public readonly setAlwaysShow = alwaysShowSelf.setValue;
-
-  /**
-   * Callback for switching between the front and back cameras.
-   */
-  public readonly switchCamera$: Behavior<(() => void) | null> =
-    this.scope.behavior(
-      platform === "desktop"
-        ? of(null)
-        : this.videoTrack$.pipe(
-            map((track) => {
-              if (track === null) return null;
-              const facingMode = facingModeFromLocalTrack(track).facingMode;
-              // If the camera isn't front or back-facing, don't provide a switch
-              // camera shortcut at all
-              if (facingMode !== "user" && facingMode !== "environment")
-                return null;
-              // Restart the track with a camera facing the opposite direction
-              return (): void =>
-                void track
-                  .restartTrack({
-                    facingMode: facingMode === "user" ? "environment" : "user",
-                  })
-                  .then(() => {
-                    // Inform the MediaDevices which camera was chosen
-                    const deviceId =
-                      track.mediaStreamTrack.getSettings().deviceId;
-                    if (deviceId !== undefined)
-                      this.mediaDevices.videoInput.select(deviceId);
-                  })
-                  .catch((e) =>
-                    logger.error("Failed to switch camera", facingMode, e),
-                  );
-            }),
-          ),
-    );
-
-  public constructor(
-    scope: ObservableScope,
-    id: string,
-    userId: string,
-    rtcBackendIdentity: string,
-    participant$: Behavior<LocalParticipant | null>,
-    encryptionSystem: EncryptionSystem,
-    livekitRoom$: Behavior<LivekitRoom | undefined>,
-    focusUrl$: Behavior<string | undefined>,
-    private readonly mediaDevices: MediaDevices,
-    displayName$: Behavior<string>,
-    mxcAvatarUrl$: Behavior<string | undefined>,
-    handRaised$: Behavior<Date | null>,
-    reaction$: Behavior<ReactionOption | null>,
-  ) {
-    super(
-      scope,
-      id,
-      userId,
-      rtcBackendIdentity,
-      participant$,
-      encryptionSystem,
-      livekitRoom$,
-      focusUrl$,
-      displayName$,
-      mxcAvatarUrl$,
-      handRaised$,
-      reaction$,
-    );
-  }
-
-  public audioStreamStats$ = combineLatest([
-    this.participant$,
-    showConnectionStats.value$,
-  ]).pipe(
-    switchMap(([p, showConnectionStats]) => {
-      if (!p || !showConnectionStats) return of(undefined);
-      return observeOutboundRtpStreamStats$(p, Track.Source.Microphone);
-    }),
-  );
-
-  public videoStreamStats$ = combineLatest([
-    this.participant$,
-    showConnectionStats.value$,
-  ]).pipe(
-    switchMap(([p, showConnectionStats]) => {
-      if (!p || !showConnectionStats) return of(undefined);
-      return observeOutboundRtpStreamStats$(p, Track.Source.Camera);
-    }),
-  );
-}
-
-/**
- * A remote participant's user media.
- */
-export class RemoteUserMediaViewModel extends BaseUserMediaViewModel {
-  /**
-   * Whether we are waiting for this user's LiveKit participant to exist. This
-   * could be because either we or the remote party are still connecting.
-   */
-  public readonly waitingForMedia$ = this.scope.behavior<boolean>(
-    combineLatest(
-      [this.livekitRoom$, this.participant$],
-      (livekitRoom, participant) =>
-        // If livekitRoom is undefined, the user is not attempting to publish on
-        // any transport and so we shouldn't expect a participant. (They might
-        // be a subscribe-only bot for example.)
-        livekitRoom !== undefined && participant === null,
-    ),
-  );
-
-  // This private field is used to override the value from the superclass
-  private __speaking$: Behavior<boolean>;
-  public get speaking$(): Behavior<boolean> {
-    return this.__speaking$;
-  }
-
-  private readonly locallyMutedToggle$ = new Subject<void>();
-  private readonly localVolumeAdjustment$ = new Subject<number>();
-  private readonly localVolumeCommit$ = new Subject<void>();
-
-  /**
-   * The volume to which this participant's audio is set, as a scalar
-   * multiplier.
-   */
-  public readonly localVolume$ = this.scope.behavior<number>(
-    merge(
-      this.locallyMutedToggle$.pipe(map(() => "toggle mute" as const)),
-      this.localVolumeAdjustment$,
-      this.localVolumeCommit$.pipe(map(() => "commit" as const)),
-    ).pipe(
-      accumulate({ volume: 1, committedVolume: 1 }, (state, event) => {
-        switch (event) {
-          case "toggle mute":
-            return {
-              ...state,
-              volume: state.volume === 0 ? state.committedVolume : 0,
-            };
-          case "commit":
-            // Dragging the slider to zero should have the same effect as
-            // muting: keep the original committed volume, as if it were never
-            // dragged
-            return {
-              ...state,
-              committedVolume:
-                state.volume === 0 ? state.committedVolume : state.volume,
-            };
-          default:
-            // Volume adjustment
-            return { ...state, volume: event };
-        }
-      }),
-      map(({ volume }) => volume),
-    ),
-  );
-
-  // This private field is used to override the value from the superclass
-  private __videoEnabled$: Behavior<boolean>;
-  public get videoEnabled$(): Behavior<boolean> {
-    return this.__videoEnabled$;
-  }
-
-  /**
-   * Whether this participant's audio is disabled.
-   */
-  public readonly locallyMuted$ = this.scope.behavior<boolean>(
-    this.localVolume$.pipe(map((volume) => volume === 0)),
-  );
-
-  public constructor(
-    scope: ObservableScope,
-    id: string,
-    userId: string,
-    rtcBackendIdentity: string,
-    participant$: Observable<RemoteParticipant | null>,
-    encryptionSystem: EncryptionSystem,
-    livekitRoom$: Behavior<LivekitRoom | undefined>,
-    focusUrl$: Behavior<string | undefined>,
-    private readonly pretendToBeDisconnected$: Behavior<boolean>,
-    displayName$: Behavior<string>,
-    mxcAvatarUrl$: Behavior<string | undefined>,
-    handRaised$: Behavior<Date | null>,
-    reaction$: Behavior<ReactionOption | null>,
-  ) {
-    super(
-      scope,
-      id,
-      userId,
-      rtcBackendIdentity,
-      participant$,
-      encryptionSystem,
-      livekitRoom$,
-      focusUrl$,
-      displayName$,
-      mxcAvatarUrl$,
-      handRaised$,
-      reaction$,
-    );
-
-    this.__speaking$ = this.scope.behavior(
-      pretendToBeDisconnected$.pipe(
-        switchMap((disconnected) =>
-          disconnected ? of(false) : super.speaking$,
-        ),
-      ),
-    );
-
-    this.__videoEnabled$ = this.scope.behavior(
-      pretendToBeDisconnected$.pipe(
-        switchMap((disconnected) =>
-          disconnected ? of(false) : super.videoEnabled$,
-        ),
-      ),
-    );
-
-    // Sync the local volume with LiveKit
-    combineLatest([
-      participant$,
-      // The local volume, taking into account whether we're supposed to pretend
-      // that the audio stream is disconnected (since we don't necessarily want
-      // that to modify the UI state).
-      this.pretendToBeDisconnected$.pipe(
-        switchMap((disconnected) => (disconnected ? of(0) : this.localVolume$)),
-        this.scope.bind(),
-      ),
-    ]).subscribe(([p, volume]) => p?.setVolume(volume));
-  }
-
-  public toggleLocallyMuted(): void {
-    this.locallyMutedToggle$.next();
-  }
-
-  public setLocalVolume(value: number): void {
-    this.localVolumeAdjustment$.next(value);
-  }
-
-  public commitLocalVolume(): void {
-    this.localVolumeCommit$.next();
-  }
-
-  public audioStreamStats$ = combineLatest([
-    this.participant$,
-    showConnectionStats.value$,
-  ]).pipe(
-    switchMap(([p, showConnectionStats]) => {
-      if (!p || !showConnectionStats) return of(undefined);
-      return observeInboundRtpStreamStats$(p, Track.Source.Microphone);
-    }),
-  );
-
-  public videoStreamStats$ = combineLatest([
-    this.participant$,
-    showConnectionStats.value$,
-  ]).pipe(
-    switchMap(([p, showConnectionStats]) => {
-      if (!p || !showConnectionStats) return of(undefined);
-      return observeInboundRtpStreamStats$(p, Track.Source.Camera);
-    }),
-  );
-}
-
-/**
- * Some participant's screen share media.
- */
-export class ScreenShareViewModel extends BaseMediaViewModel {
-  /**
-   * Whether this screen share's video should be displayed.
-   */
-  public readonly videoEnabled$ = this.scope.behavior(
-    this.pretendToBeDisconnected$.pipe(map((disconnected) => !disconnected)),
-  );
-
-  public constructor(
-    scope: ObservableScope,
-    id: string,
-    userId: string,
-    rtcBackendIdentity: string,
-    participant$: Observable<LocalParticipant | RemoteParticipant>,
-    encryptionSystem: EncryptionSystem,
-    livekitRoom$: Behavior<LivekitRoom | undefined>,
-    focusUrl$: Behavior<string | undefined>,
-    private readonly pretendToBeDisconnected$: Behavior<boolean>,
-    displayName$: Behavior<string>,
-    mxcAvatarUrl$: Behavior<string | undefined>,
-    public readonly local: boolean,
-  ) {
-    super(
-      scope,
-      id,
-      userId,
-      rtcBackendIdentity,
-      participant$,
-      encryptionSystem,
-      Track.Source.ScreenShareAudio,
-      Track.Source.ScreenShare,
-      livekitRoom$,
-      focusUrl$,
-      displayName$,
-      mxcAvatarUrl$,
-    );
-  }
-}

src/state/PipLayout.ts

@@ -5,6 +5,7 @@ SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
 Please see LICENSE in the repository root for full details.
 */
 
+import { platform } from "../Platform.ts";
 import { type PipLayout, type PipLayoutMedia } from "./layout-types.ts";
 import { type TileStore } from "./TileStore";
 
@@ -16,7 +17,11 @@ export function pipLayout(
   prevTiles: TileStore,
 ): [PipLayout, TileStore] {
   const update = prevTiles.from(0);
-  update.registerSpotlight(media.spotlight, true);
+  // Dont maximise in pip on EW since we want the rounded corners and the footer
+  update.registerSpotlight(
+    media.spotlight,
+    platform === "desktop" ? false : true,
+  );
   const tiles = update.build();
   return [
     {

src/state/ScreenShare.ts

@@ -1,55 +0,0 @@
-/*
-Copyright 2025 New Vector Ltd.
-
-SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
-Please see LICENSE in the repository root for full details.
-*/
-import { of } from "rxjs";
-import {
-  type LocalParticipant,
-  type RemoteParticipant,
-  type Room as LivekitRoom,
-} from "livekit-client";
-
-import { type ObservableScope } from "./ObservableScope.ts";
-import { ScreenShareViewModel } from "./MediaViewModel.ts";
-import type { EncryptionSystem } from "../e2ee/sharedKeyManagement.ts";
-import type { Behavior } from "./Behavior.ts";
-
-/**
- * A screen share media item to be presented in a tile. This is a thin wrapper
- * around ScreenShareViewModel which essentially just establishes an
- * ObservableScope for behaviors that the view model depends on.
- */
-export class ScreenShare {
-  public readonly vm: ScreenShareViewModel;
-
-  public constructor(
-    private readonly scope: ObservableScope,
-    id: string,
-    userId: string,
-    rtcBackendIdentity: string,
-    participant: LocalParticipant | RemoteParticipant,
-    encryptionSystem: EncryptionSystem,
-    livekitRoom$: Behavior<LivekitRoom | undefined>,
-    focusUrl$: Behavior<string | undefined>,
-    pretendToBeDisconnected$: Behavior<boolean>,
-    displayName$: Behavior<string>,
-    mxcAvatarUrl$: Behavior<string | undefined>,
-  ) {
-    this.vm = new ScreenShareViewModel(
-      this.scope,
-      id,
-      userId,
-      rtcBackendIdentity,
-      of(participant),
-      encryptionSystem,
-      livekitRoom$,
-      focusUrl$,
-      pretendToBeDisconnected$,
-      displayName$,
-      mxcAvatarUrl$,
-      participant.isLocal,
-    );
-  }
-}

src/state/TileStore.ts

@@ -8,10 +8,11 @@ Please see LICENSE in the repository root for full details.
 import { BehaviorSubject } from "rxjs";
 import { logger } from "matrix-js-sdk/lib/logger";
 
-import { type MediaViewModel, type UserMediaViewModel } from "./MediaViewModel";
 import { GridTileViewModel, SpotlightTileViewModel } from "./TileViewModel";
 import { fillGaps } from "../utils/iter";
 import { debugTileLayout } from "../settings/settings";
+import { type MediaViewModel } from "./media/MediaViewModel";
+import { type UserMediaViewModel } from "./media/UserMediaViewModel";
 
 function debugEntries(entries: GridTileData[]): string[] {
   return entries.map((e) => e.media.displayName$.value);

src/state/TileViewModel.ts

@@ -5,8 +5,9 @@ SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
 Please see LICENSE in the repository root for full details.
 */
 
