Possible fix for avatars showing stale image data

By giving React a different key for different users' avatars, this hopefully ensures that a new <img> element will be created when the avatar component switches to a different user.

locales/en/app.json

@@ -249,6 +249,8 @@
   "version": "{{productName}} version: {{version}}",
   "video_tile": {
     "always_show": "Always show",
+    "call_ended": "Call ended",
+    "calling": "Calling…",
     "camera_starting": "Video loading...",
     "collapse": "Collapse",
     "expand": "Expand",

playwright/mobile/create-call-mobile.spec.ts

@@ -100,8 +100,16 @@ mobileTest(
         { id: "earpiece", name: "Handset", isEarpiece: true },
         { id: "headphones", name: "Headphones" },
       ]);
-      window.controls.setAudioDevice("earpiece");
     });
+
+    // Open settings to select earpiece
+    await guestPage.getByRole("button", { name: "Settings" }).click();
+    await guestPage.getByText("Handset", { exact: true }).click();
+
+    // dismiss settings
+    await guestPage.locator("#root").getByLabel("Settings").press("Escape");
+
+    await guestPage.pause();
     await expect(
       guestPage.getByRole("heading", { name: "Handset Mode" }),
     ).toBeVisible();

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

@@ -66,7 +66,7 @@ widgetTest("Footer interaction in PiP", async ({ addUser, browserName }) => {
     await iFrame.getByTestId("videoTile").hover();
 
     await expect(audioBtn).toHaveAccessibleName("Unmute microphone");
-    await expect(audioBtn).toBeChecked();
+    await expect(audioBtn).not.toBeChecked();
     await expect(videoBtn).toHaveAccessibleName("Start video");
     await expect(videoBtn).not.toBeChecked();
   }

playwright/widget/test-helpers.ts

@@ -152,6 +152,22 @@ export class TestHelpers {
     }
   }
 
+  public static async maybeDismissKeyBackupToast(page: Page): Promise<void> {
+    const toast = page
+      .locator(".mx_Toast_toast")
+      .getByText("Back up your chats");
+
+    try {
+      await expect(toast).toBeVisible({ timeout: 700 });
+      await page
+        .locator(".mx_Toast_toast")
+        .getByRole("button", { name: "Dismiss" })
+        .click();
+    } catch {
+      // toast not visible, continue as normal
+    }
+  }
+
   public static async createRoom(
     name: string,
     page: Page,
@@ -167,6 +183,7 @@ export class TestHelpers {
     await page.getByRole("button", { name: "Create room" }).click();
     await expect(page.getByText("You created this room.")).toBeVisible();
     await expect(page.getByText("Encryption enabled")).toBeVisible();
+    await TestHelpers.maybeDismissKeyBackupToast(page);
 
     // Invite users if any
     if (andInvite.length > 0) {
@@ -201,6 +218,7 @@ export class TestHelpers {
     await expect(
       page.getByRole("main").getByRole("heading", { name: roomName }),
     ).toBeVisible();
+    await TestHelpers.maybeDismissKeyBackupToast(page);
   }
 
   /**

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

@@ -34,9 +34,12 @@ widgetTest(
       .locator('iframe[title="Element Call"]')
       .contentFrame();
 
-    // We should show a ringing overlay, let's check for that
+    // We should show a ringing tile, let's check for that
     await expect(
-      brooksFrame.getByText(`Waiting for ${whistler.displayName} to join…`),
+      brooksFrame
+        .getByTestId("videoTile")
+        .filter({ has: brooksFrame.getByText(whistler.displayName) })
+        .filter({ has: brooksFrame.getByText("Calling…") }),
     ).toBeVisible();
 
     await expect(whistler.page.getByText("Incoming voice call")).toBeVisible();
@@ -124,9 +127,12 @@ widgetTest(
       .locator('iframe[title="Element Call"]')
       .contentFrame();
 
-    // We should show a ringing overlay, let's check for that
+    // We should show a ringing tile, let's check for that
     await expect(
-      brooksFrame.getByText(`Waiting for ${whistler.displayName} to join…`),
+      brooksFrame
+        .getByTestId("videoTile")
+        .filter({ has: brooksFrame.getByText(whistler.displayName) })
+        .filter({ has: brooksFrame.getByText("Calling…") }),
     ).toBeVisible();
 
     await expect(whistler.page.getByText("Incoming video call")).toBeVisible();
@@ -208,9 +214,12 @@ widgetTest(
       .locator('iframe[title="Element Call"]')
       .contentFrame();
 
-    // We should show a ringing overlay, let's check for that
+    // We should show a ringing tile, let's check for that
     await expect(
-      brooksFrame.getByText(`Waiting for ${whistler.displayName} to join…`),
+      brooksFrame
+        .getByTestId("videoTile")
+        .filter({ has: brooksFrame.getByText(whistler.displayName) })
+        .filter({ has: brooksFrame.getByText("Calling…") }),
     ).toBeVisible();
 
     await expect(whistler.page.getByText("Incoming video call")).toBeVisible();

src/Avatar.test.tsx

@@ -9,14 +9,18 @@ import { afterEach, expect, test, vi } from "vitest";
 import { render, screen } from "@testing-library/react";
 import { type MatrixClient } from "matrix-js-sdk";
 import { type FC, type PropsWithChildren } from "react";
+import { type WidgetApi } from "matrix-widget-api";
 
 import { ClientContextProvider } from "./ClientContext";
 import { Avatar } from "./Avatar";
 import { mockMatrixRoomMember, mockRtcMembership } from "./utils/test";
+import { widget } from "./widget";
 
 const TestComponent: FC<
-  PropsWithChildren<{ client: MatrixClient; supportsThumbnails?: boolean }>
-> = ({ client, children, supportsThumbnails }) => {
+  PropsWithChildren<{
+    client: MatrixClient;
+  }>
+> = ({ client, children }) => {
   return (
     <ClientContextProvider
       value={{
@@ -24,7 +28,6 @@ const TestComponent: FC<
         disconnected: false,
         supportedFeatures: {
           reactions: true,
-          thumbnails: supportsThumbnails ?? true,
         },
         setClient: vi.fn(),
         authenticated: {
@@ -40,6 +43,12 @@ const TestComponent: FC<
   );
 };
 
+vi.mock("./widget", () => ({
+  widget: {
+    api: null, // Ideally we'd only mock this in the as a widget test so the whole module is otherwise null, but just nulling `api` by default works well enough
+  },
+}));
+
 afterEach(() => {
   vi.unstubAllGlobals();
 });
@@ -73,36 +82,7 @@ test("should just render a placeholder when the user has no avatar", () => {
   expect(client.mxcUrlToHttp).toBeCalledTimes(0);
 });
 
-test("should just render a placeholder when thumbnails are not supported", () => {
-  const client = vi.mocked<MatrixClient>({
-    getAccessToken: () => "my-access-token",
-    mxcUrlToHttp: () => vi.fn(),
-  } as unknown as MatrixClient);
-
-  vi.spyOn(client, "mxcUrlToHttp");
-  const member = mockMatrixRoomMember(
-    mockRtcMembership("@alice:example.org", "AAAA"),
-    {
-      getMxcAvatarUrl: () => "mxc://example.org/alice-avatar",
-    },
-  );
-  const displayName = "Alice";
-  render(
-    <TestComponent client={client} supportsThumbnails={false}>
-      <Avatar
-        id={member.userId}
-        name={displayName}
-        size={96}
-        src={member.getMxcAvatarUrl()}
-      />
-    </TestComponent>,
-  );
-  const element = screen.getByRole("img", { name: "@alice:example.org" });
-  expect(element.tagName).toEqual("SPAN");
-  expect(client.mxcUrlToHttp).toBeCalledTimes(0);
-});
-
-test("should attempt to fetch authenticated media", async () => {
+test("should attempt to fetch authenticated media from the server", async () => {
   const expectedAuthUrl = "http://example.org/media/alice-avatar";
   const expectedObjectURL = "my-object-url";
   const accessToken = "my-access-token";
@@ -154,3 +134,47 @@ test("should attempt to fetch authenticated media", async () => {
     headers: { Authorization: `Bearer ${accessToken}` },
   });
 });
+
+test("should attempt to use widget API if running as a widget", async () => {
+  const expectedMXCUrl = "mxc://example.org/alice-avatar";
+  const expectedObjectURL = "my-object-url";
+  const theBlob = new Blob([]);
+
+  // vitest doesn't have a implementation of create/revokeObjectURL, so we need
+  // to delete the property. It's a bit odd, but it works.
+  Reflect.deleteProperty(global.window.URL, "createObjectURL");
+  globalThis.URL.createObjectURL = vi.fn().mockReturnValue(expectedObjectURL);
+  Reflect.deleteProperty(global.window.URL, "revokeObjectURL");
+  globalThis.URL.revokeObjectURL = vi.fn();
+
+  const client = vi.mocked<MatrixClient>({
+    getAccessToken: () => undefined,
+  } as unknown as MatrixClient);
+
+  widget!.api = { downloadFile: vi.fn() } as unknown as WidgetApi;
+  vi.spyOn(widget!.api, "downloadFile").mockResolvedValue({ file: theBlob });
+  const member = mockMatrixRoomMember(
+    mockRtcMembership("@alice:example.org", "AAAA"),
+    {
+      getMxcAvatarUrl: () => expectedMXCUrl,
+    },
+  );
+  const displayName = "Alice";
+  render(
+    <TestComponent client={client}>
+      <Avatar
+        id={member.userId}
+        name={displayName}
+        size={96}
+        src={member.getMxcAvatarUrl()}
+      />
+    </TestComponent>,
+  );
+
+  // Fetch is asynchronous, so wait for this to resolve.
+  await vi.waitUntil(() =>
+    document.querySelector(`img[src='${expectedObjectURL}']`),
+  );
+
+  expect(widget!.api.downloadFile).toBeCalledWith(expectedMXCUrl);
+});

src/Avatar.tsx

@@ -14,8 +14,10 @@ import {
 } from "react";
 import { Avatar as CompoundAvatar } from "@vector-im/compound-web";
 import { type MatrixClient } from "matrix-js-sdk";
+import { type WidgetApi } from "matrix-widget-api";
 
 import { useClientState } from "./ClientContext";
+import { widget } from "./widget";
 
 export enum Size {
   XS = "xs",
@@ -78,50 +80,54 @@ export const Avatar: FC<Props> = ({
   const sizePx = useMemo(
     () =>
       Object.values(Size).includes(size as Size)
-        ? sizes.get(size as Size)
+        ? sizes.get(size as Size)!
         : (size as number),
     [size],
   );
 
   const [avatarUrl, setAvatarUrl] = useState<string | undefined>(undefined);
 
+  // In theory, a change in `clientState` or `sizePx` could run extra getAvatarFromWidgetAPI calls, but in practice they should be stable long before this code runs.
   useEffect(() => {
-    if (clientState?.state !== "valid") {
-      return;
-    }
-    const { authenticated, supportedFeatures } = clientState;
-    const client = authenticated?.client;
-
-    if (!client || !src || !sizePx || !supportedFeatures.thumbnails) {
+    if (!src) {
+      setAvatarUrl(undefined);
       return;
     }
 
-    const token = client.getAccessToken();
-    if (!token) {
-      return;
-    }
-    const resolveSrc = getAvatarUrl(client, src, sizePx);
-    if (!resolveSrc) {
+    let blob: Promise<Blob>;
+
+    if (widget?.api) {
+      blob = getAvatarFromWidgetAPI(widget.api, src);
+    } else if (
+      clientState?.state === "valid" &&
+      clientState.authenticated?.client &&
+      sizePx
+    ) {
+      blob = getAvatarFromServer(clientState.authenticated.client, src, sizePx);
+    } else {
       setAvatarUrl(undefined);
       return;
     }
 
     let objectUrl: string | undefined;
-    fetch(resolveSrc, {
-      headers: {
-        Authorization: `Bearer ${token}`,
-      },
-    })
-      .then(async (req) => req.blob())
+    let stale = false;
+    blob
       .then((blob) => {
+        if (stale) {
+          return;
+        }
         objectUrl = URL.createObjectURL(blob);
         setAvatarUrl(objectUrl);
       })
       .catch((ex) => {
+        if (stale) {
+          return;
+        }
         setAvatarUrl(undefined);
       });
 
     return (): void => {
+      stale = true;
       if (objectUrl) {
         URL.revokeObjectURL(objectUrl);
       }
@@ -130,6 +136,7 @@ export const Avatar: FC<Props> = ({
 
   return (
     <CompoundAvatar
+      key={id}
       className={className}
       id={id}
       name={name}
@@ -140,3 +147,44 @@ export const Avatar: FC<Props> = ({
     />
   );
 };
+
+async function getAvatarFromServer(
+  client: MatrixClient,
+  src: string,
+  sizePx: number,
+): Promise<Blob> {
+  const httpSrc = getAvatarUrl(client, src, sizePx);
+  if (!httpSrc) {
+    throw new Error("Failed to get http avatar URL");
+  }
+
+  const token = client.getAccessToken();
+  if (!token) {
+    throw new Error("Failed to get access token");
+  }
+
+  const request = await fetch(httpSrc, {
+    headers: {
+      Authorization: `Bearer ${token}`,
+    },
+  });
+
+  const blob = await request.blob();
+
+  return blob;
+}
+
+async function getAvatarFromWidgetAPI(
+  api: WidgetApi,
+  src: string,
+): Promise<Blob> {
+  const response = await api.downloadFile(src);
+  const file = response.file;
+
+  // element-web sends a Blob, and the MSC4039 is considering changing the spec to strictly Blob, so only handling that
+  if (!(file instanceof Blob)) {
+    throw new Error("Downloaded file is not a Blob");
+  }
+
+  return file;
+}

src/ClientContext.tsx

@@ -48,7 +48,6 @@ export type ValidClientState = {
   disconnected: boolean;
   supportedFeatures: {
     reactions: boolean;
-    thumbnails: boolean;
   };
   setClient: (client: MatrixClient, session: Session) => void;
 };
@@ -249,7 +248,6 @@ export const ClientProvider: FC<Props> = ({ children }) => {
 
   const [isDisconnected, setIsDisconnected] = useState(false);
   const [supportsReactions, setSupportsReactions] = useState(false);
-  const [supportsThumbnails, setSupportsThumbnails] = useState(false);
 
   const state: ClientState | undefined = useMemo(() => {
     if (alreadyOpenedErr) {
@@ -275,7 +273,6 @@ export const ClientProvider: FC<Props> = ({ children }) => {
       disconnected: isDisconnected,
       supportedFeatures: {
         reactions: supportsReactions,
-        thumbnails: supportsThumbnails,
       },
     };
   }, [
@@ -286,7 +283,6 @@ export const ClientProvider: FC<Props> = ({ children }) => {
     setClient,
     isDisconnected,
     supportsReactions,
-    supportsThumbnails,
   ]);
 
   const onSync = useCallback(
@@ -312,8 +308,6 @@ export const ClientProvider: FC<Props> = ({ children }) => {
     }
 
     if (initClientState.widgetApi) {
-      // There is currently no widget API for authenticated media thumbnails.
-      setSupportsThumbnails(false);
       const reactSend = initClientState.widgetApi.hasCapability(
         "org.matrix.msc2762.send.event:m.reaction",
       );
@@ -335,7 +329,6 @@ export const ClientProvider: FC<Props> = ({ children }) => {
       }
     } else {
       setSupportsReactions(true);
-      setSupportsThumbnails(true);
     }
 
     return (): void => {

src/__snapshots__/AppBar.test.tsx.snap

@@ -14,7 +14,7 @@ exports[`AppBar > renders 1`] = `
       >
         <button
           aria-labelledby="_r_0_"
-          class="_icon-button_1pz9o_8"
+          class="_icon-button_1215g_8"
           data-kind="primary"
           role="button"
           style="--cpd-icon-button-size: 32px;"

src/controls.ts

@@ -33,12 +33,38 @@ export interface Controls {
   showNativeOutputDevicePicker?: () => void;
 }
 
+/**
+ * Output Audio device when using the controlled audio output mode (mobile).
+ */
 export interface OutputDevice {
   id: string;
   name: string;
+  /**
+   * `forEarpiece` in an iOS only flag, that will be set on the default speaker device.
+   * The default speaker device will be used for the earpiece mode by
+   * using a stereo pan and reducing the volume significantly. (in combination this is similar to a dedicated earpiece mode)
+   *  - on iOS this is true if output is routed to speaker.
+   *    In that case then ElementCalls manually appends an earpiece device with id `EARPIECE_CONFIG_ID` and `{ type: "earpiece" }`
+   *  - on Android this is unused.
+   */
   forEarpiece?: boolean;
+  /**
+   * Is the device the OS earpiece audio configuration?
+   * - on iOS always undefined
+   * - on Android true for the `TYPE_BUILTIN_EARPIECE`
+   */
   isEarpiece?: boolean;
+  /**
+   *  Is the device the OS default speaker:
+   * - on iOS always true if output is routed to speaker. In other case iOS on declare a `dummy` id device.
+   * - on Android true for the `TYPE_BUILTIN_SPEAKER`
+   */
   isSpeaker?: boolean;
+  /**
+   * Is the device the OS default external headset (bluetooth):
+   * - on iOS always undefined.
+   * - on Android true for the `TYPE_BLUETOOTH_SCO`
+   */
   isExternalHeadset?: boolean;
 }
 
@@ -47,8 +73,16 @@ export interface OutputDevice {
  */
 export const setPipEnabled$ = new Subject<boolean>();
 
+/**
+ * Stores the list of available controlled audio output devices.
+ * This is set when the native code calls `setAvailableAudioDevices` with the list of available audio output devices.
+ */
 export const availableOutputDevices$ = new Subject<OutputDevice[]>();
 
+/**
+ * Stores the current audio output device id.
+ * This is set when the native code calls `setAudioDevice`
+ */
 export const outputDevice$ = new Subject<string>();
 
 /**
@@ -80,16 +114,41 @@ window.controls = {
     setPipEnabled$.next(false);
   },
 
