Bump patch

* Support Conversion.init(hardwareAcceleration) option passed down to the VideoEncodingConfig.

fix #253

* Few changes

---------

Co-authored-by: Raphaël Droz <[email protected]>
Co-authored-by: Vanilagy <[email protected]>

README.md

@@ -17,7 +17,10 @@ Mediabunny is a JavaScript library for reading, writing, and converting media fi
 
 <div align="center">
   <a href="https://remotion.dev/" target="_blank" rel="sponsored">
-    <img src="./docs/public/sponsors/remotion.png" width="60" height="60" alt="Remotion">
+    <picture>
+      <source srcset="./docs/public/sponsors/remotion-dark.png" media="(prefers-color-scheme: dark)">
+      <img src="./docs/public/sponsors/remotion-light.png" width="60" height="60" alt="Remotion">
+    </picture>
   </a>
   &nbsp;&nbsp;&nbsp;&nbsp;
   <a href="https://www.gling.ai/" target="_blank" rel="sponsored">

dev/convert.html

@@ -100,9 +100,10 @@
 			},
 			*/
 			video: () => ({
-				width: 720,
-				frameRate: 30,
-				bitrate: Mediabunny.QUALITY_VERY_LOW,
+				allowRotationMetadata: false,
+				//width: 720,
+				//frameRate: 30,
+				//bitrate: Mediabunny.QUALITY_VERY_LOW,
 				//discard: true,
 				/*
 				process: (sample) => {
@@ -180,7 +181,7 @@
 			},
 			trim: {
 				start: 0,
-				end: 20
+				end: 4
 			},
 		});
 		console.log(conversion);

dev/live.html

@@ -25,29 +25,41 @@
             target: new Mediabunny.BufferTarget(),
             format: new Mediabunny.Mp4OutputFormat(),
         });
+		let videoSource = null;
+		let audioSource = null;
         if (videoTrack) {
-            const source = new Mediabunny.MediaStreamVideoTrackSource(videoTrack, {
+            videoSource = new Mediabunny.MediaStreamVideoTrackSource(videoTrack, {
                 codec: 'avc',
                 bitrate: Mediabunny.QUALITY_MEDIUM
             });
 
-            source.errorPromise.catch((d) => console.log("Hello?????", d));
+            videoSource.errorPromise.catch((d) => console.log("Hello?????", d));
 
-            output.addVideoTrack(source);
+            output.addVideoTrack(videoSource);
         }
         if (audioTrack) {
-            const source = new Mediabunny.MediaStreamAudioTrackSource(audioTrack, {
+            audioSource = new Mediabunny.MediaStreamAudioTrackSource(audioTrack, {
                 codec: 'mp3',
                 bitrate: Mediabunny.QUALITY_MEDIUM
             });
 
-            source.errorPromise.catch((d) => console.log("Hello!!???", d));
+            audioSource.errorPromise.catch((d) => console.log("Hello!!???", d));
 
-            output.addAudioTrack(source);
+            output.addAudioTrack(audioSource);
         }
 
         await output.start();
 
+		setTimeout(() => {
+			videoSource?.pause();
+			audioSource?.pause();
+
+			setTimeout(() => {
+				videoSource?.resume();
+				audioSource?.resume();
+			}, 1000);
+		}, 1000);
+
         await new Promise(resolve => setTimeout(resolve, 5000));
 
         await output.finalize();

docs/guide/converting-media-files.md

@@ -120,11 +120,13 @@ type ConversionVideoOptions = {
 	height?: number;
 	fit?: 'fill' | 'contain' | 'cover';
 	rotate?: 0 | 90 | 180 | 270;
+	allowRotationMetadata?: boolean;
 	crop?: { left: number; top: number; width: number; height: number };
 	frameRate?: number;
 	codec?: VideoCodec;
 	bitrate?: number | Quality;
 	alpha?: 'discard' | 'keep'; // Defaults to 'discard'
+	hardwareAcceleration?: 'no-preference' | 'prefer-hardware' | 'prefer-software';
 	keyFrameInterval?: number;
 	forceTranscode?: boolean;
 	process?: (sample: VideoSample) => MaybePromise<
@@ -175,6 +177,8 @@ In the rare case that the input video changes size over time, the `fit` field ca
 
 `rotation` rotates the video by the specified number of degrees clockwise. This rotation is applied on top of any rotation metadata in the original input file and happens before cropping and resizing.
 
+By default, Mediabunny will try to make use of rotation metadata in the output file to perform the rotation whenever possible. However, if you don't want this to happen, or you want to use Mediabunny to strip all rotation metadata from a file, you can set `allowRotationMetadata` to `false`.
+
 ### Cropping video
 
 `crop` can be used to extract a rectangular region from the original video. The rectangle is specified using `left`, `top`, `width` and `height` and is clamped to the dimensions of the video. Cropping is applied after rotation but before resizing.
@@ -190,9 +194,10 @@ Use the `codec` property to control the codec of the output track. This should b
 Use the `bitrate` property to control the bitrate of the output video. For example, you can use this field to compress the video track. Accepted values are the number of bits per second or a [subjective quality](./media-sources#subjective-qualities). If this property is set, transcoding will always happen. If this property is not set but transcoding is still required, `QUALITY_HIGH` will be used as the value.
 
 Use the `keyFrameInterval` property to control the maximum interval in seconds between key frames in the output video. Setting this fields forces a transcode.
-
 If you want to prevent direct copying of media data and force a transcoding step, use `forceTranscode: true`.
 
+Use the `hardwareAcceleration` property to control whether hardware or software acceleration is used for video transcoding.
+
 ### Processing video
 