-import { type MediaViewModel, type UserMediaViewModel } from "./MediaViewModel";
 import { type Behavior } from "./Behavior";
+import { type MediaViewModel } from "./media/MediaViewModel";
+import { type UserMediaViewModel } from "./media/UserMediaViewModel";
 
 let nextId = 0;
 function createId(): string {

src/state/UserMedia.ts

@@ -1,210 +0,0 @@
-/*
-Copyright 2025 New Vector Ltd.
-
-SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
-Please see LICENSE in the repository root for full details.
-*/
-
-import { combineLatest, map, type Observable, of, switchMap } from "rxjs";
-import {
-  type LocalParticipant,
-  ParticipantEvent,
-  type RemoteParticipant,
-  type Room as LivekitRoom,
-} from "livekit-client";
-import { observeParticipantEvents } from "@livekit/components-core";
-
-import { type ObservableScope } from "./ObservableScope.ts";
-import {
-  LocalUserMediaViewModel,
-  RemoteUserMediaViewModel,
-  type UserMediaViewModel,
-} from "./MediaViewModel.ts";
-import type { Behavior } from "./Behavior.ts";
-import type { EncryptionSystem } from "../e2ee/sharedKeyManagement.ts";
-import type { MediaDevices } from "./MediaDevices.ts";
-import type { ReactionOption } from "../reactions";
-import { observeSpeaker$ } from "./observeSpeaker.ts";
-import { generateItems } from "../utils/observable.ts";
-import { ScreenShare } from "./ScreenShare.ts";
-import { type TaggedParticipant } from "./CallViewModel/remoteMembers/MatrixLivekitMembers.ts";
-
-/**
- * Sorting bins defining the order in which media tiles appear in the layout.
- */
-enum SortingBin {
-  /**
-   * Yourself, when the "always show self" option is on.
-   */
-  SelfAlwaysShown,
-  /**
-   * Participants that are sharing their screen.
-   */
-  Presenters,
-  /**
-   * Participants that have been speaking recently.
-   */
-  Speakers,
-  /**
-   * Participants that have their hand raised.
-   */
-  HandRaised,
-  /**
-   * Participants with video.
-   */
-  Video,
-  /**
-   * Participants not sharing any video.
-   */
-  NoVideo,
-  /**
-   * Yourself, when the "always show self" option is off.
-   */
-  SelfNotAlwaysShown,
-}
-
-/**
- * A user media item to be presented in a tile. This is a thin wrapper around
- * UserMediaViewModel which additionally determines the media item's sorting bin
- * for inclusion in the call layout and tracks associated screen shares.
- */
-export class UserMedia {
-  public readonly vm: UserMediaViewModel =
-    this.participant.type === "local"
-      ? new LocalUserMediaViewModel(
-          this.scope,
-          this.id,
-          this.userId,
-          this.rtcBackendIdentity,
-          this.participant.value$,
-          this.encryptionSystem,
-          this.livekitRoom$,
-          this.focusUrl$,
-          this.mediaDevices,
-          this.displayName$,
-          this.mxcAvatarUrl$,
-          this.scope.behavior(this.handRaised$),
-          this.scope.behavior(this.reaction$),
-        )
-      : new RemoteUserMediaViewModel(
-          this.scope,
-          this.id,
-          this.userId,
-          this.rtcBackendIdentity,
-          this.participant.value$,
-          this.encryptionSystem,
-          this.livekitRoom$,
-          this.focusUrl$,
-          this.pretendToBeDisconnected$,
-          this.displayName$,
-          this.mxcAvatarUrl$,
-          this.scope.behavior(this.handRaised$),
-          this.scope.behavior(this.reaction$),
-        );
-
-  private readonly speaker$ = this.scope.behavior(
-    observeSpeaker$(this.vm.speaking$),
-  );
-
-  // TypeScript needs this widening of the type to happen in a separate statement
-  private readonly participant$: Behavior<
-    LocalParticipant | RemoteParticipant | null
-  > = this.participant.value$;
-
-  /**
-   * All screen share media associated with this user media.
-   */
-  public readonly screenShares$ = this.scope.behavior(
-    this.participant$.pipe(
-      switchMap((p) =>
-        p === null
-          ? of([])
-          : observeParticipantEvents(
-              p,
-              ParticipantEvent.TrackPublished,
-              ParticipantEvent.TrackUnpublished,
-              ParticipantEvent.LocalTrackPublished,
-              ParticipantEvent.LocalTrackUnpublished,
-            ).pipe(
-              // Technically more than one screen share might be possible... our
-              // MediaViewModels don't support it though since they look for a unique
-              // track for the given source. So generateItems here is a bit overkill.
-              generateItems(
-                `${this.id} screenShares$`,
-                function* (p) {
-                  if (p.isScreenShareEnabled)
-                    yield {
-                      keys: ["screen-share"],
-                      data: undefined,
-                    };
-                },
-                (scope, _data$, key) =>
-                  new ScreenShare(
-                    scope,
-                    `${this.id}:${key}`,
-                    this.userId,
-                    this.rtcBackendIdentity,
-                    p,
-                    this.encryptionSystem,
-                    this.livekitRoom$,
-                    this.focusUrl$,
-                    this.pretendToBeDisconnected$,
-                    this.displayName$,
-                    this.mxcAvatarUrl$,
-                  ),
-              ),
-            ),
-      ),
-    ),
-  );
-
-  private readonly presenter$ = this.scope.behavior(
-    this.screenShares$.pipe(map((screenShares) => screenShares.length > 0)),
-  );
-
-  /**
-   * Which sorting bin the media item should be placed in.
-   */
-  // This is exposed here rather than by UserMediaViewModel because it's only
-  // relevant to the layout algorithms; the MediaView component should be
-  // ignorant of this value.
-  public readonly bin$ = combineLatest(
-    [
-      this.speaker$,
-      this.presenter$,
-      this.vm.videoEnabled$,
-      this.vm.handRaised$,
-      this.vm instanceof LocalUserMediaViewModel
-        ? this.vm.alwaysShow$
-        : of(false),
-    ],
-    (speaker, presenter, video, handRaised, alwaysShow) => {
-      if (this.vm.local)
-        return alwaysShow
-          ? SortingBin.SelfAlwaysShown
-          : SortingBin.SelfNotAlwaysShown;
-      else if (presenter) return SortingBin.Presenters;
-      else if (speaker) return SortingBin.Speakers;
-      else if (handRaised) return SortingBin.HandRaised;
-      else if (video) return SortingBin.Video;
-      else return SortingBin.NoVideo;
-    },
-  );
-
-  public constructor(
-    private readonly scope: ObservableScope,
-    public readonly id: string,
-    private readonly userId: string,
-    private readonly rtcBackendIdentity: string,
-    private readonly participant: TaggedParticipant,
-    private readonly encryptionSystem: EncryptionSystem,
-    private readonly livekitRoom$: Behavior<LivekitRoom | undefined>,
-    private readonly focusUrl$: Behavior<string | undefined>,
-    private readonly mediaDevices: MediaDevices,
-    private readonly pretendToBeDisconnected$: Behavior<boolean>,
-    private readonly displayName$: Behavior<string>,
-    private readonly mxcAvatarUrl$: Behavior<string | undefined>,
-    private readonly handRaised$: Observable<Date | null>,
-    private readonly reaction$: Observable<ReactionOption | null>,
-  ) {}
-}

src/state/VolumeControls.ts

@@ -0,0 +1,101 @@
+/*
+Copyright 2026 Element Software Ltd.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import { combineLatest, map, merge, of, Subject, switchMap } from "rxjs";
+
+import { type Behavior } from "./Behavior";
+import { type ObservableScope } from "./ObservableScope";
+import { accumulate } from "../utils/observable";
+
+/**
+ * Controls for audio playback volume.
+ */
+export interface VolumeControls {
+  /**
+   * The volume to which the audio is set, as a scalar multiplier.
+   */
+  playbackVolume$: Behavior<number>;
+  /**
+   * Whether playback of this audio is disabled.
+   */
+  playbackMuted$: Behavior<boolean>;
+  togglePlaybackMuted: () => void;
+  adjustPlaybackVolume: (value: number) => void;
+  commitPlaybackVolume: () => void;
+}
+
+interface VolumeControlsInputs {
+  pretendToBeDisconnected$: Behavior<boolean>;
+  /**
+   * The callback to run to notify the module performing audio playback of the
+   * requested volume.
+   */
+  sink$: Behavior<(volume: number) => void>;
+}
+
+/**
+ * Creates a set of controls for audio playback volume and syncs this with the
+ * audio playback module for the duration of the scope.
+ */
+export function createVolumeControls(
+  scope: ObservableScope,
+  { pretendToBeDisconnected$, sink$ }: VolumeControlsInputs,
+): VolumeControls {
+  const toggleMuted$ = new Subject<"toggle mute">();
+  const adjustVolume$ = new Subject<number>();
+  const commitVolume$ = new Subject<"commit">();
+
+  const playbackVolume$ = scope.behavior<number>(
+    merge(toggleMuted$, adjustVolume$, commitVolume$).pipe(
+      accumulate({ volume: 1, committedVolume: 1 }, (state, event) => {
+        switch (event) {
+          case "toggle mute":
+            return {
+              ...state,
+              volume: state.volume === 0 ? state.committedVolume : 0,
+            };
+          case "commit":
+            // Dragging the slider to zero should have the same effect as
+            // muting: keep the original committed volume, as if it were never
+            // dragged
+            return {
+              ...state,
+              committedVolume:
+                state.volume === 0 ? state.committedVolume : state.volume,
+            };
+          default:
+            // Volume adjustment
+            return { ...state, volume: event };
+        }
+      }),
+      map(({ volume }) => volume),
+    ),
+  );
+
+  // Sync the requested volume with the audio playback module
+  combineLatest([
+    sink$,
+    // The playback volume, taking into account whether we're supposed to
+    // pretend that the audio stream is disconnected (since we don't necessarily
+    // want that to modify the UI state).
+    pretendToBeDisconnected$.pipe(
+      switchMap((disconnected) => (disconnected ? of(0) : playbackVolume$)),
+    ),
+  ])
+    .pipe(scope.bind())
+    .subscribe(([sink, volume]) => sink(volume));
+
+  return {
+    playbackVolume$,
+    playbackMuted$: scope.behavior<boolean>(
+      playbackVolume$.pipe(map((volume) => volume === 0)),
+    ),
+    togglePlaybackMuted: () => toggleMuted$.next("toggle mute"),
+    adjustPlaybackVolume: (value: number) => adjustVolume$.next(value),
+    commitPlaybackVolume: () => commitVolume$.next("commit"),
+  };
+}

src/state/layout-types.ts

@@ -5,14 +5,14 @@ SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
 Please see LICENSE in the repository root for full details.
 */
 
+import { type LocalUserMediaViewModel } from "./media/LocalUserMediaViewModel.ts";
+import { type MediaViewModel } from "./media/MediaViewModel.ts";
+import { type RemoteUserMediaViewModel } from "./media/RemoteUserMediaViewModel.ts";
+import { type UserMediaViewModel } from "./media/UserMediaViewModel.ts";
 import {
   type GridTileViewModel,
   type SpotlightTileViewModel,
 } from "./TileViewModel.ts";
-import {
-  type MediaViewModel,
-  type UserMediaViewModel,
-} from "./MediaViewModel.ts";
 
 export interface GridLayoutMedia {
   type: "grid";
@@ -40,8 +40,8 @@ export interface SpotlightExpandedLayoutMedia {
 
 export interface OneOnOneLayoutMedia {
   type: "one-on-one";
-  local: UserMediaViewModel;
-  remote: UserMediaViewModel;
+  local: LocalUserMediaViewModel;
+  remote: RemoteUserMediaViewModel;
 }
 
 export interface PipLayoutMedia {

src/state/media/LocalScreenShareViewModel.ts

@@ -0,0 +1,32 @@
+/*
+Copyright 2023, 2024 New Vector Ltd.
+Copyright 2026 Element Creations Ltd.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import { type LocalParticipant } from "livekit-client";
+
+import { type Behavior } from "../Behavior";
+import {
+  type BaseScreenShareInputs,
+  type BaseScreenShareViewModel,
+  createBaseScreenShare,
+} from "./ScreenShareViewModel";
+import { type ObservableScope } from "../ObservableScope";
+
+export interface LocalScreenShareViewModel extends BaseScreenShareViewModel {
+  local: true;
+}
+
+export interface LocalScreenShareInputs extends BaseScreenShareInputs {
+  participant$: Behavior<LocalParticipant | null>;
+}
+
+export function createLocalScreenShare(
+  scope: ObservableScope,
+  inputs: LocalScreenShareInputs,
+): LocalScreenShareViewModel {
+  return { ...createBaseScreenShare(scope, inputs), local: true };
+}

src/state/media/LocalUserMediaViewModel.ts

@@ -0,0 +1,137 @@
+/*
+Copyright 2023, 2024 New Vector Ltd.
+Copyright 2026 Element Creations Ltd.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import {
+  facingModeFromLocalTrack,
+  type LocalParticipant,
+  LocalVideoTrack,
+  TrackEvent,
+} from "livekit-client";
+import {
+  fromEvent,
+  map,
+  merge,
+  type Observable,
+  of,
+  startWith,
+  switchMap,
+} from "rxjs";
+import { logger } from "matrix-js-sdk/lib/logger";
+
+import { type Behavior } from "../Behavior";
+import {
+  type BaseUserMediaInputs,
+  type BaseUserMediaViewModel,
+  createBaseUserMedia,
+} from "./UserMediaViewModel";
+import { type ObservableScope } from "../ObservableScope";
+import { alwaysShowSelf } from "../../settings/settings";
+import { platform } from "../../Platform";
+import { type MediaDevices } from "../MediaDevices";
+
+export interface LocalUserMediaViewModel extends BaseUserMediaViewModel {
+  local: true;
+  /**
+   * Whether the video should be mirrored.
+   */
+  mirror$: Behavior<boolean>;
+  /**
+   * Whether to show this tile in a highly visible location near the start of
+   * the grid.
+   */
+  alwaysShow$: Behavior<boolean>;
+  setAlwaysShow: (value: boolean) => void;
+  switchCamera$: Behavior<(() => void) | null>;
+}
+
+export interface LocalUserMediaInputs extends Omit<
+  BaseUserMediaInputs,
+  "statsType"
+> {
+  participant$: Behavior<LocalParticipant | null>;
+  mediaDevices: MediaDevices;
+}
+
+export function createLocalUserMedia(
+  scope: ObservableScope,
+  { mediaDevices, ...inputs }: LocalUserMediaInputs,
+): LocalUserMediaViewModel {
+  const baseUserMedia = createBaseUserMedia(scope, {
+    ...inputs,
+    statsType: "outbound-rtp",
+  });
+
+  /**
+   * The local video track as an observable that emits whenever the track
+   * changes, the camera is switched, or the track is muted.
+   */
+  const videoTrack$: Observable<LocalVideoTrack | null> =
+    baseUserMedia.video$.pipe(
+      switchMap((v) => {
+        const track = v?.publication.track;
+        if (!(track instanceof LocalVideoTrack)) return of(null);
+        return merge(
+          // Watch for track restarts because they indicate a camera switch.
+          // This event is also emitted when unmuting the track object.
+          fromEvent(track, TrackEvent.Restarted).pipe(
+            startWith(null),
+            map(() => track),
+          ),
+          // When the track object is muted, reset it to null.
+          fromEvent(track, TrackEvent.Muted).pipe(map(() => null)),
+        );
+      }),
+    );
+
+  return {
+    ...baseUserMedia,
+    local: true,
+    mirror$: scope.behavior(
+      videoTrack$.pipe(
+        // Mirror only front-facing cameras (those that face the user)
+        map(
+          (track) =>
+            track !== null &&
+            facingModeFromLocalTrack(track).facingMode === "user",
+        ),
+      ),
+    ),
+    alwaysShow$: alwaysShowSelf.value$,
+    setAlwaysShow: alwaysShowSelf.setValue,
+    switchCamera$: scope.behavior(
+      platform === "desktop"
+        ? of(null)
+        : videoTrack$.pipe(
+            map((track) => {
+              if (track === null) return null;
+              const facingMode = facingModeFromLocalTrack(track).facingMode;
+              // If the camera isn't front or back-facing, don't provide a switch
+              // camera shortcut at all
+              if (facingMode !== "user" && facingMode !== "environment")
+                return null;
+              // Restart the track with a camera facing the opposite direction
+              return (): void =>
+                void track
+                  .restartTrack({
+                    facingMode: facingMode === "user" ? "environment" : "user",
+                  })
+                  .then(() => {
+                    // Inform the MediaDevices which camera was chosen
+                    const deviceId =
+                      track.mediaStreamTrack.getSettings().deviceId;
+                    if (deviceId !== undefined)
+                      mediaDevices.videoInput.select(deviceId);
+                  })
+                  .catch((e) =>
+                    logger.error("Failed to switch camera", facingMode, e),
+                  );
+            }),
+          ),
+    ),
+  };
+}

