Bump patch

* fix tenc fallback

* Add test case

---------

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

.gitignore

@@ -2,7 +2,7 @@ node_modules
 /dist
 /dist-docs
 .DS_Store
-/docs/.vitepress/cache
+docs/.vitepress/cache
 /docs/api
 *.tsbuildinfo
 

dev/demux.html

@@ -18,8 +18,10 @@
 		});
 
 		const track = await input.getPrimaryVideoTrack();
-		console.log(await track.computeFrameRateMetrics());
+		const sink = new Mediabunny.EncodedPacketSink(track);
 
+		const packet = await sink.getFirstPacket();
+		console.log(packet);
 		/*
 
 		for await (const packet of packetSink.packets()) {

package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "mediabunny",
-  "version": "1.55.1",
+  "version": "1.55.5",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "mediabunny",
-      "version": "1.55.1",
+      "version": "1.55.5",
       "license": "MPL-2.0",
       "workspaces": [
         ".",
@@ -12958,7 +12958,7 @@
     },
     "packages/aac-encoder": {
       "name": "@mediabunny/aac-encoder",
-      "version": "1.55.1",
+      "version": "1.55.5",
       "license": "MPL-2.0",
       "devDependencies": {
         "@types/emscripten": "^1.40.1"
@@ -12973,7 +12973,7 @@
     },
     "packages/ac3": {
       "name": "@mediabunny/ac3",
-      "version": "1.55.1",
+      "version": "1.55.5",
       "license": "MPL-2.0",
       "devDependencies": {
         "@types/emscripten": "^1.40.1"
@@ -12988,7 +12988,7 @@
     },
     "packages/dts": {
       "name": "@mediabunny/dts",
-      "version": "1.55.1",
+      "version": "1.55.5",
       "license": "MPL-2.0",
       "devDependencies": {
         "@types/emscripten": "^1.40.1"
@@ -13003,7 +13003,7 @@
     },
     "packages/flac-encoder": {
       "name": "@mediabunny/flac-encoder",
-      "version": "1.55.1",
+      "version": "1.55.5",
       "license": "MPL-2.0",
       "devDependencies": {
         "@types/emscripten": "^1.40.1"
@@ -13018,7 +13018,7 @@
     },
     "packages/mp3-encoder": {
       "name": "@mediabunny/mp3-encoder",
-      "version": "1.55.1",
+      "version": "1.55.5",
       "license": "MPL-2.0",
       "devDependencies": {
         "@types/emscripten": "^1.40.1"
@@ -13033,7 +13033,7 @@
     },
     "packages/prores": {
       "name": "@mediabunny/prores",
-      "version": "1.55.1",
+      "version": "1.55.5",
       "license": "MPL-2.0",
       "dependencies": {
         "turbores": "^1.2.2"
@@ -13048,10 +13048,10 @@
     },
     "packages/server": {
       "name": "@mediabunny/server",
-      "version": "1.55.1",
+      "version": "1.55.5",
       "license": "MPL-2.0",
       "dependencies": {
-        "@mediabunny/prores": "^1.55.1",
+        "@mediabunny/prores": "^1.55.5",
         "node-av": "^6.0.0"
       },
       "funding": {

package.json

@@ -1,7 +1,7 @@
 {
   "name": "mediabunny",
   "author": "Vanilagy",
-  "version": "1.55.1",
+  "version": "1.55.5",
   "description": "Pure TypeScript media toolkit for reading, writing, and converting media files, directly in the browser.",
   "type": "module",
   "workspaces": [

packages/aac-encoder/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@mediabunny/aac-encoder",
   "author": "Vanilagy",
-  "version": "1.55.1",
+  "version": "1.55.5",
   "description": "AAC encoder extension for Mediabunny, based on libavcodec.",
   "main": "./dist/bundles/mediabunny-aac-encoder.mjs",
   "module": "./dist/bundles/mediabunny-aac-encoder.mjs",

packages/ac3/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@mediabunny/ac3",
   "author": "Vanilagy",
-  "version": "1.55.1",
+  "version": "1.55.5",
   "description": "AC-3 and E-AC-3 (Dolby Digital) decoder and encoder extension for Mediabunny, based on libavcodec.",
   "main": "./dist/bundles/mediabunny-ac3.mjs",
   "module": "./dist/bundles/mediabunny-ac3.mjs",

packages/dts/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@mediabunny/dts",
   "author": "Vanilagy",
-  "version": "1.55.1",
+  "version": "1.55.5",
   "description": "DTS decoder and encoder extension for Mediabunny, based on libavcodec.",
   "main": "./dist/bundles/mediabunny-dts.mjs",
   "module": "./dist/bundles/mediabunny-dts.mjs",

packages/flac-encoder/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@mediabunny/flac-encoder",
   "author": "Vanilagy",
-  "version": "1.55.1",
+  "version": "1.55.5",
   "description": "FLAC encoder extension for Mediabunny, based on libFLAC.",
   "main": "./dist/bundles/mediabunny-flac-encoder.mjs",
   "module": "./dist/bundles/mediabunny-flac-encoder.mjs",

packages/mp3-encoder/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@mediabunny/mp3-encoder",
   "author": "Vanilagy",
-  "version": "1.55.1",
+  "version": "1.55.5",
   "description": "MP3 encoder extension for Mediabunny, based on LAME.",
   "main": "./dist/bundles/mediabunny-mp3-encoder.mjs",
   "module": "./dist/bundles/mediabunny-mp3-encoder.mjs",

packages/prores/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@mediabunny/prores",
   "author": "Vanilagy",
-  "version": "1.55.1",
+  "version": "1.55.5",
   "description": "Apple ProRes decoder extension for Mediabunny, based on TurboRes.",
   "main": "./dist/bundles/mediabunny-prores.mjs",
   "module": "./dist/bundles/mediabunny-prores.mjs",

packages/server/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@mediabunny/server",
   "author": "Vanilagy",
-  "version": "1.55.1",
+  "version": "1.55.5",
   "description": "Adds full video and audio decoder and encoder support to Mediabunny for use in server-side environments (Node, Bun, Deno). Based on NodeAV.",
   "main": "./dist/bundles/mediabunny-server.cjs",
   "module": "./dist/bundles/mediabunny-server.mjs",
@@ -35,7 +35,7 @@
   },
   "dependencies": {
     "node-av": "^6.0.0",
-    "@mediabunny/prores": "^1.55.1"
+    "@mediabunny/prores": "^1.55.5"
   },
   "peerDependencies": {
     "mediabunny": "^1.45.0"

shared/aac-misc.ts

@@ -10,10 +10,11 @@ import { Bitstream } from './bitstream';
 
 export type AacAudioSpecificConfig = {
 	objectType: number;
+	coreObjectType: number;
 	frequencyIndex: number;
-	sampleRate: number | null;
 	channelConfiguration: number;
-	numberOfChannels: number | null;
+	outputSampleRate: number | null;
+	outputNumberOfChannels: number | null;
 };
 
 export const aacFrequencyTable = [
@@ -30,20 +31,8 @@ export const parseAacAudioSpecificConfig = (bytes: Uint8Array | null): AacAudioS
 
 	const bitstream = new Bitstream(bytes);
 
-	let objectType = bitstream.readBits(5);
-	if (objectType === 31) {
-		objectType = 32 + bitstream.readBits(6);
-	}
-
-	const frequencyIndex = bitstream.readBits(4);
-	let sampleRate: number | null = null;
-	if (frequencyIndex === 15) {
-		sampleRate = bitstream.readBits(24);
-	} else {
-		if (frequencyIndex < aacFrequencyTable.length) {
-			sampleRate = aacFrequencyTable[frequencyIndex]!;
-		}
-	}
+	const objectType = readAacObjectType(bitstream);
+	const { frequencyIndex, sampleRate } = readAacSamplingFrequency(bitstream);
 
 	const channelConfiguration = bitstream.readBits(4);
 	let numberOfChannels: number | null = null;
@@ -51,61 +40,155 @@ export const parseAacAudioSpecificConfig = (bytes: Uint8Array | null): AacAudioS
 		numberOfChannels = aacChannelMap[channelConfiguration]!;
 	}
 
+	let coreObjectType = objectType;
+	let psPresent = false;
+	let outputSampleRate = sampleRate;
+
+	if (objectType === 5 || objectType === 29) {
+		// Explicit hierarchical signaling: everything read so far describes the core coder, and the rate the
+		// decoder actually outputs follows right here
+		psPresent = objectType === 29;
+
+		outputSampleRate = readAacSamplingFrequency(bitstream).sampleRate;
+		coreObjectType = readAacObjectType(bitstream);
+
+		if (coreObjectType === 22) {
+			bitstream.skipBits(4); // extensionChannelConfiguration
+		}
+	} else {
+		// There may be SBR/PS flags sitting behind a sync word after the config of the core coder. We find them by
+		// scanning for the sync word, just like FFmpeg does.
+		while (bitstream.getBitsLeft() > 15) {
+			const searchStart = bitstream.pos;
+
+			if (bitstream.readBits(11) !== 0x2b7) {
+				bitstream.pos = searchStart + 1;
+				continue;
+			}
+
+			if (readAacObjectType(bitstream) === 5 && bitstream.readBits(1)) {
+				outputSampleRate = readAacSamplingFrequency(bitstream).sampleRate;
+
+				if (bitstream.getBitsLeft() > 11 && bitstream.readBits(11) === 0x548) {
+					psPresent = !!bitstream.readBits(1);
+				}
+			}
+
+			break;
+		}
+	}
+
+	if (numberOfChannels !== null && numberOfChannels > 1) {
+		// PS only ever upmixes a mono core. A stereo signal means whatever we read was bogus
+		psPresent = false;
+	}
+
 	return {
 		objectType,
+		coreObjectType,
 		frequencyIndex,
-		sampleRate,
 		channelConfiguration,
-		numberOfChannels,
+		outputSampleRate,
+		outputNumberOfChannels: psPresent && numberOfChannels === 1
+			? 2
+			: numberOfChannels,
+	};
+};
+
+const readAacObjectType = (bitstream: Bitstream) => {
+	const objectType = bitstream.readBits(5);
+	return objectType === 31 ? 32 + bitstream.readBits(6) : objectType;
+};
+
+const readAacSamplingFrequency = (bitstream: Bitstream) => {
+	const frequencyIndex = bitstream.readBits(4);
+
+	if (frequencyIndex === 15) {
+		return {
+			frequencyIndex,
+			sampleRate: bitstream.readBits(24),
+		};
+	}
+
+	return {
+		frequencyIndex,
+		sampleRate: frequencyIndex < aacFrequencyTable.length
+			? aacFrequencyTable[frequencyIndex]!
+			: null,
 	};
 };
 
 export const buildAacAudioSpecificConfig = (config: {
 	objectType: number;
-	sampleRate: number;
-	numberOfChannels: number;
+	outputSampleRate: number;
+	outputNumberOfChannels: number;
 }) => {
-	let frequencyIndex = aacFrequencyTable.indexOf(config.sampleRate);
-	let customSampleRate: number | null = null;
+	const usesSbr = config.objectType === 5 || config.objectType === 29;
+	const usesPs = config.objectType === 29;
 