 The `process` property can be used to define a custom video sample processing function, e.g. for [applying overlays](./quick-start#add-a-video-overlay), color transformations, or timestamp modifications. You are expected to perform this processing yourself, for example using the Canvas API.
@@ -445,4 +450,4 @@ On the flip side, you can always query which input tracks made it into the outpu
 ```ts
 const conversion = await Conversion.init({ input, output });
 conversion.utilizedTracks; // => InputTrack[]
-```
+```

docs/guide/media-sources.md

@@ -187,6 +187,16 @@ videoTrackSource.errorPromise.catch((error) => ...);
 
 This source requires no additional method calls; data will automatically be captured and piped to the output file as soon as `start()` is called on the `Output`. Make sure to `stop()` on `videoTrack` after finalizing the `Output` if you don't need the user's media anymore.
 
+If you want to temporarily stop capturing video frames from this source, you can use the `pause()` and `resume()` methods:
+```ts
+videoTrackSource.pause();
+
+// Later:
+videoTrackSource.resume();
+```
+
+While paused, video frames emitted by the stream will be ignored. When resumed, video frames are let through again, offset in timestamp such that the result plays back continuously with no gap in playback. Note that pausing does *not* stop the underlying media stream.
+
 ::: info
 If this source is the only MediaStreamTrack source in the `Output`, then the first video sample added by it starts at timestamp 0. If there are multiple, then the earliest media sample across all tracks starts at timestamp 0, and all tracks will be perfectly synchronized with each other.
 :::
@@ -344,6 +354,16 @@ audioTrackSource.errorPromise.catch((error) => ...);
 
 This source requires no additional method calls; data will automatically be captured and piped to the output file as soon as `start()` is called on the `Output`. Make sure to `stop()` on `audioTrack` after finalizing the `Output` if you don't need the user's media anymore.
 
+If you want to temporarily stop capturing audio data from this source, you can use the `pause()` and `resume()` methods:
+```ts
+audioTrackSource.pause();
+
+// Later:
+audioTrackSource.resume();
+```
+
+While paused, audio data emitted by the stream will be ignored. When resumed, audio data are let through again, offset in timestamp such that the result plays back continuously with no gap in playback. Note that pausing does *not* stop the underlying media stream.
+
 ::: info
 If this source is the only MediaStreamTrack source in the `Output`, then the first audio sample added by it starts at timestamp 0. If there are multiple, then the earliest media sample across all tracks starts at timestamp 0, and all tracks will be perfectly synchronized with each other.
 :::

docs/guide/packets-and-samples.md

@@ -378,13 +378,11 @@ const bytesNeeded = videoSample.allocationSize(); // => number
 Then, use `copyTo` to copy the pixel data into the destination buffer:
 ```ts
 const bytes = new Uint8Array(bytesNeeded);
-videoSample.copyTo(bytes);
+const planeLayout = await videoSample.copyTo(bytes);
 ```
 
 ::: info
-The data will always be in the pixel format specified in the `format` field.
-
-To convert the data into a different pixel format, or to extract only a section of the frame, please use the [`allocationSize`](https://developer.mozilla.org/en-US/docs/Web/API/VideoFrame/allocationSize) and [`copyTo`](https://developer.mozilla.org/en-US/docs/Web/API/VideoFrame/copyTo) methods on `VideoFrame` instead. Get a `VideoFrame` by running `videoSample.toVideoFrame()`.
+You can pass additional options to `allocationSize` and `copyTo` to extract data in a different pixel format.
 :::
 
 ---

docs/index.md

@@ -91,7 +91,7 @@ const bundleSizes = [
 
 const sponsors = {
 	gold: [
-		{ image: '/sponsors/remotion.png', name: 'Remotion', url: 'https://remotion.dev/' },
+		{ image: '/sponsors/remotion-light.png', name: 'Remotion', url: 'https://remotion.dev/' },
 		{ image: '/sponsors/gling.svg', name: 'Gling AI', url: 'https://www.gling.ai/' },
 		{ image: '/sponsors/diffusionstudio.png', name: 'Diffusion Studio', url: 'https://diffusion.studio/' },
 		{ image: '/sponsors/kino.jpg', name: 'Kino', url: 'https://kino.ai/' },
@@ -106,6 +106,7 @@ const sponsors = {
 		{ image: '/sponsors/jellypod.png', name: 'Jellypod', url: 'https://jellypod.ai/' },
 	],
 	individual: [
+		{ image: 'https://avatars.githubusercontent.com/u/489051', name: 'Roman Rädle', url: 'https://github.com/raedle' },
 		{ image: 'https://avatars.githubusercontent.com/u/197597', name: 'Christopher Chedeau', url: 'https://github.com/vjeux' },
 		{ image: 'https://avatars.githubusercontent.com/u/84167135', name: 'Memenome', url: 'https://github.com/memenome' },
 		{ image: 'https://avatars.githubusercontent.com/u/5913254', name: 'Brandon McConnell', url: 'https://github.com/brandonmcconnell' },
@@ -114,6 +115,7 @@ const sponsors = {
 		{ image: 'https://avatars.githubusercontent.com/u/2698271', name: 'Matthew Gardner', url: 'https://github.com/spheric' },
 		{ image: 'https://avatars.githubusercontent.com/u/5475819', name: 'AJ Funk', url: 'https://github.com/AJFunk' },
 		{ image: 'https://avatars.githubusercontent.com/u/30229596', name: 'Pablo Bonilla', url: 'https://github.com/devPablo' },
+		{ image: 'https://avatars.githubusercontent.com/u/56988069', name: 'SyhabouthAlex', url: 'https://github.com/SyhabouthAlex' },
 		{ image: 'https://avatars.githubusercontent.com/u/38181164', name: 'wcw', url: 'https://github.com/asd55667' },
 		{ image: 'https://avatars.githubusercontent.com/u/1836701', name: 'Bean Deng', url: 'https://github.com/HADB' },
 		{ image: 'https://avatars.githubusercontent.com/u/63088713', name: 'taf2000', url: 'https://github.com/taf2000' },
@@ -123,6 +125,7 @@ const sponsors = {
 		{ image: 'https://avatars.githubusercontent.com/u/97225946', name: '808vita', url: 'https://github.com/808vita' },
 		{ image: 'https://avatars.githubusercontent.com/u/3709646', name: 'Rodrigo Belfiore', url: 'https://github.com/roprgm' },
 		{ image: 'https://avatars.githubusercontent.com/u/31102694', name: 'Aiden Liu', url: 'https://github.com/aidenlx' },
+		{ image: 'https://avatars.githubusercontent.com/u/41021374', name: 'arthco', url: 'https://github.com/arthtyagi' },
 	],
 };
 </script>

package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "mediabunny",
-  "version": "1.25.7",
+  "version": "1.27.2",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "mediabunny",
-      "version": "1.25.7",
+      "version": "1.27.2",
       "license": "MPL-2.0",
       "workspaces": [
         "packages/*"
@@ -7739,9 +7739,9 @@
       }
     },
     "node_modules/mediabunny": {
-      "version": "1.25.6",
-      "resolved": "https://registry.npmjs.org/mediabunny/-/mediabunny-1.25.6.tgz",
-      "integrity": "sha512-0CUgi9oHbyzPHZttn6aCQrWZLUpSAb7m1QKSSjr780LU0bgF2rgsYXAzPwt0V9ScK0t5CPuVhlmxF8MVe5BFHw==",
+      "version": "1.27.1",
+      "resolved": "https://registry.npmjs.org/mediabunny/-/mediabunny-1.27.1.tgz",
+      "integrity": "sha512-XdgXNox1hx2ynQy3jcCbzX1YsJt9en9VnZVidNXTNnQUdl6q2twSoV1GjgTb7Wlw9evCYXcKkgjE8KNU7K5trg==",
       "license": "MPL-2.0",
       "peer": true,
       "workspaces": [
@@ -12065,7 +12065,7 @@
     },
     "packages/mp3-encoder": {
       "name": "@mediabunny/mp3-encoder",
-      "version": "1.25.7",
+      "version": "1.27.2",
       "license": "MPL-2.0",
       "devDependencies": {
         "@types/emscripten": "^1.40.1"

package.json

@@ -1,7 +1,7 @@
 {
   "name": "mediabunny",
   "author": "Vanilagy",
-  "version": "1.25.7",
+  "version": "1.27.2",
   "description": "Pure TypeScript media toolkit for reading, writing, and converting media files, directly in the browser.",
   "type": "module",
   "workspaces": [

packages/mp3-encoder/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@mediabunny/mp3-encoder",
   "author": "Vanilagy",
-  "version": "1.25.7",
+  "version": "1.27.2",
   "description": "MP3 encoder extension for Mediabunny, based on LAME.",
   "main": "./dist/bundles/mediabunny-mp3-encoder.mjs",
   "module": "./dist/bundles/mediabunny-mp3-encoder.mjs",

src/codec-data.ts

@@ -22,6 +22,9 @@ import {
 	textEncoder,
 	toDataView,
 	toUint8Array,
+	getChromiumVersion,
+	isChromium,
+	setUint24,
 } from './misc';
 import { PacketType } from './packet';
 import { MetadataTags } from './metadata';
@@ -34,6 +37,7 @@ import { MetadataTags } from './metadata';
 
 export enum AvcNalUnitType {
 	IDR = 5,
+	SEI = 6,
 	SPS = 7,
 	PPS = 8,
 	SPS_EXT = 13,
@@ -158,38 +162,54 @@ const removeEmulationPreventionBytes = (data: Uint8Array) => {
 	return new Uint8Array(result);
 };
 
-/** Converts an AVC packet in Annex B format to length-prefixed format. */
-export const transformAnnexBToLengthPrefixed = (packetData: Uint8Array) => {