src/state/media/MediaItem.ts

@@ -0,0 +1,198 @@
+/*
+Copyright 2025-2026 Element Software Ltd.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import { combineLatest, map, of, switchMap } from "rxjs";
+import {
+  type LocalParticipant,
+  ParticipantEvent,
+  type RemoteParticipant,
+} from "livekit-client";
+import { observeParticipantEvents } from "@livekit/components-core";
+
+import { type ObservableScope } from "../ObservableScope.ts";
+import type { Behavior } from "../Behavior.ts";
+import type { MediaDevices } from "../MediaDevices.ts";
+import { observeSpeaker$ } from "./observeSpeaker.ts";
+import { generateItems } from "../../utils/observable.ts";
+import { type TaggedParticipant } from "../CallViewModel/remoteMembers/MatrixLivekitMembers.ts";
+import { type UserMediaViewModel } from "./UserMediaViewModel.ts";
+import { type ScreenShareViewModel } from "./ScreenShareViewModel.ts";
+import {
+  createLocalUserMedia,
+  type LocalUserMediaInputs,
+} from "./LocalUserMediaViewModel.ts";
+import {
+  createRemoteUserMedia,
+  type RemoteUserMediaInputs,
+} from "./RemoteUserMediaViewModel.ts";
+import { createLocalScreenShare } from "./LocalScreenShareViewModel.ts";
+import { createRemoteScreenShare } from "./RemoteScreenShareViewModel.ts";
+
+/**
+ * Sorting bins defining the order in which media tiles appear in the layout.
+ */
+enum SortingBin {
+  /**
+   * Yourself, when the "always show self" option is on.
+   */
+  SelfAlwaysShown,
+  /**
+   * Participants that are sharing their screen.
+   */
+  Presenters,
+  /**
+   * Participants that have been speaking recently.
+   */
+  Speakers,
+  /**
+   * Participants that have their hand raised.
+   */
+  HandRaised,
+  /**
+   * Participants with video.
+   */
+  Video,
+  /**
+   * Participants not sharing any video.
+   */
+  NoVideo,
+  /**
+   * Yourself, when the "always show self" option is off.
+   */
+  SelfNotAlwaysShown,
+}
+
+/**
+ * A user media item to be presented in a tile. This is a thin wrapper around
+ * UserMediaViewModel which additionally carries data relevant to the tile
+ * layout algorithms (data which the MediaView component should be ignorant of).
+ */
+export type WrappedUserMediaViewModel = UserMediaViewModel & {
+  /**
+   * All screen share media associated with this user media.
+   */
+  screenShares$: Behavior<ScreenShareViewModel[]>;
+  /**
+   * Which sorting bin the media item should be placed in.
+   */
+  bin$: Behavior<SortingBin>;
+};
+
+interface WrappedUserMediaInputs extends Omit<
+  LocalUserMediaInputs & RemoteUserMediaInputs,
+  "participant$"
+> {
+  participant: TaggedParticipant;
+  mediaDevices: MediaDevices;
+  pretendToBeDisconnected$: Behavior<boolean>;
+}
+
+export function createWrappedUserMedia(
+  scope: ObservableScope,
+  {
+    participant,
+    mediaDevices,
+    pretendToBeDisconnected$,
+    ...inputs
+  }: WrappedUserMediaInputs,
+): WrappedUserMediaViewModel {
+  const userMedia =
+    participant.type === "local"
+      ? createLocalUserMedia(scope, {
+          participant$: participant.value$,
+          mediaDevices,
+          ...inputs,
+        })
+      : createRemoteUserMedia(scope, {
+          participant$: participant.value$,
+          pretendToBeDisconnected$,
+          ...inputs,
+        });
+
+  // TypeScript needs this widening of the type to happen in a separate statement
+  const participant$: Behavior<LocalParticipant | RemoteParticipant | null> =
+    participant.value$;
+
+  const screenShares$ = scope.behavior(
+    participant$.pipe(
+      switchMap((p) =>
+        p === null
+          ? of([])
+          : observeParticipantEvents(
+              p,
+              ParticipantEvent.TrackPublished,
+              ParticipantEvent.TrackUnpublished,
+              ParticipantEvent.LocalTrackPublished,
+              ParticipantEvent.LocalTrackUnpublished,
+            ).pipe(
+              // Technically more than one screen share might be possible... our
+              // MediaViewModels don't support it though since they look for a unique
+              // track for the given source. So generateItems here is a bit overkill.
+              generateItems(
+                `${inputs.id} screenShares$`,
+                function* (p) {
+                  if (p.isScreenShareEnabled)
+                    yield {
+                      keys: ["screen-share"],
+                      data: undefined,
+                    };
+                },
+                (scope, _data$, key) => {
+                  const id = `${inputs.id}:${key}`;
+                  return participant.type === "local"
+                    ? createLocalScreenShare(scope, {
+                        ...inputs,
+                        id,
+                        participant$: participant.value$,
+                      })
+                    : createRemoteScreenShare(scope, {
+                        ...inputs,
+                        id,
+                        participant$: participant.value$,
+                        pretendToBeDisconnected$,
+                      });
+                },
+              ),
+            ),
+      ),
+    ),
+  );
+
+  const speaker$ = scope.behavior(observeSpeaker$(userMedia.speaking$));
+  const presenter$ = scope.behavior(
+    screenShares$.pipe(map((screenShares) => screenShares.length > 0)),
+  );
+
+  return {
+    ...userMedia,
+    screenShares$,
+    bin$: scope.behavior(
+      combineLatest(
+        [
+          speaker$,
+          presenter$,
+          userMedia.videoEnabled$,
+          userMedia.handRaised$,
+          userMedia.local ? userMedia.alwaysShow$ : of<boolean | null>(null),
+        ],
+        (speaker, presenter, video, handRaised, alwaysShow) => {
+          if (alwaysShow !== null)
+            return alwaysShow
+              ? SortingBin.SelfAlwaysShown
+              : SortingBin.SelfNotAlwaysShown;
+          else if (presenter) return SortingBin.Presenters;
+          else if (speaker) return SortingBin.Speakers;
+          else if (handRaised) return SortingBin.HandRaised;
+          else if (video) return SortingBin.Video;
+          else return SortingBin.NoVideo;
+        },
+      ),
+    ),
+  };
+}
+
+export type MediaItem = WrappedUserMediaViewModel | ScreenShareViewModel;

src/state/media/MediaViewModel.test.ts

@@ -9,6 +9,7 @@ import { expect, onTestFinished, test, vi } from "vitest";
 import {
   type LocalTrackPublication,
   LocalVideoTrack,
+  Track,
   TrackEvent,
 } from "livekit-client";
 import { waitFor } from "@testing-library/dom";
@@ -17,13 +18,13 @@ import {
   mockLocalParticipant,
   mockMediaDevices,
   mockRtcMembership,
-  createLocalMedia,
-  createRemoteMedia,
+  mockLocalMedia,
+  mockRemoteMedia,
   withTestScheduler,
   mockRemoteParticipant,
-} from "../utils/test";
-import { getValue } from "../utils/observable";
-import { constant } from "./Behavior";
+  mockRemoteScreenShare,
+} from "../../utils/test";
+import { constant } from "../Behavior";
 
 global.MediaStreamTrack = class {} as unknown as {
   new (): MediaStreamTrack;
@@ -35,7 +36,7 @@ global.MediaStream = class {} as unknown as {
 };
 
 const platformMock = vi.hoisted(() => vi.fn(() => "desktop"));
-vi.mock("../Platform", () => ({
+vi.mock("../../Platform", () => ({
   get platform(): string {
     return platformMock();
   },
@@ -45,7 +46,7 @@ const rtcMembership = mockRtcMembership("@alice:example.org", "AAAA");
 
 test("control a participant's volume", () => {
   const setVolumeSpy = vi.fn();
-  const vm = createRemoteMedia(
+  const vm = mockRemoteMedia(
     rtcMembership,
     {},
     mockRemoteParticipant({ setVolume: setVolumeSpy }),
@@ -54,33 +55,33 @@ test("control a participant's volume", () => {
     schedule("-ab---c---d|", {
       a() {
         // Try muting by toggling
-        vm.toggleLocallyMuted();
+        vm.togglePlaybackMuted();
         expect(setVolumeSpy).toHaveBeenLastCalledWith(0);
       },
       b() {
         // Try unmuting by dragging the slider back up
-        vm.setLocalVolume(0.6);
-        vm.setLocalVolume(0.8);
-        vm.commitLocalVolume();
+        vm.adjustPlaybackVolume(0.6);
+        vm.adjustPlaybackVolume(0.8);
+        vm.commitPlaybackVolume();
         expect(setVolumeSpy).toHaveBeenCalledWith(0.6);
         expect(setVolumeSpy).toHaveBeenLastCalledWith(0.8);
       },
       c() {
         // Try muting by dragging the slider back down
-        vm.setLocalVolume(0.2);
-        vm.setLocalVolume(0);
-        vm.commitLocalVolume();
+        vm.adjustPlaybackVolume(0.2);
+        vm.adjustPlaybackVolume(0);
+        vm.commitPlaybackVolume();
         expect(setVolumeSpy).toHaveBeenCalledWith(0.2);
         expect(setVolumeSpy).toHaveBeenLastCalledWith(0);
       },
       d() {
         // Try unmuting by toggling
-        vm.toggleLocallyMuted();
+        vm.togglePlaybackMuted();
         // The volume should return to the last non-zero committed volume
         expect(setVolumeSpy).toHaveBeenLastCalledWith(0.8);
       },
     });
-    expectObservable(vm.localVolume$).toBe("ab(cd)(ef)g", {
+    expectObservable(vm.playbackVolume$).toBe("ab(cd)(ef)g", {
       a: 1,
       b: 0,
       c: 0.6,
@@ -92,23 +93,75 @@ test("control a participant's volume", () => {
   });
 });
 
-test("toggle fit/contain for a participant's video", () => {
-  const vm = createRemoteMedia(rtcMembership, {}, mockRemoteParticipant({}));
+test("control a participant's screen share volume", () => {
+  const setVolumeSpy = vi.fn();
+  const vm = mockRemoteScreenShare(
+    rtcMembership,
+    {},
+    mockRemoteParticipant({ setVolume: setVolumeSpy }),
+  );
   withTestScheduler(({ expectObservable, schedule }) => {
-    schedule("-ab|", {
-      a: () => vm.toggleFitContain(),
-      b: () => vm.toggleFitContain(),
+    schedule("-ab---c---d|", {
+      a() {
+        // Try muting by toggling
+        vm.togglePlaybackMuted();
+        expect(setVolumeSpy).toHaveBeenLastCalledWith(
+          0,
+          Track.Source.ScreenShareAudio,
+        );
+      },
+      b() {
+        // Try unmuting by dragging the slider back up
+        vm.adjustPlaybackVolume(0.6);
+        vm.adjustPlaybackVolume(0.8);
+        vm.commitPlaybackVolume();
+        expect(setVolumeSpy).toHaveBeenCalledWith(
+          0.6,
+          Track.Source.ScreenShareAudio,
+        );
+        expect(setVolumeSpy).toHaveBeenLastCalledWith(
+          0.8,
+          Track.Source.ScreenShareAudio,
+        );
+      },
+      c() {
+        // Try muting by dragging the slider back down
+        vm.adjustPlaybackVolume(0.2);
+        vm.adjustPlaybackVolume(0);
+        vm.commitPlaybackVolume();
+        expect(setVolumeSpy).toHaveBeenCalledWith(
+          0.2,
+          Track.Source.ScreenShareAudio,
+        );
+        expect(setVolumeSpy).toHaveBeenLastCalledWith(
+          0,
+          Track.Source.ScreenShareAudio,
+        );
+      },
+      d() {
+        // Try unmuting by toggling
+        vm.togglePlaybackMuted();
+        // The volume should return to the last non-zero committed volume
+        expect(setVolumeSpy).toHaveBeenLastCalledWith(
+          0.8,
+          Track.Source.ScreenShareAudio,
+        );
+      },
     });
-    expectObservable(vm.cropVideo$).toBe("abc", {
-      a: true,
-      b: false,
-      c: true,
+    expectObservable(vm.playbackVolume$).toBe("ab(cd)(ef)g", {
+      a: 1,
+      b: 0,
+      c: 0.6,
+      d: 0.8,
+      e: 0.2,
+      f: 0,
+      g: 0.8,
     });
   });
 });
 
 test("local media remembers whether it should always be shown", () => {
-  const vm1 = createLocalMedia(
+  const vm1 = mockLocalMedia(
     rtcMembership,
     {},
     mockLocalParticipant({}),
@@ -120,7 +173,7 @@ test("local media remembers whether it should always be shown", () => {
   });
 
   // Next local media should start out *not* always shown
-  const vm2 = createLocalMedia(
+  const vm2 = mockLocalMedia(
     rtcMembership,
     {},
     mockLocalParticipant({}),
@@ -166,7 +219,7 @@ test("switch cameras", async () => {
 
   const selectVideoInput = vi.fn();
 
-  const vm = createLocalMedia(
+  const vm = mockLocalMedia(
     rtcMembership,
     {},
     mockLocalParticipant({
@@ -184,7 +237,7 @@ test("switch cameras", async () => {
   );
 
   // Switch to back camera
-  getValue(vm.switchCamera$)!();
+  vm.switchCamera$.value!();
   expect(restartTrack).toHaveBeenCalledExactlyOnceWith({
     facingMode: "environment",
   });
@@ -195,7 +248,7 @@ test("switch cameras", async () => {
   expect(deviceId).toBe("back camera");
 