-	if (frequencyIndex === -1) {
-		frequencyIndex = 15;
-		customSampleRate = config.sampleRate;
-	}
+	// SBR runs the core coder at half the output rate, and PS upmixes a mono core to stereo
+	const coreSampleRate = usesSbr ? config.outputSampleRate / 2 : config.outputSampleRate;
+	const coreNumberOfChannels = usesPs ? 1 : config.outputNumberOfChannels;
 
-	const channelConfiguration = aacChannelMap.indexOf(config.numberOfChannels);
+	const channelConfiguration = aacChannelMap.indexOf(coreNumberOfChannels);
 	if (channelConfiguration === -1) {
-		throw new TypeError(`Unsupported number of channels: ${config.numberOfChannels}`);
+		throw new TypeError(`Unsupported number of channels: ${config.outputNumberOfChannels}`);
 	}
 
-	let bitCount = 5 + 4 + 4;
+	// Object type, sampling frequency, channel configuration, then a bare GASpecificConfig
+	let bitCount = 5 + 4 + 4 + 3;
 	if (config.objectType >= 32) {
 		bitCount += 6;
 	}
-	if (frequencyIndex === 15) {
+	if (findAacFrequencyIndex(coreSampleRate) === 15) {
 		bitCount += 24;
 	}
+	if (usesSbr) {
+		bitCount += 4 + 5; // Extension sampling frequency and the object type of the core coder
+		if (findAacFrequencyIndex(config.outputSampleRate) === 15) {
+			bitCount += 24;
+		}
+	}
 
 	const byteCount = Math.ceil(bitCount / 8);
 	const bytes = new Uint8Array(byteCount);
 	const bitstream = new Bitstream(bytes);
 
-	if (config.objectType < 32) {
-		bitstream.writeBits(5, config.objectType);
-	} else {
-		bitstream.writeBits(5, 31);
-		bitstream.writeBits(6, config.objectType - 32);
+	writeAacObjectType(bitstream, config.objectType);
+	writeAacSamplingFrequency(bitstream, coreSampleRate);
+	bitstream.writeBits(4, channelConfiguration);
+
+	if (usesSbr) {
+		writeAacSamplingFrequency(bitstream, config.outputSampleRate);
+		writeAacObjectType(bitstream, 2); // AAC-LC underneath
 	}
 
+	bitstream.writeBits(3, 0); // frameLengthFlag, dependsOnCoreCoder, extensionFlag
+
+	return bytes;
+};
+
+const writeAacObjectType = (bitstream: Bitstream, objectType: number) => {
+	if (objectType < 32) {
+		bitstream.writeBits(5, objectType);
+	} else {
+		bitstream.writeBits(5, 31);
+		bitstream.writeBits(6, objectType - 32);
+	}
+};
+
+const writeAacSamplingFrequency = (bitstream: Bitstream, sampleRate: number) => {
+	const frequencyIndex = findAacFrequencyIndex(sampleRate);
 	bitstream.writeBits(4, frequencyIndex);
 
 	if (frequencyIndex === 15) {
-		bitstream.writeBits(24, customSampleRate!);
+		bitstream.writeBits(24, sampleRate);
 	}
+};
 
-	bitstream.writeBits(4, channelConfiguration);
-
-	return bytes;
+const findAacFrequencyIndex = (sampleRate: number) => {
+	const index = aacFrequencyTable.indexOf(sampleRate);
+	return index === -1 ? 15 : index;
 };
 
 export type AdtsHeaderTemplate = {
@@ -117,8 +200,8 @@ export const buildAdtsHeaderTemplate = (config: AacAudioSpecificConfig): AdtsHea
 	const header = new Uint8Array(7);
 	const bitstream = new Bitstream(header);
 
-	const { objectType, frequencyIndex, channelConfiguration } = config;
-	const profile = objectType - 1;
+	const { coreObjectType, frequencyIndex, channelConfiguration } = config;
+	const profile = coreObjectType - 1;
 
 	bitstream.writeBits(12, 0b1111_11111111); // Syncword
 	bitstream.writeBits(1, 0); // MPEG Version

src/adts/adts-demuxer.ts

@@ -23,6 +23,7 @@ import {
 	AsyncMutex,
 	binarySearchExact,
 	binarySearchLessOrEqual,
+	isThenable,
 	UNDETERMINED_LANGUAGE,
 } from '../misc';
 import { EncodedPacket, PLACEHOLDER_DATA } from '../packet';
@@ -84,7 +85,7 @@ export class AdtsDemuxer extends Demuxer {
 			// Skip all ID3v2 tags at the start of the file
 			while (true) {
 				let slice = this.reader.requestSlice(this.lastLoadedPos, ID3_V2_HEADER_SIZE);
-				if (slice instanceof Promise) slice = await slice;
+				if (isThenable(slice)) slice = await slice;
 
 				if (!slice) {
 					this.lastSampleLoaded = true;
@@ -105,7 +106,7 @@ export class AdtsDemuxer extends Demuxer {
 			MIN_ADTS_FRAME_HEADER_SIZE,
 			MAX_ADTS_FRAME_HEADER_SIZE,
 		);
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 		if (!slice) {
 			this.lastSampleLoaded = true;
 			return;
@@ -167,7 +168,7 @@ export class AdtsDemuxer extends Demuxer {
 
 			while (true) {
 				let headerSlice = this.reader.requestSlice(currentPos, ID3_V2_HEADER_SIZE);
-				if (headerSlice instanceof Promise) headerSlice = await headerSlice;
+				if (isThenable(headerSlice)) headerSlice = await headerSlice;
 				if (!headerSlice) break;
 
 				const id3V2Header = readId3V2Header(headerSlice);
@@ -176,7 +177,7 @@ export class AdtsDemuxer extends Demuxer {
 				}
 
 				let contentSlice = this.reader.requestSlice(headerSlice.filePos, id3V2Header.size);
-				if (contentSlice instanceof Promise) contentSlice = await contentSlice;
+				if (isThenable(contentSlice)) contentSlice = await contentSlice;
 				if (!contentSlice) break;
 
 				parseId3V2Tag(contentSlice, id3V2Header, this.metadataTags);
@@ -306,7 +307,7 @@ class AdtsAudioTrackBacking implements InputAudioTrackBacking {
 			data = PLACEHOLDER_DATA;
 		} else {
 			let slice = this.demuxer.reader.requestSlice(rawSample.dataStart, rawSample.dataSize);
-			if (slice instanceof Promise) slice = await slice;
+			if (isThenable(slice)) slice = await slice;
 
 			if (!slice) {
 				return null; // Data didn't fit into the rest of the file

src/aes.ts

@@ -6,7 +6,7 @@
  * file, You can obtain one at https://mozilla.org/MPL/2.0/.
  */
 
-import { assert, MaybePromise } from './misc';
+import { assert, isThenable, MaybePromise } from './misc';
 import { readBytes, Reader } from './reader';
 
 // Inspired in part by https://github.com/halloweeks/AES-128-CBC/blob/main/AES_128_CBC.h
@@ -258,7 +258,7 @@ export const createAes128CbcDecryptStream = (
 			const requestedLength = CHUNK_SIZE + BLOCK_SIZE;
 
 			let nextSlice = reader.requestSliceRange(pos, 0, requestedLength);
-			if (nextSlice instanceof Promise) nextSlice = await nextSlice;
+			if (isThenable(nextSlice)) nextSlice = await nextSlice;
 			if (!nextSlice || nextSlice.length === 0) {
 				// Due to padding, this should never happen
 				throw new Error('Invalid ciphertext.');

src/codec-data.ts

@@ -22,7 +22,6 @@ import {
 	textEncoder,
 	toDataView,
 	toUint8Array,
-	getChromiumVersion,
 	isChromium,
 	popcount,
 	setUint24,
@@ -1727,6 +1726,16 @@ export type Vp9CodecInfo = {
 	matrixCoefficients: number;
 };
 
+// The color spaces from section 7.2.2 of the VP9 spec The matrix column is taken from FFmpeg's libavcodec/vp9.c
+const VP9_COLOR_SPACE_TABLE = {
+	1: { colourPrimaries: 5, transferCharacteristics: 6, matrixCoefficients: 5 }, // CS_BT_601
+	2: { colourPrimaries: 1, transferCharacteristics: 1, matrixCoefficients: 1 }, // CS_BT_709
+	3: { colourPrimaries: 6, transferCharacteristics: 6, matrixCoefficients: 6 }, // CS_SMPTE_170
+	4: { colourPrimaries: 7, transferCharacteristics: 7, matrixCoefficients: 7 }, // CS_SMPTE_240
+	5: { colourPrimaries: 9, transferCharacteristics: 14, matrixCoefficients: 9 }, // CS_BT_2020
+	7: { colourPrimaries: 1, transferCharacteristics: 13, matrixCoefficients: 0 }, // CS_RGB (sRGB)
+};
+
 export const extractVp9CodecInfoFromPacket = (
 	packet: Uint8Array,
 ): Vp9CodecInfo | null => {
@@ -1835,26 +1844,10 @@ export const extractVp9CodecInfoFromPacket = (
 		}
 	}
 
-	// Map color_space to standard values
-	const matrixCoefficients = colorSpace === 7
-		? 0
-		: colorSpace === 2
-			? 1
-			: colorSpace === 1
-				? 6
-				: 2;
-
-	const colourPrimaries = colorSpace === 2
-		? 1
-		: colorSpace === 1
-			? 6
-			: 2;
-
-	const transferCharacteristics = colorSpace === 2
-		? 1
-		: colorSpace === 1
-			? 6
-			: 2;
+	const colorSpaceValues = VP9_COLOR_SPACE_TABLE[colorSpace as keyof typeof VP9_COLOR_SPACE_TABLE];
+	const colourPrimaries = colorSpaceValues?.colourPrimaries ?? 2;
+	const transferCharacteristics = colorSpaceValues?.transferCharacteristics ?? 2;
+	const matrixCoefficients = colorSpaceValues?.matrixCoefficients ?? 2;
 
 	return {
 		profile,
@@ -1868,6 +1861,12 @@ export const extractVp9CodecInfoFromPacket = (
 	};
 };
 
