Merge pull request #3756 from element-hq/valere/auto_fit_based_on_video_ratio

Auto fit video in tile depending on orientation of the tile and the source stream

locales/en/app.json

@@ -250,7 +250,6 @@
   "video_tile": {
     "always_show": "Always show",
     "camera_starting": "Video loading...",
-    "change_fit_contain": "Fit to frame",
     "collapse": "Collapse",
     "expand": "Expand",
     "mute_for_me": "Mute for me",

src/grid/TileWrapper.tsx

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

src/state/media/MediaViewModel.test.ts

@@ -160,21 +160,6 @@ test("control a participant's screen share volume", () => {
   });
 });
 
-test("toggle fit/contain for a participant's video", () => {
-  const vm = mockRemoteMedia(rtcMembership, {}, mockRemoteParticipant({}));
-  withTestScheduler(({ expectObservable, schedule }) => {
-    schedule("-ab|", {
-      a: () => vm.toggleCropVideo(),
-      b: () => vm.toggleCropVideo(),
-    });
-    expectObservable(vm.cropVideo$).toBe("abc", {
-      a: true,
-      b: false,
-      c: true,
-    });
-  });
-});
-
 test("local media remembers whether it should always be shown", () => {
   const vm1 = mockLocalMedia(
     rtcMembership,

src/state/media/UserMediaViewModel.ts

@@ -7,6 +7,7 @@ Please see LICENSE in the repository root for full details.
 */
 
 import {
+  BehaviorSubject,
   combineLatest,
   map,
   type Observable,
@@ -30,9 +31,9 @@ import {
 } from "./MemberMediaViewModel";
 import { type RemoteUserMediaViewModel } from "./RemoteUserMediaViewModel";
 import { type ObservableScope } from "../ObservableScope";
-import { createToggle$ } from "../../utils/observable";
 import { showConnectionStats } from "../../settings/settings";
 import { observeRtpStreamStats$ } from "./observeRtpStreamStats";
+import { videoFit$, videoSizeFromParticipant$ } from "../../utils/videoFit.ts";
 
 /**
  * A participant's user media (i.e. their microphone and camera feed).
@@ -46,7 +47,7 @@ export interface BaseUserMediaViewModel extends MemberMediaViewModel {
   speaking$: Behavior<boolean>;
   audioEnabled$: Behavior<boolean>;
   videoEnabled$: Behavior<boolean>;
-  cropVideo$: Behavior<boolean>;
+  videoFit$: Behavior<"cover" | "contain">;
   toggleCropVideo: () => void;
   /**
    * The expected identity of the LiveKit participant. Exposed for debugging.
@@ -60,6 +61,13 @@ export interface BaseUserMediaViewModel extends MemberMediaViewModel {
   videoStreamStats$: Observable<
     RTCInboundRtpStreamStats | RTCOutboundRtpStreamStats | undefined
   >;
+  /**
+   * Set the target dimensions of the HTML element (final dimension after anim).
+   * This can be used to determine the best video fit (fit to frame / keep ratio).
+   * @param targetWidth - The target width of the HTML element displaying the video.
+   * @param targetHeight - The target height of the HTML element displaying the video.
+   */
+  setTargetDimensions: (targetWidth: number, targetHeight: number) => void;
 }
 
 export interface BaseUserMediaInputs extends Omit<
@@ -90,6 +98,12 @@ export function createBaseUserMedia(
   );
   const toggleCropVideo$ = new Subject<void>();
 
+  // The target size of the video element, used to determine the best video fit.
+  // The target size is the final size of the HTML element after any animations have completed.
+  const targetSize$ = new BehaviorSubject<
+    { width: number; height: number } | undefined
+  >(undefined);
+
   return {
     ...createMemberMedia(scope, {
       ...inputs,
@@ -115,7 +129,11 @@ export function createBaseUserMedia(
     videoEnabled$: scope.behavior(
       media$.pipe(map((m) => m?.cameraTrack?.isMuted === false)),
     ),
-    cropVideo$: createToggle$(scope, true, toggleCropVideo$),
+    videoFit$: videoFit$(
+      scope,
+      videoSizeFromParticipant$(participant$),
+      targetSize$,
+    ),
     toggleCropVideo: () => toggleCropVideo$.next(),
     rtcBackendIdentity,
     handRaised$,
@@ -139,5 +157,8 @@ export function createBaseUserMedia(
         return observeRtpStreamStats$(p, Track.Source.Camera, statsType);
       }),
     ),
+    setTargetDimensions: (targetWidth: number, targetHeight: number): void => {
+      targetSize$.next({ width: targetWidth, height: targetHeight });
+    },
   };
 }

src/state/media/observeRtpStreamStats.ts

@@ -67,3 +67,12 @@ export function observeInboundRtpStreamStats$(
     map((x) => x as RTCInboundRtpStreamStats | undefined),
   );
 }
+
+export function observeOutboundRtpStreamStats$(
+  participant: Participant,
+  source: Track.Source,
+): Observable<RTCOutboundRtpStreamStats | undefined> {
+  return observeRtpStreamStats$(participant, source, "outbound-rtp").pipe(
+    map((x) => x as RTCOutboundRtpStreamStats | undefined),
+  );
+}