-	const NAL_UNIT_LENGTH_SIZE = 4;
+const ANNEX_B_START_CODE = new Uint8Array([0, 0, 0, 1]);
 
-	const nalUnits = findNalUnitsInAnnexB(packetData);
-
-	if (nalUnits.length === 0) {
-		// If no NAL units were found, it's not valid Annex B data
-		return null;
-	}
-
-	let totalSize = 0;
-	for (const nalUnit of nalUnits) {
-		totalSize += NAL_UNIT_LENGTH_SIZE + nalUnit.byteLength;
-	}
-
-	const avccData = new Uint8Array(totalSize);
-	const dataView = new DataView(avccData.buffer);
+export const concatNalUnitsInAnnexB = (nalUnits: Uint8Array[]) => {
+	const totalLength = nalUnits.reduce((a, b) => a + ANNEX_B_START_CODE.byteLength + b.byteLength, 0);
+	const result = new Uint8Array(totalLength);
 	let offset = 0;
 
-	// Write each NAL unit with its length prefix
 	for (const nalUnit of nalUnits) {
-		const length = nalUnit.byteLength;
+		result.set(ANNEX_B_START_CODE, offset);
+		offset += ANNEX_B_START_CODE.byteLength;
 
-		dataView.setUint32(offset, length, false);
-		offset += 4;
-
-		avccData.set(nalUnit, offset);
+		result.set(nalUnit, offset);
 		offset += nalUnit.byteLength;
 	}
 
-	return avccData;
+	return result;
+};
+
+export const concatNalUnitsInLengthPrefixed = (nalUnits: Uint8Array[], lengthSize: 1 | 2 | 3 | 4) => {
+	const totalLength = nalUnits.reduce((a, b) => a + lengthSize + b.byteLength, 0);
+	const result = new Uint8Array(totalLength);
+	let offset = 0;
+
+	for (const nalUnit of nalUnits) {
+		const dataView = new DataView(result.buffer, result.byteOffset, result.byteLength);
+
+		switch (lengthSize) {
+			case 1:
+				dataView.setUint8(offset, nalUnit.byteLength);
+				break;
+			case 2:
+				dataView.setUint16(offset, nalUnit.byteLength, false);
+				break;
+			case 3:
+				setUint24(dataView, offset, nalUnit.byteLength, false);
+				break;
+			case 4:
+				dataView.setUint32(offset, nalUnit.byteLength, false);
+				break;
+		}
+
+		offset += lengthSize;
+
+		result.set(nalUnit, offset);
+		offset += nalUnit.byteLength;
+	}
+
+	return result;
 };
 
 // Data specified in ISO 14496-15
@@ -224,7 +244,22 @@ export const extractAvcNalUnits = (packetData: Uint8Array, decoderConfig: VideoD
 	}
 };
 
-const extractNalUnitTypeForAvc = (data: Uint8Array) => {
+export const concatAvcNalUnits = (nalUnits: Uint8Array[], decoderConfig: VideoDecoderConfig) => {
+	if (decoderConfig.description) {
+		// Stream is length-prefixed. Let's extract the size of the length prefix from the decoder config
+
+		const bytes = toUint8Array(decoderConfig.description);
+		const lengthSizeMinusOne = bytes[4]! & 0b11;
+		const lengthSize = (lengthSizeMinusOne + 1) as 1 | 2 | 3 | 4;
+
+		return concatNalUnitsInLengthPrefixed(nalUnits, lengthSize);
+	} else {
+		// Stream is in Annex B format
+		return concatNalUnitsInAnnexB(nalUnits);
+	}
+};
+
+export const extractNalUnitTypeForAvc = (data: Uint8Array) => {
 	return data[0]! & 0x1F;
 };
 
@@ -1714,7 +1749,68 @@ export const determineVideoPacketType = (
 	switch (codec) {
 		case 'avc': {
 			const nalUnits = extractAvcNalUnits(packetData, decoderConfig);
-			const isKeyframe = nalUnits.some(x => extractNalUnitTypeForAvc(x) === AvcNalUnitType.IDR);
+			let isKeyframe = nalUnits.some(x => extractNalUnitTypeForAvc(x) === AvcNalUnitType.IDR);
+
+			if (!isKeyframe && (!isChromium() || getChromiumVersion()! >= 144)) {
+				// In addition to IDR, Recovery Point SEI also counts as a valid H.264 keyframe by current consensus.
+				// See https://github.com/w3c/webcodecs/issues/650 for the relevant discussion. WebKit and Firefox have
+				// always supported them, but Chromium hasn't, therefore the (admittedly dirty) version check.
+
+				for (const nalUnit of nalUnits) {
+					const type = extractNalUnitTypeForAvc(nalUnit);
+					if (type !== AvcNalUnitType.SEI) {
+						continue;
+					}
+
+					const bytes = removeEmulationPreventionBytes(nalUnit);
+					let pos = 1; // Skip NALU header
+
+					// sei_rbsp()
+					do {
+						// sei_message()
+						let payloadType = 0;
+						while (true) {
+							const nextByte = bytes[pos++];
+							if (nextByte === undefined) break;
+							payloadType += nextByte;
+
+							if (nextByte < 255) {
+								break;
+							}
+						}
+
+						let payloadSize = 0;
+						while (true) {
+							const nextByte = bytes[pos++];
+							if (nextByte === undefined) break;
+							payloadSize += nextByte;
+
+							if (nextByte < 255) {
+								break;
+							}
+						}
+
+						// sei_payload()
+						const PAYLOAD_TYPE_RECOVERY_POINT = 6;
+						if (payloadType === PAYLOAD_TYPE_RECOVERY_POINT) {
+							const bitstream = new Bitstream(bytes);
+							bitstream.pos = 8 * pos;
+
+							const recoveryFrameCount = readExpGolomb(bitstream);
+							const exactMatchFlag = bitstream.readBits(1);
+
+							if (recoveryFrameCount === 0 && exactMatchFlag === 1) {
+								// https://github.com/w3c/webcodecs/pull/910
+								// "recovery_frame_cnt == 0 and exact_match_flag=1 in the SEI recovery payload"
+								isKeyframe = true;
+								break;
+							}
+						}
+
+						pos += payloadSize;
+					} while (pos < bytes.length - 1);
+				}
+			}
 
 			return isKeyframe ? 'key' : 'delta';
 		};

src/conversion.ts

@@ -143,6 +143,12 @@ export type ConversionVideoOptions = {
 	 * This rotation is _in addition to_ the natural rotation of the input video as specified in input file's metadata.
 	 */
 	rotate?: Rotation;
+	/**
+	 * Defaults to `true`. When enabaled, Mediabunny will use the rotation metadata in the output file to perform video
+	 * rotation whenever possible. Set this field to `false` if you want to ensure the output file does not make use of
+	 * rotation metadata and that any rotation is baked into the video frames directly.
+	 */
+	allowRotationMetadata?: boolean;
 	/**
 	 * Specifies the rectangular region of the input video to crop to. The crop region will automatically be clamped to
 	 * the dimensions of the input video track. Cropping is performed after rotation but before resizing.
@@ -180,6 +186,11 @@ export type ConversionVideoOptions = {
 	 * Setting this fields forces a transcode.
 	 */
 	keyFrameInterval?: number;
+	/**
+	 * A hint that configures the hardware acceleration method used when transcoding. This is best left on
+	 * `'no-preference'`, the default.
+	 */
+	hardwareAcceleration?: 'no-preference' | 'prefer-hardware' | 'prefer-software';
 	/** When `true`, video will always be re-encoded instead of directly copying over the encoded samples. */
 	forceTranscode?: boolean;
 	/**
@@ -304,6 +315,9 @@ const validateVideoOptions = (videoOptions: ConversionVideoOptions | undefined)
 	if (videoOptions?.rotate !== undefined && ![0, 90, 180, 270].includes(videoOptions.rotate)) {
 		throw new TypeError('options.video.rotate, when provided, must be 0, 90, 180 or 270.');
 	}
+	if (videoOptions?.allowRotationMetadata !== undefined && typeof videoOptions.allowRotationMetadata !== 'boolean') {
+		throw new TypeError('options.video.allowRotationMetadata, when provided, must be a boolean.');
+	}
 	if (videoOptions?.crop !== undefined) {
 		validateCropRectangle(videoOptions.crop, 'options.video.');
 	}
@@ -337,6 +351,15 @@ const validateVideoOptions = (videoOptions: ConversionVideoOptions | undefined)
 	) {
 		throw new TypeError('options.video.processedHeight, when provided, must be a positive integer.');
 	}
+	if (
+		videoOptions?.hardwareAcceleration !== undefined
+		&& !['no-preference', 'prefer-hardware', 'prefer-software'].includes(videoOptions.hardwareAcceleration)
+	) {
+		throw new TypeError(
+			'options.video.hardwareAcceleration, when provided, must be \'no-preference\', \'prefer-hardware\' or'
+			+ ' \'prefer-software\'.',
+		);
+	}
 };
 
 const validateAudioOptions = (audioOptions: ConversionAudioOptions | undefined) => {
@@ -838,7 +861,8 @@ export class Conversion {
 		let videoSource: VideoSource;
 
 		const totalRotation = normalizeRotation(track.rotation + (trackOptions.rotate ?? 0));
-		const outputSupportsRotation = this.output.format.supportsVideoRotationMetadata;
+		const canUseRotationMetadata = this.output.format.supportsVideoRotationMetadata
+			&& (trackOptions.allowRotationMetadata ?? true);
 
 		const [rotatedWidth, rotatedHeight] = totalRotation % 180 === 0
 			? [track.codedWidth, track.codedHeight]
@@ -883,7 +907,7 @@ export class Conversion {
 			// TODO This is suboptimal: Forcing a rerender when both rotation and process are set is not
 			// performance-optimal, but right now there's no other way because we can't change the track rotation
 			// metadata after the output has already started. Should be possible with API changes in v2, though!
-			|| (totalRotation !== 0 && (!outputSupportsRotation || trackOptions.process !== undefined))
+			|| (totalRotation !== 0 && (!canUseRotationMetadata || trackOptions.process !== undefined))
 			|| !!crop;
 
 		const alpha = trackOptions.alpha ?? 'discard';