+export const vp9CodecInfoHasColorInfo = (info: Vp9CodecInfo) => {
+	return info.colourPrimaries !== 2
+		|| info.transferCharacteristics !== 2
+		|| info.matrixCoefficients !== 2;
+};
+
 export type Av1CodecInfo = {
 	profile: number;
 	level: number;
@@ -1877,6 +1876,10 @@ export type Av1CodecInfo = {
 	chromaSubsamplingX: number;
 	chromaSubsamplingY: number;
 	chromaSamplePosition: number;
+	videoFullRangeFlag: number;
+	colourPrimaries: number;
+	transferCharacteristics: number;
+	matrixCoefficients: number;
 };
 
 /** Iterates over all OBUs in an AV1 packet bitstream. */
@@ -1891,7 +1894,8 @@ export const iterateAv1PacketObus = function* (packet: Uint8Array) {
 		for (let i = 0; i < 8; i++) {
 			const byte = bitstream.readAlignedByte();
 
-			value |= ((byte & 0x7f) << (i * 7));
+			// No bit shift since the value can get big
+			value += (byte & 0x7f) * 2 ** (i * 7);
 
 			if (!(byte & 0x80)) {
 				break;
@@ -1904,7 +1908,7 @@ export const iterateAv1PacketObus = function* (packet: Uint8Array) {
 		}
 
 		// Spec requirement
-		if (value >= 2 ** 32 - 1) {
+		if (value > 2 ** 32 - 1) {
 			return null;
 		}
 
@@ -1978,43 +1982,43 @@ export const extractAv1CodecInfoFromPacket = (
 		if (reducedStillPictureHeader) {
 			seqLevel = bitstream.readBits(5);
 		} else {
-			// Parse timing_info_present_flag
 			const timingInfoPresentFlag = bitstream.readBits(1);
 
+			let decoderModelInfoPresentFlag = 0;
+
 			if (timingInfoPresentFlag) {
-				// Skip timing info (num_units_in_display_tick, time_scale, equal_picture_interval)
 				bitstream.skipBits(32); // num_units_in_display_tick
 				bitstream.skipBits(32); // time_scale
 				const equalPictureInterval = bitstream.readBits(1);
 
 				if (equalPictureInterval) {
-					// Skip num_ticks_per_picture_minus_1 (uvlc)
-					// Since this is variable length, we'd need to implement uvlc reading
-					// For now, we'll return null as this is rare
-					return null;
+					// num_ticks_per_picture_minus_1 is uvlc-coded, so we need to seek past it manually
+					let leadingZeros = 0;
+					while (leadingZeros < 32 && !bitstream.readBits(1)) {
+						leadingZeros++;
+					}
+
+					if (leadingZeros < 32) {
+						bitstream.skipBits(leadingZeros);
+					}
+				}
+
+				decoderModelInfoPresentFlag = bitstream.readBits(1);
+
+				if (decoderModelInfoPresentFlag) {
+					bufferDelayLengthMinus1 = bitstream.readBits(5);
+					bitstream.skipBits(32); // num_units_in_decoding_tick
+					bitstream.skipBits(5); // buffer_removal_time_length_minus_1
+					bitstream.skipBits(5); // frame_presentation_time_length_minus_1
 				}
 			}
 
-			// Parse decoder_model_info_present_flag
-			const decoderModelInfoPresentFlag = bitstream.readBits(1);
-
-			if (decoderModelInfoPresentFlag) {
-				// Store buffer_delay_length_minus_1 instead of just skipping
-				bufferDelayLengthMinus1 = bitstream.readBits(5);
-				bitstream.skipBits(32); // num_units_in_decoding_tick
-				bitstream.skipBits(5); // buffer_removal_time_length_minus_1
-				bitstream.skipBits(5); // frame_presentation_time_length_minus_1
-			}
-
-			// Parse operating_points_cnt_minus_1
+			const initialDisplayDelayPresentFlag = bitstream.readBits(1);
 			const operatingPointsCntMinus1 = bitstream.readBits(5);
 
-			// For each operating point
 			for (let i = 0; i <= operatingPointsCntMinus1; i++) {
-				// operating_point_idc[i]
-				bitstream.skipBits(12);
+				bitstream.skipBits(12); // operating_point_idc[i]
 
-				// seq_level_idx[i]
 				const seqLevelIdx = bitstream.readBits(5);
 
 				if (i === 0) {
@@ -2022,7 +2026,6 @@ export const extractAv1CodecInfoFromPacket = (
 				}
 
 				if (seqLevelIdx > 7) {
-					// seq_tier[i]
 					const seqTierTemp = bitstream.readBits(1);
 					if (i === 0) {
 						seqTier = seqTierTemp;
@@ -2030,7 +2033,6 @@ export const extractAv1CodecInfoFromPacket = (
 				}
 
 				if (decoderModelInfoPresentFlag) {
-					// decoder_model_present_for_this_op[i]
 					const decoderModelPresentForThisOp = bitstream.readBits(1);
 
 					if (decoderModelPresentForThisOp) {
@@ -2041,12 +2043,12 @@ export const extractAv1CodecInfoFromPacket = (
 					}
 				}
 
-				// initial_display_delay_present_flag
-				const initialDisplayDelayPresentFlag = bitstream.readBits(1);
-
 				if (initialDisplayDelayPresentFlag) {
-					// initial_display_delay_minus_1[i]
-					bitstream.skipBits(4);
+					const initialDisplayDelayPresentForThisOp = bitstream.readBits(1);
+
+					if (initialDisplayDelayPresentForThisOp) {
+						bitstream.skipBits(4); // initial_display_delay_minus_1[i]
+					}
 				}
 			}
 		}
@@ -2129,11 +2131,32 @@ export const extractAv1CodecInfoFromPacket = (
 			monochrome = bitstream.readBits(1);
 		}
 
+		let colourPrimaries = 2; // CP_UNSPECIFIED
+		let transferCharacteristics = 2; // TC_UNSPECIFIED
+		let matrixCoefficients = 2; // MC_UNSPECIFIED
+
+		const colorDescriptionPresentFlag = bitstream.readBits(1);
+		if (colorDescriptionPresentFlag) {
+			colourPrimaries = bitstream.readBits(8);
+			transferCharacteristics = bitstream.readBits(8);
+			matrixCoefficients = bitstream.readBits(8);
+		}
+
+		let videoFullRangeFlag = 0;
 		let chromaSubsamplingX = 1;
 		let chromaSubsamplingY = 1;
-		let chromaSamplePosition = 0;
+		let chromaSamplePosition = 0; // CSP_UNKNOWN
+
+		if (monochrome) {
+			videoFullRangeFlag = bitstream.readBits(1);
+		} else if (colourPrimaries === 1 && transferCharacteristics === 13 && matrixCoefficients === 0) {
+			// sRGB with an identity matrix, which is always full range 4:4:4
+			videoFullRangeFlag = 1;
+			chromaSubsamplingX = 0;
+			chromaSubsamplingY = 0;
+		} else {
+			videoFullRangeFlag = bitstream.readBits(1);
 
-		if (!monochrome) {
 			if (seqProfile === 0) {
 				chromaSubsamplingX = 1;
 				chromaSubsamplingY = 1;
@@ -2143,9 +2166,10 @@ export const extractAv1CodecInfoFromPacket = (
 			} else {
 				if (bitDepth === 12) {
 					chromaSubsamplingX = bitstream.readBits(1);
-					if (chromaSubsamplingX) {
-						chromaSubsamplingY = bitstream.readBits(1);
-					}
+					chromaSubsamplingY = chromaSubsamplingX ? bitstream.readBits(1) : 0;
+				} else {
+					chromaSubsamplingX = 1;
+					chromaSubsamplingY = 0;
 				}
 			}
 
@@ -2163,12 +2187,56 @@ export const extractAv1CodecInfoFromPacket = (
 			chromaSubsamplingX,
 			chromaSubsamplingY,
 			chromaSamplePosition,
+			videoFullRangeFlag,
+			colourPrimaries,
+			transferCharacteristics,
+			matrixCoefficients,
 		};
 	}
 
 	return null;
 };
 
+export const av1CodecInfoHasColorInfo = (info: Av1CodecInfo) => {
+	return info.colourPrimaries !== 2
+		|| info.transferCharacteristics !== 2
+		|| info.matrixCoefficients !== 2;
+};
+
+export type ProresCodecInfo = {
+	colourPrimaries: number;
+	transferCharacteristics: number;
+	matrixCoefficients: number;
+	fullRange: boolean;
+};
+
+export const extractProresCodecInfoFromPacket = (packet: Uint8Array): ProresCodecInfo | null => {
+	// https://wiki.multimedia.cx/index.php/Apple_ProRes
+
+	const frameHeaderStart = 8;
+	if (packet.length < frameHeaderStart + 28) {
+		return null;
+	}
+
+	const view = toDataView(packet);
+
+	if (view.getUint32(4) !== 0x69637066) { // 'icpf'
+		return null;
+	}
+
+	const headerSize = view.getUint16(frameHeaderStart);
+	if (headerSize < 28) {
+		return null;
+	}
+
+	return {
+		fullRange: false, // ProRes is always limited range
+		colourPrimaries: view.getUint8(frameHeaderStart + 14),
+		transferCharacteristics: view.getUint8(frameHeaderStart + 15),
+		matrixCoefficients: view.getUint8(frameHeaderStart + 16),
+	};
+};
+
 export const parseOpusIdentificationHeader = (bytes: Uint8Array) => {
 	const view = toDataView(bytes);
 