+  /**
+   * Reverse engineered:
+   *
+   * - on iOS:
+   *  This always a list of one thing. If current route output is speaker it returns
+   *  the single `{"id":"Speaker","name":"Speaker","forEarpiece":true,"isSpeaker":true}` Notice that EC will
+   *  also manually add a virtual earpiece device with id `EARPIECE_CONFIG_ID` and `{ type: "earpiece" }`.
+   *  If the route output is not speaker then it will be `{id: 'dummy', name: 'dummy'}`
+   *
+   *
+   *  - on Android:
+   *  This is a list of all available output audio devices. The `id` is the Android AudioDeviceInfo.getId()
+   *  and the `name` is based the Android AudioDeviceInfo.productName (mapped to static strings for known types)
+   *  The `isEarpiece`, `isSpeaker` and `isExternalHeadset` are set based on the Android AudioDeviceInfo.type
+   *  matching the corresponding types for earpiece, speaker and bluetooth headset.
+   */
   setAvailableAudioDevices(devices: OutputDevice[]): void {
-    logger.info("setAvailableAudioDevices called from native:", devices);
+    logger.info(
+      "[MediaDevices controls] setAvailableAudioDevices called from native:",
+      devices,
+    );
     availableOutputDevices$.next(devices);
   },
   setAudioDevice(id: string): void {
-    logger.info("setAudioDevice called from native", id);
+    logger.info(
+      "[MediaDevices controls] setAudioDevice called from native",
+      id,
+    );
     outputDevice$.next(id);
   },
   setAudioEnabled(enabled: boolean): void {
-    logger.info("setAudioEnabled called from native:", enabled);
+    logger.info(
+      "[MediaDevices controls] setAudioEnabled called from native:",
+      enabled,
+    );
     if (!setAudioEnabled$.observed)
       throw new Error(
         "Output controls are disabled. No setAudioEnabled$ observer",

src/grid/OneOnOneLayout.tsx

@@ -51,15 +51,15 @@ export const makeOneOnOneLayout: CallLayout<OneOnOneLayoutModel> = ({
     return (
       <div ref={ref} className={styles.layer}>
         <Slot
-          id={model.remote.id}
-          model={model.remote}
+          id={model.spotlight.id}
+          model={model.spotlight}
           className={styles.container}
           style={{ width: tileWidth, height: tileHeight }}
         >
           <Slot
             className={classNames(styles.slot, styles.local)}
-            id={model.local.id}
-            model={model.local}
+            id={model.pip.id}
+            model={model.pip}
             onDrag={onDragLocalTile}
             data-block-alignment={pipAlignmentValue.block}
             data-inline-alignment={pipAlignmentValue.inline}

src/main.tsx

@@ -67,6 +67,6 @@ Initializer.initBeforeReact()
     );
   })
   .catch((e) => {
-    logger.error("Failed to initialize app", e);
+    logger.error(`Failed to initialize app ${e.message}`, e);
     root.render(e.message);
   });

src/room/InCallView.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 { IconButton, Text, Tooltip } from "@vector-im/compound-web";
+import { IconButton, Tooltip } from "@vector-im/compound-web";
 import { type MatrixClient, type Room as MatrixRoom } from "matrix-js-sdk";
 import {
   type FC,
@@ -98,8 +98,6 @@ import { useAppBarHidden, useAppBarSecondaryButton } from "../AppBar.tsx";
 import { useBehavior } from "../useBehavior.ts";
 import { Toast } from "../Toast.tsx";
 import overlayStyles from "../Overlay.module.css";
-import { Avatar, Size as AvatarSize } from "../Avatar";
-import waitingStyles from "./WaitingForJoin.module.css";
 import { prefetchSounds } from "../soundUtils";
 import { useAudioContext } from "../useAudioContext";
 import ringtoneMp3 from "../sound/ringtone.mp3?url";
@@ -107,6 +105,7 @@ import ringtoneOgg from "../sound/ringtone.ogg?url";
 import { useTrackProcessorObservable$ } from "../livekit/TrackProcessorContext.tsx";
 import { type Layout } from "../state/layout-types.ts";
 import { ObservableScope } from "../state/ObservableScope.ts";
+import { useLatest } from "../useLatest.ts";
 
 const logger = rootLogger.getChild("[InCallView]");
 
@@ -224,8 +223,6 @@ export const InCallView: FC<InCallViewProps> = ({
   const { showControls } = useUrlParams();
 
   const muteAllAudio = useBehavior(muteAllAudio$);
-  // Call pickup state and display names are needed for waiting overlay/sounds
-  const callPickupState = useBehavior(vm.callPickupState$);
 
   // Preload a waiting and decline sounds
   const pickupPhaseSoundCache = useInitial(async () => {
@@ -239,6 +236,7 @@ export const InCallView: FC<InCallViewProps> = ({
     latencyHint: "interactive",
     muted: muteAllAudio,
   });
+  const latestPickupPhaseAudio = useLatest(pickupPhaseAudio);
 
   const audioEnabled = useBehavior(muteStates.audio.enabled$);
   const videoEnabled = useBehavior(muteStates.video.enabled$);
@@ -257,6 +255,7 @@ export const InCallView: FC<InCallViewProps> = ({
     () => void toggleRaisedHand(),
   );
 
+  const ringing = useBehavior(vm.ringing$);
   const audioParticipants = useBehavior(vm.livekitRoomItems$);
   const participantCount = useBehavior(vm.participantCount$);
   const reconnecting = useBehavior(vm.reconnecting$);
@@ -271,7 +270,6 @@ export const InCallView: FC<InCallViewProps> = ({
   const audioOutputSwitcher = useBehavior(vm.audioOutputSwitcher$);
   const sharingScreen = useBehavior(vm.sharingScreen$);
 
-  const ringOverlay = useBehavior(vm.ringOverlay$);
   const fatalCallError = useBehavior(vm.fatalError$);
   // Stop the rendering and throw for the error boundary
   if (fatalCallError) {
@@ -279,58 +277,21 @@ export const InCallView: FC<InCallViewProps> = ({
     throw fatalCallError;
   }
 
-  // We need to set the proper timings on the animation based upon the sound length.
-  const ringDuration = pickupPhaseAudio?.soundDuration["waiting"] ?? 1;
-  useEffect((): (() => void) => {
-    // The CSS animation includes the delay, so we must double the length of the sound.
-    window.document.body.style.setProperty(
-      "--call-ring-duration-s",
-      `${ringDuration * 2}s`,
-    );
-    window.document.body.style.setProperty(
-      "--call-ring-delay-s",
-      `${ringDuration}s`,
-    );
-    // Remove properties when we unload.
-    return () => {
-      window.document.body.style.removeProperty("--call-ring-duration-s");
-      window.document.body.style.removeProperty("--call-ring-delay-s");
-    };
-  }, [pickupPhaseAudio?.soundDuration, ringDuration]);
-
-  // When waiting for pickup, loop a waiting sound
+  // While ringing, loop the ringtone
   useEffect((): void | (() => void) => {
-    if (callPickupState !== "ringing" || !pickupPhaseAudio) return;
-    const endSound = pickupPhaseAudio.playSoundLooping("waiting", ringDuration);
-    return () => {
-      void endSound().catch((e) => {
-        logger.error("Failed to stop ringing sound", e);
-      });
-    };
-  }, [callPickupState, pickupPhaseAudio, ringDuration]);
-
-  // Waiting UI overlay
-  const waitingOverlay: JSX.Element | null = useMemo(() => {
-    return ringOverlay ? (
-      <div className={classNames(overlayStyles.bg, waitingStyles.overlay)}>
-        <div
-          className={classNames(overlayStyles.content, waitingStyles.content)}
-        >
-          <div className={waitingStyles.pulse}>
-            <Avatar
-              id={ringOverlay.idForAvatar}
-              name={ringOverlay.name}
-              src={ringOverlay.avatarMxc}
-              size={AvatarSize.XL}
-            />
-          </div>
-          <Text size="md" className={waitingStyles.text}>
-            {ringOverlay.text}
-          </Text>
-        </div>
-      </div>
-    ) : null;
-  }, [ringOverlay]);
+    const audio = latestPickupPhaseAudio.current;
+    if (ringing && audio) {
+      const endSound = audio.playSoundLooping(
+        "waiting",
+        audio.soundDuration["waiting"] ?? 1,
+      );
+      return () => {
+        void endSound().catch((e) => {
+          logger.error("Failed to stop ringing sound", e);
+        });
+      };
+    }
+  }, [ringing, latestPickupPhaseAudio]);
 
   const onViewClick = useCallback(
     (e: ReactMouseEvent) => {
@@ -764,7 +725,6 @@ export const InCallView: FC<InCallViewProps> = ({
       {reconnectingToast}
       {earpieceOverlay}
       <ReactionsOverlay vm={vm} />
-      {waitingOverlay}
       {footer}
       {layout.type !== "pip" && (
         <>

src/room/WaitingForJoin.module.css

@@ -1,61 +0,0 @@
-.overlay {
-  position: absolute;
-  inset: 0;
-  display: flex;
-  align-items: center;
-  justify-content: center;
-  pointer-events: none;
-}
-
-.content {
-  display: flex;
-  flex-direction: column;
-  align-items: center;
-  gap: 16px;
-}
-
-.pulse {
-  position: relative;
-  height: 90px;
-}
-
-.pulse::before {
-  content: "";
-  position: absolute;
-  inset: -12px;
-  border-radius: 9999px;
-  border: 12px solid rgba(255, 255, 255, 0.6);
-  animation: pulse var(--call-ring-duration-s) ease-out infinite;
-  animation-delay: 1s;
-  opacity: 0;
-}
-
-.text {
-  color: var(--cpd-color-text-on-solid-primary);
-}
-
-@keyframes pulse {
-  0% {
-    transform: scale(0.95);
-    opacity: 0.7;
-    transform: scale(0);
-    opacity: 1;
-  }
-  35% {
-    transform: scale(1.15);
-    opacity: 0.15;
-  }
-  50% {
-    transform: scale(1.2);
-    opacity: 0;
-  }
-  50.01% {
-    transform: scale(0);
-  }
-  85% {
-    transform: scale(0);
-  }
-  100% {
-    transform: scale(0);
-  }
-}

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

@@ -128,24 +128,11 @@ exports[`InCallView > rendering > renders 1`] = `
           width="1em"
           xmlns="http://www.w3.org/2000/svg"
         >
-          <g
-            clip-path="url(#a)"
-          >
-            <path
-              clip-rule="evenodd"
-              d="M8.929 15.1a13.6 13.6 0 0 0 4.654 3.066q2.62 1.036 5.492.923h.008l.003-.004.003-.002-.034-3.124-3.52-.483-1.791 1.792-.645-.322a13.5 13.5 0 0 1-3.496-2.52 13.4 13.4 0 0 1-2.52-3.496l-.322-.644 1.792-1.792-.483-3.519-3.123-.034-.003.002-.003.004v.002a13.65 13.65 0 0 0 .932 5.492A13.4 13.4 0 0 0 8.93 15.1m3.92 4.926a15.6 15.6 0 0 1-5.334-3.511 15.4 15.4 0 0 1-3.505-5.346 15.6 15.6 0 0 1-1.069-6.274 1.93 1.93 0 0 1 .589-1.366c.366-.366.84-.589 1.386-.589h.01l3.163.035a1.96 1.96 0 0 1 1.958 1.694v.005l.487 3.545v.003c.043.297.025.605-.076.907a2 2 0 0 1-.485.773l-.762.762a11.4 11.4 0 0 0 3.206 3.54q.457.33.948.614l.762-.761a2 2 0 0 1 .774-.486c.302-.1.61-.118.907-.076l3.553.487a1.96 1.96 0 0 1 1.694 1.958l.034 3.174c0 .546-.223 1.02-.588 1.386-.361.36-.827.582-1.363.588a15.3 15.3 0 0 1-6.29-1.062"
-              fill-rule="evenodd"
-            />
-          </g>
-          <defs>
-            <clippath
-              id="a"
-            >
-              <path
-                d="M0 0h24v24H0z"
-              />
-            </clippath>
-          </defs>
+          <path
+            clip-rule="evenodd"
+            d="M8.929 15.1a13.6 13.6 0 0 0 4.654 3.066q2.62 1.036 5.492.923h.008l.003-.004.003-.002-.034-3.124-3.52-.483-1.791 1.792-.645-.322a13.5 13.5 0 0 1-3.496-2.52 13.4 13.4 0 0 1-2.52-3.496l-.322-.645 1.792-1.791-.483-3.52-3.123-.033-.003.002-.003.004v.002a13.65 13.65 0 0 0 .932 5.492A13.4 13.4 0 0 0 8.93 15.1m3.92 4.926a15.6 15.6 0 0 1-5.334-3.511 15.4 15.4 0 0 1-3.505-5.346 15.6 15.6 0 0 1-1.069-6.274 1.93 1.93 0 0 1 .589-1.366c.366-.366.84-.589 1.386-.589h.01l3.163.035a1.96 1.96 0 0 1 1.958 1.694v.005l.487 3.545v.003c.043.297.025.605-.076.907a2 2 0 0 1-.485.773l-.762.762a11.3 11.3 0 0 0 1.806 2.348 11.4 11.4 0 0 0 2.348 1.806l.762-.762a2 2 0 0 1 .774-.485c.302-.1.61-.118.907-.076l3.553.487a1.96 1.96 0 0 1 1.694 1.958l.034 3.174c0 .546-.223 1.02-.588 1.386-.36.36-.827.582-1.363.588a15.3 15.3 0 0 1-6.29-1.062"
+            fill-rule="evenodd"
+          />
         </svg>
       </div>
       <h2

src/settings/__snapshots__/DeveloperSettingsTab.test.tsx.snap

@@ -234,12 +234,12 @@ exports[`DeveloperSettingsTab > renders and matches snapshot 1`] = `
         class="_inline-field-control_19upo_44"
       >
         <div
-          class="_container_1qhtc_10"
+          class="_container_1ug7n_10"
         >
           <input
             aria-describedby="radix-_r_9_ radix-_r_b_ radix-_r_d_"
             checked=""
-            class="_input_1qhtc_18"
+            class="_input_1ug7n_18"
             id="radix-_r_8_"
             name="_r_0_"
             title=""
@@ -247,7 +247,7 @@ exports[`DeveloperSettingsTab > renders and matches snapshot 1`] = `
             value="legacy"
           />
           <div
-            class="_ui_1qhtc_19"
+            class="_ui_1ug7n_19"
           />
         </div>
       </div>
@@ -275,11 +275,11 @@ exports[`DeveloperSettingsTab > renders and matches snapshot 1`] = `
         class="_inline-field-control_19upo_44"
       >
         <div
-          class="_container_1qhtc_10"
+          class="_container_1ug7n_10"
         >
           <input
             aria-describedby="radix-_r_9_ radix-_r_b_ radix-_r_d_"
-            class="_input_1qhtc_18"
+            class="_input_1ug7n_18"
             id="radix-_r_a_"
             name="_r_0_"
             title=""
@@ -287,7 +287,7 @@ exports[`DeveloperSettingsTab > renders and matches snapshot 1`] = `
             value="compatibility"
           />
           <div
-            class="_ui_1qhtc_19"
+            class="_ui_1ug7n_19"
           />
         </div>
       </div>
@@ -315,11 +315,11 @@ exports[`DeveloperSettingsTab > renders and matches snapshot 1`] = `
         class="_inline-field-control_19upo_44"
       >
         <div
-          class="_container_1qhtc_10"
+          class="_container_1ug7n_10"
         >
           <input
             aria-describedby="radix-_r_9_ radix-_r_b_ radix-_r_d_"
-            class="_input_1qhtc_18"
+            class="_input_1ug7n_18"
             id="radix-_r_c_"
             name="_r_0_"
             title=""
@@ -327,7 +327,7 @@ exports[`DeveloperSettingsTab > renders and matches snapshot 1`] = `
             value="matrix_2_0"
           />
           <div
-            class="_ui_1qhtc_19"
+            class="_ui_1ug7n_19"
           />
         </div>
       </div>

src/settings/useSubmitRageshake.test.tsx

@@ -78,7 +78,6 @@ function renderWithMockClient(
         disconnected: false,
         supportedFeatures: {
           reactions: true,
-          thumbnails: true,
         },
         setClient: vi.fn(),
         authenticated: {