@@ -974,6 +998,7 @@ export class Conversion {
 				keyFrameInterval: trackOptions.keyFrameInterval,
 				sizeChangeBehavior: trackOptions.fit ?? 'passThrough',
 				alpha,
+				hardwareAcceleration: trackOptions.hardwareAcceleration,
 			};
 
 			const source = new VideoSampleSource(encodingConfig);

src/index.ts

@@ -168,7 +168,10 @@ export {
 	AudioSampleCopyToOptions,
 	VideoSample,
 	VideoSampleInit,
+	VideoSamplePixelFormat,
+	VideoSampleColorSpace,
 	CropRectangle,
+	VIDEO_SAMPLE_PIXEL_FORMATS,
 } from './sample';
 export {
 	AudioBufferSink,

src/isobmff/isobmff-muxer.ts

@@ -25,11 +25,12 @@ import {
 import { BufferTarget } from '../target';
 import { EncodedPacket, PacketType } from '../packet';
 import {
+	concatNalUnitsInLengthPrefixed,
 	extractAvcDecoderConfigurationRecord,
 	extractHevcDecoderConfigurationRecord,
+	findNalUnitsInAnnexB,
 	serializeAvcDecoderConfigurationRecord,
 	serializeHevcDecoderConfigurationRecord,
-	transformAnnexBToLengthPrefixed,
 } from '../codec-data';
 import { buildIsobmffMimeType } from './isobmff-misc';
 import { MAX_BOX_HEADER_SIZE, MIN_BOX_HEADER_SIZE } from './isobmff-reader';
@@ -464,15 +465,18 @@ export class IsobmffMuxer extends Muxer {
 
 			let packetData = packet.data;
 			if (trackData.info.requiresAnnexBTransformation) {
-				const transformedData = transformAnnexBToLengthPrefixed(packetData);
-				if (!transformedData) {
+				const nalUnits = findNalUnitsInAnnexB(packetData);
+				if (nalUnits.length === 0) {
+					// It's not valid Annex B data
 					throw new Error(
 						'Failed to transform packet data. Make sure all packets are provided in Annex B format, as'
 						+ ' specified in ITU-T-REC-H.264 and ITU-T-REC-H.265.',
 					);
 				}
 
-				packetData = transformedData;
+				// We don't strip things like SPS or PPS NALUs here, mainly because they can also appear in the middle
+				// of a stream and potentially modify the parameters of it. So, let's just leave them in to be sure.
+				packetData = concatNalUnitsInLengthPrefixed(nalUnits, 4);
 			}
 
 			const timestamp = this.validateAndNormalizeTimestamp(

src/media-sink.ts

@@ -8,9 +8,12 @@
 
 import { parsePcmCodec, PCM_AUDIO_CODECS, PcmAudioCodec, VideoCodec, AudioCodec } from './codec';
 import {
+	concatAvcNalUnits,
 	deserializeAvcDecoderConfigurationRecord,
 	determineVideoPacketType,
+	extractAvcNalUnits,
 	extractHevcNalUnits,
+	extractNalUnitTypeForAvc,
 	extractNalUnitTypeForHevc,
 	HevcNalUnitType,
 	parseAvcSps,
@@ -429,7 +432,6 @@ export abstract class BaseMediaSampleSink<
 
 		// The following is the "pump" process that keeps pumping packets into the decoder
 		(async () => {
-			const decoderError = new Error();
 			const decoder = await this._createDecoder((sample) => {
 				onQueueDequeue();
 				if (sample.timestamp >= endTimestamp) {
@@ -467,7 +469,6 @@ export abstract class BaseMediaSampleSink<
 				}
 			}, (error) => {
 				if (!outOfBandError) {
-					error.stack = decoderError.stack; // Provide a more useful stack trace
 					outOfBandError = error;
 					onQueueNotEmpty();
 				}
@@ -613,7 +614,6 @@ export abstract class BaseMediaSampleSink<
 
 		// The following is the "pump" process that keeps pumping packets into the decoder
 		(async () => {
-			const decoderError = new Error();
 			const decoder = await this._createDecoder((sample) => {
 				onQueueDequeue();
 
@@ -641,7 +641,6 @@ export abstract class BaseMediaSampleSink<
 				}
 			}, (error) => {
 				if (!outOfBandError) {
-					error.stack = decoderError.stack; // Provide a more useful stack trace
 					outOfBandError = error;
 					onQueueNotEmpty();
 				}
@@ -891,6 +890,8 @@ class VideoDecoderWrapper extends DecoderWrapper<VideoSample> {
 				}
 			}
 
+			const stack = new Error('Decoding error').stack;
+
 			this.decoder = new VideoDecoder({
 				output: (frame) => {
 					try {
@@ -899,7 +900,10 @@ class VideoDecoderWrapper extends DecoderWrapper<VideoSample> {
 						this.onError(error as Error);
 					}
 				},
-				error: onError,
+				error: (error) => {
+					error.stack = stack; // Provide a more useful stack trace, the default one sucks
+					this.onError(error);
+				},
 			});
 			this.decoder.configure(this.decoderConfig);
 		}
@@ -927,8 +931,6 @@ class VideoDecoderWrapper extends DecoderWrapper<VideoSample> {
 			this.raslSkipped = true;
 		}
 
-		this.currentPacketIndex++;
-
 		if (this.customDecoder) {
 			this.customDecoderQueueSize++;
 			void this.customDecoderCallSerializer
@@ -941,9 +943,24 @@ class VideoDecoderWrapper extends DecoderWrapper<VideoSample> {
 				insertSorted(this.inputTimestamps, packet.timestamp, x => x);
 			}
 
+			// Workaround for https://issues.chromium.org/issues/470109459
+			if (isChromium() && this.currentPacketIndex === 0 && this.codec === 'avc') {
+				const nalUnits = extractAvcNalUnits(packet.data, this.decoderConfig);
+				const filteredNalUnits = nalUnits.filter((x) => {
+					const type = extractNalUnitTypeForAvc(x);
+					// These trip up Chromium's key frame detection, so let's strip them
+					return !(type >= 20 && type <= 31);
+				});
+
+				const newData = concatAvcNalUnits(filteredNalUnits, this.decoderConfig);
+				packet = new EncodedPacket(newData, packet.type, packet.timestamp, packet.duration);
+			}
+
 			this.decoder.decode(packet.toEncodedVideoChunk());
 			this.decodeAlphaData(packet);
 		}
+
+		this.currentPacketIndex++;
 	}
 
 	decodeAlphaData(packet: EncodedPacket) {
@@ -999,6 +1016,8 @@ class VideoDecoderWrapper extends DecoderWrapper<VideoSample> {
 				}
 			};
 
+			const stack = new Error('Decoding error').stack;
+
 			this.alphaDecoder = new VideoDecoder({
 				output: (frame) => {
 					try {
@@ -1007,7 +1026,10 @@ class VideoDecoderWrapper extends DecoderWrapper<VideoSample> {
 						this.onError(error as Error);
 					}
 				},
-				error: this.onError,
+				error: (error) => {
+					error.stack = stack; // Provide a more useful stack trace, the default one sucks
+					this.onError(error);
+				},
 			});
 			this.alphaDecoder.configure(this.decoderConfig);
 		}
@@ -1763,6 +1785,8 @@ class AudioDecoderWrapper extends DecoderWrapper<AudioSample> {
 
 			void this.customDecoderCallSerializer.call(() => this.customDecoder!.init());
 		} else {
+			const stack = new Error('Decoding error').stack;
+
 			this.decoder = new AudioDecoder({
 				output: (data) => {
 					try {
@@ -1771,7 +1795,10 @@ class AudioDecoderWrapper extends DecoderWrapper<AudioSample> {
 						this.onError(error as Error);
 					}
 				},
-				error: onError,
+				error: (error) => {
+					error.stack = stack; // Provide a more useful stack trace, the default one sucks
+					this.onError(error);
+				},
 			});
 			this.decoder.configure(decoderConfig);
 		}

src/media-source.ts

@@ -1055,6 +1055,12 @@ export class MediaStreamVideoTrackSource extends VideoSource {
 	private _promiseWithResolvers = promiseWithResolvers();
 	/** @internal */
 	private _errorPromiseAccessed = false;
+	/** @internal */
+	private _paused = false;
+	/** @internal */
+	private _lastSampleTimestamp: number | null = null;
+	/** @internal */
+	private _pauseOffset = 0;
 
 	/** A promise that rejects upon any error within this source. This promise never resolves. */
 	get errorPromise() {
@@ -1062,6 +1068,11 @@ export class MediaStreamVideoTrackSource extends VideoSource {
 		return this._promiseWithResolvers.promise;
 	}
 
+	/** Whether this source is currently paused as a result of calling `.pause()`. */
+	get paused() {
+		return this._paused;
+	}
+
 	/**
 	 * Creates a new {@link MediaStreamVideoTrackSource} from a
 	 * [`MediaStreamVideoTrack`](https://developer.mozilla.org/en-US/docs/Web/API/MediaStreamTrack), which will pull
@@ -1103,8 +1114,29 @@ export class MediaStreamVideoTrackSource extends VideoSource {
 				return;
 			}
 
+			const currentTimestamp = videoFrame.timestamp / 1e6;
+
+			if (this._paused) {
+				const frameSeen = firstVideoFrameTimestamp !== null;
+				if (frameSeen) {
+					if (this._lastSampleTimestamp !== null) {
+						// In addition to dropping this frame, let's also keep track of the time we have lost due to the
+						// pause. Doing it like this instead of simply keeping track of the paused time is better since
+						// it retains the frame rate of the underlying source.
+						const timeDelta = currentTimestamp - this._lastSampleTimestamp;
+						// We modify this field instead of _timestampOffset since we still might have data in flight
+						// in the encoder, with which we don't want to mess.
+						this._pauseOffset -= timeDelta;
+					}
+					this._lastSampleTimestamp = currentTimestamp;
+				}
+
+				videoFrame.close();
+				return;
+			}
+
 			if (firstVideoFrameTimestamp === null) {
-				firstVideoFrameTimestamp = videoFrame.timestamp / 1e6;
+				firstVideoFrameTimestamp = currentTimestamp;
 