@@ -2337,8 +2405,8 @@ export const determineVideoPacketType = (
 
 				// 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.
-				if (type === AvcNalUnitType.SEI && (!isChromium() || getChromiumVersion()! >= 144)) {
+				// always supported them, but Chromium hasn't.
+				if (type === AvcNalUnitType.SEI && !isChromium()) {
 					const nalUnit = packetData.subarray(loc.offset, loc.offset + loc.length);
 					const bytes = removeEmulationPreventionBytes(nalUnit);
 					let pos = 1; // Skip NALU header

src/codec.ts

@@ -10,13 +10,22 @@ import { parseAacAudioSpecificConfig } from '../shared/aac-misc';
 import {
 	Av1CodecInfo,
 	AvcDecoderConfigurationRecord,
+	deserializeAvcDecoderConfigurationRecord,
+	deserializeHevcDecoderConfigurationRecord,
 	HevcDecoderConfigurationRecord,
+	HevcNalUnitType,
+	parseAvcSps,
+	parseHevcSps,
+	ProresCodecInfo,
 	Vp9CodecInfo,
 } from './codec-data';
 import {
 	COLOR_PRIMARIES_MAP,
+	COLOR_PRIMARIES_MAP_INVERSE,
 	MATRIX_COEFFICIENTS_MAP,
+	MATRIX_COEFFICIENTS_MAP_INVERSE,
 	TRANSFER_CHARACTERISTICS_MAP,
+	TRANSFER_CHARACTERISTICS_MAP_INVERSE,
 	assert,
 	assertNever,
 	base64ToBytes,
@@ -370,7 +379,7 @@ export const generateAv1CodecConfigurationFromCodecString = (codecString: string
 	const tier = levelAndTier.slice(-1) === 'H' ? 1 : 0;
 	const bitDepth = Number(parts[3]);
 	const highBitDepth = bitDepth === 8 ? 0 : 1;
-	const twelveBit = 0;
+	const twelveBit = bitDepth === 12 ? 1 : 0;
 	const monochrome = parts[4] ? Number(parts[4]) : 0;
 	const chromaSubsamplingX = parts[5] ? Number(parts[5][0]) : 1;
 	const chromaSubsamplingY = parts[5] ? Number(parts[5][1]) : 1;
@@ -578,6 +587,132 @@ export const extractVideoCodecString = (trackInfo: {
 	throw new TypeError(`Unhandled codec '${codec}'.`);
 };
 
+export const extractColorSpace = (info: {
+	codec: VideoCodec | null;
+	codecDescription: Uint8Array | null;
+	avcCodecInfo: AvcDecoderConfigurationRecord | null;
+	hevcCodecInfo: HevcDecoderConfigurationRecord | null;
+	vp9CodecInfo: Vp9CodecInfo | null;
+	av1CodecInfo: Av1CodecInfo | null;
+	proresCodecInfo: ProresCodecInfo | null;
+}): VideoColorSpaceInit => {
+	switch (info.codec) {
+		case 'avc': {
+			let spsData = info.avcCodecInfo?.sequenceParameterSets[0];
+			if (!spsData && info.codecDescription) {
+				spsData = deserializeAvcDecoderConfigurationRecord(info.codecDescription)?.sequenceParameterSets[0];
+			}
+
+			if (spsData) {
+				const spsInfo = parseAvcSps(spsData);
+				if (spsInfo) {
+					return {
+						primaries:
+							COLOR_PRIMARIES_MAP_INVERSE[spsInfo.colourPrimaries] as
+							VideoColorPrimaries | undefined,
+						transfer:
+							TRANSFER_CHARACTERISTICS_MAP_INVERSE[spsInfo.transferCharacteristics] as
+							VideoTransferCharacteristics | undefined,
+						matrix:
+							MATRIX_COEFFICIENTS_MAP_INVERSE[spsInfo.matrixCoefficients] as
+							VideoMatrixCoefficients | undefined,
+						fullRange: !!spsInfo.fullRangeFlag,
+					};
+				}
+			}
+		}; break;
+
+		case 'hevc': {
+			let spsData = info.hevcCodecInfo?.arrays
+				.find(x => x.nalUnitType === HevcNalUnitType.SPS_NUT)?.nalUnits[0];
+			if (!spsData && info.codecDescription) {
+				spsData = deserializeHevcDecoderConfigurationRecord(info.codecDescription)?.arrays
+					.find(x => x.nalUnitType === HevcNalUnitType.SPS_NUT)?.nalUnits[0];
+			}
+
+			if (spsData) {
+				const spsInfo = parseHevcSps(spsData);
+				if (spsInfo) {
+					return {
+						primaries:
+							COLOR_PRIMARIES_MAP_INVERSE[spsInfo.colourPrimaries] as
+							VideoColorPrimaries | undefined,
+						transfer:
+							TRANSFER_CHARACTERISTICS_MAP_INVERSE[spsInfo.transferCharacteristics] as
+							VideoTransferCharacteristics | undefined,
+						matrix:
+							MATRIX_COEFFICIENTS_MAP_INVERSE[spsInfo.matrixCoefficients] as
+							VideoMatrixCoefficients | undefined,
+						fullRange: !!spsInfo.fullRangeFlag,
+					};
+				}
+			}
+		}; break;
+
+		case 'vp8': {
+			// The situation is fucky; do nothing for now.
+		}; break;
+
+		case 'vp9': {
+			if (info.vp9CodecInfo) {
+				return {
+					primaries:
+						COLOR_PRIMARIES_MAP_INVERSE[info.vp9CodecInfo.colourPrimaries] as
+						VideoColorPrimaries | undefined,
+					transfer:
+						TRANSFER_CHARACTERISTICS_MAP_INVERSE[info.vp9CodecInfo.transferCharacteristics] as
+						VideoTransferCharacteristics | undefined,
+					matrix:
+						MATRIX_COEFFICIENTS_MAP_INVERSE[info.vp9CodecInfo.matrixCoefficients] as
+						VideoMatrixCoefficients | undefined,
+					fullRange: !!info.vp9CodecInfo.videoFullRangeFlag,
+				};
+			}
+		}; break;
+
+		case 'av1': {
+			if (info.av1CodecInfo) {
+				return {
+					primaries:
+						COLOR_PRIMARIES_MAP_INVERSE[info.av1CodecInfo.colourPrimaries] as
+						VideoColorPrimaries | undefined,
+					transfer:
+						TRANSFER_CHARACTERISTICS_MAP_INVERSE[info.av1CodecInfo.transferCharacteristics] as
+						VideoTransferCharacteristics | undefined,
+					matrix:
+						MATRIX_COEFFICIENTS_MAP_INVERSE[info.av1CodecInfo.matrixCoefficients] as
+						VideoMatrixCoefficients | undefined,
+					fullRange: !!info.av1CodecInfo.videoFullRangeFlag,
+				};
+			}
+		}; break;
+
+		case 'prores': {
+			if (info.proresCodecInfo) {
+				return {
+					primaries:
+						COLOR_PRIMARIES_MAP_INVERSE[info.proresCodecInfo.colourPrimaries] as
+						VideoColorPrimaries | undefined,
+					transfer:
+						TRANSFER_CHARACTERISTICS_MAP_INVERSE[info.proresCodecInfo.transferCharacteristics] as
+						VideoTransferCharacteristics | undefined,
+					matrix:
+						MATRIX_COEFFICIENTS_MAP_INVERSE[info.proresCodecInfo.matrixCoefficients] as
+						VideoMatrixCoefficients | undefined,
+					fullRange: info.proresCodecInfo.fullRange,
+				};
+			}
+		}; break;
+	}
+
+	return {
+		primaries: undefined,
+		transfer: undefined,
+		matrix: undefined,
+		fullRange: undefined,
+	};
+};
+
 export const buildAudioCodecString = (codec: AudioCodec, numberOfChannels: number, sampleRate: number) => {
 	if (codec === 'aac') {
 		// If stereo or higher channels and lower sample rate, likely using HE-AAC v2 with PS

src/flac/flac-demuxer.ts

@@ -15,6 +15,7 @@ import {
 	assert,
 	AsyncMutex,
 	binarySearchLessOrEqual,
+	isThenable,
 	textDecoder,
 	UNDETERMINED_LANGUAGE,
 } from '../misc';
@@ -108,7 +109,7 @@ export class FlacDemuxer extends Demuxer {
 			let currentPos = 0;
 			while (true) {
 				let headerSlice = this.reader.requestSlice(currentPos, ID3_V2_HEADER_SIZE);
-				if (headerSlice instanceof Promise) headerSlice = await headerSlice;
+				if (isThenable(headerSlice)) headerSlice = await headerSlice;
 
 				if (!headerSlice) {
 					this.lastSampleLoaded = true;
@@ -121,7 +122,7 @@ export class FlacDemuxer extends Demuxer {
 				}
 
 				let contentSlice = this.reader.requestSlice(headerSlice.filePos, id3V2Header.size);
-				if (contentSlice instanceof Promise) contentSlice = await contentSlice;
+				if (isThenable(contentSlice)) contentSlice = await contentSlice;
 				assert(contentSlice);
 
 				parseId3V2Tag(contentSlice, id3V2Header, this.metadataTags);
@@ -136,7 +137,7 @@ export class FlacDemuxer extends Demuxer {
 				|| currentPos < this.reader.fileSize
 			) {
 				let sizeSlice = this.reader.requestSlice(currentPos, 4);
-				if (sizeSlice instanceof Promise) sizeSlice = await sizeSlice;
+				if (isThenable(sizeSlice)) sizeSlice = await sizeSlice;
 				currentPos += 4;
 
 				if (sizeSlice === null) {
@@ -160,7 +161,7 @@ export class FlacDemuxer extends Demuxer {
 							currentPos,
 							size,
 						);
-						if (streamInfoBlock instanceof Promise) streamInfoBlock = await streamInfoBlock;
+						if (isThenable(streamInfoBlock)) streamInfoBlock = await streamInfoBlock;
 
 						assert(streamInfoBlock);
 						if (streamInfoBlock === null) {
@@ -219,7 +220,7 @@ export class FlacDemuxer extends Demuxer {
 							currentPos,
 							size,
 						);
-						if (vorbisCommentBlock instanceof Promise) vorbisCommentBlock = await vorbisCommentBlock;
+						if (isThenable(vorbisCommentBlock)) vorbisCommentBlock = await vorbisCommentBlock;
 
 						assert(vorbisCommentBlock);
 
@@ -237,7 +238,7 @@ export class FlacDemuxer extends Demuxer {
 							currentPos,
 							size,
 						);
-						if (pictureBlock instanceof Promise) pictureBlock = await pictureBlock;
+						if (isThenable(pictureBlock)) pictureBlock = await pictureBlock;
 
 						assert(pictureBlock);
 						const pictureType = readU32Be(pictureBlock);
@@ -775,7 +776,7 @@ class FlacAudioTrackBacking implements InputAudioTrackBacking {
 				rawSample.byteOffset,
 				rawSample.byteSize,
 			);
-			if (slice instanceof Promise) slice = await slice;
+			if (isThenable(slice)) slice = await slice;
 
 			if (!slice) {
 				return null; // Data didn't fit into the rest of the file

src/input-format.ts

@@ -29,14 +29,14 @@ import { OggDemuxer } from './ogg/ogg-demuxer';
 import { WaveDemuxer } from './wave/wave-demuxer';
 import { MAX_ADTS_FRAME_HEADER_SIZE, MIN_ADTS_FRAME_HEADER_SIZE, readAdtsFrameHeader } from './adts/adts-reader';
 import { AdtsDemuxer } from './adts/adts-demuxer';
-import { readAscii, readBytes, readU32Be } from './reader';
+import { readAscii, readBytes, readU32Be, readU64Be } from './reader';
 import { FlacDemuxer } from './flac/flac-demuxer';
 import { MpegTsDemuxer } from './mpeg-ts/mpeg-ts-demuxer';
 import { TS_PACKET_SIZE } from './mpeg-ts/mpeg-ts-misc';
 import { HlsDemuxer } from './hls/hls-demuxer';
 import { HLS_MIME_TYPE } from './hls/hls-misc';
 import { PathedSource } from './source';
-import { MaybePromise } from './misc';
+import { isThenable, MaybePromise } from './misc';
 