src/state/AndroidControlledAudioOutput.test.ts

@@ -0,0 +1,563 @@
+/*
+Copyright 2026 Element Corp.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import { it, vi, expect, beforeEach, afterEach, describe } from "vitest";
+import { firstValueFrom, of, Subject, take, toArray } from "rxjs";
+import { type RTCCallIntent } from "matrix-js-sdk/lib/matrixrtc";
+
+import { AndroidControlledAudioOutput } from "./AndroidControlledAudioOutput.ts";
+import type { Controls, OutputDevice } from "../controls";
+import { ObservableScope } from "./ObservableScope";
+import { withTestScheduler } from "../utils/test";
+
+// All the following device types are real device types that have been observed in the wild on Android devices,
+// gathered from logs.
+// There are no BT Speakers because they are currently filtered out by EXA (native layer)
+
+// A device type describing the speaker system (i.e. a mono speaker or stereo speakers) built in a device.
+const SPEAKER_DEVICE: OutputDevice = {
+  id: "3",
+  name: "Built-in speaker",
+  isEarpiece: false,
+  isSpeaker: true,
+  isExternalHeadset: false,
+};
+
+// A device type describing the attached earphone speaker.
+const EARPIECE_DEVICE: OutputDevice = {
+  id: "2",
+  name: "Built-in earpiece",
+  isEarpiece: true,
+  isSpeaker: false,
+  isExternalHeadset: false,
+};
+
+// A device type describing a Bluetooth device typically used for telephony
+const BT_HEADSET_DEVICE: OutputDevice = {
+  id: "2226",
+  name: "Bluetooth - OpenMove by Shokz",
+  isEarpiece: false,
+  isSpeaker: false,
+  isExternalHeadset: true,
+};
+
+// A device type describing a USB audio headset.
+const USB_HEADSET_DEVICE: OutputDevice = {
+  id: "29440",
+  name: "USB headset - USB-Audio - AB13X USB Audio",
+  isEarpiece: false,
+  isSpeaker: false,
+  isExternalHeadset: false,
+};
+
+// A device type describing a headset, which is the combination of a headphones and microphone
+const WIRED_HEADSET_DEVICE: OutputDevice = {
+  id: "54509",
+  name: "Wired headset - 23117RA68G",
+  isEarpiece: false,
+  isSpeaker: false,
+  isExternalHeadset: false,
+};
+
+// A device type describing a pair of wired headphones
+const WIRED_HEADPHONE_DEVICE: OutputDevice = {
+  id: "679",
+  name: "Wired headphones - TB02",
+  isEarpiece: false,
+  isSpeaker: false,
+  isExternalHeadset: false,
+};
+
+/**
+ * The base device list that is always present on Android devices.
+ * This list is ordered by the OS, the speaker is listed before the earpiece.
+ */
+const BASE_DEVICE_LIST = [SPEAKER_DEVICE, EARPIECE_DEVICE];
+
+const BT_HEADSET_BASE_DEVICE_LIST = [BT_HEADSET_DEVICE, ...BASE_DEVICE_LIST];
+
+const WIRED_HEADSET_BASE_DEVICE_LIST = [
+  WIRED_HEADSET_DEVICE,
+  ...BASE_DEVICE_LIST,
+];
+
+/**
+ * A full device list containing all the observed device types in the wild on Android devices.
+ * Ordered as they would be ordered by the OS.
+ */
+const FULL_DEVICE_LIST = [
+  BT_HEADSET_DEVICE,
+  USB_HEADSET_DEVICE,
+  WIRED_HEADSET_DEVICE,
+  WIRED_HEADPHONE_DEVICE,
+  ...BASE_DEVICE_LIST,
+];
+
+let testScope: ObservableScope;
+let mockControls: Controls;
+
+beforeEach(() => {
+  testScope = new ObservableScope();
+  mockControls = {
+    onAudioDeviceSelect: vi.fn(),
+    onOutputDeviceSelect: vi.fn(),
+  } as unknown as Controls;
+});
+
+afterEach(() => {
+  testScope.end();
+});
+
+describe("Default selection", () => {
+  it("Default to speaker for video calls", async () => {
+    const controlledAudioOutput = new AndroidControlledAudioOutput(
+      of(BASE_DEVICE_LIST),
+      testScope,
+      "video",
+      mockControls,
+    );
+
+    const emissions = await firstValueFrom(
+      controlledAudioOutput.selected$.pipe(take(1), toArray()),
+    );
+
+    expect(emissions).toEqual([
+      { id: SPEAKER_DEVICE.id, virtualEarpiece: false },
+    ]);
+
+    [
+      mockControls.onAudioDeviceSelect,
+      mockControls.onOutputDeviceSelect,
+    ].forEach((mockFn) => {
+      expect(mockFn).toHaveBeenCalledTimes(1);
+      expect(mockFn).toHaveBeenCalledWith(SPEAKER_DEVICE.id);
+    });
+  });
+
+  it("Default to earpiece for audio calls for base config", async () => {
+    const controlledAudioOutput = new AndroidControlledAudioOutput(
+      of(BASE_DEVICE_LIST),
+      testScope,
+      "audio",
+      mockControls,
+    );
+
+    const emissions = await firstValueFrom(
+      controlledAudioOutput.selected$.pipe(take(1), toArray()),
+    );
+
+    expect(emissions).toEqual([
+      { id: EARPIECE_DEVICE.id, virtualEarpiece: false },
+    ]);
+
+    [
+      mockControls.onAudioDeviceSelect,
+      mockControls.onOutputDeviceSelect,
+    ].forEach((mockFn) => {
+      expect(mockFn).toHaveBeenCalledTimes(1);
+      expect(mockFn).toHaveBeenCalledWith(EARPIECE_DEVICE.id);
+    });
+  });
+
+  ["audio", "video"].forEach((callIntent) => {
+    it(`Default to BT headset for ${callIntent} calls if present`, async () => {
+      const controlledAudioOutput = new AndroidControlledAudioOutput(
+        of(BT_HEADSET_BASE_DEVICE_LIST),
+        testScope,
+        callIntent,
+        mockControls,
+      );
+
+      const emissions = await firstValueFrom(
+        controlledAudioOutput.selected$.pipe(take(1), toArray()),
+      );
+
+      expect(emissions).toEqual([
+        { id: BT_HEADSET_DEVICE.id, virtualEarpiece: false },
+      ]);
+
+      [
+        mockControls.onAudioDeviceSelect,
+        mockControls.onOutputDeviceSelect,
+      ].forEach((mockFn) => {
+        expect(mockFn).toHaveBeenCalledTimes(1);
+        expect(mockFn).toHaveBeenCalledWith(BT_HEADSET_DEVICE.id);
+      });
+    });
+  });
+
+  ["audio", "video"].forEach((callIntent) => {
+    it(`Default to wired headset for ${callIntent} calls if present`, async () => {
+      const controlledAudioOutput = new AndroidControlledAudioOutput(
+        of(WIRED_HEADSET_BASE_DEVICE_LIST),
+        testScope,
+        callIntent,
+        mockControls,
+      );
+
+      const emissions = await firstValueFrom(
+        controlledAudioOutput.selected$.pipe(take(1), toArray()),
+      );
+
+      expect(emissions).toEqual([
+        { id: WIRED_HEADSET_DEVICE.id, virtualEarpiece: false },
+      ]);
+
+      expect(mockControls.onAudioDeviceSelect).toHaveBeenCalledExactlyOnceWith(
+        WIRED_HEADSET_DEVICE.id,
+      );
+      expect(mockControls.onOutputDeviceSelect).toHaveBeenCalledExactlyOnceWith(
+        WIRED_HEADSET_DEVICE.id,
+      );
+    });
+  });
+});
+
+describe("Test mappings", () => {
+  it("Should map output device to correct AudioDeviceLabel", async () => {
+    const controlledAudioOutput = new AndroidControlledAudioOutput(
+      of(FULL_DEVICE_LIST),
+      testScope,
+      undefined,
+      mockControls,
+    );
+
+    const availableDevices = await firstValueFrom(
+      controlledAudioOutput.available$.pipe(take(1)),
+    );
+
+    expect(availableDevices).toEqual(
+      new Map([
+        [BT_HEADSET_DEVICE.id, { type: "name", name: BT_HEADSET_DEVICE.name }],
+        [
+          USB_HEADSET_DEVICE.id,
+          { type: "name", name: USB_HEADSET_DEVICE.name },
+        ],
+        [
+          WIRED_HEADSET_DEVICE.id,
+          { type: "name", name: WIRED_HEADSET_DEVICE.name },
+        ],
+        [
+          WIRED_HEADPHONE_DEVICE.id,
+          { type: "name", name: WIRED_HEADPHONE_DEVICE.name },
+        ],
+        [SPEAKER_DEVICE.id, { type: "speaker" }],
+        [EARPIECE_DEVICE.id, { type: "earpiece" }],
+      ]),
+    );
+  });
+});
+
+describe("Test select a device", () => {
+  it(`Switch to correct device `, () => {
+    withTestScheduler(({ cold, schedule, expectObservable, flush }) => {
+      const controlledAudioOutput = new AndroidControlledAudioOutput(
+        cold("a", { a: FULL_DEVICE_LIST }),
+        testScope,
+        undefined,
+        mockControls,
+      );
+
+      schedule("-abc", {
+        a: () => controlledAudioOutput.select(EARPIECE_DEVICE.id),
+        b: () => controlledAudioOutput.select(USB_HEADSET_DEVICE.id),
+        c: () => controlledAudioOutput.select(SPEAKER_DEVICE.id),
+      });
+
+      expectObservable(controlledAudioOutput.selected$).toBe("abcd", {
+        // virtualEarpiece is always false on android.
+        // Initially the BT_HEADSET is selected.
+        a: { id: BT_HEADSET_DEVICE.id, virtualEarpiece: false },
+        b: { id: EARPIECE_DEVICE.id, virtualEarpiece: false },
+        c: { id: USB_HEADSET_DEVICE.id, virtualEarpiece: false },
+        d: { id: SPEAKER_DEVICE.id, virtualEarpiece: false },
+      });
+
+      flush();
+
+      [
+        mockControls.onOutputDeviceSelect,
+        mockControls.onAudioDeviceSelect,
+      ].forEach((mockFn) => {
+        expect(mockFn).toHaveBeenCalledTimes(4);
+        expect(mockFn).toHaveBeenNthCalledWith(1, BT_HEADSET_DEVICE.id);
+        expect(mockFn).toHaveBeenNthCalledWith(2, EARPIECE_DEVICE.id);
+        expect(mockFn).toHaveBeenNthCalledWith(3, USB_HEADSET_DEVICE.id);
+        expect(mockFn).toHaveBeenNthCalledWith(4, SPEAKER_DEVICE.id);
+      });
+    });
+  });
+
+  it(`manually switch then a bt headset is added`, () => {
+    withTestScheduler(({ cold, schedule, expectObservable, flush }) => {
+      const controlledAudioOutput = new AndroidControlledAudioOutput(
+        cold("a--b", {
+          a: BASE_DEVICE_LIST,
+          b: BT_HEADSET_BASE_DEVICE_LIST,
+        }),
+        testScope,
+        "audio",
+        mockControls,
+      );
+
+      // Default was earpiece (audio call), let's switch to speaker
+      schedule("-a--", {
+        a: () => controlledAudioOutput.select(SPEAKER_DEVICE.id),
+      });
+
+      expectObservable(controlledAudioOutput.selected$).toBe("ab-c", {
+        // virtualEarpiece is always false on android.
+        // Initially the BT_HEADSET is selected.
+        a: { id: EARPIECE_DEVICE.id, virtualEarpiece: false },
+        b: { id: SPEAKER_DEVICE.id, virtualEarpiece: false },
+        c: { id: BT_HEADSET_DEVICE.id, virtualEarpiece: false },
+      });
+
+      flush();
+
+      [
+        mockControls.onOutputDeviceSelect,
+        mockControls.onAudioDeviceSelect,
+      ].forEach((mockFn) => {
+        expect(mockFn).toHaveBeenCalledTimes(3);
+        expect(mockFn).toHaveBeenNthCalledWith(1, EARPIECE_DEVICE.id);
+        expect(mockFn).toHaveBeenNthCalledWith(2, SPEAKER_DEVICE.id);
+        expect(mockFn).toHaveBeenNthCalledWith(3, BT_HEADSET_DEVICE.id);
+      });
+    });
+  });
+
+  it(`Go back to the previously selected after the auto-switch device goes away`, () => {
+    withTestScheduler(({ cold, schedule, expectObservable, flush }) => {
+      const controlledAudioOutput = new AndroidControlledAudioOutput(
+        cold("a--b-c", {
+          a: BASE_DEVICE_LIST,
+          b: BT_HEADSET_BASE_DEVICE_LIST,
+          c: BASE_DEVICE_LIST,
+        }),
+        testScope,
+        "audio",
+        mockControls,
+      );
+
+      // Default was earpiece (audio call), let's switch to speaker
+      schedule("-a---", {
+        a: () => controlledAudioOutput.select(SPEAKER_DEVICE.id),
+      });
+
+      expectObservable(controlledAudioOutput.selected$).toBe("ab-c-d", {
+        // virtualEarpiece is always false on android.
+        // Initially the BT_HEADSET is selected.
+        a: { id: EARPIECE_DEVICE.id, virtualEarpiece: false },
+        b: { id: SPEAKER_DEVICE.id, virtualEarpiece: false },
+        c: { id: BT_HEADSET_DEVICE.id, virtualEarpiece: false },
+        d: { id: SPEAKER_DEVICE.id, virtualEarpiece: false },
+      });
+
+      flush();
+
+      [
+        mockControls.onOutputDeviceSelect,
+        mockControls.onAudioDeviceSelect,
+      ].forEach((mockFn) => {
+        expect(mockFn).toHaveBeenCalledTimes(4);
+        expect(mockFn).toHaveBeenNthCalledWith(1, EARPIECE_DEVICE.id);
+        expect(mockFn).toHaveBeenNthCalledWith(2, SPEAKER_DEVICE.id);
+        expect(mockFn).toHaveBeenNthCalledWith(3, BT_HEADSET_DEVICE.id);
+        expect(mockFn).toHaveBeenNthCalledWith(4, SPEAKER_DEVICE.id);
+      });
+    });
+  });
+});
+
+describe("Available device changes", () => {
+  let availableSource$: Subject<OutputDevice[]>;
+
+  const createAudioControlledOutput = (
+    intent: RTCCallIntent,
+  ): AndroidControlledAudioOutput => {
+    return new AndroidControlledAudioOutput(
+      availableSource$,
+      testScope,
+      intent,
+      mockControls,
+    );
+  };
+
+  beforeEach(() => {
+    availableSource$ = new Subject<OutputDevice[]>();
+  });
+
+  it("When a BT headset is added, control should switch to use it", () => {
+    createAudioControlledOutput("video");
+
+    // Emit the base device list, the speaker should be selected
+    availableSource$.next(BASE_DEVICE_LIST);
+    // Initially speaker would be selected
+    [
+      mockControls.onOutputDeviceSelect,
+      mockControls.onAudioDeviceSelect,
+    ].forEach((mockFn) => {
+      expect(mockFn).toHaveBeenCalledTimes(1);
+      expect(mockFn).toHaveBeenCalledWith(SPEAKER_DEVICE.id);
+    });
+
+    // Emit a new device list with a BT device, the control should switch to it
+    availableSource$.next([BT_HEADSET_DEVICE, ...BASE_DEVICE_LIST]);
+    [
+      mockControls.onOutputDeviceSelect,
+      mockControls.onAudioDeviceSelect,
+    ].forEach((mockFn) => {
+      expect(mockFn).toHaveBeenCalledTimes(2);
+      expect(mockFn).toHaveBeenLastCalledWith(BT_HEADSET_DEVICE.id);
+    });
+  });
+
+  // Android does not set `isExternalHeadset` to true for wired headphones, so we can't test this case.'
+  it.skip("When a wired headset is added, control should switch to use it", async () => {
+    const controlledAudioOutput = createAudioControlledOutput("video");
+
+    // Emit the base device list, the speaker should be selected
+    availableSource$.next(BASE_DEVICE_LIST);
+
+    await firstValueFrom(controlledAudioOutput.selected$.pipe(take(1)));
+    // Initially speaker would be selected
+    [
+      mockControls.onOutputDeviceSelect,
+      mockControls.onAudioDeviceSelect,
+    ].forEach((mockFn) => {
+      expect(mockFn).toHaveBeenCalledTimes(1);
+      expect(mockFn).toHaveBeenCalledWith(SPEAKER_DEVICE.id);
+    });
+
+    // Emit a new device list with a wired headset, the control should switch to it
+    availableSource$.next([WIRED_HEADPHONE_DEVICE, ...BASE_DEVICE_LIST]);
+    [
+      mockControls.onOutputDeviceSelect,
+      mockControls.onAudioDeviceSelect,
+    ].forEach((mockFn) => {
+      expect(mockFn).toHaveBeenCalledTimes(2);
+      expect(mockFn).toHaveBeenLastCalledWith(WIRED_HEADPHONE_DEVICE.id);
+    });
+  });
+
+  it("When the active bt headset is removed on audio call, control should switch to earpiece", () => {
+    createAudioControlledOutput("audio");
+
+    // Emit the BT headset device list, the BT headset should be selected
+    availableSource$.next(BT_HEADSET_BASE_DEVICE_LIST);
+    // Initially speaker would be selected
+    [
+      mockControls.onOutputDeviceSelect,
+      mockControls.onAudioDeviceSelect,
+    ].forEach((mockFn) => {
+      expect(mockFn).toHaveBeenCalledTimes(1);
+      expect(mockFn).toHaveBeenCalledWith(BT_HEADSET_DEVICE.id);
+    });
+
+    // Emit a new device list without the BT headset, the control should switch to the earpiece for
+    // audio calls
+    availableSource$.next(BASE_DEVICE_LIST);
+    [
+      mockControls.onOutputDeviceSelect,
+      mockControls.onAudioDeviceSelect,
+    ].forEach((mockFn) => {
+      expect(mockFn).toHaveBeenCalledTimes(2);
+      expect(mockFn).toHaveBeenLastCalledWith(EARPIECE_DEVICE.id);
+    });
+  });
+
+  it("When the active bt headset is removed on video call, control should switch to speaker", () => {
+    createAudioControlledOutput("video");
+
+    availableSource$.next(BT_HEADSET_BASE_DEVICE_LIST);
+
+    // Initially bt headset would be selected
+    [
+      mockControls.onOutputDeviceSelect,
+      mockControls.onAudioDeviceSelect,
+    ].forEach((mockFn) => {
+      expect(mockFn).toHaveBeenCalledTimes(1);
+      expect(mockFn).toHaveBeenCalledWith(BT_HEADSET_DEVICE.id);
+    });
+
+    // Emit a new device list without the BT headset, the control should switch to speaker for video call
+    availableSource$.next(BASE_DEVICE_LIST);
+    [
+      mockControls.onOutputDeviceSelect,
+      mockControls.onAudioDeviceSelect,
+    ].forEach((mockFn) => {
+      expect(mockFn).toHaveBeenCalledTimes(2);
+      expect(mockFn).toHaveBeenLastCalledWith(SPEAKER_DEVICE.id);
+    });
+  });
+
+  it("Do not repeatidly set the same device", () => {
+    createAudioControlledOutput("video");
+
+    availableSource$.next(BT_HEADSET_BASE_DEVICE_LIST);
+    availableSource$.next(BT_HEADSET_BASE_DEVICE_LIST);
+    availableSource$.next(BT_HEADSET_BASE_DEVICE_LIST);
+    availableSource$.next(BT_HEADSET_BASE_DEVICE_LIST);
+    availableSource$.next(BT_HEADSET_BASE_DEVICE_LIST);
+
+    // Initially bt headset would be selected
+    [
+      mockControls.onOutputDeviceSelect,
+      mockControls.onAudioDeviceSelect,
+    ].forEach((mockFn) => {
+      expect(mockFn).toHaveBeenCalledTimes(1);
+      expect(mockFn).toHaveBeenCalledWith(BT_HEADSET_DEVICE.id);
+    });
+  });
+});
+
+describe("Scope management", () => {
+  it("Should stop emitting when scope ends", () => {
+    const aScope = new ObservableScope();
+    const controlledAudioOutput = new AndroidControlledAudioOutput(
+      of(BASE_DEVICE_LIST),
+      aScope,
+      undefined,
+      mockControls,
+    );
+
+    expect(mockControls.onAudioDeviceSelect).toHaveBeenCalledOnce();
+
+    aScope.end();
+
+    controlledAudioOutput.select(EARPIECE_DEVICE.id);
+
+    expect(mockControls.onAudioDeviceSelect).not.toHaveBeenCalledTimes(2);
+    expect(mockControls.onAudioDeviceSelect).toHaveBeenCalledOnce();
+  });
+
+  it("Should stop updating when scope ends", () => {
+    const aScope = new ObservableScope();
+    const availableSource$ = new Subject<OutputDevice[]>();
+    new AndroidControlledAudioOutput(
+      availableSource$,
+      aScope,
+      undefined,
+      mockControls,
+    );
+
+    availableSource$.next(BT_HEADSET_BASE_DEVICE_LIST);
+    expect(mockControls.onAudioDeviceSelect).toHaveBeenCalledOnce();
+    expect(mockControls.onAudioDeviceSelect).toHaveBeenCalledWith(
+      BT_HEADSET_DEVICE.id,
+    );
+
+    aScope.end();
+
+    availableSource$.next(BASE_DEVICE_LIST);
+
+    expect(mockControls.onAudioDeviceSelect).not.toHaveBeenCalledTimes(2);
+    // Should have been called only once with the initial BT_HEADSET_DEVICE.id
+    expect(mockControls.onAudioDeviceSelect).toHaveBeenCalledOnce();
+  });
+});