 				const muxer = this._connectedTrack!.output._muxer;
 				if (muxer.firstMediaStreamTimestamp === null) {
@@ -1116,13 +1148,19 @@ export class MediaStreamVideoTrackSource extends VideoSource {
 				}
 			}
 
+			this._lastSampleTimestamp = currentTimestamp;
+
 			if (this._encoder.getQueueSize() >= 4) {
 				// Drop frames if the encoder is overloaded
 				videoFrame.close();
 				return;
 			}
 
-			void this._encoder.add(new VideoSample(videoFrame), true)
+			const sample = new VideoSample(videoFrame, {
+				timestamp: currentTimestamp + this._pauseOffset,
+			});
+
+			void this._encoder.add(sample, true)
 				.catch((error) => {
 					errored = true;
 
@@ -1182,6 +1220,19 @@ export class MediaStreamVideoTrackSource extends VideoSource {
 		}
 	}
 
+	/**
+	 * Pauses the capture of video frames - any video frames emitted by the underlying media stream will be ignored
+	 * while paused. This does *not* close the underlying `MediaStreamVideoTrack`, it just ignores its output.
+	 */
+	pause() {
+		this._paused = true;
+	}
+
+	/** Resumes the capture of video frames after being paused. */
+	resume() {
+		this._paused = false;
+	}
+
 	/** @internal */
 	override async _flushAndClose(forceClose: boolean) {
 		if (this._abortController) {
@@ -1857,6 +1908,12 @@ export class MediaStreamAudioTrackSource extends AudioSource {
 	private _promiseWithResolvers = promiseWithResolvers();
 	/** @internal */
 	private _errorPromiseAccessed = false;
+	/** @internal */
+	private _paused = false;
+	/** @internal */
+	private _lastSampleTimestamp: number | null = null;
+	/** @internal */
+	private _pauseOffset = 0;
 
 	/** A promise that rejects upon any error within this source. This promise never resolves. */
 	get errorPromise() {
@@ -1864,6 +1921,11 @@ export class MediaStreamAudioTrackSource extends AudioSource {
 		return this._promiseWithResolvers.promise;
 	}
 
+	/** Whether this source is currently paused as a result of calling `.pause()`. */
+	get paused() {
+		return this._paused;
+	}
+
 	/**
 	 * Creates a new {@link MediaStreamAudioTrackSource} from a `MediaStreamAudioTrack`, which will pull audio samples
 	 * from the stream in real time and encode them according to {@link AudioEncodingConfig}.
@@ -1890,38 +1952,74 @@ export class MediaStreamAudioTrackSource extends AudioSource {
 
 		this._abortController = new AbortController();
 
+		let firstAudioDataTimestamp: number | null = null;
+		let errored = false;
+
+		const onAudioSample = (audioSample: AudioSample) => {
+			if (errored) {
+				audioSample.close();
+				return;
+			}
+
+			const currentTimestamp = audioSample.timestamp;
+
+			if (this._paused) {
+				const dataSeen = firstAudioDataTimestamp !== null;
+				if (dataSeen) {
+					if (this._lastSampleTimestamp !== null) {
+						// In addition to dropping this sample, let's also keep track of the time we have lost due to
+						// the pause. Doing it like this instead of simply keeping track of the paused time is better
+						// since it retains the sample rate of the underlying source.
+						const timeDelta = currentTimestamp - this._lastSampleTimestamp;
+						// We modify this field instead of _timestampOffset since we still might have data in flight
+						// in the encoder, with which we don't want to mess.
+						this._pauseOffset -= timeDelta;
+					}
+					this._lastSampleTimestamp = currentTimestamp;
+				}
+
+				audioSample.close();
+				return;
+			}
+
+			if (firstAudioDataTimestamp === null) {
+				firstAudioDataTimestamp = audioSample.timestamp;
+
+				const muxer = this._connectedTrack!.output._muxer;
+				if (muxer.firstMediaStreamTimestamp === null) {
+					muxer.firstMediaStreamTimestamp = performance.now() / 1000;
+					this._timestampOffset = -firstAudioDataTimestamp;
+				} else {
+					this._timestampOffset = (performance.now() / 1000 - muxer.firstMediaStreamTimestamp)
+						- firstAudioDataTimestamp;
+				}
+			}
+
+			this._lastSampleTimestamp = currentTimestamp;
+
+			if (this._encoder.getQueueSize() >= 4) {
+				// Drop data if the encoder is overloaded
+				audioSample.close();
+				return;
+			}
+
+			audioSample.setTimestamp(currentTimestamp + this._pauseOffset);
+
+			void this._encoder.add(audioSample, true)
+				.catch((error) => {
+					errored = true;
+
+					this._abortController?.abort();
+					this._promiseWithResolvers.reject(error);
+					void this._audioContext?.suspend();
+				});
+		};
+
 		if (typeof MediaStreamTrackProcessor !== 'undefined') {
 			// Great, MediaStreamTrackProcessor is supported, this is the preferred way of doing things
-			let firstAudioDataTimestamp: number | null = null;
-
 			const processor = new MediaStreamTrackProcessor({ track: this._track });
 			const consumer = new WritableStream<AudioData>({
-				write: (audioData) => {
-					if (firstAudioDataTimestamp === null) {
-						firstAudioDataTimestamp = audioData.timestamp / 1e6;
-
-						const muxer = this._connectedTrack!.output._muxer;
-						if (muxer.firstMediaStreamTimestamp === null) {
-							muxer.firstMediaStreamTimestamp = performance.now() / 1000;
-							this._timestampOffset = -firstAudioDataTimestamp;
-						} else {
-							this._timestampOffset = (performance.now() / 1000 - muxer.firstMediaStreamTimestamp)
-								- firstAudioDataTimestamp;
-						}
-					}
-
-					if (this._encoder.getQueueSize() >= 4) {
-						// Drop data if the encoder is overloaded
-						audioData.close();
-						return;
-					}
-
-					void this._encoder.add(new AudioSample(audioData), true)
-						.catch((error) => {
-							this._abortController?.abort();
-							this._promiseWithResolvers.reject(error);
-						});
-				},
+				write: audioData => onAudioSample(new AudioSample(audioData)),
 			});
 
 			processor.readable.pipeTo(consumer, {
@@ -1949,7 +2047,6 @@ export class MediaStreamAudioTrackSource extends AudioSource {
 			sourceNode.connect(this._scriptProcessorNode);
 			this._scriptProcessorNode.connect(this._audioContext.destination);
 
-			let audioReceived = false;
 			let totalDuration = 0;
 
 			this._scriptProcessorNode.onaudioprocess = (event) => {
@@ -1957,33 +2054,25 @@ export class MediaStreamAudioTrackSource extends AudioSource {
 				totalDuration += event.inputBuffer.duration;
 
 				for (const audioSample of iterator) {
-					if (!audioReceived) {
-						audioReceived = true;
-
-						const muxer = this._connectedTrack!.output._muxer;
-						if (muxer.firstMediaStreamTimestamp === null) {
-							muxer.firstMediaStreamTimestamp = performance.now() / 1000;
-						} else {
-							this._timestampOffset = performance.now() / 1000 - muxer.firstMediaStreamTimestamp;
-						}
-					}
-
-					if (this._encoder.getQueueSize() >= 4) {
-						// Drop data if the encoder is overloaded
-						audioSample.close();
-						continue;
-					}
-
-					void this._encoder.add(audioSample, true)
-						.catch((error) => {
-							void this._audioContext!.suspend();
-							this._promiseWithResolvers.reject(error);
-						});
+					onAudioSample(audioSample);
 				}
 			};
 		}
 	}
 
+	/**
+	 * Pauses the capture of audio data - any audio data emitted by the underlying media stream will be ignored
+	 * while paused. This does *not* close the underlying `MediaStreamAudioTrack`, it just ignores its output.
+	 */
+	pause() {
+		this._paused = true;
+	}
+
+	/** Resumes the capture of audio data after being paused. */
+	resume() {
+		this._paused = false;
+	}
+
 	/** @internal */
 	override async _flushAndClose(forceClose: boolean) {
 		if (this._abortController) {

src/misc.ts

@@ -679,7 +679,15 @@ export const isWebKit = () => {
 	}
 
 	// This even returns true for WebKit-wrapping browsers such as Chrome on iOS
-	return isWebKitCache = !!(typeof navigator !== 'undefined' && navigator.vendor?.match(/apple/i));
+	return isWebKitCache = !!(
+		typeof navigator !== 'undefined'
+		&& (
+			navigator.vendor?.match(/apple/i)
+			// Or, in workers:
+			|| (/AppleWebKit/.test(navigator.userAgent) && !/Chrome/.test(navigator.userAgent))
+			|| /\b(iPad|iPhone|iPod)\b/.test(navigator.userAgent)
+		)
+	);
 };
 
 let isFirefoxCache: boolean | null = null;
@@ -697,7 +705,28 @@ export const isChromium = () => {
 		return isChromiumCache;
 	}
 
-	return isChromiumCache = !!(typeof navigator !== 'undefined' && navigator.vendor?.includes('Google Inc'));
+	return isChromiumCache = !!(
+		typeof navigator !== 'undefined'
+		&& (navigator.vendor?.includes('Google Inc') || /Chrome/.test(navigator.userAgent))
+	);
+};
+
+let chromiumVersionCache: number | null = null;
+export const getChromiumVersion = () => {
+	if (chromiumVersionCache !== null) {
+		return chromiumVersionCache;
+	}
+
+	if (typeof navigator === 'undefined') {
+		return null;
+	}
+
+	const match = /\bChrome\/(\d+)/.exec(navigator.userAgent);
+	if (!match) {
+		return null;
+	}
+
+	return chromiumVersionCache = Number(match[1]!);
 };
 