 /**
  * Base class representing an input media file format.
@@ -75,7 +75,7 @@ export abstract class IsobmffInputFormat extends InputFormat {
 	/** @internal */
 	protected async _getMajorBrand(input: Input) {
 		let slice = input._reader.requestSlice(0, 12);
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 		if (!slice) return null;
 
 		slice.skip(4);
@@ -116,12 +116,41 @@ export class Mp4InputFormat extends IsobmffInputFormat {
 			return majorBrand !== 'qt  ';
 		}
 
-		let slice = input._reader.requestSlice(4, 4);
-		if (slice instanceof Promise) slice = await slice;
-		if (!slice) return false;
+		let pos = 0;
+		for (let iter = 0; iter < 10; iter++) { // Bounded loop
+			let slice = input._reader.requestSlice(pos, 8);
+			if (isThenable(slice)) slice = await slice;
+			if (!slice) return false;
 
-		const fourCc = readAscii(slice, 4);
-		return fourCc === 'moof' || fourCc === 'sidx'; // Seen in HLS for example
+			let size = readU32Be(slice);
+			let headerSize = 8;
+			if (size === 1) {
+				let sizeExtensionSlice = input._reader.requestSlice(pos + 8, 8);
+				if (isThenable(sizeExtensionSlice)) sizeExtensionSlice = await sizeExtensionSlice;
+				if (!sizeExtensionSlice) return false;
+
+				size = readU64Be(sizeExtensionSlice);
+				headerSize = 16;
+			}
+
+			if (size < headerSize) {
+				return false; // Hardly a box is it
+			}
+
+			const fourCc = readAscii(slice, 4);
+			if (fourCc === 'moof' || fourCc === 'sidx') { // Seen in HLS for example
+				return true;
+			} else if (fourCc === 'emsg' || fourCc === 'prft' || fourCc === 'free') {
+				// These are boxes that can appear before the first moof in DASH Media Segments. They're a strong signal
+				// that this is indeed an fMP4 segment, but just to be sure, let's walk the box structure to make sure
+				// it stays sane.
+				pos += size;
+			} else {
+				return false;
+			}
+		}
+
+		return false;
 	}
 
 	get name() {
@@ -169,7 +198,7 @@ export class MatroskaInputFormat extends InputFormat {
 	/** @internal */
 	protected async isSupportedEBMLOfDocType(input: Input, desiredDocType: string) {
 		let headerSlice = input._reader.requestSlice(0, MAX_HEADER_SIZE);
-		if (headerSlice instanceof Promise) headerSlice = await headerSlice;
+		if (isThenable(headerSlice)) headerSlice = await headerSlice;
 		if (!headerSlice) return false;
 
 		const varIntSize = readVarIntSize(headerSlice);
@@ -192,7 +221,7 @@ export class MatroskaInputFormat extends InputFormat {
 		}
 
 		let dataSlice = input._reader.requestSlice(headerSlice.filePos, dataSize);
-		if (dataSlice instanceof Promise) dataSlice = await dataSlice;
+		if (isThenable(dataSlice)) dataSlice = await dataSlice;
 		if (!dataSlice) return false;
 
 		const startPos = headerSlice.filePos;
@@ -295,7 +324,7 @@ export class Mp3InputFormat extends InputFormat {
 
 		while (true) {
 			let slice = input._reader.requestSlice(currentPos, ID3_V2_HEADER_SIZE);
-			if (slice instanceof Promise) slice = await slice;
+			if (isThenable(slice)) slice = await slice;
 			if (!slice) break;
 
 			const id3V2Header = readId3V2Header(slice);
@@ -315,7 +344,7 @@ export class Mp3InputFormat extends InputFormat {
 		const xingOffset = getXingOffset(firstHeader.mpegVersionId, firstHeader.channel);
 
 		let slice = input._reader.requestSlice(firstResult.startPos + xingOffset, 4);
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 		if (!slice) return false;
 
 		const word = readU32Be(slice);
@@ -376,7 +405,7 @@ export class WaveInputFormat extends InputFormat {
 	/** @internal */
 	async _canReadInput(input: Input) {
 		let slice = input._reader.requestSlice(0, 12);
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 		if (!slice) return false;
 
 		const riffType = readAscii(slice, 4);
@@ -416,7 +445,7 @@ export class OggInputFormat extends InputFormat {
 	/** @internal */
 	async _canReadInput(input: Input) {
 		let slice = input._reader.requestSlice(0, 4);
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 		if (!slice) return false;
 
 		return readAscii(slice, 4) === 'OggS';
@@ -451,7 +480,7 @@ export class FlacInputFormat extends InputFormat {
 		// There might be ID3v2 headers at the start, skip 'em
 		while (true) {
 			let slice = input._reader.requestSlice(currentPos, ID3_V2_HEADER_SIZE);
-			if (slice instanceof Promise) slice = await slice;
+			if (isThenable(slice)) slice = await slice;
 			if (!slice) break;
 
 			const id3V2Header = readId3V2Header(slice);
@@ -463,7 +492,7 @@ export class FlacInputFormat extends InputFormat {
 		}
 
 		let slice = input._reader.requestSlice(currentPos, 4);
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 		if (!slice) return false;
 
 		return readAscii(slice, 4) === 'fLaC';
@@ -498,7 +527,7 @@ export class AdtsInputFormat extends InputFormat {
 
 		while (true) {
 			let slice = input._reader.requestSlice(currentPos, ID3_V2_HEADER_SIZE);
-			if (slice instanceof Promise) slice = await slice;
+			if (isThenable(slice)) slice = await slice;
 			if (!slice) break;
 
 			const id3V2Header = readId3V2Header(slice);
@@ -514,7 +543,7 @@ export class AdtsInputFormat extends InputFormat {
 			MIN_ADTS_FRAME_HEADER_SIZE,
 			MAX_ADTS_FRAME_HEADER_SIZE,
 		);
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 		if (!slice) return false;
 
 		const firstHeader = readAdtsFrameHeader(slice);
@@ -529,7 +558,7 @@ export class AdtsInputFormat extends InputFormat {
 			MIN_ADTS_FRAME_HEADER_SIZE,
 			MAX_ADTS_FRAME_HEADER_SIZE,
 		);
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 		if (!slice) return false;
 
 		const secondHeader = readAdtsFrameHeader(slice);
@@ -573,7 +602,7 @@ export class MpegTsInputFormat extends InputFormat {
 	async _canReadInput(input: Input) {
 		const lengthToCheck = TS_PACKET_SIZE + 16 + 1;
 		let slice = input._reader.requestSlice(0, lengthToCheck);
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 		if (!slice) return false;
 
 		const bytes = readBytes(slice, lengthToCheck);
@@ -618,7 +647,7 @@ export class HlsInputFormat extends InputFormat {
 	/** @internal */
 	async _canReadInput(input: Input) {
 		let slice = input._reader.requestSlice(0, 7);
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 		if (!slice) return false;
 
 		const isM3u8 = readAscii(slice, 7) === '#EXTM3U';

src/input-track.ts

@@ -12,7 +12,16 @@ import { customAudioDecoders, customVideoDecoders } from './custom-coder';
 import { Input } from './input';
 import { Logging } from './logging';
 import { EncodedPacketSink, PacketRetrievalOptions } from './media-sink';
-import { assert, MaybePromise, Rational, Rotation, roundToDivisor, simplifyRational } from './misc';
+import {
+	assert,
+	isNumber,
+	isThenable,
+	MaybePromise,
+	Rational,
+	Rotation,
+	roundToDivisor,
+	simplifyRational,
+} from './misc';
 import { TrackType } from './output';
 import { EncodedPacket, PacketType } from './packet';
 import { TrackDisposition } from './metadata';
@@ -528,7 +537,7 @@ export abstract class InputTrack {
 }
 
 const requireSync = <T>(value: MaybePromise<T>, getterName: string, asyncName: string): T => {
-	if (value instanceof Promise) {
+	if (isThenable(value)) {
 		throw new Error(
 			`'${getterName}' is deprecated and not available synchronously for this track. Use the preferred`
 			+ ` '${asyncName}()' instead.`,
@@ -554,7 +563,7 @@ const toValidatedPredicate = <T extends InputTrack>(
 				};
 
 				const result = predicate(track);
-				if (result instanceof Promise) {
+				if (isThenable(result)) {
 					return result.then(handle);
 				}
 				return handle(result);
@@ -869,7 +878,7 @@ export class InputVideoTrack extends InputTrack {
 		}
 		if (
 			options.targetPacketCount !== undefined
-			&& (!Number.isFinite(options.targetPacketCount) || options.targetPacketCount < 0)
+			&& (!isNumber(options.targetPacketCount) || options.targetPacketCount < 0)
 		) {
 			throw new TypeError('options.targetPacketCount must be a non-negative number.');
 		}
@@ -1225,7 +1234,7 @@ export const toValidatedInputTrackQuery = <T extends InputTrack>(
 					};
 
 					const result = query.filter!(track);
-					if (result instanceof Promise) {
+					if (isThenable(result)) {
 						return result.then(handle);
 					} else {
 						return handle(result);
@@ -1248,7 +1257,7 @@ export const toValidatedInputTrackQuery = <T extends InputTrack>(
 					};
 
 					const result = query.sortBy!(track);
-					if (result instanceof Promise) {
+					if (isThenable(result)) {
 						return result.then(handle);
 					} else {
 						return handle(result);
@@ -1274,7 +1283,7 @@ export const mergeInputTrackQueries = <T extends InputTrack>(
 						return queryB?.filter?.(track) ?? true;
 					};
 
-					if (resultA instanceof Promise) {
+					if (isThenable(resultA)) {
 						return resultA.then(handleResultA);
 					} else {
 						return handleResultA(resultA);
@@ -1294,7 +1303,7 @@ export const mergeInputTrackQueries = <T extends InputTrack>(
 						];
 					};
 