src/state/AndroidControlledAudioOutput.ts

@@ -0,0 +1,360 @@
+/*
+Copyright 2026 Element Corp.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+import { logger as rootLogger } from "matrix-js-sdk/lib/logger";
+import {
+  distinctUntilChanged,
+  map,
+  merge,
+  type Observable,
+  scan,
+  startWith,
+  Subject,
+  tap,
+} from "rxjs";
+
+import {
+  type AudioOutputDeviceLabel,
+  type MediaDevice,
+  type SelectedAudioOutputDevice,
+} from "./MediaDevices.ts";
+import type { ObservableScope } from "./ObservableScope.ts";
+import type { RTCCallIntent } from "matrix-js-sdk/lib/matrixrtc";
+import { type Controls, type OutputDevice } from "../controls.ts";
+import { type Behavior } from "./Behavior.ts";
+
+type ControllerState = {
+  /**
+   * The list of available output devices, ordered by preference order (most preferred first).
+   */
+  devices: OutputDevice[];
+  /**
+   * Explicit user preference for the selected device.
+   */
+  preferredDeviceId: string | undefined;
+  /**
+   * The effective selected device, always valid against available devices.
+   */
+  selectedDeviceId: string | undefined;
+};
+
+/**
+ * The possible actions that can be performed on the controller,
+ * either by the user or by the system.
+ */
+type ControllerAction =
+  | { type: "selectDevice"; deviceId: string | undefined }
+  | { type: "deviceUpdated"; devices: OutputDevice[] };
+/**
+ * The implementation of the audio output media device for Android when using the controlled audio output mode.
+ *
+ * In this mode, the hosting application (e.g. Element Mobile) is responsible for providing the list of available audio output devices.
+ * There are some android specific logic compared to others:
+ *  - AndroidControlledAudioOutput is the only one responsible for selecting the best output device.
+ *  - On android, we don't listen to the selected device from native code (control.setAudioDevice).
+ *  - If a new device is added or removed, this controller will determine the new selected device based
+ *    on the available devices (that is ordered by preference order) and the user's selection (if any).
+ *
+ * Given the differences in how the native code is handling the audio routing on Android compared to iOS,
+ * we have this separate implementation. It allows us to have proper testing and avoid side effects
+ * from platform specific logic breaking the other platform's implementation.
+ */
+export class AndroidControlledAudioOutput implements MediaDevice<
+  AudioOutputDeviceLabel,
+  SelectedAudioOutputDevice
+> {
+  private logger = rootLogger.getChild(
+    "[MediaDevices AndroidControlledAudioOutput]",
+  );
+
+  /**
+   * STATE stream: the current state of the controller, including the list of available devices and the selected device.
+   */
+  private readonly controllerState$: Behavior<ControllerState>;
+
+  /**
+   * @inheritdoc
+   */
+  public readonly available$: Behavior<Map<string, AudioOutputDeviceLabel>>;
+
+  /**
+   * Effective selected device, always valid against available devices.
+   *
+   * On android, we don't listen to the selected device from native code (control.setAudioDevice).
+   * Instead, we determine the selected device ourselves based on the available devices and the user's selection (if any).
+   */
+  public readonly selected$: Behavior<SelectedAudioOutputDevice | undefined>;
+
+  // COMMAND stream: user asks to select a device
+  private readonly selectDeviceCommand$ = new Subject<string | undefined>();
+
+  public select(id: string): void {
+    this.logger.info(`select device: ${id}`);
+    this.selectDeviceCommand$.next(id);
+  }
+
+  /**
+   * Creates an instance of AndroidControlledAudioOutput.
+   *
+   * @constructor
+   * @param controlledDevices$ - The list of available output devices coming from the hosting application, ordered by preference order (most preferred first).
+   * @param scope - The ObservableScope to create the Behaviors in.
+   * @param initialIntent - The initial call intent (e.g. "audio" or "video") that can be used to determine the default audio routing (e.g. default to earpiece for audio calls and speaker for video calls).
+   * @param controls - The controls provided by the hosting application to control the audio routing and notify of user actions.
+   */
+  public constructor(
+    private readonly controlledDevices$: Observable<OutputDevice[]>,
+    private readonly scope: ObservableScope,
+    private initialIntent: RTCCallIntent | undefined = undefined,
+    controls: Controls,
+  ) {
+    this.controllerState$ = this.startObservingState$();
+
+    this.selected$ = this.effectiveSelectionFromState$(this.controllerState$);
+
+    this.available$ = scope.behavior(
+      this.controllerState$.pipe(
+        map((state) => {
+          this.logger.info("available devices updated:", state.devices);
+
+          return new Map<string, AudioOutputDeviceLabel>(
+            state.devices.map((outputDevice) => {
+              return [outputDevice.id, mapDeviceToLabel(outputDevice)];
+            }),
+          );
+        }),
+      ),
+    );
+
+    // Effect 1: notify host when effective selection changes
+    this.selected$
+      // It is a behavior so it has built-in distinct until change
+      .pipe(scope.bind())
+      .subscribe((device) => {
+        // Let the hosting application know which output device has been selected.
+        if (device !== undefined) {
+          this.logger.info("onAudioDeviceSelect called:", device);
+          controls.onAudioDeviceSelect?.(device.id);
+          // Also invoke the deprecated callback for backward compatibility
+          // TODO: it appears that on Android the hosting application is only using the deprecated callback (onOutputDeviceSelect)
+          // and not the new one (onAudioDeviceSelect), we should clean this up and only have one callback for audio device selection.
+          controls.onOutputDeviceSelect?.(device.id);
+        }
+      });
+  }
+
+  private startObservingState$(): Behavior<ControllerState> {
+    const initialState: ControllerState = {
+      devices: [],
+      preferredDeviceId: undefined,
+      selectedDeviceId: undefined,
+    };
+
+    // Merge the two possible inputs observable as a single
+    // stream of actions that will update the state of the controller.
+    const actions$: Observable<ControllerAction> = merge(
+      this.controlledDevices$.pipe(
+        map(
+          (devices) =>
+            ({ type: "deviceUpdated", devices }) satisfies ControllerAction,
+        ),
+      ),
+      this.selectDeviceCommand$.pipe(
+        map(
+          (deviceId) =>
+            ({ type: "selectDevice", deviceId }) satisfies ControllerAction,
+        ),
+      ),
+    );
+
+    const initialAction: ControllerAction = {
+      type: "deviceUpdated",
+      devices: [],
+    };
+
+    return this.scope.behavior(
+      actions$.pipe(
+        startWith(initialAction),
+        scan((state, action): ControllerState => {
+          switch (action.type) {
+            case "deviceUpdated": {
+              const chosenDevice = this.chooseEffectiveSelection({
+                previousDevices: state.devices,
+                availableDevices: action.devices,
+                currentSelectedId: state.selectedDeviceId,
+                preferredDeviceId: state.preferredDeviceId,
+              });
+
+              return {
+                ...state,
+                devices: action.devices,
+                selectedDeviceId: chosenDevice,
+              };
+            }
+            case "selectDevice": {
+              const chosenDevice = this.chooseEffectiveSelection({
+                previousDevices: state.devices,
+                availableDevices: state.devices,
+                currentSelectedId: state.selectedDeviceId,
+                preferredDeviceId: action.deviceId,
+              });
+
+              return {
+                ...state,
+                preferredDeviceId: action.deviceId,
+                selectedDeviceId: chosenDevice,
+              };
+            }
+          }
+        }, initialState),
+      ),
+    );
+  }
+
+  private effectiveSelectionFromState$(
+    state$: Observable<ControllerState>,
+  ): Behavior<SelectedAudioOutputDevice | undefined> {
+    return this.scope.behavior(
+      state$
+        .pipe(
+          map((state) => {
+            if (state.selectedDeviceId) {
+              return {
+                id: state.selectedDeviceId,
+                /** This is an iOS thing, always false for android*/
+                virtualEarpiece: false,
+              };
+            }
+            return undefined;
+          }),
+          distinctUntilChanged((a, b) => a?.id === b?.id),
+        )
+        .pipe(
+          tap((selected) => {
+            this.logger.debug(`selected device: ${selected?.id}`);
+          }),
+        ),
+    );
+  }
+
+  private chooseEffectiveSelection(args: {
+    previousDevices: OutputDevice[];
+    availableDevices: OutputDevice[];
+    currentSelectedId: string | undefined;
+    preferredDeviceId: string | undefined;
+  }): string | undefined {
+    const {
+      previousDevices,
+      availableDevices,
+      currentSelectedId,
+      preferredDeviceId,
+    } = args;
+
+    this.logger.debug(`chooseEffectiveSelection with args:`, args);
+
+    // Take preferredDeviceId in priority or default to the last effective selection.
+    const activeSelectedDeviceId = preferredDeviceId || currentSelectedId;
+    const isAvailable = availableDevices.some(
+      (device) => device.id === activeSelectedDeviceId,
+    );
+
+    // If there is no current device, or it is not available anymore,
+    // choose the default device selection logic.
+    if (activeSelectedDeviceId === undefined || !isAvailable) {
+      this.logger.debug(
+        `No current device or it is not available, using default selection logic.`,
+      );
+      // use the default selection logic
+      return this.chooseDefaultDeviceId(availableDevices);
+    }
+
+    // Is there a new added device?
+    // If a device is added, we might want to switch to it if it's more preferred than the currently selected device.
+    const newDeviceWasAdded = availableDevices.some(
+      (device) => !previousDevices.some((d) => d.id === device.id),
+    );
+
+    if (newDeviceWasAdded) {
+      // TODO only want to check from the added device, not all devices.?
+      // check if the currently selected device is the most preferred one, if not switch to the most preferred one.
+      const mostPreferredDevice = availableDevices[0];
+      this.logger.debug(
+        `A new device was added, checking if we should switch to it.`,
+        mostPreferredDevice,
+      );
+      if (mostPreferredDevice.id !== activeSelectedDeviceId) {
+        // Given this is automatic switching, we want to be careful and only switch to a more private device
+        // (e.g. from speaker to a BT headset) but not switch from a more private device to a less private one
+        // (e.g. from a BT headset to the speaker), as that can be disruptive for the user if it happens unexpectedly.
+        if (mostPreferredDevice.isExternalHeadset == true) {
+          this.logger.info(
+            `The currently selected device ${mostPreferredDevice.id} is not the most preferred one, switching to the most preferred one ${activeSelectedDeviceId} instead.`,
+          );
+          // Let's switch as it is a more private device.
+          return mostPreferredDevice.id;
+        }
+      }
+    }
+
+    // no changes
+    return activeSelectedDeviceId;
+  }
+
+  /**
+   *   The logic for the default is different based on the call type.
+   *   For example for a voice call we want to default to the earpiece if it's available,
+   *   but for a video call we want to default to the speaker.
+   *   If the user is using a BT headset we want to default to that, as it's likely what they want to use for both video and voice calls.
+   *
+   *   @param available the available audio output devices to choose from, keyed by their id, sorted by likelihood of it being used for communication.
+   *
+   */
+  private chooseDefaultDeviceId(available: OutputDevice[]): string | undefined {
+    this.logger.debug(
+      `Android routing logic intent: ${this.initialIntent} finding best default...`,
+    );
+    if (this.initialIntent === "audio") {
+      const systemProposed = available[0];
+      // If no headset is connected, android will route to the speaker by default,
+      // but for a voice call we want to route to the earpiece instead,
+      // so override the system proposed routing in that case.
+      if (systemProposed?.isSpeaker == true) {
+        // search for the earpiece
+        const earpieceDevice = available.find(
+          (device) => device.isEarpiece == true,
+        );
+        if (earpieceDevice) {
+          this.logger.debug(
+            `Android routing: Switch to earpiece instead of speaker for voice call`,
+          );
+          return earpieceDevice.id;
+        } else {
+          this.logger.debug(
+            `Android routing: no earpiece found, cannot switch, use system proposed routing`,
+          );
+          return systemProposed.id;
+        }
+      } else {
+        this.logger.debug(
+          `Android routing: Use system proposed routing `,
+          systemProposed,
+        );
+        return systemProposed?.id;
+      }
+    } else {
+      // Use the system best proposed best routing.
+      return available[0]?.id;
+    }
+  }
+}
+
+// Utilities
+function mapDeviceToLabel(device: OutputDevice): AudioOutputDeviceLabel {
+  const { name, isEarpiece, isSpeaker } = device;
+  if (isEarpiece) return { type: "earpiece" };
+  else if (isSpeaker) return { type: "speaker" };
+  else return { type: "name", name };
+}

src/state/AudioOutput.test.ts

@@ -0,0 +1,193 @@
+/*
+Copyright 2026 Element Corp.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import { afterEach, beforeEach, describe, vi, it } from "vitest";
+import * as ComponentsCore from "@livekit/components-core";
+
+import { ObservableScope } from "./ObservableScope";
+import { AudioOutput } from "./MediaDevices";
+import { withTestScheduler } from "../utils/test";
+
+const BT_SPEAKER = {
+  deviceId: "f9fc8f5f94578fe3abd89e086c1e78c08477aa564dd9e917950f0e7ebb37a6a2",
+  kind: "audiooutput",
+  label: "JBL (Bluetooth)",
+  groupId: "309a5c086cd8eb885a164046db6ec834c349be01d86448d02c1a5279456ff9e4",
+} as unknown as MediaDeviceInfo;
+
+const BUILT_IN_SPEAKER = {
+  deviceId: "acdbb8546ea6fa85ba2d861e9bcc0e71810d03bbaf6d1712c69e8d9c0c6c2e0a",
+  kind: "audiooutput",
+  label: "MacBook Speakers (Built-in)",
+  groupId: "08a5a3a486473aaa898eb81cda3113f3e21053fb8b84155f4e612fe3f8db5d17",
+} as unknown as MediaDeviceInfo;
+
+const BT_HEADSET = {
+  deviceId: "ff8e6edb4ebb512b2b421335bfd14994a5b4c7192b3e84a8696863d83cf46d12",
+  kind: "audiooutput",
+  label: "OpenMove (Bluetooth)",
+  groupId: "c2893c2438c44248368e0533300245c402764991506f42cd73818dc8c3ee9c88",
+} as unknown as MediaDeviceInfo;
+
+const AMAC_DEVICE_LIST = [BT_SPEAKER, BUILT_IN_SPEAKER];
+
+const AMAC_DEVICE_LIST_WITH_DEFAULT = [
+  asDefault(BUILT_IN_SPEAKER),
+  ...AMAC_DEVICE_LIST,
+];
+
+const AMAC_HS_DEVICE_LIST = [
+  asDefault(BT_HEADSET),
+  BT_SPEAKER,
+  BT_HEADSET,
+  BUILT_IN_SPEAKER,
+];
+
+const LAPTOP_SPEAKER = {
+  deviceId: "EcUxTMu8He2wz+3Y8m/u0fy6M92pUk=",
+  kind: "audiooutput",
+  label: "Raptor AVS Speaker",
+  groupId: "kSrdanhpEDLg3vN8z6Z9MJ1EdanB8zI+Q1dxA=",
+} as unknown as MediaDeviceInfo;
+
+const MONITOR_SPEAKER = {
+  deviceId: "gBryZdAdC8I/rrJpr9r6R+rZzKkoIK5cpU=",
+  kind: "audiooutput",
+  label: "Raptor AVS HDMI / DisplayPort 1 Output",
+  groupId: "kSrdanhpEDLg3vN8z6Z9MJ1EdanB8zI+Q1dxA=",
+} as unknown as MediaDeviceInfo;
+
+const DEVICE_LIST_B = [LAPTOP_SPEAKER, MONITOR_SPEAKER];
+
+// On chrome, there is an additional synthetic device called "Default - <device name>",
+// it represents what the OS default is now.
+function asDefault(device: MediaDeviceInfo): MediaDeviceInfo {
+  return {
+    ...device,
+    deviceId: "default",
+    label: `Default - ${device.label}`,
+  };
+}
+// When the authorization is not yet granted, every device is still listed
+// but only with empty/blank labels and ids.
+// This is a transition state.
+function toBlankDevice(device: MediaDeviceInfo): MediaDeviceInfo {
+  return {
+    ...device,
+    deviceId: "",
+    label: "",
+    groupId: "",
+  };
+}
+
+vi.mock("@livekit/components-core", () => ({
+  createMediaDeviceObserver: vi.fn(),
+}));
+
+describe("AudioOutput Tests", () => {
+  let testScope: ObservableScope;
+
+  beforeEach(() => {
+    testScope = new ObservableScope();
+  });
+
+  afterEach(() => {
+    testScope.end();
+  });
+
+  it("should select the default audio output device", () => {
+    // In a real life setup there would be first a blanked list
+    // then the real one.
+    withTestScheduler(({ behavior, cold, expectObservable }) => {
+      vi.mocked(ComponentsCore.createMediaDeviceObserver).mockReturnValue(
+        cold("ab", {
+          // In a real life setup there would be first a blanked list
+          // then the real one.
+          a: AMAC_DEVICE_LIST_WITH_DEFAULT.map(toBlankDevice),
+          b: AMAC_DEVICE_LIST_WITH_DEFAULT,
+        }),
+      );
+
+      const audioOutput = new AudioOutput(
+        behavior("a", { a: true }),
+        testScope,
+      );
+
+      expectObservable(audioOutput.selected$).toBe("ab", {
+        a: undefined,
+        b: { id: "default", virtualEarpiece: false },
+      });
+    });
+  });
+
+  it("Select the correct device when requested", () => {
+    // In a real life setup there would be first a blanked list
+    // then the real one.
+    withTestScheduler(({ behavior, cold, schedule, expectObservable }) => {
+      vi.mocked(ComponentsCore.createMediaDeviceObserver).mockReturnValue(
+        cold("ab", {
+          // In a real life setup there would be first a blanked list
+          // then the real one.
+          a: DEVICE_LIST_B.map(toBlankDevice),
+          b: DEVICE_LIST_B,
+        }),
+      );
+
+      const audioOutput = new AudioOutput(
+        behavior("a", { a: true }),
+        testScope,
+      );
+
+      schedule("--abc", {
+        a: () => audioOutput.select(MONITOR_SPEAKER.deviceId),
+        b: () => audioOutput.select(LAPTOP_SPEAKER.deviceId),
+        c: () => audioOutput.select(MONITOR_SPEAKER.deviceId),
+      });
+
+      expectObservable(audioOutput.selected$).toBe("abcde", {
+        a: undefined,
+        b: { id: LAPTOP_SPEAKER.deviceId, virtualEarpiece: false },
+        c: { id: MONITOR_SPEAKER.deviceId, virtualEarpiece: false },
+        d: { id: LAPTOP_SPEAKER.deviceId, virtualEarpiece: false },
+        e: { id: MONITOR_SPEAKER.deviceId, virtualEarpiece: false },
+      });
+    });
+  });
+
+  it("Test mappings", () => {
+    // In a real life setup there would be first a blanked list
+    // then the real one.
+    withTestScheduler(({ behavior, cold, schedule, expectObservable }) => {
+      vi.mocked(ComponentsCore.createMediaDeviceObserver).mockReturnValue(
+        cold("a", {
+          // In a real life setup there would be first a blanked list
+          // then the real one.
+          a: AMAC_HS_DEVICE_LIST,
+        }),
+      );
+
+      const audioOutput = new AudioOutput(
+        behavior("a", { a: true }),
+        testScope,
+      );
+
+      const expectedMappings = new Map([
+        [`default`, { type: "name", name: asDefault(BT_HEADSET).label }],
+        [BT_SPEAKER.deviceId, { type: "name", name: BT_SPEAKER.label }],
+        [BT_HEADSET.deviceId, { type: "name", name: BT_HEADSET.label }],
+        [
+          BUILT_IN_SPEAKER.deviceId,
+          { type: "name", name: BUILT_IN_SPEAKER.label },
+        ],
+      ]);
+
+      expectObservable(audioOutput.available$).toBe("a", {
+        a: expectedMappings,
+      });
+    });
+  });
+});

src/state/CallViewModel/CallNotificationLifecycle.ts

@@ -89,7 +89,6 @@ export interface Props {
  * `callPickupState$` The current call pickup state of the call.
  *  - "unknown": The client has not yet sent the notification event. We don't know if it will because it first needs to send its own membership.
  *     Then we can conclude if we were the first one to join or not.
- *     This may also be set if we are disconnected.
  *  - "ringing": The call is ringing on other devices in this room (This client should give audiovisual feedback that this is happening).
  *  - "timeout": No-one picked up in the defined time this call should be ringing on others devices.
  *     The call failed. If desired this can be used as a trigger to exit the call.
@@ -131,15 +130,9 @@ export function createCallNotificationLifecycle$({
   ) as Behavior<Epoch<boolean>>;
 