src/tile/GridTile.tsx

@@ -11,6 +11,7 @@ import {
   type ReactNode,
   type Ref,
   useCallback,
+  useEffect,
   useRef,
   useState,
 } from "react";
@@ -26,7 +27,6 @@ import {
   VolumeOffIcon,
   VisibilityOnIcon,
   UserProfileIcon,
-  ExpandIcon,
   VolumeOffSolidIcon,
   SwitchCameraSolidIcon,
 } from "@vector-im/compound-design-tokens/assets/web/icons";
@@ -87,6 +87,8 @@ const UserMediaTile: FC<UserMediaTileProps> = ({
   displayName,
   mxcAvatarUrl,
   focusable,
+  targetWidth,
+  targetHeight,
   ...props
 }) => {
   const { toggleRaisedHand } = useReactionsSender();
@@ -103,18 +105,19 @@ const UserMediaTile: FC<UserMediaTileProps> = ({
   const audioEnabled = useBehavior(vm.audioEnabled$);
   const videoEnabled = useBehavior(vm.videoEnabled$);
   const speaking = useBehavior(vm.speaking$);
-  const cropVideo = useBehavior(vm.cropVideo$);
+  const videoFit = useBehavior(vm.videoFit$);
-  const onSelectFitContain = useCallback(
+
-    (e: Event) => {
-      e.preventDefault();
-      vm.toggleCropVideo();
-    },
-    [vm],
-  );
   const rtcBackendIdentity = vm.rtcBackendIdentity;
   const handRaised = useBehavior(vm.handRaised$);
   const reaction = useBehavior(vm.reaction$);
 
+  // Whenever bounds change, inform the viewModel
+  useEffect(() => {
+    if (targetWidth > 0 && targetHeight > 0) {
+      vm.setTargetDimensions(targetWidth, targetHeight);
+    }
+  }, [targetWidth, targetHeight, vm]);
+
   const AudioIcon = playbackMuted
     ? VolumeOffSolidIcon
     : audioEnabled
@@ -130,12 +133,10 @@ const UserMediaTile: FC<UserMediaTileProps> = ({
   const menu = (
     <>
       {menuStart}
-      <ToggleMenuItem
+      {/*
-        Icon={ExpandIcon}
+       No additional menu item (used to be the manual fit to frame.
-        label={t("video_tile.change_fit_contain")}
+       Placeholder for future menu items that should be placed here.
-        checked={cropVideo}
+       */}
-        onSelect={onSelectFitContain}
-      />
       {menuEnd}
     </>
   );
@@ -154,7 +155,7 @@ const UserMediaTile: FC<UserMediaTileProps> = ({
       unencryptedWarning={unencryptedWarning}
       encryptionStatus={encryptionStatus}
       videoEnabled={videoEnabled}
-      videoFit={cropVideo ? "cover" : "contain"}
+      videoFit={videoFit}
       className={classNames(className, styles.tile, {
         [styles.speaking]: showSpeaking,
         [styles.handRaised]: !showSpeaking && handRaised,
@@ -200,6 +201,8 @@ const UserMediaTile: FC<UserMediaTileProps> = ({
       audioStreamStats={audioStreamStats}
       videoStreamStats={videoStreamStats}
       rtcBackendIdentity={rtcBackendIdentity}
+      targetWidth={targetWidth}
+      targetHeight={targetHeight}
       {...props}
     />
   );

src/tile/SpotlightTile.tsx

@@ -111,12 +111,12 @@ const SpotlightUserMediaItem: FC<SpotlightUserMediaItemProps> = ({
   vm,
   ...props
 }) => {
-  const cropVideo = useBehavior(vm.cropVideo$);
+  const videoFit = useBehavior(vm.videoFit$);
   const videoEnabled = useBehavior(vm.videoEnabled$);
 
   const baseProps: SpotlightUserMediaItemBaseProps &
     RefAttributes<HTMLDivElement> = {
-    videoFit: cropVideo ? "cover" : "contain",
+    videoFit,
     videoEnabled,
     ...props,
   };
@@ -158,7 +158,13 @@ const SpotlightRemoteScreenShareItem: FC<
 interface SpotlightItemProps {
   ref?: Ref<HTMLDivElement>;
   vm: MediaViewModel;
+  /**
+   * The width this tile will have once its animations have settled.
+   */
   targetWidth: number;
+  /**
+   * The height this tile will have once its animations have settled.
+   */
   targetHeight: number;
   focusable: boolean;
   intersectionObserver$: Observable<IntersectionObserver>;
@@ -180,6 +186,16 @@ const SpotlightItem: FC<SpotlightItemProps> = ({
   "aria-hidden": ariaHidden,
 }) => {
   const ourRef = useRef<HTMLDivElement | null>(null);
+
+  // Whenever target bounds change, inform the viewModel
+  useEffect(() => {
+    if (targetWidth > 0 && targetHeight > 0) {
+      if (vm.type != "screen share") {
+        vm.setTargetDimensions(targetWidth, targetHeight);
+      }
+    }
+  }, [targetWidth, targetHeight, vm]);
+
   const ref = useMergedRefs(ourRef, theirRef);
   const focusUrl = useBehavior(vm.focusUrl$);
   const displayName = useBehavior(vm.displayName$);

src/utils/videoFit.test.ts

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

src/utils/videoFit.ts

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