-					if (resultA instanceof Promise || resultB instanceof Promise) {
+					if (isThenable(resultA) || isThenable(resultB)) {
 						return Promise.all([resultA, resultB]).then(([resultA, resultB]) => {
 							return join(resultA, resultB);
 						});
@@ -1313,7 +1322,7 @@ export const queryInputTracks = async <T extends InputTrack>(
 	let matched = tracks;
 	if (query?.filter) {
 		const filterMatches = tracks.map(t => query.filter!(t));
-		const hasAsyncFilter = filterMatches.some(x => x instanceof Promise);
+		const hasAsyncFilter = filterMatches.some(x => isThenable(x));
 		if (hasAsyncFilter) {
 			// eslint-disable-next-line @typescript-eslint/await-thenable
 			const resolvedFilterMatches = await Promise.all(filterMatches);
@@ -1328,7 +1337,7 @@ export const queryInputTracks = async <T extends InputTrack>(
 	}
 
 	const sortValues = matched.map(t => query.sortBy!(t));
-	const hasAsyncSort = sortValues.some(x => x instanceof Promise);
+	const hasAsyncSort = sortValues.some(x => isThenable(x));
 	const resolvedSortValues = hasAsyncSort
 		// eslint-disable-next-line @typescript-eslint/await-thenable
 		? await Promise.all(sortValues)

src/isobmff/isobmff-boxes.ts

@@ -16,7 +16,7 @@ import {
 	COLOR_PRIMARIES_MAP,
 	TRANSFER_CHARACTERISTICS_MAP,
 	MATRIX_COEFFICIENTS_MAP,
-	colorSpaceIsComplete,
+	colorSpaceIsEmpty,
 	UNDETERMINED_LANGUAGE,
 	assertNever,
 	keyValueIterator,
@@ -750,7 +750,9 @@ export const videoSampleDescription = (
 ], [
 	VIDEO_CODEC_TO_CONFIGURATION_BOX[trackData.track.source._codec]?.(trackData) ?? null,
 	pasp(trackData),
-	colorSpaceIsComplete(trackData.info.decoderConfig.colorSpace) ? colr(trackData) : null,
+	colorSpaceIsEmpty(trackData.info.decoderConfig.colorSpace)
+		? null
+		: colr(trackData),
 ]);
 
 /** Pixel Aspect Ratio Box: Specifies pixel width:height spacing for non-square pixels. */
@@ -766,15 +768,28 @@ export const pasp = (trackData: IsobmffVideoTrackData) => {
 };
 
 /** Colour Information Box: Specifies the color space of the video. */
-export const colr = (trackData: IsobmffVideoTrackData) => box('colr', [
-	ascii(trackData.muxer.isQuickTime ? 'nclc' : 'nclx'), // Colour type
-	u16(COLOR_PRIMARIES_MAP[trackData.info.decoderConfig.colorSpace!.primaries!]), // Colour primaries
-	u16(TRANSFER_CHARACTERISTICS_MAP[trackData.info.decoderConfig.colorSpace!.transfer!]), // Transfer characteristics
-	u16(MATRIX_COEFFICIENTS_MAP[trackData.info.decoderConfig.colorSpace!.matrix!]), // Matrix coefficients
-	trackData.muxer.isQuickTime
-		? [] // Doesn't have it
-		: u8((trackData.info.decoderConfig.colorSpace!.fullRange ? 1 : 0) << 7), // Full range flag
-]);
+export const colr = (trackData: IsobmffVideoTrackData) => {
+	const colorSpace = trackData.info.decoderConfig.colorSpace;
+
+	return box('colr', [
+		// Colour type
+		ascii(trackData.muxer.isQuickTime ? 'nclc' : 'nclx'),
+
+		// Colour primaries
+		u16(colorSpace?.primaries != null ? COLOR_PRIMARIES_MAP[colorSpace.primaries] : 2),
+
+		// Transfer characteristics
+		u16(colorSpace?.transfer != null ? TRANSFER_CHARACTERISTICS_MAP[colorSpace.transfer] : 2),
+
+		// Matrix coefficients
+		u16(colorSpace?.matrix != null ? MATRIX_COEFFICIENTS_MAP[colorSpace.matrix] : 2),
+
+		// Full range flag
+		trackData.muxer.isQuickTime
+			? [] // Doesn't have it
+			: u8((colorSpace?.fullRange ? 1 : 0) << 7),
+	]);
+};
 
 /** AVC Configuration Box: Provides additional information to the decoder. */
 export const avcC = (trackData: IsobmffVideoTrackData) => trackData.info.decoderConfig && box('avcC', [
@@ -811,17 +826,17 @@ export const vpcC = (trackData: IsobmffVideoTrackData) => {
 		? Number(parts[5])
 		: decoderConfig.colorSpace?.primaries
 			? COLOR_PRIMARIES_MAP[decoderConfig.colorSpace.primaries]
-			: 2; // Default to undetermined
+			: 1;
 	const transferCharacteristics = parts[6]
 		? Number(parts[6])
 		: decoderConfig.colorSpace?.transfer
 			? TRANSFER_CHARACTERISTICS_MAP[decoderConfig.colorSpace.transfer]
-			: 2;
+			: 1;
 	const matrixCoefficients = parts[7]
 		? Number(parts[7])
 		: decoderConfig.colorSpace?.matrix
 			? MATRIX_COEFFICIENTS_MAP[decoderConfig.colorSpace.matrix]
-			: 2;
+			: 1;
 
 	return fullBox('vpcC', 1, 0, [
 		u8(profile), // Profile

src/isobmff/isobmff-demuxer.ts

@@ -14,6 +14,7 @@ import {
 	DTS_FOURCCS,
 	DtsFourCc,
 	extractAudioCodecString,
+	extractColorSpace,
 	extractVideoCodecString,
 	MediaCodec,
 	OPUS_SAMPLE_RATE,
@@ -26,12 +27,16 @@ import {
 } from '../codec';
 import {
 	Av1CodecInfo,
+	av1CodecInfoHasColorInfo,
 	AvcDecoderConfigurationRecord,
 	extractAv1CodecInfoFromPacket,
+	extractProresCodecInfoFromPacket,
 	extractVp9CodecInfoFromPacket,
 	FlacBlockType,
 	HevcDecoderConfigurationRecord,
+	ProresCodecInfo,
 	Vp9CodecInfo,
+	vp9CodecInfoHasColorInfo,
 	parseEac3Config,
 	getEac3SampleRate,
 	getEac3ChannelCount,
@@ -39,6 +44,8 @@ import {
 	parseDtsSpecificBox,
 	DTS_SPECIFIC_BOX_SIZE,
 	AC3_ACMOD_CHANNEL_COUNTS,
+	extractAvcDecoderConfigurationRecord,
+	extractHevcDecoderConfigurationRecord,
 } from '../codec-data';
 import { Demuxer } from '../demuxer';
 import { Input } from '../input';
@@ -56,6 +63,7 @@ import {
 	COLOR_PRIMARIES_MAP_INVERSE,
 	findLastIndex,
 	isIso639Dash2LanguageCode,
+	isThenable,
 	last,
 	MATRIX_COEFFICIENTS_MAP_INVERSE,
 	normalizeRotation,
@@ -69,6 +77,8 @@ import {
 	roundIfAlmostInteger,
 	hexStringToBytes,
 	HEX_STRING_REGEX,
+	EMPTY_COLOR_SPACE,
+	colorSpaceIsComplete,
 } from '../misc';
 import { EncodedPacket, PLACEHOLDER_DATA } from '../packet';
 import { buildIsobmffMimeType, parsePsshBoxContents, psshBoxesAreEqual, PsshBox } from './isobmff-misc';
@@ -148,12 +158,13 @@ type InternalTrack = {
 		squarePixelHeight: number;
 		codec: VideoCodec | null;
 		codecDescription: Uint8Array | null;
-		colorSpace: VideoColorSpaceInit | null;
+		colorSpace: VideoColorSpaceInit;
 		avcType: 1 | 3 | null;
 		avcCodecInfo: AvcDecoderConfigurationRecord | null;
 		hevcCodecInfo: HevcDecoderConfigurationRecord | null;
 		vp9CodecInfo: Vp9CodecInfo | null;
 		av1CodecInfo: Av1CodecInfo | null;
+		proresCodecInfo: ProresCodecInfo | null;
 		proresFormat: ProresFourCc | null;
 	};
 } | {
@@ -362,7 +373,7 @@ export class IsobmffDemuxer extends Demuxer {
 
 			while (true) {
 				let slice = this.reader.requestSliceRange(currentPos, MIN_BOX_HEADER_SIZE, MAX_BOX_HEADER_SIZE);
-				if (slice instanceof Promise) slice = await slice;
+				if (isThenable(slice)) slice = await slice;
 				if (!slice) break;
 
 				const startPos = currentPos;
@@ -378,7 +389,7 @@ export class IsobmffDemuxer extends Demuxer {
 					// Found moov, load it
 
 					let moovSlice = this.reader.requestSlice(slice.filePos, boxInfo.contentSize);
-					if (moovSlice instanceof Promise) moovSlice = await moovSlice;
+					if (isThenable(moovSlice)) moovSlice = await moovSlice;
 					if (!moovSlice) break;
 
 					this.moovSlice = moovSlice;
@@ -429,7 +440,7 @@ export class IsobmffDemuxer extends Demuxer {
 