 /**

src/sample.ts

@@ -17,6 +17,7 @@ import {
 	SetRequired,
 	isFirefox,
 	polyfillSymbolDispose,
+	assertNever,
 } from './misc';
 
 polyfillSymbolDispose();
@@ -69,6 +70,57 @@ if (typeof FinalizationRegistry !== 'undefined') {
 	});
 }
 
+/**
+ * The list of {@link VideoSample} pixel formats.
+ * @group Samples
+ * @public
+ */
+export const VIDEO_SAMPLE_PIXEL_FORMATS = [
+	// 4:2:0 Y, U, V
+	'I420',
+	'I420P10',
+	'I420P12',
+	// 4:2:0 Y, U, V, A
+	'I420A',
+	'I420AP10',
+	'I420AP12',
+	// 4:2:2 Y, U, V
+	'I422',
+	'I422P10',
+	'I422P12',
+	// 4:2:2 Y, U, V, A
+	'I422A',
+	'I422AP10',
+	'I422AP12',
+	// 4:4:4 Y, U, V
+	'I444',
+	'I444P10',
+	'I444P12',
+	// 4:4:4 Y, U, V, A
+	'I444A',
+	'I444AP10',
+	'I444AP12',
+	// 4:2:0 Y, UV
+	'NV12',
+	// 4:4:4 RGBA
+	'RGBA',
+	// 4:4:4 RGBX (opaque)
+	'RGBX',
+	// 4:4:4 BGRA
+	'BGRA',
+	// 4:4:4 BGRX (opaque)
+	'BGRX',
+] as const;
+const VIDEO_SAMPLE_PIXEL_FORMATS_SET = new Set(VIDEO_SAMPLE_PIXEL_FORMATS);
+
+/**
+ * The internal pixel format with which a {@link VideoSample} is stored.
+ * [See pixel formats](https://www.w3.org/TR/webcodecs/#pixel-format) for more.
+ * @group Samples
+ * @public
+ */
+export type VideoSamplePixelFormat = typeof VIDEO_SAMPLE_PIXEL_FORMATS[number];
+
 /**
  * Metadata used for VideoSample initialization.
  * @group Samples
@@ -77,9 +129,9 @@ if (typeof FinalizationRegistry !== 'undefined') {
 export type VideoSampleInit = {
 	/**
 	 * The internal pixel format in which the frame is stored.
-	 * [See pixel formats](https://developer.mozilla.org/en-US/docs/Web/API/VideoFrame/format)
+	 * [See pixel formats](https://www.w3.org/TR/webcodecs/#pixel-format)
 	 */
-	format?: VideoPixelFormat;
+	format?: VideoSamplePixelFormat;
 	/** The width of the frame in pixels. */
 	codedWidth?: number;
 	/** The height of the frame in pixels. */
@@ -92,6 +144,8 @@ export type VideoSampleInit = {
 	duration?: number;
 	/** The color space of the frame. */
 	colorSpace?: VideoColorSpaceInit;
+	/** The byte layout of the planes of the frame. */
+	layout?: PlaneLayout[];
 };
 
 /**
@@ -103,14 +157,20 @@ export type VideoSampleInit = {
 export class VideoSample implements Disposable {
 	/** @internal */
 	_data!: VideoFrame | OffscreenCanvas | Uint8Array | null;
+	/**
+	 * Used for the ArrayBuffer-backed case.
+	 * @internal
+	 */
+	_layout!: PlaneLayout[] | null;
 	/** @internal */
 	_closed: boolean = false;
 
 	/**
-	 * The internal pixel format in which the frame is stored.
-	 * [See pixel formats](https://developer.mozilla.org/en-US/docs/Web/API/VideoFrame/format)
+	 * The internal pixel format in which the frame is stored. Will be `null` if it's using an arbitrary internal
+	 * format not representable by `VideoPixelFormat`.
+	 * [See pixel formats](https://www.w3.org/TR/webcodecs/#pixel-format)
 	 */
-	readonly format!: VideoPixelFormat | null;
+	readonly format!: VideoSamplePixelFormat | null;
 	/** The width of the frame in pixels. */
 	readonly codedWidth!: number;
 	/** The height of the frame in pixels. */
@@ -125,7 +185,7 @@ export class VideoSample implements Disposable {
 	/** The duration of the frame in seconds. */
 	readonly duration!: number;
 	/** The color space of the frame. */
-	readonly colorSpace!: VideoColorSpace;
+	readonly colorSpace!: VideoSampleColorSpace;
 
 	/** The width of the frame in pixels after rotation. */
 	get displayWidth() {
@@ -189,8 +249,8 @@ export class VideoSample implements Disposable {
 			if (!init || typeof init !== 'object') {
 				throw new TypeError('init must be an object.');
 			}
-			if (!('format' in init) || typeof init.format !== 'string') {
-				throw new TypeError('init.format must be a string.');
+			if (init.format === undefined || !VIDEO_SAMPLE_PIXEL_FORMATS_SET.has(init.format)) {
+				throw new TypeError('init.format must be one of: ' + VIDEO_SAMPLE_PIXEL_FORMATS.join(', '));
 			}
 			if (!Number.isInteger(init.codedWidth) || init.codedWidth! <= 0) {
 				throw new TypeError('init.codedWidth must be a positive integer.');
@@ -209,6 +269,7 @@ export class VideoSample implements Disposable {
 			}
 
 			this._data = toUint8Array(data).slice(); // Copy it
+			this._layout = init.layout ?? createDefaultPlaneLayout(init.format, init.codedWidth!, init.codedHeight!);
 
 			this.format = init.format;
 			this.codedWidth = init.codedWidth!;
@@ -216,7 +277,7 @@ export class VideoSample implements Disposable {
 			this.rotation = init.rotation ?? 0;
 			this.timestamp = init.timestamp!;
 			this.duration = init.duration ?? 0;
-			this.colorSpace = new VideoColorSpace(init.colorSpace);
+			this.colorSpace = new VideoSampleColorSpace(init.colorSpace);
 		} else if (typeof VideoFrame !== 'undefined' && data instanceof VideoFrame) {
 			if (init?.rotation !== undefined && ![0, 90, 180, 270].includes(init.rotation)) {
 				throw new TypeError('init.rotation, when provided, must be 0, 90, 180, or 270.');
@@ -229,6 +290,7 @@ export class VideoSample implements Disposable {
 			}
 
 			this._data = data;
+			this._layout = null;
 
 			this.format = data.format;
 			// Copying the display dimensions here, assuming no innate VideoFrame rotation
@@ -239,7 +301,7 @@ export class VideoSample implements Disposable {
 			this.rotation = init?.rotation ?? 0;
 			this.timestamp = init?.timestamp ?? data.timestamp / 1e6;
 			this.duration = init?.duration ?? (data.duration ?? 0) / 1e6;
-			this.colorSpace = data.colorSpace;
+			this.colorSpace = new VideoSampleColorSpace(data.colorSpace);
 		} else if (
 			(typeof HTMLImageElement !== 'undefined' && data instanceof HTMLImageElement)
 			|| (typeof SVGImageElement !== 'undefined' && data instanceof SVGImageElement)
@@ -301,6 +363,7 @@ export class VideoSample implements Disposable {
 			// Draw it to a canvas
 			context.drawImage(data, 0, 0);
 			this._data = canvas;
+			this._layout = null;
 
 			this.format = 'RGBX';
 			this.codedWidth = width;
@@ -308,7 +371,7 @@ export class VideoSample implements Disposable {
 			this.rotation = init.rotation ?? 0;
 			this.timestamp = init.timestamp!;
 			this.duration = init.duration ?? 0;
-			this.colorSpace = new VideoColorSpace({
+			this.colorSpace = new VideoSampleColorSpace({
 				matrix: 'rgb',
 				primaries: 'bt709',
 				transfer: 'iec61966-2-1',
@@ -336,8 +399,11 @@ export class VideoSample implements Disposable {
 				rotation: this.rotation,
 			});
 		} else if (this._data instanceof Uint8Array) {
-			return new VideoSample(this._data.slice(), {
+			assert(this._layout);
+
+			return new VideoSample(this._data, {
 				format: this.format!,
+				layout: this._layout,
 				codedWidth: this.codedWidth,
 				codedHeight: this.codedHeight,
 				timestamp: this.timestamp,
@@ -378,16 +444,43 @@ export class VideoSample implements Disposable {
 		this._closed = true;
 	}
 
-	/** Returns the number of bytes required to hold this video sample's pixel data. */
-	allocationSize() {
+	/**
+	 * Returns the number of bytes required to hold this video sample's pixel data. Throws if `format` is `null`;
+	 * specify an explicit RGB format in the options in this case.
+	 */
+	allocationSize(options: VideoFrameCopyToOptions = {}): number {
+		validateVideoFrameCopyToOptions(options);
+
 		if (this._closed) {
 			throw new Error('VideoSample is closed.');
 		}
+		if ((options.format ?? this.format) === null) {
+			throw new Error(
+				'Cannot get allocation size when format is null. Please manually provide an RGB pixel format in the'
+				+ ' options instead.',
+			);
+		}
 
 		assert(this._data !== null);
 
+		if (!isVideoFrame(this._data)) {
+			if (
+				options.colorSpace
+				|| (options.format && options.format !== this.format)
+				|| options.layout
+				|| options.rect
+			) {
+				// Temporarily convert to VideoFrame to get it done
+				const videoFrame = this.toVideoFrame();
+				const size = videoFrame.allocationSize(options);
+				videoFrame.close();
+
+				return size;
+			}
+		}
+
 		if (isVideoFrame(this._data)) {
-			return this._data.allocationSize();
+			return this._data.allocationSize(options);
 		} else if (this._data instanceof Uint8Array) {
 			return this._data.byteLength;
 		} else {
@@ -395,23 +488,54 @@ export class VideoSample implements Disposable {
 		}
 	}
 
-	/** Copies this video sample's pixel data to an ArrayBuffer or ArrayBufferView. */
-	async copyTo(destination: AllowSharedBufferSource) {
+	/**
+	 * Copies this video sample's pixel data to an ArrayBuffer or ArrayBufferView. Throws if `format` is `null`;
+	 * specify an explicit RGB format in the options in this case.
+	 * @returns The byte layout of the planes of the copied data.
+	 */
+	async copyTo(destination: AllowSharedBufferSource, options: VideoFrameCopyToOptions = {}): Promise<PlaneLayout[]> {
 		if (!isAllowSharedBufferSource(destination)) {
 			throw new TypeError('destination must be an ArrayBuffer or an ArrayBuffer view.');