   // Switch to front camera
-  getValue(vm.switchCamera$)!();
+  vm.switchCamera$.value!();
   expect(restartTrack).toHaveBeenCalledTimes(2);
   expect(restartTrack).toHaveBeenLastCalledWith({ facingMode: "user" });
   await waitFor(() => {
@@ -206,17 +259,17 @@ test("switch cameras", async () => {
 });
 
 test("remote media is in waiting state when participant has not yet connected", () => {
-  const vm = createRemoteMedia(rtcMembership, {}, null); // null participant
+  const vm = mockRemoteMedia(rtcMembership, {}, null); // null participant
   expect(vm.waitingForMedia$.value).toBe(true);
 });
 
 test("remote media is not in waiting state when participant is connected", () => {
-  const vm = createRemoteMedia(rtcMembership, {}, mockRemoteParticipant({}));
+  const vm = mockRemoteMedia(rtcMembership, {}, mockRemoteParticipant({}));
   expect(vm.waitingForMedia$.value).toBe(false);
 });
 
 test("remote media is not in waiting state when participant is connected with no publications", () => {
-  const vm = createRemoteMedia(
+  const vm = mockRemoteMedia(
     rtcMembership,
     {},
     mockRemoteParticipant({
@@ -228,7 +281,7 @@ test("remote media is not in waiting state when participant is connected with no
 });
 
 test("remote media is not in waiting state when user does not intend to publish anywhere", () => {
-  const vm = createRemoteMedia(
+  const vm = mockRemoteMedia(
     rtcMembership,
     {},
     mockRemoteParticipant({}),

src/state/media/MediaViewModel.ts

@@ -0,0 +1,44 @@
+/*
+Copyright 2023, 2024 New Vector Ltd.
+Copyright 2026 Element Creations Ltd.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import { type Behavior } from "../Behavior";
+import { type ScreenShareViewModel } from "./ScreenShareViewModel";
+import { type UserMediaViewModel } from "./UserMediaViewModel";
+
+/**
+ * A participant's media.
+ */
+export type MediaViewModel = UserMediaViewModel | ScreenShareViewModel;
+
+/**
+ * Properties which are common to all MediaViewModels.
+ */
+export interface BaseMediaViewModel {
+  /**
+   * An opaque identifier for this media.
+   */
+  id: string;
+  /**
+   * The Matrix user to which this media belongs.
+   */
+  userId: string;
+  displayName$: Behavior<string>;
+  mxcAvatarUrl$: Behavior<string | undefined>;
+}
+
+export type BaseMediaInputs = BaseMediaViewModel;
+
+// All this function does is strip out superfluous data from the input object
+export function createBaseMedia({
+  id,
+  userId,
+  displayName$,
+  mxcAvatarUrl$,
+}: BaseMediaInputs): BaseMediaViewModel {
+  return { id, userId, displayName$, mxcAvatarUrl$ };
+}

src/state/media/MemberMediaViewModel.ts

@@ -0,0 +1,272 @@
+/*
+Copyright 2023, 2024 New Vector Ltd.
+Copyright 2026 Element Creations Ltd.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import {
+  type Room as LivekitRoom,
+  RoomEvent as LivekitRoomEvent,
+  type Participant,
+  type Track,
+} from "livekit-client";
+import {
+  type AudioSource,
+  roomEventSelector,
+  type TrackReference,
+  type VideoSource,
+} from "@livekit/components-core";
+import { type LocalParticipant, type RemoteParticipant } from "livekit-client";
+import {
+  combineLatest,
+  distinctUntilChanged,
+  filter,
+  map,
+  type Observable,
+  of,
+  startWith,
+  switchMap,
+  throttleTime,
+} from "rxjs";
+
+import { type Behavior } from "../Behavior";
+import { type BaseMediaViewModel, createBaseMedia } from "./MediaViewModel";
+import { type EncryptionSystem } from "../../e2ee/sharedKeyManagement";
+import { type ObservableScope } from "../ObservableScope";
+import { observeTrackReference$ } from "../observeTrackReference";
+import { E2eeType } from "../../e2ee/e2eeType";
+import { observeInboundRtpStreamStats$ } from "./observeRtpStreamStats";
+
+// TODO: Encryption status is kinda broken and thus unused right now. Remove?
+export enum EncryptionStatus {
+  Connecting,
+  Okay,
+  KeyMissing,
+  KeyInvalid,
+  PasswordInvalid,
+}
+
+/**
+ * Media belonging to an active member of the RTC session.
+ */
+export interface MemberMediaViewModel extends BaseMediaViewModel {
+  /**
+   * The LiveKit video track for this media.
+   */
+  video$: Behavior<TrackReference | undefined>;
+  /**
+   * The URL of the LiveKit focus on which this member should be publishing.
+   * Exposed for debugging.
+   */
+  focusUrl$: Behavior<string | undefined>;
+  /**
+   * Whether there should be a warning that this media is unencrypted.
+   */
+  unencryptedWarning$: Behavior<boolean>;
+  encryptionStatus$: Behavior<EncryptionStatus>;
+}
+
+export interface MemberMediaInputs extends BaseMediaViewModel {
+  participant$: Behavior<LocalParticipant | RemoteParticipant | null>;
+  livekitRoom$: Behavior<LivekitRoom | undefined>;
+  audioSource: AudioSource;
+  videoSource: VideoSource;
+  focusUrl$: Behavior<string | undefined>;
+  encryptionSystem: EncryptionSystem;
+}
+
+export function createMemberMedia(
+  scope: ObservableScope,
+  {
+    participant$,
+    livekitRoom$,
+    audioSource,
+    videoSource,
+    focusUrl$,
+    encryptionSystem,
+    ...inputs
+  }: MemberMediaInputs,
+): MemberMediaViewModel {
+  const trackBehavior$ = (
+    source: Track.Source,
+  ): Behavior<TrackReference | undefined> =>
+    scope.behavior(
+      participant$.pipe(
+        switchMap((p) =>
+          !p ? of(undefined) : observeTrackReference$(p, source),
+        ),
+      ),
+    );
+
+  const audio$ = trackBehavior$(audioSource);
+  const video$ = trackBehavior$(videoSource);
+
+  return {
+    ...createBaseMedia(inputs),
+    video$,
+    focusUrl$,
+    unencryptedWarning$: scope.behavior(
+      combineLatest(
+        [audio$, video$],
+        (a, v) =>
+          encryptionSystem.kind !== E2eeType.NONE &&
+          (a?.publication.isEncrypted === false ||
+            v?.publication.isEncrypted === false),
+      ),
+    ),
+    encryptionStatus$: scope.behavior(
+      participant$.pipe(
+        switchMap((participant): Observable<EncryptionStatus> => {
+          if (!participant) {
+            return of(EncryptionStatus.Connecting);
+          } else if (
+            participant.isLocal ||
+            encryptionSystem.kind === E2eeType.NONE
+          ) {
+            return of(EncryptionStatus.Okay);
+          } else if (encryptionSystem.kind === E2eeType.PER_PARTICIPANT) {
+            return combineLatest([
+              encryptionErrorObservable$(
+                livekitRoom$,
+                participant,
+                encryptionSystem,
+                "MissingKey",
+              ),
+              encryptionErrorObservable$(
+                livekitRoom$,
+                participant,
+                encryptionSystem,
+                "InvalidKey",
+              ),
+              observeRemoteTrackReceivingOkay$(participant, audioSource),
+              observeRemoteTrackReceivingOkay$(participant, videoSource),
+            ]).pipe(
+              map(([keyMissing, keyInvalid, audioOkay, videoOkay]) => {
+                if (keyMissing) return EncryptionStatus.KeyMissing;
+                if (keyInvalid) return EncryptionStatus.KeyInvalid;
+                if (audioOkay || videoOkay) return EncryptionStatus.Okay;
+                return undefined; // no change
+              }),
+              filter((x) => !!x),
+              startWith(EncryptionStatus.Connecting),
+            );
+          } else {
+            return combineLatest([
+              encryptionErrorObservable$(
+                livekitRoom$,
+                participant,
+                encryptionSystem,
+                "InvalidKey",
+              ),
+              observeRemoteTrackReceivingOkay$(participant, audioSource),
+              observeRemoteTrackReceivingOkay$(participant, videoSource),
+            ]).pipe(
+              map(
+                ([keyInvalid, audioOkay, videoOkay]):
+                  | EncryptionStatus
+                  | undefined => {
+                  if (keyInvalid) return EncryptionStatus.PasswordInvalid;
+                  if (audioOkay || videoOkay) return EncryptionStatus.Okay;
+                  return undefined; // no change
+                },
+              ),
+              filter((x) => !!x),
+              startWith(EncryptionStatus.Connecting),
+            );
+          }
+        }),
+      ),
+    ),
+  };
+}
+
+function encryptionErrorObservable$(
+  room$: Behavior<LivekitRoom | undefined>,
+  participant: Participant,
+  encryptionSystem: EncryptionSystem,
+  criteria: string,
+): Observable<boolean> {
+  return room$.pipe(
+    switchMap((room) => {
+      if (room === undefined) return of(false);
+      return roomEventSelector(room, LivekitRoomEvent.EncryptionError).pipe(
+        map((e) => {
+          const [err] = e;
+          if (encryptionSystem.kind === E2eeType.PER_PARTICIPANT) {
+            return (
+              // Ideally we would pull the participant identity from the field on the error.
+              // However, it gets lost in the serialization process between workers.
+              // So, instead we do a string match
+              (err?.message.includes(participant.identity) &&
+                err?.message.includes(criteria)) ??
+              false
+            );
+          } else if (encryptionSystem.kind === E2eeType.SHARED_KEY) {
+            return !!err?.message.includes(criteria);
+          }
+
+          return false;
+        }),
+      );
+    }),
+    distinctUntilChanged(),
+    throttleTime(1000), // Throttle to avoid spamming the UI
+    startWith(false),
+  );
+}
+
+function observeRemoteTrackReceivingOkay$(
+  participant: Participant,
+  source: Track.Source,
+): Observable<boolean | undefined> {
+  let lastStats: {
+    framesDecoded: number | undefined;
+    framesDropped: number | undefined;
+    framesReceived: number | undefined;
+  } = {
+    framesDecoded: undefined,
+    framesDropped: undefined,
+    framesReceived: undefined,
+  };
+
+  return observeInboundRtpStreamStats$(participant, source).pipe(
+    map((stats) => {
+      if (!stats) return undefined;
+      const { framesDecoded, framesDropped, framesReceived } = stats;
+      return {
+        framesDecoded,
+        framesDropped,
+        framesReceived,
+      };
+    }),
+    filter((newStats) => !!newStats),
+    map((newStats): boolean | undefined => {
+      const oldStats = lastStats;
+      lastStats = newStats;
+      if (
+        typeof newStats.framesReceived === "number" &&
+        typeof oldStats.framesReceived === "number" &&
+        typeof newStats.framesDecoded === "number" &&
+        typeof oldStats.framesDecoded === "number"
+      ) {
+        const framesReceivedDelta =
+          newStats.framesReceived - oldStats.framesReceived;
+        const framesDecodedDelta =
+          newStats.framesDecoded - oldStats.framesDecoded;
+
+        // if we received >0 frames and managed to decode >0 frames then we treat that as success
+
+        if (framesReceivedDelta > 0) {
+          return framesDecodedDelta > 0;
+        }
+      }
+
+      // no change
+      return undefined;
+    }),
+    filter((x) => typeof x === "boolean"),
+    startWith(undefined),
+  );
+}

src/state/media/RemoteScreenShareViewModel.ts

@@ -0,0 +1,72 @@
+/*
+Copyright 2023, 2024 New Vector Ltd.
+Copyright 2026 Element Creations Ltd.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import { Track, type RemoteParticipant } from "livekit-client";
+import { map, of, switchMap } from "rxjs";
+
+import { type Behavior } from "../Behavior";
+import {
+  type BaseScreenShareInputs,
+  type BaseScreenShareViewModel,
+  createBaseScreenShare,
+} from "./ScreenShareViewModel";
+import { type ObservableScope } from "../ObservableScope";
+import { createVolumeControls, type VolumeControls } from "../VolumeControls";
+import { observeTrackReference$ } from "../observeTrackReference";
+
+export interface RemoteScreenShareViewModel
+  extends BaseScreenShareViewModel, VolumeControls {
+  local: false;
+  /**
+   * Whether this screen share's video should be displayed.
+   */
+  videoEnabled$: Behavior<boolean>;
+  /**
+   * Whether this screen share should be considered to have an audio track.
+   */
+  audioEnabled$: Behavior<boolean>;
+}
+
+export interface RemoteScreenShareInputs extends BaseScreenShareInputs {
+  participant$: Behavior<RemoteParticipant | null>;
+  pretendToBeDisconnected$: Behavior<boolean>;
+}
+
+export function createRemoteScreenShare(
+  scope: ObservableScope,
+  { pretendToBeDisconnected$, ...inputs }: RemoteScreenShareInputs,
+): RemoteScreenShareViewModel {
+  return {
+    ...createBaseScreenShare(scope, inputs),
+    ...createVolumeControls(scope, {
+      pretendToBeDisconnected$,
+      sink$: scope.behavior(
+        inputs.participant$.pipe(
+          map(
+            (p) => (volume) =>
+              p?.setVolume(volume, Track.Source.ScreenShareAudio),
+          ),
+        ),
+      ),
+    }),
+    local: false,
+    videoEnabled$: scope.behavior(
+      pretendToBeDisconnected$.pipe(map((disconnected) => !disconnected)),
+    ),
+    audioEnabled$: scope.behavior(
+      inputs.participant$.pipe(
+        switchMap((p) =>
+          p
+            ? observeTrackReference$(p, Track.Source.ScreenShareAudio)
+            : of(null),
+        ),
+        map(Boolean),
+      ),
+    ),
+  };
+}

src/state/media/RemoteUserMediaViewModel.ts

@@ -0,0 +1,82 @@
+/*
+Copyright 2023, 2024 New Vector Ltd.
+Copyright 2026 Element Creations Ltd.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import { type RemoteParticipant } from "livekit-client";
+import { combineLatest, map, of, switchMap } from "rxjs";
+
+import { type Behavior } from "../Behavior";
+import { createVolumeControls, type VolumeControls } from "../VolumeControls";
+import {
+  type BaseUserMediaInputs,
+  type BaseUserMediaViewModel,
+  createBaseUserMedia,
+} from "./UserMediaViewModel";
+import { type ObservableScope } from "../ObservableScope";
+
+export interface RemoteUserMediaViewModel
+  extends BaseUserMediaViewModel, VolumeControls {
+  local: false;
+  /**
+   * Whether we are waiting for this user's LiveKit participant to exist. This
+   * could be because either we or the remote party are still connecting.
+   */
+  waitingForMedia$: Behavior<boolean>;
+}
+
+export interface RemoteUserMediaInputs extends Omit<
+  BaseUserMediaInputs,
+  "statsType"
+> {
+  participant$: Behavior<RemoteParticipant | null>;
+  pretendToBeDisconnected$: Behavior<boolean>;
+}
+
+export function createRemoteUserMedia(
+  scope: ObservableScope,
+  { pretendToBeDisconnected$, ...inputs }: RemoteUserMediaInputs,
+): RemoteUserMediaViewModel {
+  const baseUserMedia = createBaseUserMedia(scope, {
+    ...inputs,
+    statsType: "inbound-rtp",
+  });
+
+  return {
+    ...baseUserMedia,
+    ...createVolumeControls(scope, {
+      pretendToBeDisconnected$,
+      sink$: scope.behavior(
+        inputs.participant$.pipe(map((p) => (volume) => p?.setVolume(volume))),
+      ),
+    }),
+    local: false,
+    speaking$: scope.behavior(
+      pretendToBeDisconnected$.pipe(
+        switchMap((disconnected) =>
+          disconnected ? of(false) : baseUserMedia.speaking$,
+        ),
+      ),
+    ),
+    videoEnabled$: scope.behavior(
+      pretendToBeDisconnected$.pipe(
+        switchMap((disconnected) =>
+          disconnected ? of(false) : baseUserMedia.videoEnabled$,
+        ),
+      ),
+    ),
+    waitingForMedia$: scope.behavior(
+      combineLatest(
+        [inputs.livekitRoom$, inputs.participant$],
+        (livekitRoom, participant) =>
+          // If livekitRoom is undefined, the user is not attempting to publish on
+          // any transport and so we shouldn't expect a participant. (They might
+          // be a subscribe-only bot for example.)
+          livekitRoom !== undefined && participant === null,
+      ),
+    ),
+  };
+}