   /**
-   * Whenever the RTC session tells us that it intends to ring the remote
-   * participant's devices, this emits an Observable tracking the current state of
-   * that ringing process.
+   * The state of the current ringing attempt, if the RTC session is indeed
+   * ringing the remote participant's devices. Otherwise `null`.
    */
-  // This is a behavior since we need to store the latest state for when we subscribe to this after `didSendCallNotification$`
-  // has already emitted but we still need the latest observable with a timeout timer that only gets created on after receiving `notificationEvent`.
-  // A behavior will emit the latest observable with the running timer to new subscribers.
-  // see also: callPickupState$ and in particular the line: `return this.ring$.pipe(mergeAll());` here we otherwise might get an EMPTY observable if
-  // `ring$` would not be a behavior.
   const remoteRingState$: Behavior<"ringing" | "timeout" | "decline" | null> =
     scope.behavior(
       sentCallNotification$.pipe(

src/state/CallViewModel/CallViewModel.test.ts

@@ -46,9 +46,11 @@ import {
 } from "../../utils/test.ts";
 import { E2eeType } from "../../e2ee/e2eeType.ts";
 import {
+  alice,
   aliceId,
   aliceParticipant,
   aliceRtcMember,
+  aliceUserId,
   bobId,
   bobRtcMember,
   local,
@@ -140,8 +142,8 @@ export interface SpotlightExpandedLayoutSummary {
 
 export interface OneOnOneLayoutSummary {
   type: "one-on-one";
-  local: string;
-  remote: string;
+  spotlight: string;
+  pip: string;
 }
 
 export interface PipLayoutSummary {
@@ -194,11 +196,11 @@ function summarizeLayout$(l$: Observable<Layout>): Observable<LayoutSummary> {
           );
         case "one-on-one":
           return combineLatest(
-            [l.local.media$, l.remote.media$],
-            (local, remote) => ({
+            [l.spotlight.media$, l.pip.media$],
+            (spotlight, pip) => ({
               type: l.type,
-              local: local.id,
-              remote: remote.id,
+              spotlight: spotlight.id,
+              pip: pip.id,
             }),
           );
         case "pip":
@@ -537,8 +539,8 @@ describe.each([
               b: {
                 // In a larger window, expect the normal one-on-one layout
                 type: "one-on-one",
-                local: `${localId}:0`,
-                remote: `${aliceId}:0`,
+                pip: `${localId}:0`,
+                spotlight: `${aliceId}:0`,
               },
               c: {
                 // In a PiP-sized window, we of course expect a PiP layout
@@ -840,8 +842,8 @@ describe.each([
               },
               b: {
                 type: "one-on-one",
-                local: `${localId}:0`,
-                remote: `${aliceId}:0`,
+                pip: `${localId}:0`,
+                spotlight: `${aliceId}:0`,
               },
               c: {
                 type: "grid",
@@ -883,8 +885,8 @@ describe.each([
               },
               b: {
                 type: "one-on-one",
-                local: `${localId}:0`,
-                remote: `${aliceId}:0`,
+                pip: `${localId}:0`,
+                spotlight: `${aliceId}:0`,
               },
               c: {
                 type: "grid",
@@ -893,8 +895,8 @@ describe.each([
               },
               d: {
                 type: "one-on-one",
-                local: `${localId}:0`,
-                remote: `${daveId}:0`,
+                pip: `${localId}:0`,
+                spotlight: `${daveId}:0`,
               },
             },
           );
@@ -1087,83 +1089,81 @@ describe.each([
     });
   });
 
-  describe("waitForCallPickup$", () => {
-    it.skip("regression test: does stop ringing in case livekitConnectionState$ emits after didSendCallNotification$ has already emitted", () => {
-      withTestScheduler(({ schedule, expectObservable, behavior }) => {
-        withCallViewModel(
-          {
-            livekitConnectionState$: behavior("d 9ms c", {
-              d: ConnectionState.Disconnected,
-              c: ConnectionState.Connected,
-            }),
-          },
-          (vm, rtcSession) => {
-            // Fire a call notification IMMEDIATELY (its important for this test, that this happens before the livekitConnectionState$ emits)
-            schedule("n", {
-              n: () => {
-                rtcSession.emit(
-                  MatrixRTCSessionEvent.DidSendCallNotification,
-                  mockRingEvent("$notif1", 30),
-                );
-              },
-            });
+  test("recipient has placeholder tile while ringing or timed out", () => {
+    withTestScheduler(({ schedule, expectObservable }) => {
+      withCallViewModel(
+        {
+          roomMembers: [alice, local], // Simulate a DM
+        },
+        (vm, rtcSession) => {
+          // Fire a ringing notification
+          schedule("n", {
+            n: () => {
+              rtcSession.emit(
+                MatrixRTCSessionEvent.DidSendCallNotification,
+                mockRingEvent("$notif1", 30),
+              );
+            },
+          });
 
-            expectObservable(vm.callPickupState$).toBe("a 9ms b 19ms c", {
-              a: "unknown",
-              b: "ringing",
-              c: "timeout",
-            });
-          },
-          {
-            waitForCallPickup: true,
-            encryptionSystem: { kind: E2eeType.PER_PARTICIPANT },
-          },
-        );
-      });
+          // Should ring for 30ms and then time out
+          expectObservable(vm.ringing$).toBe("(ny) 26ms n", yesNo);
+          // Layout should show placeholder media for the participant we're
+          // ringing the entire time (even once timed out)
+          expectObservable(summarizeLayout$(vm.layout$)).toBe("a", {
+            a: {
+              type: "one-on-one",
+              spotlight: `${localId}:0`,
+              pip: `ringing:${aliceUserId}`,
+            },
+          });
+        },
+        { waitForCallPickup: true },
+      );
     });
+  });
 
-    it.skip("ringing -> unknown if we get disconnected", () => {
-      withTestScheduler(({ behavior, schedule, expectObservable }) => {
-        const connectionState$ = new BehaviorSubject(ConnectionState.Connected);
-        // Someone joins at 20ms (both LiveKit participant and MatrixRTC member)
-        withCallViewModel(
-          {
-            remoteParticipants$: behavior("a 19ms b", {
-              a: [],
-              b: [aliceParticipant],
-            }),
-            rtcMembers$: behavior("a 19ms b", {
-              a: [localRtcMember],
-              b: [localRtcMember, aliceRtcMember],
-            }),
-            livekitConnectionState$: connectionState$,
-          },
-          (vm, rtcSession) => {
-            // Notify at 5ms so we enter ringing, then get disconnected 5ms later
-            schedule("          5ms r 5ms d", {
-              r: () => {
-                rtcSession.emit(
-                  MatrixRTCSessionEvent.DidSendCallNotification,
-                  mockRingEvent("$notif2", 100),
-                );
-              },
-              d: () => {
-                connectionState$.next(ConnectionState.Disconnected);
-              },
-            });
+  test("recipient's placeholder tile is replaced by their real tile once they answer", () => {
+    withTestScheduler(({ behavior, schedule, expectObservable }) => {
+      withCallViewModel(
+        {
+          // Alice answers after 20ms
+          rtcMembers$: behavior("a 20ms b", {
+            a: [localRtcMember],
+            b: [localRtcMember, aliceRtcMember],
+          }),
+          roomMembers: [alice, local], // Simulate a DM
+        },
+        (vm, rtcSession) => {
+          // Fire a ringing notification
+          schedule("n", {
+            n: () => {
+              rtcSession.emit(
+                MatrixRTCSessionEvent.DidSendCallNotification,
+                mockRingEvent("$notif1", 30),
+              );
+            },
+          });
 
-            expectObservable(vm.callPickupState$).toBe("a 4ms b 5ms c", {
-              a: "unknown",
-              b: "ringing",
-              c: "unknown",
-            });
-          },
-          {
-            waitForCallPickup: true,
-            encryptionSystem: { kind: E2eeType.PER_PARTICIPANT },
-          },
-        );
-      });
+          // Should ring until Alice joins
+          expectObservable(vm.ringing$).toBe("(ny) 17ms n", yesNo);
+          // Layout should show placeholder media for the participant we're
+          // ringing the entire time
+          expectObservable(summarizeLayout$(vm.layout$)).toBe("a 20ms b", {
+            a: {
+              type: "one-on-one",
+              spotlight: `${localId}:0`,
+              pip: `ringing:${aliceUserId}`,
+            },
+            b: {
+              type: "one-on-one",
+              spotlight: `${aliceId}:0`,
+              pip: `${localId}:0`,
+            },
+          });
+        },
+        { waitForCallPickup: true },
+      );
     });
   });
 

src/state/CallViewModel/CallViewModel.ts

@@ -128,7 +128,6 @@ import {
   createSentCallNotification$,
 } from "./CallNotificationLifecycle.ts";
 import {
-  createDMMember$,
   createMatrixMemberMetadata$,
   createRoomMembers$,
 } from "./remoteMembers/MatrixMemberMetadata.ts";
@@ -137,12 +136,17 @@ import { type Connection } from "./remoteMembers/Connection.ts";
 import { createLayoutModeSwitch } from "./LayoutSwitch.ts";
 import {
   createWrappedUserMedia,
-  type MediaItem,
   type WrappedUserMediaViewModel,
-} from "../media/MediaItem.ts";
+} from "../media/WrappedUserMediaViewModel.ts";
 import { type ScreenShareViewModel } from "../media/ScreenShareViewModel.ts";
 import { type UserMediaViewModel } from "../media/UserMediaViewModel.ts";
 import { type MediaViewModel } from "../media/MediaViewModel.ts";
+import { type LocalUserMediaViewModel } from "../media/LocalUserMediaViewModel.ts";
+import { type RemoteUserMediaViewModel } from "../media/RemoteUserMediaViewModel.ts";
+import {
+  createRingingMedia,
+  type RingingMediaViewModel,
+} from "../media/RingingMediaViewModel.ts";
 
 const logger = rootLogger.getChild("[CallViewModel]");
 //TODO
@@ -210,11 +214,10 @@ export type LivekitRoomItem = {
 export interface CallViewModel {
   // lifecycle
   autoLeave$: Observable<AutoLeaveReason>;
-  // TODO if we are in "unknown" state we need a loading rendering (or empty screen)
-  // Otherwise it looks like we already connected and only than the ringing starts which is weird.
-  callPickupState$: Behavior<
-    "unknown" | "ringing" | "timeout" | "decline" | "success" | null
-  >;
+  /**
+   * Whether we are ringing a call recipient.
+   */
+  ringing$: Behavior<boolean>;
   /** Observable that emits when the user should leave the call (hangup pressed, widget action, error).
    * THIS DOES NOT LEAVE THE CALL YET. The only way to leave the call (send the hangup event) is
    *  - by ending the scope
@@ -289,13 +292,6 @@ export interface CallViewModel {
   /** List of reactions. Keys are: membership.membershipId (currently predefined as: `${membershipEvent.userId}:${membershipEvent.deviceId}`)*/
   reactions$: Behavior<Record<string, ReactionOption>>;
 
-  ringOverlay$: Behavior<null | {
-    name: string;
-    /** roomId or userId for the avatar generation. */
-    idForAvatar: string;
-    text: string;
-    avatarMxc?: string;
-  }>;
   // sounds and events
   joinSoundEffect$: Observable<void>;
   leaveSoundEffect$: Observable<void>;
@@ -611,40 +607,6 @@ export function createCallViewModel$(
     matrixRoomMembers$,
   );
 
-  const dmMember$ = createDMMember$(scope, matrixRoomMembers$, matrixRoom);
-  const noUserToCallInRoom$ = scope.behavior(
-    matrixRoomMembers$.pipe(
-      map(
-        (roomMembersMap) =>
-          roomMembersMap.size === 1 && roomMembersMap.get(userId) !== undefined,
-      ),
-    ),
-  );
-
-  const ringOverlay$ = scope.behavior(
-    combineLatest([noUserToCallInRoom$, dmMember$, callPickupState$]).pipe(
-      map(([noUserToCallInRoom, dmMember, callPickupState]) => {
-        // No overlay if not in ringing state
-        if (callPickupState !== "ringing" || noUserToCallInRoom) return null;
-
-        const name = dmMember ? dmMember.rawDisplayName : matrixRoom.name;
-        const id = dmMember ? dmMember.userId : matrixRoom.roomId;
-        const text = dmMember
-          ? `Waiting for ${name} to join…`
-          : "Waiting for other participants…";
-        const avatarMxc = dmMember
-          ? (dmMember.getMxcAvatarUrl?.() ?? undefined)
-          : (matrixRoom.getMxcAvatarUrl() ?? undefined);
-        return {
-          name: name ?? id,
-          idForAvatar: id,
-          text,
-          avatarMxc,
-        };
-      }),
-    ),
-  );
-
   const allConnections$ = scope.behavior(
     connectionManager.connectionManagerData$.pipe(map((d) => d.value)),
   );
@@ -720,7 +682,7 @@ export function createCallViewModel$(
       matrixLivekitMembers$,
       duplicateTiles.value$,
     ]).pipe(
-      // Generate a collection of MediaItems from the list of expected (whether
+      // Generate a collection of user media from the list of expected (whether
       // present or missing) LiveKit participants.
       generateItems(
         "CallViewModel userMedia$",
@@ -793,32 +755,67 @@ export function createCallViewModel$(
     ),
   );
 
+  const ringingMedia$ = scope.behavior<RingingMediaViewModel[]>(
+    combineLatest([userMedia$, matrixRoomMembers$, callPickupState$]).pipe(
+      generateItems(
+        "CallViewModel ringingMedia$",
+        function* ([userMedia, roomMembers, callPickupState]) {
+          if (
+            callPickupState === "ringing" ||
+            callPickupState === "timeout" ||
+            callPickupState === "decline"
+          ) {
+            for (const member of roomMembers.values()) {
+              if (!userMedia.some((vm) => vm.userId === member.userId))
+                yield {
+                  keys: [member.userId],
+                  data: callPickupState,
+                };
+            }
+          }
+        },
+        (scope, pickupState$, userId) =>
+          createRingingMedia({
+            id: `ringing:${userId}`,
+            userId,
+            displayName$: scope.behavior(
+              matrixRoomMembers$.pipe(
+                map((members) => members.get(userId)?.rawDisplayName || userId),
+              ),
+            ),
+            mxcAvatarUrl$:
+              matrixMemberMetadataStore.createAvatarUrlBehavior$(userId),
+            pickupState$,
+            muteStates,
+          }),
+      ),
+      distinctUntilChanged(shallowEquals),
+      tap((ringingMedia) => {
+        if (ringingMedia.length > 1)
+          // Warn that UI may do something unexpected in this case
+          logger.warn(
+            `Ringing more than one participant is not supported (ringing ${ringingMedia.map((vm) => vm.userId).join(", ")})`,
+          );
+      }),
+    ),
+  );
+
   /**
-   * List of all media items (user media and screen share media) that we want
-   * tiles for.
+   * All screen share media that we want to display.
    */
-  const mediaItems$ = scope.behavior<MediaItem[]>(
+  const screenShares$ = scope.behavior<ScreenShareViewModel[]>(
     userMedia$.pipe(
       switchMap((userMedia) =>
         userMedia.length === 0
           ? of([])
           : combineLatest(
               userMedia.map((m) => m.screenShares$),
-              (...screenShares) => [...userMedia, ...screenShares.flat(1)],
+              (...screenShares) => screenShares.flat(1),
             ),
       ),
     ),
   );
 
-  /**
-   * List of MediaItems that we want to display, that are of type ScreenShare
-   */
-  const screenShares$ = scope.behavior<ScreenShareViewModel[]>(
-    mediaItems$.pipe(
-      map((mediaItems) => mediaItems.filter((m) => m.type === "screen share")),
-    ),
-  );
-
   const joinSoundEffect$ = userMedia$.pipe(
     pairwise(),
     filter(
@@ -931,40 +928,20 @@ export function createCallViewModel$(
     ),
   );
 
-  const spotlight$ = scope.behavior<MediaViewModel[]>(
-    screenShares$.pipe(
-      switchMap((screenShares) => {
-        if (screenShares.length > 0) return of(screenShares);
-
-        return spotlightSpeaker$.pipe(
-          map((speaker) => (speaker ? [speaker] : [])),
+  /**
+   * Local user media suitable for displaying in a PiP (undefined if not found
+   * or if user prefers to not see themselves).
+   */
+  const localUserMediaForPip$ = scope.behavior<
+    LocalUserMediaViewModel | undefined
+  >(
+    userMedia$.pipe(
+      switchMap((userMedia) => {
+        const localUserMedia = userMedia.find(
+          (m): m is WrappedUserMediaViewModel & LocalUserMediaViewModel =>
+            m.type === "user" && m.local,
         );
-      }),
-      distinctUntilChanged<MediaViewModel[]>(shallowEquals),
-    ),
-  );
-
-  const pip$ = scope.behavior<UserMediaViewModel | undefined>(
-    combineLatest([
-      // TODO This also needs epoch logic to dedupe the screenshares and mediaItems emits
-      screenShares$,
-      spotlightSpeaker$,
-      mediaItems$,
-    ]).pipe(
-      switchMap(([screenShares, spotlight, mediaItems]) => {
-        if (screenShares.length > 0) {
-          return spotlightSpeaker$;
-        }
-        if (!spotlight || spotlight.local) {
-          return of(undefined);
-        }
-
-        const localUserMedia = mediaItems.find(
-          (m) => m.type === "user" && m.local,
-        );
-        if (!localUserMedia) {
-          return of(undefined);
-        }
+        if (!localUserMedia) return of(undefined);
         return localUserMedia.alwaysShow$.pipe(
           map((alwaysShow) => (alwaysShow ? localUserMedia : undefined)),
         );
@@ -972,6 +949,39 @@ export function createCallViewModel$(
     ),
   );
 
+  const spotlightAndPip$ = scope.behavior<{
+    spotlight: MediaViewModel[];
+    pip$: Behavior<UserMediaViewModel | undefined>;
+  }>(
+    ringingMedia$.pipe(
+      switchMap((ringingMedia) => {
+        if (ringingMedia.length > 0)
+          return of({ spotlight: ringingMedia, pip$: localUserMediaForPip$ });
+
+        return screenShares$.pipe(
+          switchMap((screenShares) => {
+            if (screenShares.length > 0)
+              return of({ spotlight: screenShares, pip$: spotlightSpeaker$ });
+
+            return spotlightSpeaker$.pipe(
+              map((speaker) => ({
+                spotlight: speaker ? [speaker] : [],
+                pip$: localUserMediaForPip$,
+              })),
+            );
+          }),
+        );
+      }),
+    ),
+  );
+
+  const spotlight$ = scope.behavior<MediaViewModel[]>(
+    spotlightAndPip$.pipe(
+      map(({ spotlight }) => spotlight),
+      distinctUntilChanged<MediaViewModel[]>(shallowEquals),
+    ),
+  );
+
   const hasRemoteScreenShares$ = scope.behavior<boolean>(
     spotlight$.pipe(
       map((spotlight) =>
@@ -1054,24 +1064,61 @@ export function createCallViewModel$(
     }));
 
   const spotlightExpandedLayoutMedia$: Observable<SpotlightExpandedLayoutMedia> =
-    combineLatest([spotlight$, pip$], (spotlight, pip) => ({
-      type: "spotlight-expanded",
-      spotlight,
-      pip: pip ?? undefined,
-    }));
+    spotlightAndPip$.pipe(
+      switchMap(({ spotlight, pip$ }) =>
+        pip$.pipe(
+          map((pip) => ({
+            type: "spotlight-expanded" as const,
+            spotlight,
+            pip: pip ?? undefined,
+          })),
+        ),
+      ),
+    );
 
   const oneOnOneLayoutMedia$: Observable<OneOnOneLayoutMedia | null> =
-    mediaItems$.pipe(
-      map((mediaItems) => {
-        if (mediaItems.length !== 2) return null;
-        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
-        if (!remote || !local) return null;
+    userMedia$.pipe(
+      switchMap((userMedia) => {
+        if (userMedia.length <= 2) {
+          const local = userMedia.find(
+            (vm): vm is WrappedUserMediaViewModel & LocalUserMediaViewModel =>
+              vm.type === "user" && vm.local,
+          );
 
-        return { type: "one-on-one", local, remote };
+          if (local !== undefined) {
+            const remote = userMedia.find(
+              (
+                vm,
+              ): vm is WrappedUserMediaViewModel & RemoteUserMediaViewModel =>
+                vm.type === "user" && !vm.local,
+            );
+
+            if (remote !== undefined)
+              return of({
+                type: "one-on-one" as const,
+                spotlight: remote,
+                pip: local,
+              });
+
+            // If there's no other user media in the call (could still happen in
+            // this branch due to the duplicate tiles option), we could possibly
+            // show ringing media instead
+            if (userMedia.length === 1)
+              return ringingMedia$.pipe(
+                map((ringingMedia) => {
+                  return ringingMedia.length === 1
+                    ? {
+                        type: "one-on-one" as const,
+                        spotlight: local,
+                        pip: ringingMedia[0],
+                      }
+                    : null;
+                }),
+              );
+          }
+        }
+
+        return of(null);
       }),
     );
 