 		}
+		validateVideoFrameCopyToOptions(options);
 
 		if (this._closed) {
 			throw new Error('VideoSample is closed.');
 		}
+		if ((options.format ?? this.format) === null) {
+			throw new Error(
+				'Cannot copy video sample data when format is null. Please manually provide an RGB pixel format in the'
+				+ ' options instead.',
+			);
+		}
 
 		assert(this._data !== null);
 
+		if (!isVideoFrame(this._data)) {
+			if (
+				options.colorSpace
+				|| (options.format && options.format !== this.format)
+				|| options.layout
+				|| options.rect
+			) {
+				// Temporarily convert to VideoFrame to get it done
+				const videoFrame = this.toVideoFrame();
+				const layout = await videoFrame.copyTo(destination, options);
+				videoFrame.close();
+
+				return layout;
+			}
+		}
+
 		if (isVideoFrame(this._data)) {
-			await this._data.copyTo(destination);
+			return this._data.copyTo(destination, options);
 		} else if (this._data instanceof Uint8Array) {
+			assert(this._layout);
+
 			const dest = toUint8Array(destination);
 			dest.set(this._data);
+
+			return this._layout;
 		} else {
 			const canvas = this._data;
 			const context = canvas.getContext('2d');
@@ -420,6 +544,11 @@ export class VideoSample implements Disposable {
 			const imageData = context.getImageData(0, 0, this.codedWidth, this.codedHeight);
 			const dest = toUint8Array(destination);
 			dest.set(imageData.data);
+
+			return [{
+				offset: 0,
+				stride: 4 * this.codedWidth,
+			}];
 		}
 	}
 
@@ -441,7 +570,7 @@ export class VideoSample implements Disposable {
 			});
 		} else if (this._data instanceof Uint8Array) {
 			return new VideoFrame(this._data, {
-				format: this.format!,
+				format: this.format! as VideoPixelFormat,
 				codedWidth: this.codedWidth,
 				codedHeight: this.codedHeight,
 				timestamp: this.microsecondTimestamp,
@@ -805,6 +934,40 @@ export class VideoSample implements Disposable {
 	}
 }
 
+/**
+ * Describes the color space of a {@link VideoSample}. Corresponds to the WebCodecs API's VideoColorSpace.
+ * @group Samples
+ * @public
+ */
+export class VideoSampleColorSpace {
+	/** The color primaries standard used. */
+	readonly primaries: VideoColorPrimaries | null;
+	/** The transfer characteristics used. */
+	readonly transfer: VideoTransferCharacteristics | null;
+	/** The color matrix coefficients used. */
+	readonly matrix: VideoMatrixCoefficients | null;
+	/** Whether the color values use the full range or limited range. */
+	readonly fullRange: boolean | null;
+
+	/** Creates a new VideoSampleColorSpace. */
+	constructor(init?: VideoColorSpaceInit) {
+		this.primaries = init?.primaries ?? null;
+		this.transfer = init?.transfer ?? null;
+		this.matrix = init?.matrix ?? null;
+		this.fullRange = init?.fullRange ?? null;
+	}
+
+	/** Serializes the color space to a JSON object. */
+	toJSON(): VideoColorSpaceInit {
+		return {
+			primaries: this.primaries,
+			transfer: this.transfer,
+			matrix: this.matrix,
+			fullRange: this.fullRange,
+		};
+	}
+}
+
 const isVideoFrame = (x: unknown): x is VideoFrame => {
 	return typeof VideoFrame !== 'undefined' && x instanceof VideoFrame;
 };
@@ -853,7 +1016,168 @@ export const validateCropRectangle = (crop: CropRectangle, prefix: string) => {
 	}
 };
 
-const AUDIO_SAMPLE_FORMATS = new Set(
+const validateVideoFrameCopyToOptions = (options: VideoFrameCopyToOptions) => {
+	if (!options || typeof options !== 'object') {
+		throw new TypeError('options must be an object.');
+	}
+	if (options.colorSpace !== undefined && !['display-p3', 'srgb'].includes(options.colorSpace)) {
+		throw new TypeError('options.colorSpace, when provided, must be \'display-p3\' or \'srgb\'.');
+	}
+	if (options.format !== undefined && typeof options.format !== 'string') {
+		throw new TypeError('options.format, when provided, must be a string.');
+	}
+	if (options.layout !== undefined) {
+		if (!Array.isArray(options.layout)) {
+			throw new TypeError('options.layout, when provided, must be an array.');
+		}
+
+		for (const plane of options.layout) {
+			if (!plane || typeof plane !== 'object') {
+				throw new TypeError('Each entry in options.layout must be an object.');
+			}
+			if (!Number.isInteger(plane.offset) || plane.offset < 0) {
+				throw new TypeError('plane.offset must be a non-negative integer.');
+			}
+			if (!Number.isInteger(plane.stride) || plane.stride < 0) {
+				throw new TypeError('plane.stride must be a non-negative integer.');
+			}
+		}
+	}
+	if (options.rect !== undefined) {
+		if (!options.rect || typeof options.rect !== 'object') {
+			throw new TypeError('options.rect, when provided, must be an object.');
+		}
+		if (options.rect.x !== undefined && (!Number.isInteger(options.rect.x) || options.rect.x < 0)) {
+			throw new TypeError('options.rect.x, when provided, must be a non-negative integer.');
+		}
+		if (options.rect.y !== undefined && (!Number.isInteger(options.rect.y) || options.rect.y < 0)) {
+			throw new TypeError('options.rect.y, when provided, must be a non-negative integer.');
+		}
+		if (options.rect.width !== undefined && (!Number.isInteger(options.rect.width) || options.rect.width < 0)) {
+			throw new TypeError('options.rect.width, when provided, must be a non-negative integer.');
+		}
+		if (options.rect.height !== undefined && (!Number.isInteger(options.rect.height) || options.rect.height < 0)) {
+			throw new TypeError('options.rect.height, when provided, must be a non-negative integer.');
+		}
+	}
+};
+
+/** Implements logic from WebCodecs § 9.4.6 "Compute Layout and Allocation Size" */
+const createDefaultPlaneLayout = (
+	format: VideoSamplePixelFormat,
+	codedWidth: number,
+	codedHeight: number,
+): PlaneLayout[] => {
+	const planes = getPlaneConfigs(format);
+	const layouts: PlaneLayout[] = [];
+	let currentOffset = 0;
+
+	for (const plane of planes) {
+		// Per § 9.8, dimensions are usually "rounded up to the nearest integer".
+		const planeWidth = Math.ceil(codedWidth / plane.widthDivisor);
+		const planeHeight = Math.ceil(codedHeight / plane.heightDivisor);
+
+		const stride = planeWidth * plane.sampleBytes;
+
+		// Tight packing
+		const planeSize = stride * planeHeight;
+
+		layouts.push({
+			offset: currentOffset,
+			stride: stride,
+		});
+
+		currentOffset += planeSize;
+	}
+
+	return layouts;
+};
+
+type PlaneConfig = {
+	sampleBytes: number;
+	widthDivisor: number; // Horizontal sub-sampling factor
+	heightDivisor: number; // Vertical sub-sampling factor
+};
+
+/** Helper to retrieve plane configurations based on WebCodecs § 9.8 Pixel Format definitions. */
+const getPlaneConfigs = (format: VideoSamplePixelFormat): PlaneConfig[] => {
+	// Helper for standard YUV planes
+	const yuv = (
+		yBytes: number,
+		uvBytes: number,
+		subX: number,
+		subY: number,
+		hasAlpha: boolean,
+	): PlaneConfig[] => {
+		const configs: PlaneConfig[] = [
+			{ sampleBytes: yBytes, widthDivisor: 1, heightDivisor: 1 },
+			{ sampleBytes: uvBytes, widthDivisor: subX, heightDivisor: subY },
+			{ sampleBytes: uvBytes, widthDivisor: subX, heightDivisor: subY },
+		];
+
+		if (hasAlpha) {
+			// Match luma dimensions
+			configs.push({ sampleBytes: yBytes, widthDivisor: 1, heightDivisor: 1 });
+		}
+
+		return configs;
+	};
+
+	switch (format) {
+		case 'I420':
+			return yuv(1, 1, 2, 2, false);
+		case 'I420P10':
+		case 'I420P12':
+			return yuv(2, 2, 2, 2, false);
+		case 'I420A':
+			return yuv(1, 1, 2, 2, true);
+		case 'I420AP10':
+		case 'I420AP12':
+			return yuv(2, 2, 2, 2, true);
+
+		case 'I422':
+			return yuv(1, 1, 2, 1, false);
+		case 'I422P10':
+		case 'I422P12':
+			return yuv(2, 2, 2, 1, false);
+		case 'I422A':
+			return yuv(1, 1, 2, 1, true);
+		case 'I422AP10':
+		case 'I422AP12':
+			return yuv(2, 2, 2, 1, true);
+
+		case 'I444':
+			return yuv(1, 1, 1, 1, false);
+		case 'I444P10':
+		case 'I444P12':
+			return yuv(2, 2, 1, 1, false);
+		case 'I444A':
+			return yuv(1, 1, 1, 1, true);
+		case 'I444AP10':
+		case 'I444AP12':
+			return yuv(2, 2, 1, 1, true);
+
+		case 'NV12':
+			return [
+				{ sampleBytes: 1, widthDivisor: 1, heightDivisor: 1 },
+				{ sampleBytes: 2, widthDivisor: 2, heightDivisor: 2 }, // Interleaved U and V
+			];
+
+		case 'RGBA':
+		case 'RGBX':
+		case 'BGRA':
+		case 'BGRX':
+			return [
+				{ sampleBytes: 4, widthDivisor: 1, heightDivisor: 1 },
+			];
+
+		default:
+			assertNever(format);
+			assert(false);
+	}
+};
+
+const AUDIO_SAMPLE_FORMATS = new Set<AudioSampleFormat>(
 	['f32', 'f32-planar', 's16', 's16-planar', 's32', 's32-planar', 'u8', 'u8-planar'],
 );
 

src/source.ts

@@ -263,6 +263,11 @@ export class BlobSource extends Source {
 		}
 
 		worker.running = false;
+
+		if (worker.aborted) {
+			// MDN: "Calling this method signals a loss of interest in the stream by a consumer."
+			await reader?.cancel();
+		}
 	}
 
 	/** @internal */