src/state/media/ScreenShareViewModel.ts

@@ -0,0 +1,51 @@
+/*
+Copyright 2023, 2024 New Vector Ltd.
+Copyright 2026 Element Creations Ltd.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import { Track } from "livekit-client";
+
+import { type ObservableScope } from "../ObservableScope";
+import { type LocalScreenShareViewModel } from "./LocalScreenShareViewModel";
+import {
+  createMemberMedia,
+  type MemberMediaInputs,
+  type MemberMediaViewModel,
+} from "./MemberMediaViewModel";
+import { type RemoteScreenShareViewModel } from "./RemoteScreenShareViewModel";
+
+/**
+ * A participant's screen share media.
+ */
+export type ScreenShareViewModel =
+  | LocalScreenShareViewModel
+  | RemoteScreenShareViewModel;
+
+/**
+ * Properties which are common to all ScreenShareViewModels.
+ */
+export interface BaseScreenShareViewModel extends MemberMediaViewModel {
+  type: "screen share";
+}
+
+export type BaseScreenShareInputs = Omit<
+  MemberMediaInputs,
+  "audioSource" | "videoSource"
+>;
+
+export function createBaseScreenShare(
+  scope: ObservableScope,
+  inputs: BaseScreenShareInputs,
+): BaseScreenShareViewModel {
+  return {
+    ...createMemberMedia(scope, {
+      ...inputs,
+      audioSource: Track.Source.ScreenShareAudio,
+      videoSource: Track.Source.ScreenShare,
+    }),
+    type: "screen share",
+  };
+}

src/state/media/UserMediaViewModel.ts

@@ -0,0 +1,164 @@
+/*
+Copyright 2023, 2024 New Vector Ltd.
+Copyright 2026 Element Creations Ltd.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import {
+  BehaviorSubject,
+  combineLatest,
+  map,
+  type Observable,
+  of,
+  Subject,
+  switchMap,
+} from "rxjs";
+import {
+  observeParticipantEvents,
+  observeParticipantMedia,
+} from "@livekit/components-core";
+import { ParticipantEvent, Track } from "livekit-client";
+
+import { type ReactionOption } from "../../reactions";
+import { type Behavior } from "../Behavior";
+import { type LocalUserMediaViewModel } from "./LocalUserMediaViewModel";
+import {
+  createMemberMedia,
+  type MemberMediaInputs,
+  type MemberMediaViewModel,
+} from "./MemberMediaViewModel";
+import { type RemoteUserMediaViewModel } from "./RemoteUserMediaViewModel";
+import { type ObservableScope } from "../ObservableScope";
+import { showConnectionStats } from "../../settings/settings";
+import { observeRtpStreamStats$ } from "./observeRtpStreamStats";
+import { videoFit$, videoSizeFromParticipant$ } from "../../utils/videoFit.ts";
+
+/**
+ * A participant's user media (i.e. their microphone and camera feed).
+ */
+export type UserMediaViewModel =
+  | LocalUserMediaViewModel
+  | RemoteUserMediaViewModel;
+
+export interface BaseUserMediaViewModel extends MemberMediaViewModel {
+  type: "user";
+  speaking$: Behavior<boolean>;
+  audioEnabled$: Behavior<boolean>;
+  videoEnabled$: Behavior<boolean>;
+  videoFit$: Behavior<"cover" | "contain">;
+  toggleCropVideo: () => void;
+  /**
+   * The expected identity of the LiveKit participant. Exposed for debugging.
+   */
+  rtcBackendIdentity: string;
+  handRaised$: Behavior<Date | null>;
+  reaction$: Behavior<ReactionOption | null>;
+  audioStreamStats$: Observable<
+    RTCInboundRtpStreamStats | RTCOutboundRtpStreamStats | undefined
+  >;
+  videoStreamStats$: Observable<
+    RTCInboundRtpStreamStats | RTCOutboundRtpStreamStats | undefined
+  >;
+  /**
+   * Set the target dimensions of the HTML element (final dimension after anim).
+   * This can be used to determine the best video fit (fit to frame / keep ratio).
+   * @param targetWidth - The target width of the HTML element displaying the video.
+   * @param targetHeight - The target height of the HTML element displaying the video.
+   */
+  setTargetDimensions: (targetWidth: number, targetHeight: number) => void;
+}
+
+export interface BaseUserMediaInputs extends Omit<
+  MemberMediaInputs,
+  "audioSource" | "videoSource"
+> {
+  rtcBackendIdentity: string;
+  handRaised$: Behavior<Date | null>;
+  reaction$: Behavior<ReactionOption | null>;
+  statsType: "inbound-rtp" | "outbound-rtp";
+}
+
+export function createBaseUserMedia(
+  scope: ObservableScope,
+  {
+    rtcBackendIdentity,
+    handRaised$,
+    reaction$,
+    statsType,
+    ...inputs
+  }: BaseUserMediaInputs,
+): BaseUserMediaViewModel {
+  const { participant$ } = inputs;
+  const media$ = scope.behavior(
+    participant$.pipe(
+      switchMap((p) => (p && observeParticipantMedia(p)) ?? of(undefined)),
+    ),
+  );
+  const toggleCropVideo$ = new Subject<void>();
+
+  // The target size of the video element, used to determine the best video fit.
+  // The target size is the final size of the HTML element after any animations have completed.
+  const targetSize$ = new BehaviorSubject<
+    { width: number; height: number } | undefined
+  >(undefined);
+
+  return {
+    ...createMemberMedia(scope, {
+      ...inputs,
+      audioSource: Track.Source.Microphone,
+      videoSource: Track.Source.Camera,
+    }),
+    type: "user",
+    speaking$: scope.behavior(
+      participant$.pipe(
+        switchMap((p) =>
+          p
+            ? observeParticipantEvents(
+                p,
+                ParticipantEvent.IsSpeakingChanged,
+              ).pipe(map((p) => p.isSpeaking))
+            : of(false),
+        ),
+      ),
+    ),
+    audioEnabled$: scope.behavior(
+      media$.pipe(map((m) => m?.microphoneTrack?.isMuted === false)),
+    ),
+    videoEnabled$: scope.behavior(
+      media$.pipe(map((m) => m?.cameraTrack?.isMuted === false)),
+    ),
+    videoFit$: videoFit$(
+      scope,
+      videoSizeFromParticipant$(participant$),
+      targetSize$,
+    ),
+    toggleCropVideo: () => toggleCropVideo$.next(),
+    rtcBackendIdentity,
+    handRaised$,
+    reaction$,
+    audioStreamStats$: combineLatest([
+      participant$,
+      showConnectionStats.value$,
+    ]).pipe(
+      switchMap(([p, showConnectionStats]) => {
+        //
+        if (!p || !showConnectionStats) return of(undefined);
+        return observeRtpStreamStats$(p, Track.Source.Microphone, statsType);
+      }),
+    ),
+    videoStreamStats$: combineLatest([
+      participant$,
+      showConnectionStats.value$,
+    ]).pipe(
+      switchMap(([p, showConnectionStats]) => {
+        if (!p || !showConnectionStats) return of(undefined);
+        return observeRtpStreamStats$(p, Track.Source.Camera, statsType);
+      }),
+    ),
+    setTargetDimensions: (targetWidth: number, targetHeight: number): void => {
+      targetSize$.next({ width: targetWidth, height: targetHeight });
+    },
+  };
+}

src/state/media/observeRtpStreamStats.ts

@@ -0,0 +1,78 @@
+/*
+Copyright 2023, 2024 New Vector Ltd.
+Copyright 2026 Element Creations Ltd.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import {
+  LocalTrack,
+  type Participant,
+  RemoteTrack,
+  type Track,
+} from "livekit-client";
+import {
+  combineLatest,
+  interval,
+  type Observable,
+  startWith,
+  switchMap,
+  map,
+} from "rxjs";
+
+import { observeTrackReference$ } from "../observeTrackReference";
+
+export function observeRtpStreamStats$(
+  participant: Participant,
+  source: Track.Source,
+  type: "inbound-rtp" | "outbound-rtp",
+): Observable<
+  RTCInboundRtpStreamStats | RTCOutboundRtpStreamStats | undefined
+> {
+  return combineLatest([
+    observeTrackReference$(participant, source),
+    interval(1000).pipe(startWith(0)),
+  ]).pipe(
+    switchMap(async ([trackReference]) => {
+      const track = trackReference?.publication?.track;
+      if (
+        !track ||
+        !(track instanceof RemoteTrack || track instanceof LocalTrack)
+      ) {
+        return undefined;
+      }
+      const report = await track.getRTCStatsReport();
+      if (!report) {
+        return undefined;
+      }
+
+      for (const v of report.values()) {
+        if (v.type === type) {
+          return v;
+        }
+      }
+
+      return undefined;
+    }),
+    startWith(undefined),
+  );
+}
+
+export function observeInboundRtpStreamStats$(
+  participant: Participant,
+  source: Track.Source,
+): Observable<RTCInboundRtpStreamStats | undefined> {
+  return observeRtpStreamStats$(participant, source, "inbound-rtp").pipe(
+    map((x) => x as RTCInboundRtpStreamStats | undefined),
+  );
+}
+
+export function observeOutboundRtpStreamStats$(
+  participant: Participant,
+  source: Track.Source,
+): Observable<RTCOutboundRtpStreamStats | undefined> {
+  return observeRtpStreamStats$(participant, source, "outbound-rtp").pipe(
+    map((x) => x as RTCOutboundRtpStreamStats | undefined),
+  );
+}

src/state/media/observeSpeaker.test.ts

@@ -7,7 +7,7 @@ Please see LICENSE in the repository root for full details.
 
 import { describe, test } from "vitest";
 
-import { withTestScheduler } from "../utils/test";
+import { withTestScheduler } from "../../utils/test";
 import { observeSpeaker$ } from "./observeSpeaker";
 
 const yesNo = {

src/state/observeTrackReference.ts

@@ -0,0 +1,28 @@
+/*
+Copyright 2023, 2024 New Vector Ltd.
+Copyright 2026 Element Creations Ltd.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import {
+  observeParticipantMedia,
+  type TrackReference,
+} from "@livekit/components-core";
+import { type Participant, type Track } from "livekit-client";
+import { distinctUntilChanged, map, type Observable } from "rxjs";
+
+/**
+ * Reactively reads a participant's track reference for a given media source.
+ */
+export function observeTrackReference$(
+  participant: Participant,
+  source: Track.Source,
+): Observable<TrackReference | undefined> {
+  return observeParticipantMedia(participant).pipe(
+    map(() => participant.getTrackPublication(source)),
+    distinctUntilChanged(),
+    map((publication) => publication && { participant, publication, source }),
+  );
+}

src/tile/GridTile.test.tsx

@@ -14,7 +14,7 @@ import { type MatrixRTCSession } from "matrix-js-sdk/lib/matrixrtc";
 import { GridTile } from "./GridTile";
 import {
   mockRtcMembership,
-  createRemoteMedia,
+  mockRemoteMedia,
   mockRemoteParticipant,
 } from "../utils/test";
 import { GridTileViewModel } from "../state/TileViewModel";
@@ -29,7 +29,7 @@ global.IntersectionObserver = class MockIntersectionObserver {
 } as unknown as typeof IntersectionObserver;
 
 test("GridTile is accessible", async () => {
-  const vm = createRemoteMedia(
+  const vm = mockRemoteMedia(
     mockRtcMembership("@alice:example.org", "AAAA"),
     {
       rawDisplayName: "Alice",

src/tile/GridTile.tsx

@@ -11,6 +11,7 @@ import {
   type ReactNode,
   type Ref,
   useCallback,
+  useEffect,
   useRef,
   useState,
 } from "react";
@@ -26,7 +27,6 @@ import {
   VolumeOffIcon,
   VisibilityOnIcon,
   UserProfileIcon,
-  ExpandIcon,
   VolumeOffSolidIcon,
   SwitchCameraSolidIcon,
 } from "@vector-im/compound-design-tokens/assets/web/icons";
@@ -39,11 +39,6 @@ import {
 import { useObservableEagerState } from "observable-hooks";
 
 import styles from "./GridTile.module.css";
-import {
-  type UserMediaViewModel,
-  LocalUserMediaViewModel,
-  type RemoteUserMediaViewModel,
-} from "../state/MediaViewModel";
 import { Slider } from "../Slider";
 import { MediaView } from "./MediaView";
 import { useLatest } from "../useLatest";
@@ -51,6 +46,9 @@ import { type GridTileViewModel } from "../state/TileViewModel";
 import { useMergedRefs } from "../useMergedRefs";
 import { useReactionsSender } from "../reactions/useReactionsSender";
 import { useBehavior } from "../useBehavior";
+import { type LocalUserMediaViewModel } from "../state/media/LocalUserMediaViewModel";
+import { type RemoteUserMediaViewModel } from "../state/media/RemoteUserMediaViewModel";
+import { type UserMediaViewModel } from "../state/media/UserMediaViewModel";
 
 interface TileProps {
   ref?: Ref<HTMLDivElement>;
@@ -68,7 +66,7 @@ interface TileProps {
 interface UserMediaTileProps extends TileProps {
   vm: UserMediaViewModel;
   mirror: boolean;
-  locallyMuted: boolean;
+  playbackMuted: boolean;
   waitingForMedia?: boolean;
   primaryButton?: ReactNode;
   menuStart?: ReactNode;
@@ -79,7 +77,7 @@ const UserMediaTile: FC<UserMediaTileProps> = ({
   ref,
   vm,
   showSpeakingIndicators,
-  locallyMuted,
+  playbackMuted,
   waitingForMedia,
   primaryButton,
   menuStart,
@@ -89,6 +87,8 @@ const UserMediaTile: FC<UserMediaTileProps> = ({
   displayName,
   mxcAvatarUrl,
   focusable,
+  targetWidth,
+  targetHeight,
   ...props
 }) => {
   const { toggleRaisedHand } = useReactionsSender();
@@ -105,24 +105,25 @@ const UserMediaTile: FC<UserMediaTileProps> = ({
   const audioEnabled = useBehavior(vm.audioEnabled$);
   const videoEnabled = useBehavior(vm.videoEnabled$);
   const speaking = useBehavior(vm.speaking$);
-  const cropVideo = useBehavior(vm.cropVideo$);
-  const onSelectFitContain = useCallback(
-    (e: Event) => {
-      e.preventDefault();
-      vm.toggleFitContain();
-    },
-    [vm],
-  );
+  const videoFit = useBehavior(vm.videoFit$);
+
   const rtcBackendIdentity = vm.rtcBackendIdentity;
   const handRaised = useBehavior(vm.handRaised$);
   const reaction = useBehavior(vm.reaction$);
 