@@ -1482,8 +1529,9 @@ export function createCallViewModel$(
 
   return {
     autoLeave$: autoLeave$,
-    callPickupState$: callPickupState$,
-    ringOverlay$: ringOverlay$,
+    ringing$: scope.behavior(
+      callPickupState$.pipe(map((state) => state === "ringing")),
+    ),
     leave$: leave$,
     hangup: (): void => userHangup$.next(),
     join: localMembership.requestJoinAndPublish,

src/state/CallViewModel/CallViewModelTestUtils.ts

@@ -17,7 +17,7 @@ import {
 import { SyncState } from "matrix-js-sdk/lib/sync";
 import { BehaviorSubject, type Observable, map, of } from "rxjs";
 import { onTestFinished, vi } from "vitest";
-import { ClientEvent, type MatrixClient } from "matrix-js-sdk";
+import { ClientEvent, type RoomMember, type MatrixClient } from "matrix-js-sdk";
 import EventEmitter from "events";
 import * as ComponentsCore from "@livekit/components-core";
 
@@ -63,15 +63,10 @@ const carol = local;
 
 const dave = mockMatrixRoomMember(daveRTLRtcMember, { rawDisplayName: "Dave" });
 
-const roomMembers = new Map(
-  [alice, aliceDoppelganger, bob, bobZeroWidthSpace, carol, dave, daveRTL].map(
-    (p) => [p.userId, p],
-  ),
-);
-
 export interface CallViewModelInputs {
   remoteParticipants$: Behavior<RemoteParticipant[]>;
   rtcMembers$: Behavior<Partial<CallMembership>[]>;
+  roomMembers: RoomMember[];
   livekitConnectionState$: Behavior<ConnectionState>;
   speaking: Map<Participant, Observable<boolean>>;
   mediaDevices: MediaDevices;
@@ -86,6 +81,15 @@ export function withCallViewModel(mode: MatrixRTCMode) {
     {
       remoteParticipants$ = constant([]),
       rtcMembers$ = constant([localRtcMember]),
+      roomMembers = [
+        alice,
+        aliceDoppelganger,
+        bob,
+        bobZeroWidthSpace,
+        carol,
+        dave,
+        daveRTL,
+      ],
       livekitConnectionState$: connectionState$ = constant(
         ConnectionState.Connected,
       ),
@@ -128,8 +132,8 @@ export function withCallViewModel(mode: MatrixRTCMode) {
           return syncState;
         }
       })() as Partial<MatrixClient> as MatrixClient,
-      getMembers: () => Array.from(roomMembers.values()),
-      getMembersWithMembership: () => Array.from(roomMembers.values()),
+      getMembers: () => roomMembers,
+      getMembersWithMembership: () => roomMembers,
     });
     const rtcSession = new MockRTCSession(room, []).withMemberships(
       rtcMembers$,

src/state/CallViewModel/remoteMembers/MatrixMemberMetadata.ts

@@ -54,31 +54,6 @@ export function createRoomMembers$(
   );
 }
 
-/**
- * creates the member that this DM is with in case it is a DM (two members) otherwise null
- */
-export function createDMMember$(
-  scope: ObservableScope,
-  roomMembers$: Behavior<RoomMemberMap>,
-  matrixRoom: MatrixRoom,
-): Behavior<Pick<
-  RoomMember,
-  "userId" | "getMxcAvatarUrl" | "rawDisplayName"
-> | null> {
-  // We cannot use the normal direct check from matrix since we do not have access to the account data.
-  // use primitive member count === 2 check instead.
-  return scope.behavior(
-    roomMembers$.pipe(
-      map((membersMap) => {
-        // primitive appraoch do to no access to account data.
-        const isDM = membersMap.size === 2;
-        if (!isDM) return null;
-        return matrixRoom.getMember(matrixRoom.guessDMUserId());
-      }),
-    ),
-  );
-}
-
 /**
  * Displayname for each member of the call. This will disambiguate
  * any displayname that clashes with another member. Only members

src/state/IOSControlledAudioOutput.ts

@@ -0,0 +1,132 @@
+/*
+Copyright 2026 Element Corp.
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import { logger as rootLogger } from "matrix-js-sdk/lib/logger";
+import { combineLatest, merge, startWith, Subject, tap } from "rxjs";
+
+import {
+  availableOutputDevices$ as controlledAvailableOutputDevices$,
+  outputDevice$ as controlledOutputSelection$,
+} from "../controls.ts";
+import type { Behavior } from "./Behavior.ts";
+import type { ObservableScope } from "./ObservableScope.ts";
+import {
+  type AudioOutputDeviceLabel,
+  availableRawDevices$,
+  iosDeviceMenu$,
+  type MediaDevice,
+  type SelectedAudioOutputDevice,
+} from "./MediaDevices.ts";
+
+// This hardcoded id is used in EX ios! It can only be changed in coordination with
+// the ios swift team.
+const EARPIECE_CONFIG_ID = "earpiece-id";
+
+/**
+ * A special implementation of audio output that allows the hosting application
+ * to have more control over the device selection process. This is used when the
+ * `controlledAudioDevices` URL parameter is set, which is currently only true on mobile.
+ */
+export class IOSControlledAudioOutput implements MediaDevice<
+  AudioOutputDeviceLabel,
+  SelectedAudioOutputDevice
+> {
+  private logger = rootLogger.getChild("[MediaDevices ControlledAudioOutput]");
+  // We need to subscribe to the raw devices so that the OS does update the input
+  // back to what it was before. otherwise we will switch back to the default
+  // whenever we allocate a new stream.
+  public readonly availableRaw$ = availableRawDevices$(
+    "audiooutput",
+    this.usingNames$,
+    this.scope,
+    this.logger,
+  );
+
+  public readonly available$ = this.scope.behavior(
+    combineLatest(
+      [controlledAvailableOutputDevices$.pipe(startWith([])), iosDeviceMenu$],
+      (availableRaw, iosDeviceMenu) => {
+        const available = new Map<string, AudioOutputDeviceLabel>(
+          availableRaw.map(
+            ({ id, name, isEarpiece, isSpeaker /*,isExternalHeadset*/ }) => {
+              let deviceLabel: AudioOutputDeviceLabel;
+              // if (isExternalHeadset) // Do we want this?
+              if (isEarpiece) deviceLabel = { type: "earpiece" };
+              else if (isSpeaker) deviceLabel = { type: "speaker" };
+              else deviceLabel = { type: "name", name };
+              return [id, deviceLabel];
+            },
+          ),
+        );
+
+        // Create a virtual earpiece device in case a non-earpiece device is
+        // designated for this purpose
+        if (iosDeviceMenu && availableRaw.some((d) => d.forEarpiece)) {
+          this.logger.info(
+            `IOS Add virtual earpiece device with id ${EARPIECE_CONFIG_ID}`,
+          );
+          available.set(EARPIECE_CONFIG_ID, { type: "earpiece" });
+        }
+
+        return available;
+      },
+    ),
+  );
+
+  private readonly deviceSelection$ = new Subject<string>();
+
+  public select(id: string): void {
+    this.logger.info(`select device: ${id}`);
+    this.deviceSelection$.next(id);
+  }
+
+  public readonly selected$ = this.scope.behavior(
+    combineLatest(
+      [
+        this.available$,
+        merge(
+          controlledOutputSelection$.pipe(startWith(undefined)),
+          this.deviceSelection$,
+        ),
+      ],
+      (available, preferredId) => {
+        const id = preferredId ?? available.keys().next().value;
+        return id === undefined
+          ? undefined
+          : { id, virtualEarpiece: id === EARPIECE_CONFIG_ID };
+      },
+    ).pipe(
+      tap((selected) => {
+        this.logger.debug(`selected device: ${selected?.id}`);
+      }),
+    ),
+  );
+
+  public constructor(
+    private readonly usingNames$: Behavior<boolean>,
+    private readonly scope: ObservableScope,
+  ) {
+    this.selected$.subscribe((device) => {
+      // Let the hosting application know which output device has been selected.
+      // This information is probably only of interest if the earpiece mode has
+      // been selected - for example, Element X iOS listens to this to determine
+      // whether it should enable the proximity sensor.
+      if (device !== undefined) {
+        this.logger.info("onAudioDeviceSelect called:", device);
+        window.controls.onAudioDeviceSelect?.(device.id);
+        // Also invoke the deprecated callback for backward compatibility
+        window.controls.onOutputDeviceSelect?.(device.id);
+      }
+    });
+    this.available$.subscribe((available) => {
+      this.logger.debug("available devices:", available);
+    });
+    this.availableRaw$.subscribe((availableRaw) => {
+      this.logger.debug("available raw devices:", availableRaw);
+    });
+  }
+}

src/state/MediaDevices.ts

@@ -9,35 +9,28 @@ import {
   combineLatest,
   filter,
   map,
-  merge,
+  type Observable,
   pairwise,
-  startWith,
   Subject,
   switchMap,
-  type Observable,
 } from "rxjs";
 import { createMediaDeviceObserver } from "@livekit/components-core";
 import { type Logger, logger as rootLogger } from "matrix-js-sdk/lib/logger";
 
 import {
+  alwaysShowIphoneEarpiece as alwaysShowIphoneEarpieceSetting,
   audioInput as audioInputSetting,
   audioOutput as audioOutputSetting,
   videoInput as videoInputSetting,
-  alwaysShowIphoneEarpiece as alwaysShowIphoneEarpieceSetting,
 } from "../settings/settings";
 import { type ObservableScope } from "./ObservableScope";
-import {
-  outputDevice$ as controlledOutputSelection$,
-  availableOutputDevices$ as controlledAvailableOutputDevices$,
-} from "../controls";
+import { availableOutputDevices$ as controlledAvailableOutputDevices$ } from "../controls";
 import { getUrlParams } from "../UrlParams";
 import { platform } from "../Platform";
 import { switchWhen } from "../utils/observable";
 import { type Behavior, constant } from "./Behavior";
-
-// This hardcoded id is used in EX ios! It can only be changed in coordination with
-// the ios swift team.
-const EARPIECE_CONFIG_ID = "earpiece-id";
+import { AndroidControlledAudioOutput } from "./AndroidControlledAudioOutput.ts";
+import { IOSControlledAudioOutput } from "./IOSControlledAudioOutput.ts";
 
 export type DeviceLabel =
   | { type: "name"; name: string }
@@ -49,10 +42,18 @@ export type AudioOutputDeviceLabel =
   | { type: "earpiece" }
   | { type: "default"; name: string | null };
 
+/**
+ * Base selected-device value shared by all media kinds.
+ *
+ * `id` is the effective device identifier used by browser media APIs.
+ */
 export interface SelectedDevice {
   id: string;
 }
 
+/**
+ * Selected audio input value with audio-input-specific metadata.
+ */
 export interface SelectedAudioInputDevice extends SelectedDevice {
   /**
    * Emits whenever we think that this audio input device has logically changed
@@ -61,6 +62,9 @@ export interface SelectedAudioInputDevice extends SelectedDevice {
   hardwareDeviceChange$: Observable<void>;
 }
 
+/**
+ * Selected audio output value with output-routing-specific metadata.
+ */
 export interface SelectedAudioOutputDevice extends SelectedDevice {
   /**
    * Whether this device is a "virtual earpiece" device. If so, we should output
@@ -69,23 +73,42 @@ export interface SelectedAudioOutputDevice extends SelectedDevice {
   virtualEarpiece: boolean;
 }
 
+/**
+ * Common reactive contract for selectable input/output media devices (mic, speaker, camera).
+ *
+ * `Label` is the type used to represent a device in UI lists.
+ * `Selected` is the type used to represent the active selection for a device kind.
+ */
 export interface MediaDevice<Label, Selected> {
   /**
-   * A map from available device IDs to labels.
+   * Reactive map of currently available devices keyed by device ID.
+   *
+   * `Label` defines the UI-facing label data structure for each device type.
    */
   available$: Behavior<Map<string, Label>>;
+
   /**
-   * The selected device.
+   * The active device selection.
+   * Can be `undefined` when no device is yet selected.
+   *
+   * When defined, `Selected` contains the selected device ID plus any
+   * type-specific metadata.
    */
   selected$: Behavior<Selected | undefined>;
+
   /**
-   * Selects a new device.
+   * Requests selection of a device by ID.
+   *
+   * Implementations typically persist this preference and let `selected$`
+   * converge to the effective device (which may differ if the requested ID is
+   * unavailable).
    */
   select(id: string): void;
 }
 
 /**
  * An observable that represents if we should display the devices menu for iOS.
+ *
  * This implies the following
  *  - hide any input devices (they do not work anyhow on ios)
  *  - Show a button to show the native output picker instead.
@@ -95,7 +118,7 @@ export interface MediaDevice<Label, Selected> {
 export const iosDeviceMenu$ =
   platform === "ios" ? constant(true) : alwaysShowIphoneEarpieceSetting.value$;
 
-function availableRawDevices$(
+export function availableRawDevices$(
   kind: MediaDeviceKind,
   usingNames$: Behavior<boolean>,
   scope: ObservableScope,
@@ -146,16 +169,23 @@ function selectDevice$<Label>(
 ): Observable<string | undefined> {
   return combineLatest([available$, preferredId$], (available, preferredId) => {
     if (available.size) {
-      // If the preferred device is available, use it. Or if every available
-      // device ID is falsy, the browser is probably just being paranoid about
-      // fingerprinting and we should still try using the preferred device.
-      // Worst case it is not available and the browser will gracefully fall
-      // back to some other device for us when requesting the media stream.
-      // Otherwise, select the first available device.
-      return (preferredId !== undefined && available.has(preferredId)) ||
-        (available.size === 1 && available.has(""))
-        ? preferredId
-        : available.keys().next().value;
+      if (preferredId !== undefined && available.has(preferredId)) {
+        // If the preferred device is available, use it.
+        return preferredId;
+      } else if (available.size === 1 && available.has("")) {
+        // In some cases the enumerateDevices will list the devices with empty string details:
+        // `{deviceId:'', kind:'audiooutput|audioinput|videoinput', label:'', groupId:''}`
+        // This can happen when:
+        //     1. The user has not yet granted permissions to microphone/devices
+        //     2. The page is not running in a secure context (e.g. localhost or https)
+        //     3. In embedded WebViews, restrictions are often tighter, need active capture..
+        //     3. The browser is blocking access to device details for privacy reasons (?)
+        // This is most likely transitional, so keep the current device selected until we get a more accurate enumerateDevices.
+        return preferredId;
+      } else {
+        // No preferred, so pick a default.
+        return available.keys().next().value;
+      }
     }
     return undefined;
   });
@@ -212,7 +242,7 @@ class AudioInput implements MediaDevice<DeviceLabel, SelectedAudioInputDevice> {
   }
 }
 
-class AudioOutput implements MediaDevice<
+export class AudioOutput implements MediaDevice<
   AudioOutputDeviceLabel,
   SelectedAudioOutputDevice
 > {
@@ -251,14 +281,16 @@ class AudioOutput implements MediaDevice<
 
   public readonly selected$ = this.scope.behavior(
     selectDevice$(this.available$, audioOutputSetting.value$).pipe(
-      map((id) =>
-        id === undefined
-          ? undefined
-          : {
-              id,
-              virtualEarpiece: false,
-            },
-      ),
+      map((id) => {
+        if (id === undefined) {
+          return undefined;
+        } else {
+          return {
+            id,
+            virtualEarpiece: false,
+          };
+        }
+      }),
     ),
   );
   public select(id: string): void {
@@ -275,103 +307,6 @@ class AudioOutput implements MediaDevice<
   }
 }
 
-class ControlledAudioOutput implements MediaDevice<
-  AudioOutputDeviceLabel,
-  SelectedAudioOutputDevice
-> {
-  private logger = rootLogger.getChild("[MediaDevices ControlledAudioOutput]");
-  // We need to subscribe to the raw devices so that the OS does update the input
-  // back to what it was before. otherwise we will switch back to the default
-  // whenever we allocate a new stream.
-  public readonly availableRaw$ = availableRawDevices$(
-    "audiooutput",
-    this.usingNames$,
-    this.scope,
-    this.logger,
-  );
-
-  public readonly available$ = this.scope.behavior(
-    combineLatest(
-      [controlledAvailableOutputDevices$.pipe(startWith([])), iosDeviceMenu$],
-      (availableRaw, iosDeviceMenu) => {
-        const available = new Map<string, AudioOutputDeviceLabel>(
-          availableRaw.map(
-            ({ id, name, isEarpiece, isSpeaker /*,isExternalHeadset*/ }) => {
-              let deviceLabel: AudioOutputDeviceLabel;
-              // if (isExternalHeadset) // Do we want this?
-              if (isEarpiece) deviceLabel = { type: "earpiece" };
-              else if (isSpeaker) deviceLabel = { type: "speaker" };
-              else deviceLabel = { type: "name", name };
-              return [id, deviceLabel];
-            },
-          ),
-        );
-
-        // Create a virtual earpiece device in case a non-earpiece device is
-        // designated for this purpose
-        if (iosDeviceMenu && availableRaw.some((d) => d.forEarpiece))
-          available.set(EARPIECE_CONFIG_ID, { type: "earpiece" });
-
-        return available;
-      },
-    ),
-  );
-
-  private readonly deviceSelection$ = new Subject<string>();
-
-  public select(id: string): void {
-    this.deviceSelection$.next(id);
-  }
-
-  public readonly selected$ = this.scope.behavior(
-    combineLatest(
-      [
-        this.available$,
-        merge(
-          controlledOutputSelection$.pipe(startWith(undefined)),
-          this.deviceSelection$,
-        ),
-      ],
-      (available, preferredId) => {
-        const id = preferredId ?? available.keys().next().value;
-        return id === undefined
-          ? undefined
-          : { id, virtualEarpiece: id === EARPIECE_CONFIG_ID };
-      },
-    ),
-  );
-
-  public constructor(
-    private readonly usingNames$: Behavior<boolean>,
-    private readonly scope: ObservableScope,
-  ) {
-    this.selected$.subscribe((device) => {
-      // Let the hosting application know which output device has been selected.
-      // This information is probably only of interest if the earpiece mode has
-      // been selected - for example, Element X iOS listens to this to determine
-      // whether it should enable the proximity sensor.
-      if (device !== undefined) {
-        this.logger.info(
-          "[controlled-output] onAudioDeviceSelect called:",
-          device,
-        );
-        window.controls.onAudioDeviceSelect?.(device.id);
-        // Also invoke the deprecated callback for backward compatibility
-        window.controls.onOutputDeviceSelect?.(device.id);
-      }
-    });
-    this.available$.subscribe((available) => {
-      this.logger.info("[controlled-output] available devices:", available);
-    });
-    this.availableRaw$.subscribe((availableRaw) => {
-      this.logger.info(
-        "[controlled-output] available raw devices:",
-        availableRaw,
-      );
-    });
-  }
-}
-
 class VideoInput implements MediaDevice<DeviceLabel, SelectedDevice> {
   private logger = rootLogger.getChild("[MediaDevices VideoInput]");
 