 				// The last 4 bytes may contain the size of the mfra box at the end of the file
 				let lastWordSlice = this.reader.requestSlice(this.reader.fileSize - 4, 4);
-				if (lastWordSlice instanceof Promise) lastWordSlice = await lastWordSlice;
+				if (isThenable(lastWordSlice)) lastWordSlice = await lastWordSlice;
 				assert(lastWordSlice);
 
 				const lastWord = readU32Be(lastWordSlice);
@@ -441,7 +452,7 @@ export class IsobmffDemuxer extends Demuxer {
 						MIN_BOX_HEADER_SIZE,
 						MAX_BOX_HEADER_SIZE,
 					);
-					if (mfraHeaderSlice instanceof Promise) mfraHeaderSlice = await mfraHeaderSlice;
+					if (isThenable(mfraHeaderSlice)) mfraHeaderSlice = await mfraHeaderSlice;
 
 					if (mfraHeaderSlice) {
 						const boxInfo = readBoxHeader(mfraHeaderSlice);
@@ -449,7 +460,7 @@ export class IsobmffDemuxer extends Demuxer {
 						if (boxInfo && boxInfo.name === 'mfra') {
 							// We found the mfra box, allowing for much better random access. Let's parse it.
 							let mfraSlice = this.reader.requestSlice(mfraHeaderSlice.filePos, boxInfo.contentSize);
-							if (mfraSlice instanceof Promise) mfraSlice = await mfraSlice;
+							if (isThenable(mfraSlice)) mfraSlice = await mfraSlice;
 
 							if (mfraSlice) {
 								this.readContiguousBoxes(mfraSlice);
@@ -676,14 +687,14 @@ export class IsobmffDemuxer extends Demuxer {
 		}
 
 		let headerSlice = this.reader.requestSliceRange(startPos, MIN_BOX_HEADER_SIZE, MAX_BOX_HEADER_SIZE);
-		if (headerSlice instanceof Promise) headerSlice = await headerSlice;
+		if (isThenable(headerSlice)) headerSlice = await headerSlice;
 		assert(headerSlice);
 
 		const moofBoxInfo = readBoxHeader(headerSlice);
 		assert(moofBoxInfo?.name === 'moof');
 
 		let entireSlice = this.reader.requestSlice(startPos, moofBoxInfo.totalSize);
-		if (entireSlice instanceof Promise) entireSlice = await entireSlice;
+		if (isThenable(entireSlice)) entireSlice = await entireSlice;
 		assert(entireSlice);
 
 		this.traverseBox(entireSlice);
@@ -1024,12 +1035,13 @@ export class IsobmffDemuxer extends Demuxer {
 						squarePixelHeight: -1,
 						codec: null,
 						codecDescription: null,
-						colorSpace: null,
+						colorSpace: { ...EMPTY_COLOR_SPACE },
 						avcType: null,
 						avcCodecInfo: null,
 						hevcCodecInfo: null,
 						vp9CodecInfo: null,
 						av1CodecInfo: null,
+						proresCodecInfo: null,
 						proresFormat: null,
 					};
 				} else if (handlerType === 'soun') {
@@ -1390,6 +1402,11 @@ export class IsobmffDemuxer extends Demuxer {
 				}
 				assert(track.info);
 
+				if (boxInfo.contentSize === 0) {
+					// avcC box is empty, let's treat this like an Annex B stream
+					break;
+				}
+
 				track.info.codecDescription = readBytes(slice, boxInfo.contentSize);
 			}; break;
 
@@ -1400,6 +1417,11 @@ export class IsobmffDemuxer extends Demuxer {
 				}
 				assert(track.info);
 
+				if (boxInfo.contentSize === 0) {
+					// hvcC box is empty, let's treat this like an Annex B stream
+					break;
+				}
+
 				track.info.codecDescription = readBytes(slice, boxInfo.contentSize);
 			}; break;
 
@@ -1459,6 +1481,12 @@ export class IsobmffDemuxer extends Demuxer {
 				// Logic from https://aomediacodec.github.io/av1-spec/av1-spec.pdf
 				const bitDepth = profile === 2 && highBitDepth ? (twelveBit ? 12 : 10) : (highBitDepth ? 10 : 8);
 
+				slice.skip(1); // Reserved bits + initial presentation delay
+
+				// Parse config OBUs if there are any
+				const configObus = readBytes(slice, boxInfo.contentSize - 4);
+				const configObuInfo = extractAv1CodecInfoFromPacket(configObus);
+
 				track.info.av1CodecInfo = {
 					profile,
 					level,
@@ -1468,6 +1496,10 @@ export class IsobmffDemuxer extends Demuxer {
 					chromaSubsamplingX,
 					chromaSubsamplingY,
 					chromaSamplePosition,
+					videoFullRangeFlag: configObuInfo?.videoFullRangeFlag ?? 0,
+					colourPrimaries: configObuInfo?.colourPrimaries ?? 2,
+					transferCharacteristics: configObuInfo?.transferCharacteristics ?? 2,
+					matrixCoefficients: configObuInfo?.matrixCoefficients ?? 2,
 				};
 			}; break;
 
@@ -1596,11 +1628,11 @@ export class IsobmffDemuxer extends Demuxer {
 					if (track.info.codec === 'aac') {
 						// Let's try to deduce more accurate values directly from the AudioSpecificConfig:
 						const audioSpecificConfig = parseAacAudioSpecificConfig(track.info.codecDescription);
-						if (audioSpecificConfig.numberOfChannels !== null) {
-							track.info.numberOfChannels = audioSpecificConfig.numberOfChannels;
+						if (audioSpecificConfig.outputNumberOfChannels !== null) {
+							track.info.numberOfChannels = audioSpecificConfig.outputNumberOfChannels;
 						}
-						if (audioSpecificConfig.sampleRate !== null) {
-							track.info.sampleRate = audioSpecificConfig.sampleRate;
+						if (audioSpecificConfig.outputSampleRate !== null) {
+							track.info.sampleRate = audioSpecificConfig.outputSampleRate;
 						}
 					}
 				}
@@ -3132,24 +3164,31 @@ abstract class IsobmffTrackBacking implements InputTrackBacking {
 				sampleInfo.sampleOffset,
 				sampleInfo.sampleSize,
 			);
-			if (slice instanceof Promise) slice = await slice;
+			if (isThenable(slice)) slice = await slice;
 			if (!slice) {
 				return null; // Data is outside
 			}
 
 			data = readBytes(slice, sampleInfo.sampleSize);
 
-			if (this.internalTrack.encryptionAuxInfo) {
-				assert(this.internalTrack.encryptionInfo);
+			if (this.internalTrack.encryptionInfo) {
+				let sampleEncryption: SampleEncryptionInfo | null = null;
 
-				const entries = await resolveEncryptionAuxInfo(
-					this.internalTrack.demuxer.reader,
-					this.internalTrack.encryptionInfo,
-					this.internalTrack.encryptionAuxInfo,
-				);
+				if (this.internalTrack.encryptionAuxInfo) {
+					const entries = await resolveEncryptionAuxInfo(
+						this.internalTrack.demuxer.reader,
+						this.internalTrack.encryptionInfo,
+						this.internalTrack.encryptionAuxInfo,
+					);
 
-				if (sampleIndex < entries.length) {
-					data = await decryptSample(this.internalTrack, entries[sampleIndex]!, data, null);
+					if (sampleIndex < entries.length) {
+						sampleEncryption = entries[sampleIndex]!;
+					}
+				}
+
+				sampleEncryption ??= getDefaultSampleEncryption(this.internalTrack.encryptionInfo);
+				if (sampleEncryption) {
+					data = await decryptSample(this.internalTrack, sampleEncryption, data, null);
 				}
 			}
 		}
@@ -3188,15 +3227,19 @@ abstract class IsobmffTrackBacking implements InputTrackBacking {
 				fragmentSample.byteOffset,
 				fragmentSample.byteSize,
 			);
-			if (slice instanceof Promise) slice = await slice;
+			if (isThenable(slice)) slice = await slice;
 			if (!slice) {
 				return null; // Data is outside
 			}
 
 			data = readBytes(slice, fragmentSample.byteSize);
 
-			if (fragmentSample.encryption) {
-				data = await decryptSample(this.internalTrack, fragmentSample.encryption, data, fragment);
+			if (this.internalTrack.encryptionInfo) {
+				const sampleEncryption = fragmentSample.encryption
+					?? getDefaultSampleEncryption(this.internalTrack.encryptionInfo);
+				if (sampleEncryption) {
+					data = await decryptSample(this.internalTrack, sampleEncryption, data, fragment);
+				}
 			}
 		}
 
@@ -3298,7 +3341,7 @@ abstract class IsobmffTrackBacking implements InputTrackBacking {
 
 			// Load the header
 			let slice = demuxer.reader.requestSliceRange(currentPos, MIN_BOX_HEADER_SIZE, MAX_BOX_HEADER_SIZE);
-			if (slice instanceof Promise) slice = await slice;
+			if (isThenable(slice)) slice = await slice;
 			if (!slice) break;
 
 			const boxStartPos = currentPos;
@@ -3386,11 +3429,16 @@ class IsobmffVideoTrackBacking extends IsobmffTrackBacking implements InputVideo
 	}
 
 	async getColorSpace(): Promise<VideoColorSpaceInit> {
+		const decoderConfig = await this.getDecoderConfig();
+		if (!decoderConfig) {
+			return this.internalTrack.info.colorSpace;
+		}
+
 		return {
-			primaries: this.internalTrack.info.colorSpace?.primaries,
-			transfer: this.internalTrack.info.colorSpace?.transfer,
-			matrix: this.internalTrack.info.colorSpace?.matrix,
-			fullRange: this.internalTrack.info.colorSpace?.fullRange,
+			primaries: decoderConfig.colorSpace?.primaries,
+			transfer: decoderConfig.colorSpace?.transfer,
+			matrix: decoderConfig.colorSpace?.matrix,
+			fullRange: decoderConfig.colorSpace?.fullRange,
 		};
 	}
 
@@ -3407,12 +3455,64 @@ class IsobmffVideoTrackBacking extends IsobmffTrackBacking implements InputVideo
 		}
 