-  const AudioIcon = locallyMuted
+  // Whenever bounds change, inform the viewModel
+  useEffect(() => {
+    if (targetWidth > 0 && targetHeight > 0) {
+      vm.setTargetDimensions(targetWidth, targetHeight);
+    }
+  }, [targetWidth, targetHeight, vm]);
+
+  const AudioIcon = playbackMuted
     ? VolumeOffSolidIcon
     : audioEnabled
       ? MicOnSolidIcon
       : MicOffSolidIcon;
-  const audioIconLabel = locallyMuted
+  const audioIconLabel = playbackMuted
     ? t("video_tile.muted_for_me")
     : audioEnabled
       ? t("microphone_on")
@@ -132,12 +133,10 @@ const UserMediaTile: FC<UserMediaTileProps> = ({
   const menu = (
     <>
       {menuStart}
-      <ToggleMenuItem
-        Icon={ExpandIcon}
-        label={t("video_tile.change_fit_contain")}
-        checked={cropVideo}
-        onSelect={onSelectFitContain}
-      />
+      {/*
+       No additional menu item (used to be the manual fit to frame.
+       Placeholder for future menu items that should be placed here.
+       */}
       {menuEnd}
     </>
   );
@@ -156,7 +155,7 @@ const UserMediaTile: FC<UserMediaTileProps> = ({
       unencryptedWarning={unencryptedWarning}
       encryptionStatus={encryptionStatus}
       videoEnabled={videoEnabled}
-      videoFit={cropVideo ? "cover" : "contain"}
+      videoFit={videoFit}
       className={classNames(className, styles.tile, {
         [styles.speaking]: showSpeaking,
         [styles.handRaised]: !showSpeaking && handRaised,
@@ -166,7 +165,7 @@ const UserMediaTile: FC<UserMediaTileProps> = ({
           width={20}
           height={20}
           aria-label={audioIconLabel}
-          data-muted={locallyMuted || !audioEnabled}
+          data-muted={playbackMuted || !audioEnabled}
           className={styles.muteIcon}
         />
       }
@@ -202,6 +201,8 @@ const UserMediaTile: FC<UserMediaTileProps> = ({
       audioStreamStats={audioStreamStats}
       videoStreamStats={videoStreamStats}
       rtcBackendIdentity={rtcBackendIdentity}
+      targetWidth={targetWidth}
+      targetHeight={targetHeight}
       {...props}
     />
   );
@@ -245,7 +246,7 @@ const LocalUserMediaTile: FC<LocalUserMediaTileProps> = ({
     <UserMediaTile
       ref={ref}
       vm={vm}
-      locallyMuted={false}
+      playbackMuted={false}
       mirror={mirror}
       primaryButton={
         switchCamera === null ? undefined : (
@@ -295,36 +296,31 @@ const RemoteUserMediaTile: FC<RemoteUserMediaTileProps> = ({
 }) => {
   const { t } = useTranslation();
   const waitingForMedia = useBehavior(vm.waitingForMedia$);
-  const locallyMuted = useBehavior(vm.locallyMuted$);
-  const localVolume = useBehavior(vm.localVolume$);
+  const playbackMuted = useBehavior(vm.playbackMuted$);
+  const playbackVolume = useBehavior(vm.playbackVolume$);
   const onSelectMute = useCallback(
     (e: Event) => {
       e.preventDefault();
-      vm.toggleLocallyMuted();
+      vm.togglePlaybackMuted();
     },
     [vm],
   );
-  const onChangeLocalVolume = useCallback(
-    (v: number) => vm.setLocalVolume(v),
-    [vm],
-  );
-  const onCommitLocalVolume = useCallback(() => vm.commitLocalVolume(), [vm]);
 
-  const VolumeIcon = locallyMuted ? VolumeOffIcon : VolumeOnIcon;
+  const VolumeIcon = playbackMuted ? VolumeOffIcon : VolumeOnIcon;
 
   return (
     <UserMediaTile
       ref={ref}
       vm={vm}
       waitingForMedia={waitingForMedia}
-      locallyMuted={locallyMuted}
+      playbackMuted={playbackMuted}
       mirror={false}
       menuStart={
         <>
           <ToggleMenuItem
             Icon={MicOffIcon}
             label={t("video_tile.mute_for_me")}
-            checked={locallyMuted}
+            checked={playbackMuted}
             onSelect={onSelectMute}
           />
           {/* TODO: Figure out how to make this slider keyboard accessible */}
@@ -332,9 +328,9 @@ const RemoteUserMediaTile: FC<RemoteUserMediaTileProps> = ({
             <Slider
               className={styles.volumeSlider}
               label={t("video_tile.volume")}
-              value={localVolume}
-              onValueChange={onChangeLocalVolume}
-              onValueCommit={onCommitLocalVolume}
+              value={playbackVolume}
+              onValueChange={vm.adjustPlaybackVolume}
+              onValueCommit={vm.commitPlaybackVolume}
               min={0}
               max={1}
               step={0.01}
@@ -374,7 +370,7 @@ export const GridTile: FC<GridTileProps> = ({
   const displayName = useBehavior(media.displayName$);
   const mxcAvatarUrl = useBehavior(media.mxcAvatarUrl$);
 
-  if (media instanceof LocalUserMediaViewModel) {
+  if (media.local) {
     return (
       <LocalUserMediaTile
         ref={ref}

src/tile/MediaView.test.tsx

@@ -18,7 +18,7 @@ import { TrackInfo } from "@livekit/protocol";
 import { type ComponentProps } from "react";
 
 import { MediaView } from "./MediaView";
-import { EncryptionStatus } from "../state/MediaViewModel";
+import { EncryptionStatus } from "../state/media/MemberMediaViewModel";
 import { mockLocalParticipant } from "../utils/test";
 
 describe("MediaView", () => {

src/tile/MediaView.tsx

@@ -16,10 +16,10 @@ import { ErrorSolidIcon } from "@vector-im/compound-design-tokens/assets/web/ico
 
 import styles from "./MediaView.module.css";
 import { Avatar } from "../Avatar";
-import { type EncryptionStatus } from "../state/MediaViewModel";
+import { type EncryptionStatus } from "../state/media/MemberMediaViewModel";
 import { RaisedHandIndicator } from "../reactions/RaisedHandIndicator";
 import {
-  showConnectionStats,
+  showConnectionStats as showConnectionStatsSetting,
   showHandRaisedTimer,
   useSetting,
 } from "../settings/settings";
@@ -85,7 +85,7 @@ export const MediaView: FC<Props> = ({
 }) => {
   const { t } = useTranslation();
   const [handRaiseTimerVisible] = useSetting(showHandRaisedTimer);
-  const [showConnectioStats] = useSetting(showConnectionStats);
+  const [showConnectionStats] = useSetting(showConnectionStatsSetting);
 
   const avatarSize = Math.round(Math.min(targetWidth, targetHeight) / 2);
 
@@ -139,10 +139,10 @@ export const MediaView: FC<Props> = ({
         {waitingForMedia && (
           <div className={styles.status}>
             {t("video_tile.waiting_for_media")}
-            {showConnectioStats ? " " + rtcBackendIdentity : ""}
+            {showConnectionStats ? " " + rtcBackendIdentity : ""}
           </div>
         )}
-        {(audioStreamStats || videoStreamStats) && (
+        {showConnectionStats && (
           <>
             <RTCConnectionStats
               audio={audioStreamStats}

src/tile/SpotlightTile.module.css

@@ -84,7 +84,6 @@ Please see LICENSE in the repository root for full details.
 .expand {
   appearance: none;
   cursor: pointer;
-  opacity: 0;
   padding: var(--cpd-space-2x);
   border: none;
   border-radius: var(--cpd-radius-pill-effect);
@@ -108,6 +107,35 @@ Please see LICENSE in the repository root for full details.
   z-index: 1;
 }
 
+.volumeSlider {
+  width: 100%;
+  min-width: 172px;
+}
+
+/* Disable the hover effect for the screen share volume menu button */
+.volumeMenuItem:hover {
+  background: transparent;
+  cursor: default;
+}
+
+.volumeMenuItem {
+  gap: var(--cpd-space-3x);
+}
+
+.menuMuteButton {
+  appearance: none;
+  background: none;
+  border: none;
+  padding: 0;
+  cursor: pointer;
+  display: flex;
+}
+
+/* Make icons change color with the theme */
+.menuMuteButton > svg {
+  color: var(--cpd-color-icon-primary);
+}
+
 .expand > svg {
   display: block;
   color: var(--cpd-color-icon-primary);
@@ -119,17 +147,22 @@ Please see LICENSE in the repository root for full details.
   }
 }
 
-.expand:active {
+.expand:active,
+.expand[data-state="open"] {
   background: var(--cpd-color-gray-100);
 }
 
 @media (hover) {
+  .tile > div > button {
+    opacity: 0;
+  }
   .tile:hover > div > button {
     opacity: 1;
   }
 }
 
-.tile:has(:focus-visible) > div > button {
+.tile:has(:focus-visible) > div > button,
+.tile > div:has([data-state="open"]) > button {
   opacity: 1;
 }
 

src/tile/SpotlightTile.test.tsx

@@ -9,15 +9,17 @@ import { test, expect, vi } from "vitest";
 import { isInaccessible, render, screen } from "@testing-library/react";
 import { axe } from "vitest-axe";
 import userEvent from "@testing-library/user-event";
+import { TooltipProvider } from "@vector-im/compound-web";
 
 import { SpotlightTile } from "./SpotlightTile";
 import {
   mockLocalParticipant,
   mockMediaDevices,
   mockRtcMembership,
-  createLocalMedia,
-  createRemoteMedia,
+  mockLocalMedia,
+  mockRemoteMedia,
   mockRemoteParticipant,
+  mockRemoteScreenShare,
 } from "../utils/test";
 import { SpotlightTileViewModel } from "../state/TileViewModel";
 import { constant } from "../state/Behavior";
@@ -28,7 +30,7 @@ global.IntersectionObserver = class MockIntersectionObserver {
 } as unknown as typeof IntersectionObserver;
 
 test("SpotlightTile is accessible", async () => {
-  const vm1 = createRemoteMedia(
+  const vm1 = mockRemoteMedia(
     mockRtcMembership("@alice:example.org", "AAAA"),
     {
       rawDisplayName: "Alice",
@@ -37,7 +39,7 @@ test("SpotlightTile is accessible", async () => {
     mockRemoteParticipant({}),
   );
 
-  const vm2 = createLocalMedia(
+  const vm2 = mockLocalMedia(
     mockRtcMembership("@bob:example.org", "BBBB"),
     {
       rawDisplayName: "Bob",
@@ -78,3 +80,63 @@ test("SpotlightTile is accessible", async () => {
   await user.click(screen.getByRole("button", { name: "Expand" }));
   expect(toggleExpanded).toHaveBeenCalled();
 });
+
+test("Screen share volume UI is shown when screen share has audio", async () => {
+  const vm = mockRemoteScreenShare(
+    mockRtcMembership("@alice:example.org", "AAAA"),
+    {},
+    mockRemoteParticipant({}),
+  );
+
+  vi.spyOn(vm, "audioEnabled$", "get").mockReturnValue(constant(true));
+
+  const toggleExpanded = vi.fn();
+  const { container } = render(
+    <TooltipProvider>
+      <SpotlightTile
+        vm={new SpotlightTileViewModel(constant([vm]), constant(false))}
+        targetWidth={300}
+        targetHeight={200}
+        expanded={false}
+        onToggleExpanded={toggleExpanded}
+        showIndicators
+        focusable
+      />
+    </TooltipProvider>,
+  );
+
+  expect(await axe(container)).toHaveNoViolations();
+
+  // Volume menu button should exist
+  expect(screen.queryByRole("button", { name: /volume/i })).toBeInTheDocument();
+});
+
+test("Screen share volume UI is hidden when screen share has no audio", async () => {
+  const vm = mockRemoteScreenShare(
+    mockRtcMembership("@alice:example.org", "AAAA"),
+    {},
+    mockRemoteParticipant({}),
+  );
+
+  vi.spyOn(vm, "audioEnabled$", "get").mockReturnValue(constant(false));
+
+  const toggleExpanded = vi.fn();
+  const { container } = render(
+    <SpotlightTile
+      vm={new SpotlightTileViewModel(constant([vm]), constant(false))}
+      targetWidth={300}
+      targetHeight={200}
+      expanded={false}
+      onToggleExpanded={toggleExpanded}
+      showIndicators
+      focusable
+    />,
+  );
+
+  expect(await axe(container)).toHaveNoViolations();
+
+  // Volume menu button should not exist
+  expect(
+    screen.queryByRole("button", { name: /volume/i }),
+  ).not.toBeInTheDocument();
+});

src/tile/SpotlightTile.tsx

@@ -20,6 +20,10 @@ import {
   CollapseIcon,
   ChevronLeftIcon,
   ChevronRightIcon,
+  VolumeOffIcon,
+  VolumeOnIcon,
+  VolumeOffSolidIcon,
+  VolumeOnSolidIcon,
 } from "@vector-im/compound-design-tokens/assets/web/icons";
 import { animated } from "@react-spring/web";
 import { type Observable, map } from "rxjs";
@@ -27,25 +31,27 @@ import { useObservableRef } from "observable-hooks";
 import { useTranslation } from "react-i18next";
 import classNames from "classnames";
 import { type TrackReferenceOrPlaceholder } from "@livekit/components-core";
+import { Menu, MenuItem } from "@vector-im/compound-web";
 
 import FullScreenMaximiseIcon from "../icons/FullScreenMaximise.svg?react";
 import FullScreenMinimiseIcon from "../icons/FullScreenMinimise.svg?react";
 import { MediaView } from "./MediaView";
 import styles from "./SpotlightTile.module.css";
-import {
-  type EncryptionStatus,
-  LocalUserMediaViewModel,
-  type MediaViewModel,
-  ScreenShareViewModel,
-  type UserMediaViewModel,
-  type RemoteUserMediaViewModel,
-} from "../state/MediaViewModel";
 import { useInitial } from "../useInitial";
 import { useMergedRefs } from "../useMergedRefs";
 import { useReactiveState } from "../useReactiveState";
 import { useLatest } from "../useLatest";
 import { type SpotlightTileViewModel } from "../state/TileViewModel";
 import { useBehavior } from "../useBehavior";
+import { type EncryptionStatus } from "../state/media/MemberMediaViewModel";
+import { type LocalUserMediaViewModel } from "../state/media/LocalUserMediaViewModel";
+import { type RemoteUserMediaViewModel } from "../state/media/RemoteUserMediaViewModel";
+import { type UserMediaViewModel } from "../state/media/UserMediaViewModel";
+import { type ScreenShareViewModel } from "../state/media/ScreenShareViewModel";
+import { type RemoteScreenShareViewModel } from "../state/media/RemoteScreenShareViewModel";
+import { type MediaViewModel } from "../state/media/MediaViewModel";
+import { Slider } from "../Slider";
+import { platform } from "../Platform";
 
 interface SpotlightItemBaseProps {
   ref?: Ref<HTMLDivElement>;
@@ -54,7 +60,6 @@ interface SpotlightItemBaseProps {
   targetWidth: number;
   targetHeight: number;
   video: TrackReferenceOrPlaceholder | undefined;
-  videoEnabled: boolean;
   userId: string;
   unencryptedWarning: boolean;
   encryptionStatus: EncryptionStatus;
@@ -67,6 +72,7 @@ interface SpotlightItemBaseProps {
 
 interface SpotlightUserMediaItemBaseProps extends SpotlightItemBaseProps {
   videoFit: "contain" | "cover";
+  videoEnabled: boolean;
 }
 
 interface SpotlightLocalUserMediaItemProps extends SpotlightUserMediaItemBaseProps {
@@ -105,15 +111,17 @@ const SpotlightUserMediaItem: FC<SpotlightUserMediaItemProps> = ({
   vm,
   ...props
 }) => {
-  const cropVideo = useBehavior(vm.cropVideo$);
+  const videoFit = useBehavior(vm.videoFit$);
+  const videoEnabled = useBehavior(vm.videoEnabled$);
 
   const baseProps: SpotlightUserMediaItemBaseProps &
     RefAttributes<HTMLDivElement> = {
-    videoFit: cropVideo ? "cover" : "contain",
+    videoFit,
+    videoEnabled,
     ...props,
   };
 