@@ -556,7 +561,7 @@ export class UrlSource extends Source {
 				}
 
 				if (worker.aborted) {
-					break;
+					continue; // Cleanup happens in next iteration
 				}
 
 				const { done, value } = readResult;
@@ -578,14 +583,8 @@ export class UrlSource extends Source {
 				this.onread?.(worker.currentPos, worker.currentPos + value.length);
 				this._orchestrator.supplyWorkerData(worker, value);
 			}
-
-			if (worker.aborted) {
-				break;
-			}
 		}
 
-		worker.running = false;
-
 		// The previous UrlSource had logic for circumventing https://issues.chromium.org/issues/436025873; I haven't
 		// been able to observe this bug with the new UrlSource (maybe because we're using response streaming), so the
 		// logic for that has vanished for now. Leaving a comment here if this becomes relevant again.

test/browser/video-samples.test.ts

@@ -0,0 +1,220 @@
+import { expect, test } from 'vitest';
+import { VideoSample } from '../../src/sample.js';
+
+test('allocationSize', async () => {
+	{
+		const canvas = new OffscreenCanvas(1280, 720);
+		canvas.getContext('2d');
+
+		using sample = new VideoSample(canvas, { timestamp: 0 });
+
+		const size1 = sample.allocationSize();
+		expect(size1).toBe(1280 * 720 * 4);
+
+		const size2 = sample.allocationSize({
+			layout: [{
+				offset: 0,
+				stride: 1300 * 4,
+			}],
+		});
+		expect(size2).toBe(1300 * 720 * 4);
+	}
+
+	{
+		const data = new Uint8Array(1280 * 720 * 4);
+		using sample = new VideoSample(data, {
+			timestamp: 0,
+			codedWidth: 1280,
+			codedHeight: 720,
+			format: 'RGBA',
+		});
+
+		const size1 = sample.allocationSize();
+		expect(size1).toBe(1280 * 720 * 4);
+
+		const size2 = sample.allocationSize({
+			layout: [{
+				offset: 0,
+				stride: 1300 * 4,
+			}],
+		});
+		expect(size2).toBe(1300 * 720 * 4);
+	}
+
+	{
+		const data = new Uint8Array(1280 * 720 * 1.5);
+		using sample = new VideoSample(data, {
+			timestamp: 0,
+			codedWidth: 1280,
+			codedHeight: 720,
+			format: 'I420',
+		});
+
+		const size1 = sample.allocationSize();
+		expect(size1).toBe(1280 * 720 * 1.5);
+
+		const size2 = sample.allocationSize({ format: 'RGBA' });
+		expect(size2).toBe(1280 * 720 * 4);
+
+		const size3 = sample.allocationSize({
+			format: 'RGBA',
+			layout: [{
+				offset: 0,
+				stride: 1300 * 4,
+			}],
+		});
+		expect(size3).toBe(1300 * 720 * 4);
+	}
+});
+
+test('copyTo and plane layouts', async () => {
+	const buffer = new ArrayBuffer(1e7);
+
+	{
+		const canvas = new OffscreenCanvas(1280, 720);
+		canvas.getContext('2d');
+
+		using sample = new VideoSample(canvas, { timestamp: 0 });
+		const layout = await sample.copyTo(buffer);
+
+		expect(layout).toEqual([{
+			offset: 0,
+			stride: 1280 * 4,
+		}]);
+	}
+
+	{
+		const canvas = new OffscreenCanvas(1280, 720);
+		canvas.getContext('2d');
+
+		using sample = new VideoSample(canvas, { timestamp: 0 });
+		const layout = await sample.copyTo(buffer, {
+			layout: [{
+				offset: 0,
+				stride: 1300 * 4,
+			}],
+		});
+
+		expect(layout).toEqual([{
+			offset: 0,
+			stride: 1300 * 4,
+		}]);
+	}
+
+	{
+		const data = new Uint8Array(1280 * 720 * 4);
+		using sample = new VideoSample(data, {
+			timestamp: 0,
+			codedWidth: 1280,
+			codedHeight: 720,
+			format: 'RGBA',
+		});
+		const layout = await sample.copyTo(buffer);
+
+		expect(layout).toEqual([{
+			offset: 0,
+			stride: 1280 * 4,
+		}]);
+	}
+
+	{
+		const data = new Uint8Array(1280 * 720 * 4);
+		using sample = new VideoSample(data, {
+			timestamp: 0,
+			codedWidth: 1280,
+			codedHeight: 720,
+			format: 'RGBA',
+		});
+		const layout = await sample.copyTo(buffer, {
+			layout: [{
+				offset: 0,
+				stride: 1300 * 4,
+			}],
+		});
+
+		expect(layout).toEqual([{
+			offset: 0,
+			stride: 1300 * 4,
+		}]);
+	}
+
+	{
+		const data = new Uint8Array(1300 * 720 * 4);
+		using sample = new VideoSample(data, {
+			timestamp: 0,
+			codedWidth: 1280,
+			codedHeight: 720,
+			format: 'RGBA',
+			layout: [{
+				offset: 0,
+				stride: 1300 * 4,
+			}],
+		});
+		const layout = await sample.copyTo(buffer);
+
+		expect(layout).toEqual([{
+			offset: 0,
+			stride: 1300 * 4,
+		}]);
+
+		using clone = sample.clone();
+		const clonedLayout = await clone.copyTo(buffer);
+
+		expect(clonedLayout).toEqual([{
+			offset: 0,
+			stride: 1300 * 4,
+		}]);
+	}
+
+	{
+		const data = new Uint8Array(1280 * 720 * 1.5);
+		using sample = new VideoSample(data, {
+			timestamp: 0,
+			codedWidth: 1280,
+			codedHeight: 720,
+			format: 'I420',
+		});
+		const layout = await sample.copyTo(buffer);
+
+		expect(layout).toEqual([{
+			offset: 0,
+			stride: 1280,
+		}, {
+			offset: 1280 * 720,
+			stride: 1280 / 2,
+		}, {
+			offset: 1280 * 720 + (1280 / 2) * (720 / 2),
+			stride: 1280 / 2,
+		}]);
+	}
+});
+
+test('null format', async () => {
+	const canvas = new OffscreenCanvas(1280, 720);
+	canvas.getContext('2d');
+
+	const frame = new VideoFrame(canvas, { timestamp: 0 });
+	using sample = new VideoSample(frame);
+	// @ts-expect-error Sybau
+	sample.format = null;
+
+	expect(() => sample.allocationSize()).toThrow('when format is null');
+	await expect(async () => sample.copyTo(new ArrayBuffer())).rejects.toThrow('when format is null');
+
+	const size = sample.allocationSize({ format: 'RGBA' });
+	expect(size).toBe(1280 * 720 * 4);
+	const buffer = new ArrayBuffer(size);
+	const layout = await sample.copyTo(buffer, { format: 'RGBA' });
+
+	expect(layout).toEqual([{
+		offset: 0,
+		stride: 1280 * 4,
+	}]);
+
+	sample.allocationSize({ format: 'RGBX' });
+	sample.allocationSize({ format: 'BGRA' });
+	sample.allocationSize({ format: 'BGRX' });
+	await sample.copyTo(buffer, { format: 'RGBX' });
+	await sample.copyTo(buffer, { format: 'BGRA' });
+	await sample.copyTo(buffer, { format: 'BGRX' });
+});

test/node/annex-b-conversion.test.ts

@@ -0,0 +1,54 @@
+import { expect, test } from 'vitest';
+import { Input } from '../../src/input.js';
+import { BufferSource, FilePathSource } from '../../src/source.js';
+import path from 'node:path';
+import { ALL_FORMATS } from '../../src/input-format.js';
+import { Output } from '../../src/output.js';
+import { Mp4OutputFormat } from '../../src/output-format.js';
+import { BufferTarget } from '../../src/target.js';
+import { Conversion } from '../../src/conversion.js';
+import { EncodedPacketSink } from '../../src/media-sink.js';
+import { extractAvcNalUnits } from '../../src/codec-data.js';
+
+const __dirname = new URL('.', import.meta.url).pathname;
+
+test('Annex B to length-prefixed conversion, MP4', async () => {
+	using originalInput = new Input({
+		source: new FilePathSource(path.join(__dirname, '..', 'public/annex-b-avc.mkv')),
+		formats: ALL_FORMATS,
+	});
+	const originalVideoTrack = (await originalInput.getPrimaryVideoTrack())!;
+	const originalDecoderConfig = (await originalVideoTrack.getDecoderConfig())!;
+	expect(originalDecoderConfig.description).toBeUndefined();
+	expect(originalVideoTrack.codec).toBe('avc');
+
+	const originalSink = new EncodedPacketSink(originalVideoTrack);
+	const originalFirstPacket = await originalSink.getFirstPacket();
+	expect([...originalFirstPacket!.data.slice(0, 4)]).toEqual([0, 0, 0, 1]);
+
+	const originalNalUnits = extractAvcNalUnits(originalFirstPacket!.data, originalDecoderConfig);
+
+	const output = new Output({
+		format: new Mp4OutputFormat(),
+		target: new BufferTarget(),
+	});
+
+	const conversion = await Conversion.init({ input: originalInput, output });
+	await conversion.execute();
+
+	using newInput = new Input({
+		source: new BufferSource(output.target.buffer!),
+		formats: ALL_FORMATS,
+	});
+	const newVideoTrack = (await newInput.getPrimaryVideoTrack())!;
+	const newDecoderConfig = (await newVideoTrack.getDecoderConfig())!;
+	expect(newDecoderConfig.description).toBeDefined();
+	expect(newVideoTrack.codec).toBe('avc');
+
+	const newSink = new EncodedPacketSink(newVideoTrack);
+	const newFirstPacket = await newSink.getFirstPacket();
+	expect([...newFirstPacket!.data.slice(0, 4)]).not.toEqual([0, 0, 0, 1]); // Successfully converted
+
+	const newNalUnits = extractAvcNalUnits(newFirstPacket!.data, newDecoderConfig);
+	expect(newNalUnits).toEqual(originalNalUnits); // Content is the same though
+});

test/node/node-sample-creation.test.ts

@@ -0,0 +1,27 @@
+import { test } from 'vitest';
+import { AudioSample, VideoSample } from '../../src/sample.js';
+
+test('VideoSample creation from bytes', async () => {
+	const bytes = new Uint8Array(1024);
+	const sample = new VideoSample(bytes, {
+		codedWidth: 1280,
+		codedHeight: 720,
+		format: 'RGBA',
+		timestamp: 0,
+	});
+
+	sample.close();
+});
+
+test('AudioSample creation from bytes', async () => {
+	const bytes = new Uint8Array(1024);
+	const sample = new AudioSample({
+		data: bytes,
+		numberOfChannels: 2,
+		format: 's16',
+		sampleRate: 48000,
+		timestamp: 0,
+	});
+
+	sample.close();
+});