@@ -434,7 +369,14 @@ export class MediaDevices {
     AudioOutputDeviceLabel,
     SelectedAudioOutputDevice
   > = getUrlParams().controlledAudioDevices
-    ? new ControlledAudioOutput(this.usingNames$, this.scope)
+    ? platform == "android"
+      ? new AndroidControlledAudioOutput(
+          controlledAvailableOutputDevices$,
+          this.scope,
+          getUrlParams().callIntent,
+          window.controls,
+        )
+      : new IOSControlledAudioOutput(this.usingNames$, this.scope)
     : new AudioOutput(this.usingNames$, this.scope);
 
   public readonly videoInput: MediaDevice<DeviceLabel, SelectedDevice> =

src/state/OneOnOneLayout.ts

@@ -16,14 +16,14 @@ export function oneOnOneLayout(
   prevTiles: TileStore,
 ): [OneOnOneLayout, TileStore] {
   const update = prevTiles.from(2);
-  update.registerGridTile(media.local);
-  update.registerGridTile(media.remote);
+  update.registerGridTile(media.pip);
+  update.registerGridTile(media.spotlight);
   const tiles = update.build();
   return [
     {
       type: media.type,
-      local: tiles.gridTilesByMedia.get(media.local)!,
-      remote: tiles.gridTilesByMedia.get(media.remote)!,
+      spotlight: tiles.gridTilesByMedia.get(media.spotlight)!,
+      pip: tiles.gridTilesByMedia.get(media.pip)!,
     },
     tiles,
   ];

src/state/TileStore.ts

@@ -13,6 +13,7 @@ import { fillGaps } from "../utils/iter";
 import { debugTileLayout } from "../settings/settings";
 import { type MediaViewModel } from "./media/MediaViewModel";
 import { type UserMediaViewModel } from "./media/UserMediaViewModel";
+import { type RingingMediaViewModel } from "./media/RingingMediaViewModel";
 
 function debugEntries(entries: GridTileData[]): string[] {
   return entries.map((e) => e.media.displayName$.value);
@@ -48,8 +49,10 @@ class SpotlightTileData {
 }
 
 class GridTileData {
-  private readonly media$: BehaviorSubject<UserMediaViewModel>;
-  public get media(): UserMediaViewModel {
+  private readonly media$: BehaviorSubject<
+    UserMediaViewModel | RingingMediaViewModel
+  >;
+  public get media(): UserMediaViewModel | RingingMediaViewModel {
     return this.media$.value;
   }
   public set media(value: UserMediaViewModel) {
@@ -58,7 +61,7 @@ class GridTileData {
 
   public readonly vm: GridTileViewModel;
 
-  public constructor(media: UserMediaViewModel) {
+  public constructor(media: UserMediaViewModel | RingingMediaViewModel) {
     this.media$ = new BehaviorSubject(media);
     this.vm = new GridTileViewModel(this.media$);
   }
@@ -178,7 +181,9 @@ export class TileStoreBuilder {
    * Sets up a grid tile for the given media. If this is never called for some
    * media, then that media will have no grid tile.
    */
-  public registerGridTile(media: UserMediaViewModel): void {
+  public registerGridTile(
+    media: UserMediaViewModel | RingingMediaViewModel,
+  ): void {
     if (DEBUG_ENABLED)
       logger.debug(
         `[TileStore, ${this.generation}] register grid tile: ${media.displayName$.value}`,
@@ -187,7 +192,11 @@ export class TileStoreBuilder {
     if (this.spotlight !== null) {
       // We actually *don't* want spotlight speakers to appear in both the
       // spotlight and the grid, so they're filtered out here
-      if (!media.local && this.spotlight.media.includes(media)) return;
+      if (
+        !(media.type === "user" && media.local) &&
+        this.spotlight.media.includes(media)
+      )
+        return;
       // When the spotlight speaker changes, we would see one grid tile appear
       // and another grid tile disappear. This would be an undesirable layout
       // shift, so instead what we do is take the speaker's grid tile and swap

src/state/TileViewModel.ts

@@ -7,6 +7,7 @@ Please see LICENSE in the repository root for full details.
 
 import { type Behavior } from "./Behavior";
 import { type MediaViewModel } from "./media/MediaViewModel";
+import { type RingingMediaViewModel } from "./media/RingingMediaViewModel";
 import { type UserMediaViewModel } from "./media/UserMediaViewModel";
 
 let nextId = 0;
@@ -17,7 +18,11 @@ function createId(): string {
 export class GridTileViewModel {
   public readonly id = createId();
 
-  public constructor(public readonly media$: Behavior<UserMediaViewModel>) {}
+  public constructor(
+    public readonly media$: Behavior<
+      UserMediaViewModel | RingingMediaViewModel
+    >,
+  ) {}
 }
 
 export class SpotlightTileViewModel {

src/state/layout-types.ts

@@ -7,7 +7,7 @@ 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 RingingMediaViewModel } from "./media/RingingMediaViewModel.ts";
 import { type UserMediaViewModel } from "./media/UserMediaViewModel.ts";
 import {
   type GridTileViewModel,
@@ -40,8 +40,8 @@ export interface SpotlightExpandedLayoutMedia {
 
 export interface OneOnOneLayoutMedia {
   type: "one-on-one";
-  local: LocalUserMediaViewModel;
-  remote: RemoteUserMediaViewModel;
+  spotlight: UserMediaViewModel;
+  pip: LocalUserMediaViewModel | RingingMediaViewModel;
 }
 
 export interface PipLayoutMedia {
@@ -86,8 +86,8 @@ export interface SpotlightExpandedLayout {
 
 export interface OneOnOneLayout {
   type: "one-on-one";
-  local: GridTileViewModel;
-  remote: GridTileViewModel;
+  spotlight: GridTileViewModel;
+  pip: GridTileViewModel;
 }
 
 export interface PipLayout {

src/state/media/MediaViewModel.ts

@@ -7,13 +7,17 @@ Please see LICENSE in the repository root for full details.
 */
 
 import { type Behavior } from "../Behavior";
+import { type RingingMediaViewModel } from "./RingingMediaViewModel";
 import { type ScreenShareViewModel } from "./ScreenShareViewModel";
 import { type UserMediaViewModel } from "./UserMediaViewModel";
 
 /**
  * A participant's media.
  */
-export type MediaViewModel = UserMediaViewModel | ScreenShareViewModel;
+export type MediaViewModel =
+  | UserMediaViewModel
+  | ScreenShareViewModel
+  | RingingMediaViewModel;
 
 /**
  * Properties which are common to all MediaViewModels.

src/state/media/MemberMediaViewModel.ts

@@ -38,6 +38,8 @@ import { type ObservableScope } from "../ObservableScope";
 import { observeTrackReference$ } from "../observeTrackReference";
 import { E2eeType } from "../../e2ee/e2eeType";
 import { observeInboundRtpStreamStats$ } from "./observeRtpStreamStats";
+import { type UserMediaViewModel } from "./UserMediaViewModel";
+import { type ScreenShareViewModel } from "./ScreenShareViewModel";
 
 // TODO: Encryption status is kinda broken and thus unused right now. Remove?
 export enum EncryptionStatus {
@@ -49,9 +51,9 @@ export enum EncryptionStatus {
 }
 
 /**
- * Media belonging to an active member of the RTC session.
+ * Properties common to all MemberMediaViewModels.
  */
-export interface MemberMediaViewModel extends BaseMediaViewModel {
+export interface BaseMemberMediaViewModel extends BaseMediaViewModel {
   /**
    * The LiveKit video track for this media.
    */
@@ -88,7 +90,7 @@ export function createMemberMedia(
     encryptionSystem,
     ...inputs
   }: MemberMediaInputs,
-): MemberMediaViewModel {
+): BaseMemberMediaViewModel {
   const trackBehavior$ = (
     source: Track.Source,
   ): Behavior<TrackReference | undefined> =>
@@ -270,3 +272,8 @@ function observeRemoteTrackReceivingOkay$(
     startWith(undefined),
   );
 }
+
+/**
+ * Media belonging to an active member of the call.
+ */
+export type MemberMediaViewModel = UserMediaViewModel | ScreenShareViewModel;

src/state/media/RingingMediaViewModel.ts

@@ -0,0 +1,51 @@
+/*
+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 MuteStates } from "../MuteStates";
+import {
+  type BaseMediaInputs,
+  type BaseMediaViewModel,
+  createBaseMedia,
+} from "./MediaViewModel";
+
+/**
+ * Media representing a user who is not yet part of the call — one that we are
+ * *ringing*.
+ */
+export interface RingingMediaViewModel extends BaseMediaViewModel {
+  type: "ringing";
+  pickupState$: Behavior<"ringing" | "timeout" | "decline">;
+  /**
+   * Whether this media would be expected to have video, were it not simply a
+   * placeholder.
+   */
+  videoEnabled$: Behavior<boolean>;
+}
+
+export interface RingingMediaInputs extends BaseMediaInputs {
+  pickupState$: Behavior<"ringing" | "timeout" | "decline">;
+  /**
+   * The local user's own mute states.
+   */
+  muteStates: MuteStates;
+}
+
+export function createRingingMedia({
+  pickupState$,
+  muteStates,
+  ...inputs
+}: RingingMediaInputs): RingingMediaViewModel {
+  return {
+    ...createBaseMedia(inputs),
+    type: "ringing",
+    pickupState$,
+    // If our own video is enabled, then this is a video call and we would
+    // expect remote media to have video as well
+    videoEnabled$: muteStates.video.enabled$,
+  };
+}

src/state/media/ScreenShareViewModel.ts

@@ -13,7 +13,7 @@ import { type LocalScreenShareViewModel } from "./LocalScreenShareViewModel";
 import {
   createMemberMedia,
   type MemberMediaInputs,
-  type MemberMediaViewModel,
+  type BaseMemberMediaViewModel,
 } from "./MemberMediaViewModel";
 import { type RemoteScreenShareViewModel } from "./RemoteScreenShareViewModel";
 
@@ -27,7 +27,7 @@ export type ScreenShareViewModel =
 /**
  * Properties which are common to all ScreenShareViewModels.
  */
-export interface BaseScreenShareViewModel extends MemberMediaViewModel {
+export interface BaseScreenShareViewModel extends BaseMemberMediaViewModel {
   type: "screen share";
 }
 

src/state/media/UserMediaViewModel.ts

@@ -27,7 +27,7 @@ import { type LocalUserMediaViewModel } from "./LocalUserMediaViewModel";
 import {
   createMemberMedia,
   type MemberMediaInputs,
-  type MemberMediaViewModel,
+  type BaseMemberMediaViewModel,
 } from "./MemberMediaViewModel";
 import { type RemoteUserMediaViewModel } from "./RemoteUserMediaViewModel";
 import { type ObservableScope } from "../ObservableScope";
@@ -42,7 +42,7 @@ export type UserMediaViewModel =
   | LocalUserMediaViewModel
   | RemoteUserMediaViewModel;
 
-export interface BaseUserMediaViewModel extends MemberMediaViewModel {
+export interface BaseUserMediaViewModel extends BaseMemberMediaViewModel {
   type: "user";
   speaking$: Behavior<boolean>;
   audioEnabled$: Behavior<boolean>;

src/state/media/WrappedUserMediaViewModel.ts

@@ -194,5 +194,3 @@ export function createWrappedUserMedia(
     ),
   };
 }
-
-export type MediaItem = WrappedUserMediaViewModel | ScreenShareViewModel;

src/tile/GridTile.test.tsx

@@ -7,9 +7,10 @@ Please see LICENSE in the repository root for full details.
 
 import { type RemoteTrackPublication } from "livekit-client";
 import { test, expect } from "vitest";
-import { render, screen } from "@testing-library/react";
+import { act, render, screen } from "@testing-library/react";
 import { axe } from "vitest-axe";
 import { type MatrixRTCSession } from "matrix-js-sdk/lib/matrixrtc";
+import { BehaviorSubject } from "rxjs";
 
 import { GridTile } from "./GridTile";
 import {
@@ -21,6 +22,11 @@ import { GridTileViewModel } from "../state/TileViewModel";
 import { ReactionsSenderProvider } from "../reactions/useReactionsSender";
 import type { CallViewModel } from "../state/CallViewModel/CallViewModel";
 import { constant } from "../state/Behavior";
+import {
+  createRingingMedia,
+  type RingingMediaViewModel,
+} from "../state/media/RingingMediaViewModel";
+import { type MuteStates } from "../state/MuteStates";
 
 global.IntersectionObserver = class MockIntersectionObserver {
   public observe(): void {}
@@ -28,6 +34,27 @@ global.IntersectionObserver = class MockIntersectionObserver {
   public disconnect(): void {}
 } as unknown as typeof IntersectionObserver;
 
+const fakeRtcSession = {
+  on: () => {},
+  off: () => {},
+  room: {
+    on: () => {},
+    off: () => {},
+    client: {
+      getUserId: () => null,
+      getDeviceId: () => null,
+      on: () => {},
+      off: () => {},
+    },
+  },
+  memberships: [],
+} as unknown as MatrixRTCSession;
+
+const callVm = {
+  reactions$: constant({}),
+  handsRaised$: constant({}),
+} as Partial<CallViewModel> as CallViewModel;
+
 test("GridTile is accessible", async () => {
   const vm = mockRemoteMedia(
     mockRtcMembership("@alice:example.org", "AAAA"),
@@ -42,34 +69,15 @@ test("GridTile is accessible", async () => {
     }),
   );
 
-  const fakeRtcSession = {
-    on: () => {},
-    off: () => {},
-    room: {
-      on: () => {},
-      off: () => {},
-      client: {
-        getUserId: () => null,
-        getDeviceId: () => null,
-        on: () => {},
-        off: () => {},
-      },
-    },
-    memberships: [],
-  } as unknown as MatrixRTCSession;
-  const cVm = {
-    reactions$: constant({}),
-    handsRaised$: constant({}),
-  } as Partial<CallViewModel> as CallViewModel;
   const { container } = render(
-    <ReactionsSenderProvider vm={cVm} rtcSession={fakeRtcSession}>
+    <ReactionsSenderProvider vm={callVm} rtcSession={fakeRtcSession}>
       <GridTile
         vm={new GridTileViewModel(constant(vm))}
         onOpenProfile={() => {}}
         targetWidth={300}
         targetHeight={200}
         showSpeakingIndicators
-        focusable={true}
+        focusable
       />
     </ReactionsSenderProvider>,
   );
@@ -77,3 +85,40 @@ test("GridTile is accessible", async () => {
   // Name should be visible
   screen.getByText("Alice");
 });
+
+test("GridTile displays ringing media", async () => {
+  const pickupState$ = new BehaviorSubject<
+    RingingMediaViewModel["pickupState$"]["value"]
+  >("ringing");
+  const vm = createRingingMedia({
+    pickupState$,
+    muteStates: {
+      video: { enabled$: constant(false) },
+    } as unknown as MuteStates,
+    id: "test",
+    userId: "@alice:example.org",
+    displayName$: constant("Alice"),
+    mxcAvatarUrl$: constant(undefined),
+  });
+
+  const { container } = render(
+    <ReactionsSenderProvider vm={callVm} rtcSession={fakeRtcSession}>
+      <GridTile
+        vm={new GridTileViewModel(constant(vm))}
+        onOpenProfile={() => {}}
+        targetWidth={300}
+        targetHeight={200}
+        showSpeakingIndicators
+        focusable
+      />
+    </ReactionsSenderProvider>,
+  );
+  expect(await axe(container)).toHaveNoViolations();
+  // Name and status should be visible
+  screen.getByText("Alice");
+  screen.getByText("Calling…");
+
+  // Alice declines the call
+  act(() => pickupState$.next("decline"));
+  screen.getByText("Call ended");
+});

src/tile/GridTile.tsx

@@ -29,6 +29,9 @@ import {
   UserProfileIcon,
   VolumeOffSolidIcon,
   SwitchCameraSolidIcon,
+  VideoCallSolidIcon,
+  VoiceCallSolidIcon,
+  EndCallIcon,
 } from "@vector-im/compound-design-tokens/assets/web/icons";
 import {
   ContextMenu,
@@ -49,6 +52,7 @@ 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";
+import { type RingingMediaViewModel } from "../state/media/RingingMediaViewModel";
 
 interface TileProps {
   ref?: Ref<HTMLDivElement>;
@@ -56,21 +60,56 @@ interface TileProps {
   style?: ComponentProps<typeof animated.div>["style"];
   targetWidth: number;
   targetHeight: number;
-  focusUrl: string | undefined;
   displayName: string;
   mxcAvatarUrl: string | undefined;
-  showSpeakingIndicators: boolean;
   focusable: boolean;
 }
 
+interface RingingMediaTileProps extends TileProps {
+  vm: RingingMediaViewModel;
+}
+
+const RingingMediaTile: FC<RingingMediaTileProps> = ({
+  vm,
+  className,
+  ...props
+}) => {
+  const { t } = useTranslation();
+  const pickupState = useBehavior(vm.pickupState$);
+  const videoEnabled = useBehavior(vm.videoEnabled$);
+
+  return (
+    <MediaView
+      className={classNames(className, styles.tile)}
+      video={undefined}
+      userId={vm.userId}
+      unencryptedWarning={false}
+      status={
+        pickupState === "ringing"
+          ? {
+              text: t("video_tile.calling"),
+              Icon: videoEnabled ? VideoCallSolidIcon : VoiceCallSolidIcon,
+            }
+          : { text: t("video_tile.call_ended"), Icon: EndCallIcon }
+      }
+      videoEnabled={videoEnabled}
+      videoFit="cover"
+      mirror={false}
+      {...props}
+    />
+  );
+};
+
 interface UserMediaTileProps extends TileProps {
   vm: UserMediaViewModel;
+  showSpeakingIndicators: boolean;
   mirror: boolean;
   playbackMuted: boolean;
   waitingForMedia?: boolean;
   primaryButton?: ReactNode;
   menuStart?: ReactNode;
   menuEnd?: ReactNode;
+  focusUrl: string | undefined;
 }
 
 const UserMediaTile: FC<UserMediaTileProps> = ({
@@ -95,7 +134,6 @@ const UserMediaTile: FC<UserMediaTileProps> = ({
   const { t } = useTranslation();
   const video = useBehavior(vm.video$);
   const unencryptedWarning = useBehavior(vm.unencryptedWarning$);
-  const encryptionStatus = useBehavior(vm.encryptionStatus$);
   const audioStreamStats = useObservableEagerState<
     RTCInboundRtpStreamStats | RTCOutboundRtpStreamStats | undefined
   >(vm.audioStreamStats$);
@@ -153,7 +191,6 @@ const UserMediaTile: FC<UserMediaTileProps> = ({
       video={video}
       userId={vm.userId}
       unencryptedWarning={unencryptedWarning}
-      encryptionStatus={encryptionStatus}
       videoEnabled={videoEnabled}
       videoFit={videoFit}
       className={classNames(className, styles.tile, {
@@ -218,6 +255,7 @@ UserMediaTile.displayName = "UserMediaTile";
 
 interface LocalUserMediaTileProps extends TileProps {
   vm: LocalUserMediaViewModel;
+  showSpeakingIndicators: boolean;
   onOpenProfile: (() => void) | null;
 }
 
@@ -232,6 +270,7 @@ const LocalUserMediaTile: FC<LocalUserMediaTileProps> = ({
   const mirror = useBehavior(vm.mirror$);
   const alwaysShow = useBehavior(vm.alwaysShow$);
   const switchCamera = useBehavior(vm.switchCamera$);
+  const focusUrl = useBehavior(vm.focusUrl$);
 
   const latestAlwaysShow = useLatest(alwaysShow);
   const onSelectAlwaysShow = useCallback(
@@ -278,6 +317,7 @@ const LocalUserMediaTile: FC<LocalUserMediaTileProps> = ({
         )
       }
       focusable={focusable}
+      focusUrl={focusUrl}
       {...props}
     />
   );
@@ -287,6 +327,7 @@ LocalUserMediaTile.displayName = "LocalUserMediaTile";
 
 interface RemoteUserMediaTileProps extends TileProps {
   vm: RemoteUserMediaViewModel;
+  showSpeakingIndicators: boolean;
 }
 
 const RemoteUserMediaTile: FC<RemoteUserMediaTileProps> = ({
@@ -298,6 +339,8 @@ const RemoteUserMediaTile: FC<RemoteUserMediaTileProps> = ({
   const waitingForMedia = useBehavior(vm.waitingForMedia$);
   const playbackMuted = useBehavior(vm.playbackMuted$);
   const playbackVolume = useBehavior(vm.playbackVolume$);
+  const focusUrl = useBehavior(vm.focusUrl$);
+
   const onSelectMute = useCallback(
     (e: Event) => {
       e.preventDefault();
@@ -338,6 +381,7 @@ const RemoteUserMediaTile: FC<RemoteUserMediaTileProps> = ({
           </MenuItem>
         </>
       }
+      focusUrl={focusUrl}
       {...props}
     />
   );
@@ -360,23 +404,33 @@ interface GridTileProps {
 export const GridTile: FC<GridTileProps> = ({
   ref: theirRef,
   vm,
+  showSpeakingIndicators,
   onOpenProfile,
   ...props
 }) => {
   const ourRef = useRef<HTMLDivElement | null>(null);
   const ref = useMergedRefs(ourRef, theirRef);
   const media = useBehavior(vm.media$);
-  const focusUrl = useBehavior(media.focusUrl$);
   const displayName = useBehavior(media.displayName$);
   const mxcAvatarUrl = useBehavior(media.mxcAvatarUrl$);
 