-  return vm instanceof LocalUserMediaViewModel ? (
+  return vm.local ? (
     <SpotlightLocalUserMediaItem vm={vm} {...baseProps} />
   ) : (
     <SpotlightRemoteUserMediaItem vm={vm} {...baseProps} />
@@ -122,10 +130,41 @@ const SpotlightUserMediaItem: FC<SpotlightUserMediaItemProps> = ({
 
 SpotlightUserMediaItem.displayName = "SpotlightUserMediaItem";
 
+interface SpotlightScreenShareItemProps extends SpotlightItemBaseProps {
+  vm: ScreenShareViewModel;
+  videoEnabled: boolean;
+}
+
+const SpotlightScreenShareItem: FC<SpotlightScreenShareItemProps> = ({
+  vm,
+  ...props
+}) => {
+  return <MediaView videoFit="contain" mirror={false} {...props} />;
+};
+
+interface SpotlightRemoteScreenShareItemProps extends SpotlightItemBaseProps {
+  vm: RemoteScreenShareViewModel;
+}
+
+const SpotlightRemoteScreenShareItem: FC<
+  SpotlightRemoteScreenShareItemProps
+> = ({ vm, ...props }) => {
+  const videoEnabled = useBehavior(vm.videoEnabled$);
+  return (
+    <SpotlightScreenShareItem vm={vm} videoEnabled={videoEnabled} {...props} />
+  );
+};
+
 interface SpotlightItemProps {
   ref?: Ref<HTMLDivElement>;
   vm: MediaViewModel;
+  /**
+   * The width this tile will have once its animations have settled.
+   */
   targetWidth: number;
+  /**
+   * The height this tile will have once its animations have settled.
+   */
   targetHeight: number;
   focusable: boolean;
   intersectionObserver$: Observable<IntersectionObserver>;
@@ -147,12 +186,21 @@ const SpotlightItem: FC<SpotlightItemProps> = ({
   "aria-hidden": ariaHidden,
 }) => {
   const ourRef = useRef<HTMLDivElement | null>(null);
+
+  // Whenever target bounds change, inform the viewModel
+  useEffect(() => {
+    if (targetWidth > 0 && targetHeight > 0) {
+      if (vm.type != "screen share") {
+        vm.setTargetDimensions(targetWidth, targetHeight);
+      }
+    }
+  }, [targetWidth, targetHeight, vm]);
+
   const ref = useMergedRefs(ourRef, theirRef);
   const focusUrl = useBehavior(vm.focusUrl$);
   const displayName = useBehavior(vm.displayName$);
   const mxcAvatarUrl = useBehavior(vm.mxcAvatarUrl$);
   const video = useBehavior(vm.video$);
-  const videoEnabled = useBehavior(vm.videoEnabled$);
   const unencryptedWarning = useBehavior(vm.unencryptedWarning$);
   const encryptionStatus = useBehavior(vm.encryptionStatus$);
 
@@ -178,7 +226,6 @@ const SpotlightItem: FC<SpotlightItemProps> = ({
     targetWidth,
     targetHeight,
     video: video ?? undefined,
-    videoEnabled,
     userId: vm.userId,
     unencryptedWarning,
     focusUrl,
@@ -189,15 +236,84 @@ const SpotlightItem: FC<SpotlightItemProps> = ({
     "aria-hidden": ariaHidden,
   };
 
-  return vm instanceof ScreenShareViewModel ? (
-    <MediaView videoFit="contain" mirror={false} {...baseProps} />
+  if (vm.type === "user")
+    return <SpotlightUserMediaItem vm={vm} {...baseProps} />;
+  return vm.local ? (
+    <SpotlightScreenShareItem vm={vm} videoEnabled {...baseProps} />
   ) : (
-    <SpotlightUserMediaItem vm={vm} {...baseProps} />
+    <SpotlightRemoteScreenShareItem vm={vm} {...baseProps} />
   );
 };
 
 SpotlightItem.displayName = "SpotlightItem";
 
+interface ScreenShareVolumeButtonProps {
+  vm: RemoteScreenShareViewModel;
+}
+
+const ScreenShareVolumeButton: FC<ScreenShareVolumeButtonProps> = ({ vm }) => {
+  const { t } = useTranslation();
+
+  const audioEnabled = useBehavior(vm.audioEnabled$);
+  const playbackMuted = useBehavior(vm.playbackMuted$);
+  const playbackVolume = useBehavior(vm.playbackVolume$);
+
+  const VolumeIcon = playbackMuted ? VolumeOffIcon : VolumeOnIcon;
+  const VolumeSolidIcon = playbackMuted
+    ? VolumeOffSolidIcon
+    : VolumeOnSolidIcon;
+
+  const [volumeMenuOpen, setVolumeMenuOpen] = useState(false);
+  const onMuteButtonClick = useCallback(() => vm.togglePlaybackMuted(), [vm]);
+  const onVolumeChange = useCallback(
+    (v: number) => vm.adjustPlaybackVolume(v),
+    [vm],
+  );
+  const onVolumeCommit = useCallback(() => vm.commitPlaybackVolume(), [vm]);
+
+  return (
+    audioEnabled && (
+      <Menu
+        open={volumeMenuOpen}
+        onOpenChange={setVolumeMenuOpen}
+        title={t("video_tile.screen_share_volume")}
+        side="top"
+        align="end"
+        trigger={
+          <button
+            className={styles.expand}
+            aria-label={t("video_tile.screen_share_volume")}
+          >
+            <VolumeSolidIcon aria-hidden width={20} height={20} />
+          </button>
+        }
+      >
+        <MenuItem
+          as="div"
+          className={styles.volumeMenuItem}
+          onSelect={null}
+          label={null}
+          hideChevron={true}
+        >
+          <button className={styles.menuMuteButton} onClick={onMuteButtonClick}>
+            <VolumeIcon aria-hidden width={24} height={24} />
+          </button>
+          <Slider
+            className={styles.volumeSlider}
+            label={t("video_tile.volume")}
+            value={playbackVolume}
+            min={0}
+            max={1}
+            step={0.01}
+            onValueChange={onVolumeChange}
+            onValueCommit={onVolumeCommit}
+          />
+        </MenuItem>
+      </Menu>
+    )
+  );
+};
+
 interface Props {
   ref?: Ref<HTMLDivElement>;
   vm: SpotlightTileViewModel;
@@ -232,6 +348,7 @@ export const SpotlightTile: FC<Props> = ({
   const latestMedia = useLatest(media);
   const latestVisibleId = useLatest(visibleId);
   const visibleIndex = media.findIndex((vm) => vm.id === visibleId);
+  const visibleMedia = media.at(visibleIndex);
   const canGoBack = visibleIndex > 0;
   const canGoToNext = visibleIndex !== -1 && visibleIndex < media.length - 1;
 
@@ -339,16 +456,21 @@ export const SpotlightTile: FC<Props> = ({
           />
         ))}
       </div>
-      <div className={styles.bottomRightButtons}>
-        <button
-          className={classNames(styles.expand)}
-          aria-label={"maximise"}
-          onClick={onToggleFullscreen}
-          tabIndex={focusable ? undefined : -1}
-        >
-          <FullScreenIcon aria-hidden width={20} height={20} />
-        </button>
 
+      <div className={styles.bottomRightButtons}>
+        {visibleMedia?.type === "screen share" && !visibleMedia.local && (
+          <ScreenShareVolumeButton vm={visibleMedia} />
+        )}
+        {platform === "desktop" && (
+          <button
+            className={classNames(styles.expand)}
+            aria-label={"maximise"}
+            onClick={onToggleFullscreen}
+            tabIndex={focusable ? undefined : -1}
+          >
+            <FullScreenIcon aria-hidden width={20} height={20} />
+          </button>
+        )}
         {onToggleExpanded && (
           <button
             className={classNames(styles.expand)}

src/useAudioContext.test.tsx

@@ -90,11 +90,18 @@ export const testAudioContext = {
   createStereoPanner: vi.fn().mockReturnValue(panNode),
   close: vi.fn().mockResolvedValue(undefined),
 };
-export const TestAudioContextConstructor = vi.fn(() => testAudioContext);
+
+const TestAudioContext = vi.fn(
+  class {
+    public constructor() {
+      return testAudioContext;
+    }
+  },
+);
 
 let user: UserEvent;
 beforeEach(() => {
-  vi.stubGlobal("AudioContext", TestAudioContextConstructor);
+  vi.stubGlobal("AudioContext", TestAudioContext);
   user = userEvent.setup();
 });
 

src/utils/observable.ts

@@ -61,6 +61,20 @@ export function accumulate<State, Event>(
     events$.pipe(scan(update, initial), startWith(initial));
 }
 
+/**
+ * Given a source of toggle events, creates a Behavior whose value toggles
+ * between `true` and `false`.
+ */
+export function createToggle$(
+  scope: ObservableScope,
+  initialValue: boolean,
+  toggle$: Observable<void>,
+): Behavior<boolean> {
+  return scope.behavior(
+    toggle$.pipe(accumulate(initialValue, (state) => !state)),
+  );
+}
+
 const switchSymbol = Symbol("switch");
 
 /**

src/utils/test.ts

@@ -52,10 +52,6 @@ import {
 } from "matrix-js-sdk/lib/matrixrtc/IKeyTransport";
 import { type CallMembershipIdentityParts } from "matrix-js-sdk/lib/matrixrtc/EncryptionManager";
 
-import {
-  LocalUserMediaViewModel,
-  RemoteUserMediaViewModel,
-} from "../state/MediaViewModel";
 import { E2eeType } from "../e2ee/e2eeType";
 import {
   DEFAULT_CONFIG,
@@ -66,6 +62,18 @@ import { type MediaDevices } from "../state/MediaDevices";
 import { type Behavior, constant } from "../state/Behavior";
 import { ObservableScope } from "../state/ObservableScope";
 import { MuteStates } from "../state/MuteStates";
+import {
+  createLocalUserMedia,
+  type LocalUserMediaViewModel,
+} from "../state/media/LocalUserMediaViewModel";
+import {
+  createRemoteUserMedia,
+  type RemoteUserMediaViewModel,
+} from "../state/media/RemoteUserMediaViewModel";
+import {
+  createRemoteScreenShare,
+  type RemoteScreenShareViewModel,
+} from "../state/media/RemoteScreenShareViewModel";
 
 export function withFakeTimers(continuation: () => void): void {
   vi.useFakeTimers();
@@ -323,30 +331,27 @@ export function mockLocalParticipant(
   } as Partial<LocalParticipant> as LocalParticipant;
 }
 
-export function createLocalMedia(
+export function mockLocalMedia(
   rtcMember: CallMembership,
   roomMember: Partial<RoomMember>,
   localParticipant: LocalParticipant,
   mediaDevices: MediaDevices,
 ): LocalUserMediaViewModel {
   const member = mockMatrixRoomMember(rtcMember, roomMember);
-  return new LocalUserMediaViewModel(
-    testScope(),
-    "local",
-    member.userId,
-    rtcMember.rtcBackendIdentity,
-    constant(localParticipant),
-    {
-      kind: E2eeType.PER_PARTICIPANT,
-    },
-    constant(mockLivekitRoom({ localParticipant })),
-    constant("https://rtc-example.org"),
+  return createLocalUserMedia(testScope(), {
+    id: "local",
+    userId: member.userId,
+    rtcBackendIdentity: rtcMember.rtcBackendIdentity,
+    participant$: constant(localParticipant),
+    encryptionSystem: { kind: E2eeType.PER_PARTICIPANT },
+    livekitRoom$: constant(mockLivekitRoom({ localParticipant })),
+    focusUrl$: constant("https://rtc-example.org"),
     mediaDevices,
-    constant(member.rawDisplayName ?? "nodisplayname"),
-    constant(member.getMxcAvatarUrl()),
-    constant(null),
-    constant(null),
-  );
+    displayName$: constant(member.rawDisplayName ?? "nodisplayname"),
+    mxcAvatarUrl$: constant(member.getMxcAvatarUrl()),
+    handRaised$: constant(null),
+    reaction$: constant(null),
+  });
 }
 
 export function mockRemoteParticipant(
@@ -364,7 +369,7 @@ export function mockRemoteParticipant(
   } as RemoteParticipant;
 }
 