-  if (media.local) {
+  if (media.type === "ringing") {
+    return (
+      <RingingMediaTile
+        ref={ref}
+        vm={media}
+        {...props}
+        displayName={displayName}
+        mxcAvatarUrl={mxcAvatarUrl}
+      />
+    );
+  } else if (media.local) {
     return (
       <LocalUserMediaTile
         ref={ref}
         vm={media}
+        showSpeakingIndicators={showSpeakingIndicators}
         onOpenProfile={onOpenProfile}
-        focusUrl={focusUrl}
         displayName={displayName}
         mxcAvatarUrl={mxcAvatarUrl}
         {...props}
@@ -387,7 +441,7 @@ export const GridTile: FC<GridTileProps> = ({
       <RemoteUserMediaTile
         ref={ref}
         vm={media}
-        focusUrl={focusUrl}
+        showSpeakingIndicators={showSpeakingIndicators}
         displayName={displayName}
         mxcAvatarUrl={mxcAvatarUrl}
         {...props}

src/tile/MediaView.module.css

@@ -52,6 +52,11 @@ Please see LICENSE in the repository root for full details.
   pointer-events: none;
 }
 
+.translucent {
+  opacity: 50%;
+  mix-blend-mode: multiply;
+}
+
 /* CSS makes us put a condition here, even though all we want to do is
 unconditionally select the container so we can use cqmin units */
 @container mediaView (width > 0) {
@@ -71,14 +76,15 @@ unconditionally select the container so we can use cqmin units */
 
 .fg {
   position: absolute;
-  inset: var(
+  --fg-inset: var(
     --media-view-fg-inset,
     calc(var(--media-view-border-radius) - var(--cpd-space-3x))
   );
+  inset: var(--fg-inset);
   display: grid;
   grid-template-columns: 30px 1fr 30px;
   grid-template-rows: 1fr auto;
-  grid-template-areas: "reactions status ." "nameTag nameTag button";
+  grid-template-areas: "status status reactions" "nameTag nameTag button";
   gap: var(--cpd-space-1x);
   place-items: start;
 }
@@ -102,21 +108,19 @@ unconditionally select the container so we can use cqmin units */
 
 .status {
   grid-area: status;
-  justify-self: center;
-  align-self: start;
-  padding: var(--cpd-space-2x);
-  padding-block: var(--cpd-space-2x);
   color: var(--cpd-color-text-primary);
-  background-color: var(--cpd-color-bg-canvas-default);
   display: flex;
+  flex-wrap: none;
   align-items: center;
-  border-radius: var(--cpd-radius-pill-effect);
+  gap: 3px;
   user-select: none;
   overflow: hidden;
-  box-shadow: var(--small-drop-shadow);
-  box-sizing: border-box;
-  max-inline-size: 100%;
-  text-align: center;
+  margin-block-start: calc(var(--cpd-space-3x) - var(--fg-inset));
+  margin-inline-start: calc(var(--cpd-space-4x) - var(--fg-inset));
+}
+
+.status svg {
+  color: var(--cpd-color-icon-tertiary);
 }
 
 .reactions {

src/tile/MediaView.test.tsx

@@ -18,7 +18,6 @@ import { TrackInfo } from "@livekit/protocol";
 import { type ComponentProps } from "react";
 
 import { MediaView } from "./MediaView";
-import { EncryptionStatus } from "../state/media/MemberMediaViewModel";
 import { mockLocalParticipant } from "../utils/test";
 
 describe("MediaView", () => {
@@ -41,7 +40,6 @@ describe("MediaView", () => {
     videoFit: "contain",
     targetWidth: 300,
     targetHeight: 200,
-    encryptionStatus: EncryptionStatus.Connecting,
     mirror: false,
     unencryptedWarning: false,
     video: trackReference,

src/tile/MediaView.tsx

@@ -7,7 +7,13 @@ Please see LICENSE in the repository root for full details.
 
 import { type TrackReferenceOrPlaceholder } from "@livekit/components-core";
 import { animated } from "@react-spring/web";
-import { type FC, type ComponentProps, type ReactNode } from "react";
+import {
+  type FC,
+  type ComponentProps,
+  type ReactNode,
+  type ComponentType,
+  type SVGAttributes,
+} from "react";
 import { useTranslation } from "react-i18next";
 import classNames from "classnames";
 import { VideoTrack } from "@livekit/components-react";
@@ -16,7 +22,6 @@ 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/media/MemberMediaViewModel";
 import { RaisedHandIndicator } from "../reactions/RaisedHandIndicator";
 import {
   showConnectionStats as showConnectionStatsSetting,
@@ -38,7 +43,7 @@ interface Props extends ComponentProps<typeof animated.div> {
   userId: string;
   videoEnabled: boolean;
   unencryptedWarning: boolean;
-  encryptionStatus: EncryptionStatus;
+  status?: { text: string; Icon: ComponentType<SVGAttributes<SVGElement>> };
   nameTagLeadingIcon?: ReactNode;
   displayName: string;
   mxcAvatarUrl: string | undefined;
@@ -72,7 +77,7 @@ export const MediaView: FC<Props> = ({
   mxcAvatarUrl,
   focusable,
   primaryButton,
-  encryptionStatus,
+  status,
   raisedHandTime,
   currentReaction,
   raisedHandOnClick,
@@ -106,7 +111,11 @@ export const MediaView: FC<Props> = ({
           name={displayName}
           size={avatarSize}
           src={mxcAvatarUrl}
-          className={styles.avatar}
+          className={classNames(styles.avatar, {
+            // When the avatar is overlaid with a status, make it translucent
+            // for readability
+            [styles.translucent]: status,
+          })}
           style={{ display: video && videoEnabled ? "none" : "initial" }}
         />
         {video?.publication !== undefined && (
@@ -152,6 +161,14 @@ export const MediaView: FC<Props> = ({
             />
           </>
         )}
+        {status && (
+          <div className={styles.status}>
+            <status.Icon width={16} height={16} aria-hidden />
+            <Text as="span" size="sm" weight="medium">
+              {status.text}
+            </Text>
+          </div>
+        )}
         {/* TODO: Bring this back once encryption status is less broken */}
         {/*encryptionStatus !== EncryptionStatus.Okay && (
             <div className={styles.status}>

src/tile/SpotlightTile.test.tsx

@@ -6,10 +6,11 @@ Please see LICENSE in the repository root for full details.
 */
 
 import { test, expect, vi } from "vitest";
-import { isInaccessible, render, screen } from "@testing-library/react";
+import { act, 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 { BehaviorSubject } from "rxjs";
 
 import { SpotlightTile } from "./SpotlightTile";
 import {
@@ -23,6 +24,11 @@ import {
 } from "../utils/test";
 import { SpotlightTileViewModel } from "../state/TileViewModel";
 import { constant } from "../state/Behavior";
+import {
+  createRingingMedia,
+  type RingingMediaViewModel,
+} from "../state/media/RingingMediaViewModel";
+import { type MuteStates } from "../state/MuteStates";
 
 global.IntersectionObserver = class MockIntersectionObserver {
   public observe(): void {}
@@ -140,3 +146,41 @@ test("Screen share volume UI is hidden when screen share has no audio", async ()
     screen.queryByRole("button", { name: /volume/i }),
   ).not.toBeInTheDocument();
 });
+
+test("SpotlightTile displays ringing media", async () => {
+  const pickupState$ = new BehaviorSubject<
+    RingingMediaViewModel["pickupState$"]["value"]
+  >("ringing");
+  const vm = createRingingMedia({
+    pickupState$,
+    muteStates: {
+      video: { enabled$: constant(false) },
+    } as unknown as MuteStates,
+    id: "test",
+    userId: "@alice:example.org",
+    displayName$: constant("Alice"),
+    mxcAvatarUrl$: constant(undefined),
+  });
+
+  const toggleExpanded = vi.fn();
+  const { container } = render(
+    <SpotlightTile
+      vm={new SpotlightTileViewModel(constant([vm]), constant(false))}
+      targetWidth={300}
+      targetHeight={200}
+      expanded={false}
+      onToggleExpanded={toggleExpanded}
+      showIndicators
+      focusable={true}
+    />,
+  );
+
+  expect(await axe(container)).toHaveNoViolations();
+  // Alice should be in the spotlight with the right status
+  screen.getByText("Alice");
+  screen.getByText("Calling…");
+
+  // Now we time out ringing to Alice
+  act(() => pickupState$.next("timeout"));
+  screen.getByText("Call ended");
+});

src/tile/SpotlightTile.tsx

@@ -24,6 +24,9 @@ import {
   VolumeOnIcon,
   VolumeOffSolidIcon,
   VolumeOnSolidIcon,
+  VideoCallSolidIcon,
+  VoiceCallSolidIcon,
+  EndCallIcon,
 } from "@vector-im/compound-design-tokens/assets/web/icons";
 import { animated } from "@react-spring/web";
 import { type Observable, map } from "rxjs";
@@ -43,7 +46,7 @@ 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 MemberMediaViewModel } 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";
@@ -52,6 +55,7 @@ import { type RemoteScreenShareViewModel } from "../state/media/RemoteScreenShar
 import { type MediaViewModel } from "../state/media/MediaViewModel";
 import { Slider } from "../Slider";
 import { platform } from "../Platform";
+import { type RingingMediaViewModel } from "../state/media/RingingMediaViewModel";
 
 interface SpotlightItemBaseProps {
   ref?: Ref<HTMLDivElement>;
@@ -59,18 +63,20 @@ interface SpotlightItemBaseProps {
   "data-id": string;
   targetWidth: number;
   targetHeight: number;
-  video: TrackReferenceOrPlaceholder | undefined;
   userId: string;
-  unencryptedWarning: boolean;
-  encryptionStatus: EncryptionStatus;
-  focusUrl: string | undefined;
   displayName: string;
   mxcAvatarUrl: string | undefined;
   focusable: boolean;
   "aria-hidden"?: boolean;
 }
 
-interface SpotlightUserMediaItemBaseProps extends SpotlightItemBaseProps {
+interface SpotlightMemberMediaItemBaseProps extends SpotlightItemBaseProps {
+  video: TrackReferenceOrPlaceholder | undefined;
+  unencryptedWarning: boolean;
+  focusUrl: string | undefined;
+}
+
+interface SpotlightUserMediaItemBaseProps extends SpotlightMemberMediaItemBaseProps {
   videoFit: "contain" | "cover";
   videoEnabled: boolean;
 }
@@ -103,21 +109,32 @@ const SpotlightRemoteUserMediaItem: FC<SpotlightRemoteUserMediaItemProps> = ({
   );
 };
 
-interface SpotlightUserMediaItemProps extends SpotlightItemBaseProps {
+interface SpotlightUserMediaItemProps extends SpotlightMemberMediaItemBaseProps {
   vm: UserMediaViewModel;
 }
 
 const SpotlightUserMediaItem: FC<SpotlightUserMediaItemProps> = ({
   vm,
+  targetWidth,
+  targetHeight,
   ...props
 }) => {
   const videoFit = useBehavior(vm.videoFit$);
   const videoEnabled = useBehavior(vm.videoEnabled$);
 
+  // Whenever target bounds change, inform the viewModel
+  useEffect(() => {
+    if (targetWidth > 0 && targetHeight > 0) {
+      vm.setTargetDimensions(targetWidth, targetHeight);
+    }
+  }, [targetWidth, targetHeight, vm]);
+
   const baseProps: SpotlightUserMediaItemBaseProps &
     RefAttributes<HTMLDivElement> = {
     videoFit,
     videoEnabled,
+    targetWidth,
+    targetHeight,
     ...props,
   };
 
@@ -130,7 +147,7 @@ const SpotlightUserMediaItem: FC<SpotlightUserMediaItemProps> = ({
 
 SpotlightUserMediaItem.displayName = "SpotlightUserMediaItem";
 
-interface SpotlightScreenShareItemProps extends SpotlightItemBaseProps {
+interface SpotlightScreenShareItemProps extends SpotlightMemberMediaItemBaseProps {
   vm: ScreenShareViewModel;
   videoEnabled: boolean;
 }
@@ -142,7 +159,7 @@ const SpotlightScreenShareItem: FC<SpotlightScreenShareItemProps> = ({
   return <MediaView videoFit="contain" mirror={false} {...props} />;
 };
 
-interface SpotlightRemoteScreenShareItemProps extends SpotlightItemBaseProps {
+interface SpotlightRemoteScreenShareItemProps extends SpotlightMemberMediaItemBaseProps {
   vm: RemoteScreenShareViewModel;
 }
 
@@ -155,6 +172,67 @@ const SpotlightRemoteScreenShareItem: FC<
   );
 };
 
+interface SpotlightMemberMediaItemProps extends SpotlightItemBaseProps {
+  vm: MemberMediaViewModel;
+}
+
+const SpotlightMemberMediaItem: FC<SpotlightMemberMediaItemProps> = ({
+  vm,
+  ...props
+}) => {
+  const video = useBehavior(vm.video$);
+  const unencryptedWarning = useBehavior(vm.unencryptedWarning$);
+  const focusUrl = useBehavior(vm.focusUrl$);
+
+  const baseProps: SpotlightMemberMediaItemBaseProps &
+    RefAttributes<HTMLDivElement> = {
+    video: video ?? undefined,
+    unencryptedWarning,
+    focusUrl,
+    ...props,
+  };
+
+  if (vm.type === "user")
+    return <SpotlightUserMediaItem vm={vm} {...baseProps} />;
+  return vm.local ? (
+    <SpotlightScreenShareItem vm={vm} videoEnabled {...baseProps} />
+  ) : (
+    <SpotlightRemoteScreenShareItem vm={vm} {...baseProps} />
+  );
+};
+
+interface SpotlightRingingMediaItemProps extends SpotlightItemBaseProps {
+  vm: RingingMediaViewModel;
+}
+
+const SpotlightRingingMediaItem: FC<SpotlightRingingMediaItemProps> = ({
+  vm,
+  ...props
+}) => {
+  const { t } = useTranslation();
+  const pickupState = useBehavior(vm.pickupState$);
+  const videoEnabled = useBehavior(vm.videoEnabled$);
+
+  return (
+    <MediaView
+      video={undefined}
+      unencryptedWarning={false}
+      status={
+        pickupState === "ringing"
+          ? {
+              text: t("video_tile.calling"),
+              Icon: videoEnabled ? VideoCallSolidIcon : VoiceCallSolidIcon,
+            }
+          : { text: t("video_tile.call_ended"), Icon: EndCallIcon }
+      }
+      videoEnabled={false}
+      videoFit="cover"
+      mirror={false}
+      {...props}
+    />
+  );
+};
+
 interface SpotlightItemProps {
   ref?: Ref<HTMLDivElement>;
   vm: MediaViewModel;
@@ -187,22 +265,9 @@ const SpotlightItem: FC<SpotlightItemProps> = ({
 }) => {
   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 unencryptedWarning = useBehavior(vm.unencryptedWarning$);
-  const encryptionStatus = useBehavior(vm.encryptionStatus$);
 
   // Hook this item up to the intersection observer
   useEffect(() => {
@@ -225,23 +290,17 @@ const SpotlightItem: FC<SpotlightItemProps> = ({
     className: classNames(styles.item, { [styles.snap]: snap }),
     targetWidth,
     targetHeight,
-    video: video ?? undefined,
     userId: vm.userId,
-    unencryptedWarning,
-    focusUrl,
     displayName,
     mxcAvatarUrl,
     focusable,
-    encryptionStatus,
     "aria-hidden": ariaHidden,
   };
 
-  if (vm.type === "user")
-    return <SpotlightUserMediaItem vm={vm} {...baseProps} />;
-  return vm.local ? (
-    <SpotlightScreenShareItem vm={vm} videoEnabled {...baseProps} />
+  return vm.type === "ringing" ? (
+    <SpotlightRingingMediaItem vm={vm} {...baseProps} />
   ) : (
-    <SpotlightRemoteScreenShareItem vm={vm} {...baseProps} />
+    <SpotlightMemberMediaItem vm={vm} {...baseProps} />
   );
 };
 

src/widget.ts

@@ -93,6 +93,7 @@ export const initializeWidget = (
       logger.info("Widget API is available");
       const api = new WidgetApi(widgetId, parentOrigin);
       api.requestCapability(MatrixCapabilities.AlwaysOnScreen);
+      api.requestCapability(MatrixCapabilities.MSC4039DownloadFile);
 
       // Set up the lazy action emitter, but only for select actions that we
       // intend for the app to handle

yarn.lock

@@ -6236,8 +6236,8 @@ __metadata:
   linkType: hard
 
 "@vector-im/compound-design-tokens@npm:^6.0.0":
-  version: 6.6.0
-  resolution: "@vector-im/compound-design-tokens@npm:6.6.0"
+  version: 6.10.2
+  resolution: "@vector-im/compound-design-tokens@npm:6.10.2"
   peerDependencies:
     "@types/react": "*"
     react: ^17 || ^18 || ^19.0.0
@@ -6246,13 +6246,13 @@ __metadata:
       optional: true
     react:
       optional: true
-  checksum: 10c0/93b152dd1de96371f9b6b1f7dbcc381d7ab598031dbc900f52d610f015766c0d4426ae6e47d417e723bfb62d1a53099155b4d788848b78232916ba132c03c2fe
+  checksum: 10c0/bcac6d79fcfb8cc1356d65dff576bdad217edd0df189a5dea032b0fd57cef335b73ad6d8e395709245bc1c6a8c672a83144ecea48550ca560544d2399af8f2d3
   languageName: node
   linkType: hard
 
 "@vector-im/compound-web@npm:^8.0.0":
-  version: 8.3.4
-  resolution: "@vector-im/compound-web@npm:8.3.4"
+  version: 8.4.0
+  resolution: "@vector-im/compound-web@npm:8.4.0"
   dependencies:
     "@floating-ui/react": "npm:^0.27.0"
     "@radix-ui/react-context-menu": "npm:^2.2.16"
@@ -6272,7 +6272,7 @@ __metadata:
   peerDependenciesMeta:
     "@types/react":
       optional: true
-  checksum: 10c0/44764fa64b5fce2e7181e25b50ee970eda4d921cf650b92bd5e88df0eb60872f3086b8702d18f55c3e39b3751ac19f10bafda8c4306df65c3605bd44b297d95c
+  checksum: 10c0/31b73555c47b373d4250872bfe863a030b487197bf1198e3cf3a1ec344f2b02f0c72c1513bb598c1cbd7a91d3c6a334d0c8ae37bd7c90d4859c864fc223e059a
   languageName: node
   linkType: hard