-export function createRemoteMedia(
+export function mockRemoteMedia(
   rtcMember: CallMembership,
   roomMember: Partial<RoomMember>,
   participant: RemoteParticipant | null,
@@ -376,23 +381,45 @@ export function createRemoteMedia(
   ),
 ): RemoteUserMediaViewModel {
   const member = mockMatrixRoomMember(rtcMember, roomMember);
-  return new RemoteUserMediaViewModel(
-    testScope(),
-    "remote",
-    member.userId,
-    rtcMember.rtcBackendIdentity,
-    constant(participant),
+  return createRemoteUserMedia(testScope(), {
+    id: "remote",
+    userId: member.userId,
+    rtcBackendIdentity: rtcMember.rtcBackendIdentity,
+    participant$: constant(participant),
+    encryptionSystem: { kind: E2eeType.PER_PARTICIPANT },
+    livekitRoom$: constant(livekitRoom),
+    focusUrl$: constant("https://rtc-example.org"),
+    pretendToBeDisconnected$: constant(false),
+    displayName$: constant(member.rawDisplayName ?? "nodisplayname"),
+    mxcAvatarUrl$: constant(member.getMxcAvatarUrl()),
+    handRaised$: constant(null),
+    reaction$: constant(null),
+  });
+}
+
+export function mockRemoteScreenShare(
+  rtcMember: CallMembership,
+  roomMember: Partial<RoomMember>,
+  participant: RemoteParticipant | null,
+  livekitRoom: LivekitRoom | undefined = mockLivekitRoom(
+    {},
     {
-      kind: E2eeType.PER_PARTICIPANT,
+      remoteParticipants$: of(participant ? [participant] : []),
     },
-    constant(livekitRoom),
-    constant("https://rtc-example.org"),
-    constant(false),
-    constant(member.rawDisplayName ?? "nodisplayname"),
-    constant(member.getMxcAvatarUrl()),
-    constant(null),
-    constant(null),
-  );
+  ),
+): RemoteScreenShareViewModel {
+  const member = mockMatrixRoomMember(rtcMember, roomMember);
+  return createRemoteScreenShare(testScope(), {
+    id: "screenshare",
+    userId: member.userId,
+    participant$: constant(participant),
+    encryptionSystem: { kind: E2eeType.PER_PARTICIPANT },
+    livekitRoom$: constant(livekitRoom),
+    focusUrl$: constant("https://rtc-example.org"),
+    pretendToBeDisconnected$: constant(false),
+    displayName$: constant(member.rawDisplayName ?? "nodisplayname"),
+    mxcAvatarUrl$: constant(member.getMxcAvatarUrl()),
+  });
 }
 
 export function mockConfig(
@@ -416,10 +443,10 @@ export class MockRTCSession extends TypedEventEmitter<
   public asMockedSession(): MockedObject<MatrixRTCSession> {
     const session = this as unknown as MockedObject<MatrixRTCSession>;
 
-    vi.mocked(session).reemitEncryptionKeys = vi
+    session.reemitEncryptionKeys = vi
       .fn<() => void>()
       .mockReturnValue(undefined);
-    vi.mocked(session).getOldestMembership = vi
+    session.getOldestMembership = vi
       .fn<() => CallMembership | undefined>()
       .mockReturnValue(this.memberships[0]);
 

src/utils/videoFit.test.ts

@@ -0,0 +1,263 @@
+/*
+Copyright 2026 Element Creations Ltd.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import { describe, expect, test, vi } from "vitest";
+import {
+  LocalTrack,
+  type LocalTrackPublication,
+  type RemoteTrackPublication,
+  Track,
+} from "livekit-client";
+
+import { ObservableScope } from "../state/ObservableScope";
+import { videoFit$, videoSizeFromParticipant$ } from "./videoFit";
+import { constant } from "../state/Behavior";
+import {
+  flushPromises,
+  mockLocalParticipant,
+  mockRemoteParticipant,
+} from "./test";
+
+describe("videoFit$ defaults", () => {
+  test.each([
+    {
+      videoSize: { width: 1920, height: 1080 },
+      tileSize: undefined,
+    },
+    {
+      videoSize: { width: 1080, height: 1920 },
+      tileSize: undefined,
+    },
+    {
+      videoSize: undefined,
+      tileSize: { width: 1920, height: 1080 },
+    },
+    {
+      videoSize: undefined,
+      tileSize: { width: 1080, height: 1920 },
+    },
+  ])(
+    "videoFit$ returns `cover` when videoSize is $videoSize and tileSize is $tileSize",
+    ({ videoSize, tileSize }) => {
+      const scope = new ObservableScope();
+      const videoSize$ = constant(videoSize);
+      const tileSize$ = constant(tileSize);
+
+      const fit = videoFit$(scope, videoSize$, tileSize$);
+      expect(fit.value).toBe("cover");
+    },
+  );
+});
+
+const VIDEO_480_L = { width: 640, height: 480 };
+const VIDEO_720_L = { width: 1280, height: 720 };
+const VIDEO_1080_L = { width: 1920, height: 1080 };
+
+// Some sizes from real world testing, which don't match the standard video sizes exactly
+const TILE_SIZE_1_L = { width: 180, height: 135 };
+const TILE_SIZE_3_P = { width: 379, height: 542 };
+const TILE_SIZE_4_L = { width: 957, height: 542 };
+// This is the size of an iPhone Xr in portrait mode
+const TILE_SIZE_5_P = { width: 414, height: 896 };
+
+export function invertSize(size: { width: number; height: number }): {
+  width: number;
+  height: number;
+} {
+  return {
+    width: size.height,
+    height: size.width,
+  };
+}
+
+test.each([
+  {
+    videoSize: VIDEO_480_L,
+    tileSize: TILE_SIZE_1_L,
+    expected: "cover",
+  },
+  {
+    videoSize: invertSize(VIDEO_480_L),
+    tileSize: TILE_SIZE_1_L,
+    expected: "contain",
+  },
+  {
+    videoSize: VIDEO_720_L,
+    tileSize: TILE_SIZE_4_L,
+    expected: "cover",
+  },
+  {
+    videoSize: invertSize(VIDEO_720_L),
+    tileSize: TILE_SIZE_4_L,
+    expected: "contain",
+  },
+  {
+    videoSize: invertSize(VIDEO_1080_L),
+    tileSize: TILE_SIZE_3_P,
+    expected: "cover",
+  },
+  {
+    videoSize: VIDEO_1080_L,
+    tileSize: TILE_SIZE_5_P,
+    expected: "contain",
+  },
+  {
+    videoSize: invertSize(VIDEO_1080_L),
+    tileSize: TILE_SIZE_5_P,
+    expected: "cover",
+  },
+  {
+    // square video
+    videoSize: { width: 400, height: 400 },
+    tileSize: VIDEO_480_L,
+    expected: "contain",
+  },
+  {
+    // Should default to cover if the initial size is 0:0.
+    // Or else it will cause a flash of "contain" mode until the real size is loaded, which can be jarring.
+    videoSize: VIDEO_480_L,
+    tileSize: { width: 0, height: 0 },
+    expected: "cover",
+  },
+  {
+    videoSize: { width: 0, height: 0 },
+    tileSize: VIDEO_480_L,
+    expected: "cover",
+  },
+])(
+  "videoFit$ returns $expected when videoSize is $videoSize and tileSize is $tileSize",
+  ({ videoSize, tileSize, expected }) => {
+    const scope = new ObservableScope();
+    const videoSize$ = constant(videoSize);
+    const tileSize$ = constant(tileSize);
+
+    const fit = videoFit$(scope, videoSize$, tileSize$);
+    expect(fit.value).toBe(expected);
+  },
+);
+
+describe("extracting video size from participant stats", () => {
+  function createMockRtpStats(
+    isInbound: boolean,
+    props: Partial<RTCInboundRtpStreamStats | RTCOutboundRtpStreamStats> = {},
+  ): RTCInboundRtpStreamStats | RTCOutboundRtpStreamStats {
+    const baseStats = {
+      id: "mock-stats-id",
+      timestamp: Date.now(),
+      type: isInbound ? "inbound-rtp" : "outbound-rtp",
+      kind: "video",
+      ...props,
+    };
+
+    return baseStats as RTCInboundRtpStreamStats | RTCOutboundRtpStreamStats;
+  }
+
+  test("get stats for local user", async () => {
+    const localParticipant = mockLocalParticipant({
+      identity: "@local:example.org:AAAAAA",
+    });
+
+    const mockReport: RTCStatsReport = new Map([
+      [
+        "OT01V639885149",
+        createMockRtpStats(false, {
+          frameWidth: 1280,
+          frameHeight: 720,
+        }),
+      ],
+    ]);
+
+    const track = {
+      source: Track.Source.Camera,
+      getRTCStatsReport: vi
+        .fn()
+        .mockImplementation(async () => Promise.resolve(mockReport)),
+    } as Partial<LocalTrack> as LocalTrack;
+
+    // Set up the prototype chain (there is an instanceof check in getRTCStatsReport)
+    Object.setPrototypeOf(track, LocalTrack.prototype);
+
+    localParticipant.getTrackPublication = vi
+      .fn()
+      .mockImplementation((source: Track.Source) => {
+        if (source === Track.Source.Camera) {
+          return {
+            track,
+          } as unknown as LocalTrackPublication;
+        } else {
+          return undefined;
+        }
+      });
+
+    const videoDimensions$ = videoSizeFromParticipant$(
+      constant(localParticipant),
+    );
+
+    const publishedDimensions: { width: number; height: number }[] = [];
+    videoDimensions$.subscribe((dimensions) => {
+      if (dimensions) publishedDimensions.push(dimensions);
+    });
+
+    await flushPromises();
+
+    const dimension = publishedDimensions.pop();
+    expect(dimension).toEqual({ width: 1280, height: 720 });
+  });
+
+  test("get stats for remote user", async () => {
+    // vi.useFakeTimers()
+    const remoteParticipant = mockRemoteParticipant({
+      identity: "@bob:example.org:AAAAAA",
+    });
+
+    const mockReport: RTCStatsReport = new Map([
+      [
+        "OT01V639885149",
+        createMockRtpStats(true, {
+          frameWidth: 480,
+          frameHeight: 640,
+        }),
+      ],
+    ]);
+
+    const track = {
+      source: Track.Source.Camera,
+      getRTCStatsReport: vi
+        .fn()
+        .mockImplementation(async () => Promise.resolve(mockReport)),
+    } as Partial<LocalTrack> as LocalTrack;
+
+    // Set up the prototype chain (there is an instanceof check in getRTCStatsReport)
+    Object.setPrototypeOf(track, LocalTrack.prototype);
+
+    remoteParticipant.getTrackPublication = vi
+      .fn()
+      .mockImplementation((source: Track.Source) => {
+        if (source === Track.Source.Camera) {
+          return {
+            track,
+          } as unknown as RemoteTrackPublication;
+        } else {
+          return undefined;
+        }
+      });
+
+    const videoDimensions$ = videoSizeFromParticipant$(
+      constant(remoteParticipant),
+    );
+
+    const publishedDimensions: { width: number; height: number }[] = [];
+    videoDimensions$.subscribe((dimensions) => {
+      if (dimensions) publishedDimensions.push(dimensions);
+    });
+
+    await flushPromises();
+
+    const dimension = publishedDimensions.pop();
+    expect(dimension).toEqual({ width: 480, height: 640 });
+  });
+});

src/utils/videoFit.ts

@@ -0,0 +1,111 @@
+/*
+Copyright 2026 Element Creations Ltd.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import { combineLatest, map, type Observable, of, switchMap } from "rxjs";
+import {
+  type LocalParticipant,
+  type RemoteParticipant,
+  Track,
+} from "livekit-client";
+
+import { type ObservableScope } from "../state/ObservableScope.ts";
+import { type Behavior } from "../state/Behavior.ts";
+import {
+  observeInboundRtpStreamStats$,
+  observeOutboundRtpStreamStats$,
+} from "../state/media/observeRtpStreamStats";
+
+type Size = {
+  width: number;
+  height: number;
+};
+
+/**
+ * Computes the appropriate video fit mode ("cover" or "contain") based on the aspect ratios of the video and the tile.
+ * - If the video and tile have the same orientation (both landscape or both portrait), we use "cover" to fill the tile, even if it means cropping.
+ * - If the video and tile have different orientations, we use "contain" to ensure the entire video is visible, even if it means letterboxing (black bars).
+ * @param scope - the ObservableScope to create the Behavior in
+ * @param videoSize$ - an Observable of the video size (width and height) or undefined if the size is not yet known (no data yet received).
+ * @param tileSize$ - an Observable of the tile size (width and height) or undefined if the size is not yet known (not yet rendered).
+ */
+export function videoFit$(
+  scope: ObservableScope,
+  videoSize$: Observable<Size | undefined>,
+  tileSize$: Observable<Size | undefined>,
+): Behavior<"cover" | "contain"> {
+  const fit$ = combineLatest([videoSize$, tileSize$]).pipe(
+    map(([videoSize, tileSize]) => {
+      if (!videoSize || !tileSize) {
+        // If we don't have the sizes, default to cover to avoid black bars.
+        // This is a reasonable default as it will ensure the video fills the tile, even if it means cropping.
+        return "cover";
+      }
+      if (
+        videoSize.width === 0 ||
+        videoSize.height === 0 ||
+        tileSize.width === 0 ||
+        tileSize.height === 0
+      ) {
+        // If we have invalid sizes (e.g. width or height is 0), default to cover to avoid black bars.
+        return "cover";
+      }
+      const videoAspectRatio = videoSize.width / videoSize.height;
+      const tileAspectRatio = tileSize.width / tileSize.height;
+
+      // If video is landscape (ratio > 1) and tile is portrait (ratio < 1) or vice versa,
+      // we want to use "contain" (fit) mode to avoid excessive cropping
+      const videoIsLandscape = videoAspectRatio > 1;
+      const tileIsLandscape = tileAspectRatio > 1;
+
+      // If the orientations are the same, use the cover mode (Preserves the aspect ratio, and the image fills the container.)
+      // If they're not the same orientation, use the contain mode (Preserves the aspect ratio, but the image is letterboxed - black bars- to fit within the container.)
+      return videoIsLandscape === tileIsLandscape ? "cover" : "contain";
+    }),
+  );
+
+  return scope.behavior(fit$, "cover");
+}
+
+/**
+ * Helper function to get the video size from a participant.
+ * It observes the participant's video track stats and extracts the frame width and height.
+ * @param participant$ - an Observable of a LocalParticipant or RemoteParticipant, or null if no participant is selected.
+ * @returns an Observable of the video size (width and height) or undefined if the size cannot be determined.
+ */
+export function videoSizeFromParticipant$(
+  participant$: Observable<LocalParticipant | RemoteParticipant | null>,
+): Observable<{ width: number; height: number } | undefined> {
+  return participant$
+    .pipe(
+      // If we have a participant, observe their video track stats. If not, return undefined.
+      switchMap((p) => {
+        if (!p) return of(undefined);
+        if (p.isLocal) {
+          return observeOutboundRtpStreamStats$(p, Track.Source.Camera);
+        } else {
+          return observeInboundRtpStreamStats$(p, Track.Source.Camera);
+        }
+      }),
+    )
+    .pipe(
+      // Extract the frame width and height from the stats. If we don't have valid stats, return undefined.
+      map((stats) => {
+        if (!stats) return undefined;
+        if (
+          // For video tracks, frameWidth and frameHeight should be numbers. If they're not, we can't determine the size.
+          typeof stats.frameWidth !== "number" ||
+          typeof stats.frameHeight !== "number"
+        ) {
+          return undefined;
+        }
+        return {
+          width: stats.frameWidth,
+          height: stats.frameHeight,
+        };
+      }),
+    );
+}