 		return this.decoderConfigPromise ??= (async (): Promise<VideoDecoderConfig> => {
-			if (this.internalTrack.info.codec === 'vp9' && !this.internalTrack.info.vp9CodecInfo) {
+			if (this.internalTrack.info.codec === 'avc' && !this.internalTrack.info.codecDescription) {
 				const firstPacket = await this.getFirstPacket({});
-				this.internalTrack.info.vp9CodecInfo = firstPacket && extractVp9CodecInfoFromPacket(firstPacket.data);
-			} else if (this.internalTrack.info.codec === 'av1' && !this.internalTrack.info.av1CodecInfo) {
+				this.internalTrack.info.avcCodecInfo
+					= firstPacket && extractAvcDecoderConfigurationRecord(firstPacket.data);
+			} else if (this.internalTrack.info.codec === 'hevc' && !this.internalTrack.info.codecDescription) {
 				const firstPacket = await this.getFirstPacket({});
-				this.internalTrack.info.av1CodecInfo = firstPacket && extractAv1CodecInfoFromPacket(firstPacket.data);
+				this.internalTrack.info.hevcCodecInfo
+					= firstPacket && extractHevcDecoderConfigurationRecord(firstPacket.data);
+			} else if (
+				this.internalTrack.info.codec === 'vp9'
+				&& (
+					!this.internalTrack.info.vp9CodecInfo
+					// The codec info extracted from vpcC may claim the color space is "undefined"
+					|| !vp9CodecInfoHasColorInfo(this.internalTrack.info.vp9CodecInfo)
+				)
+			) {
+				const firstPacket = await this.getFirstPacket({});
+				const packetInfo = firstPacket && extractVp9CodecInfoFromPacket(firstPacket.data);
+
+				if (packetInfo) {
+					this.internalTrack.info.vp9CodecInfo = {
+						...(this.internalTrack.info.vp9CodecInfo ?? packetInfo),
+						videoFullRangeFlag: packetInfo.videoFullRangeFlag,
+						colourPrimaries: packetInfo.colourPrimaries,
+						transferCharacteristics: packetInfo.transferCharacteristics,
+						matrixCoefficients: packetInfo.matrixCoefficients,
+					};
+				}
+			} else if (
+				this.internalTrack.info.codec === 'av1'
+				&& (
+					!this.internalTrack.info.av1CodecInfo
+					// The codec info extracted from av1C may not contain color space information
+					|| !av1CodecInfoHasColorInfo(
+						this.internalTrack.info.av1CodecInfo,
+					)
+				)
+			) {
+				const firstPacket = await this.getFirstPacket({});
+				const packetInfo = firstPacket && extractAv1CodecInfoFromPacket(firstPacket.data);
+
+				if (packetInfo) {
+					this.internalTrack.info.av1CodecInfo = packetInfo;
+				}
+			} else if (this.internalTrack.info.codec === 'prores' && !this.internalTrack.info.proresCodecInfo) {
+				const firstPacket = await this.getFirstPacket({});
+				this.internalTrack.info.proresCodecInfo
+					= firstPacket && extractProresCodecInfoFromPacket(firstPacket.data);
+			}
+
+			if (!colorSpaceIsComplete(this.internalTrack.info.colorSpace)) {
+				const colorSpace = extractColorSpace(this.internalTrack.info);
+
+				// Fill the missing values
+				this.internalTrack.info.colorSpace.primaries ??= colorSpace.primaries;
+				this.internalTrack.info.colorSpace.transfer ??= colorSpace.transfer;
+				this.internalTrack.info.colorSpace.matrix ??= colorSpace.matrix;
+				this.internalTrack.info.colorSpace.fullRange ??= colorSpace.fullRange;
 			}
 
 			const config: VideoDecoderConfig = {
@@ -3420,7 +3520,7 @@ class IsobmffVideoTrackBacking extends IsobmffTrackBacking implements InputVideo
 				codedWidth: this.internalTrack.info.width,
 				codedHeight: this.internalTrack.info.height,
 				description: this.internalTrack.info.codecDescription ?? undefined,
-				colorSpace: this.internalTrack.info.colorSpace ?? undefined,
+				colorSpace: this.internalTrack.info.colorSpace,
 			};
 
 			if (
@@ -3717,7 +3817,7 @@ const resolveEncryptionAuxInfo = async (
 	}
 
 	let slice = reader.requestSlice(aux.offset, totalSize);
-	if (slice instanceof Promise) slice = await slice;
+	if (isThenable(slice)) slice = await slice;
 	if (!slice) {
 		throw new Error('Failed to read auxiliary encryption info.');
 	}
@@ -3759,6 +3859,17 @@ const resolveEncryptionAuxInfo = async (
 	return entries;
 };
 
+const getDefaultSampleEncryption = (encryptionInfo: TrackEncryptionInfo): SampleEncryptionInfo | null => {
+	if (!encryptionInfo.defaultConstantIv) {
+		return null;
+	}
+
+	return {
+		iv: encryptionInfo.defaultConstantIv,
+		subsamples: null,
+	};
+};
+
 const decryptSample = async (
 	track: InternalTrack,
 	sampleEncryption: SampleEncryptionInfo,

src/isobmff/isobmff-muxer.ts

@@ -525,8 +525,8 @@ export class IsobmffMuxer extends Muxer {
 
 			decoderConfig.description = buildAacAudioSpecificConfig({
 				objectType: adtsFrame.objectType,
-				sampleRate,
-				numberOfChannels,
+				outputSampleRate: sampleRate,
+				outputNumberOfChannels: numberOfChannels,
 			});
 			requiresAdtsStripping = true;
 		}

src/matroska/ebml.ts

@@ -7,7 +7,7 @@
  */
 
 import { MediaCodec } from '../codec';
-import { assert, assertNever, textDecoder, textEncoder } from '../misc';
+import { assert, assertNever, isThenable, textDecoder, textEncoder } from '../misc';
 import { FileSlice, readBytes, Reader, readF32Be, readF64Be, readU8 } from '../reader';
 import { Writer } from '../writer';
 
@@ -688,7 +688,7 @@ export const searchForNextElementId = async (
 
 	while (until === null || currentPos < until) {
 		let slice = reader.requestSliceRange(currentPos, MIN_HEADER_SIZE, MAX_HEADER_SIZE);
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 		if (!slice) break;
 
 		const elementHeader = readElementHeader(slice);
@@ -716,7 +716,7 @@ export const resync = async (reader: Reader, startPos: number, ids: EBMLId[], un
 
 	while (currentPos < until) {
 		let slice = reader.requestSliceRange(currentPos, 0, Math.min(CHUNK_SIZE, until - currentPos));
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 		if (!slice) break;
 		if (slice.length < MAX_VAR_INT_SIZE) break;
 

src/matroska/matroska-demuxer.ts

@@ -12,6 +12,7 @@ import {
 	extractDtsFourCcFromPacket,
 	extractAvcDecoderConfigurationRecord,
 	extractHevcDecoderConfigurationRecord,
+	extractProresCodecInfoFromPacket,
 	extractVp9CodecInfoFromPacket,
 } from '../codec-data';
 import {
@@ -19,6 +20,7 @@ import {
 	AudioCodec,
 	DtsFourCc,
 	extractAudioCodecString,
+	extractColorSpace,
 	extractVideoCodecString,
 	MediaCodec,
 	OPUS_SAMPLE_RATE,
@@ -40,8 +42,11 @@ import {
 	assert,
 	binarySearchLessOrEqual,
 	COLOR_PRIMARIES_MAP_INVERSE,
+	colorSpaceIsComplete,
+	EMPTY_COLOR_SPACE,
 	findLastIndex,
 	isIso639Dash2LanguageCode,
+	isThenable,
 	last,
 	MATRIX_COEFFICIENTS_MAP_INVERSE,
 	normalizeRotation,
@@ -219,7 +224,7 @@ type InternalTrack = {
 			rotation: Rotation;
 			codec: VideoCodec | null;
 			codecDescription: Uint8Array | null;
-			colorSpace: VideoColorSpaceInit | null;
+			colorSpace: VideoColorSpaceInit;
 			alphaMode: boolean;
 			proresFormat: ProresFourCc | null;
 		}
@@ -334,7 +339,7 @@ export class MatroskaDemuxer extends Demuxer {
 			// Loop over all top-level elements in the file
 			while (true) {
 				let slice = this.reader.requestSliceRange(currentPos, MIN_HEADER_SIZE, MAX_HEADER_SIZE);
-				if (slice instanceof Promise) slice = await slice;
+				if (isThenable(slice)) slice = await slice;
 				if (!slice) break;
 
 				const header = readElementHeader(slice);
@@ -350,7 +355,7 @@ export class MatroskaDemuxer extends Demuxer {
 					assertDefinedSize(size);
 
 					let slice = this.reader.requestSlice(dataStartPos, size);
-					if (slice instanceof Promise) slice = await slice;
+					if (isThenable(slice)) slice = await slice;
 					if (!slice) break;
 
 					this.readContiguousElements(slice);
@@ -433,7 +438,7 @@ export class MatroskaDemuxer extends Demuxer {
 
 		while (this.currentSegment.elementEndPos === null || currentPos < this.currentSegment.elementEndPos) {
 			let slice = this.reader.requestSliceRange(currentPos, MIN_HEADER_SIZE, MAX_HEADER_SIZE);
-			if (slice instanceof Promise) slice = await slice;
+			if (isThenable(slice)) slice = await slice;
 			if (!slice) break;
 
 			const elementStartPos = currentPos;
@@ -468,7 +473,7 @@ export class MatroskaDemuxer extends Demuxer {
 				assertDefinedSize(size);
 
 				let slice = this.reader.requestSlice(dataStartPos, size);
-				if (slice instanceof Promise) slice = await slice;
+				if (isThenable(slice)) slice = await slice;
 
 				if (slice) {
 					this.readContiguousElements(slice);
@@ -484,7 +489,7 @@ export class MatroskaDemuxer extends Demuxer {
 				assertDefinedSize(size);
 
 				let slice = this.reader.requestSlice(dataStartPos, size);
-				if (slice instanceof Promise) slice = await slice;
+				if (isThenable(slice)) slice = await slice;
 
 				if (slice) {
 					this.readContiguousElements(slice);
@@ -519,7 +524,7 @@ export class MatroskaDemuxer extends Demuxer {
 					MIN_HEADER_SIZE,
 					MAX_HEADER_SIZE,
 				);
-				if (slice instanceof Promise) slice = await slice;
+				if (isThenable(slice)) slice = await slice;
 				if (!slice) continue;
 
 				const header = readElementHeader(slice);
@@ -533,7 +538,7 @@ export class MatroskaDemuxer extends Demuxer {
 				this.currentSegment[target.flag] = true;
 
 				let dataSlice = this.reader.requestSlice(slice.filePos, size);
-				if (dataSlice instanceof Promise) dataSlice = await dataSlice;
+				if (isThenable(dataSlice)) dataSlice = await dataSlice;
 				if (!dataSlice) continue;
 
 				this.readContiguousElements(dataSlice);
@@ -608,7 +613,7 @@ export class MatroskaDemuxer extends Demuxer {
 		}
 
 		let headerSlice = this.reader.requestSliceRange(startPos, MIN_HEADER_SIZE, MAX_HEADER_SIZE);
-		if (headerSlice instanceof Promise) headerSlice = await headerSlice;
+		if (isThenable(headerSlice)) headerSlice = await headerSlice;
 		assert(headerSlice);
 
 		const elementStartPos = startPos;
@@ -637,7 +642,7 @@ export class MatroskaDemuxer extends Demuxer {
 
 		// Load the entire cluster
 		let dataSlice = this.reader.requestSlice(dataStartPos, size);
-		if (dataSlice instanceof Promise) dataSlice = await dataSlice;
+		if (isThenable(dataSlice)) dataSlice = await dataSlice;
 
 		const cluster: Cluster = {
 			segment,
@@ -894,7 +899,7 @@ export class MatroskaDemuxer extends Demuxer {
 				MIN_HEADER_SIZE,
 				MAX_HEADER_SIZE,
 			);
-			if (slice instanceof Promise) slice = await slice;
+			if (isThenable(slice)) slice = await slice;
 			if (!slice) continue;
 
 			const header = readElementHeader(slice);
@@ -907,7 +912,7 @@ export class MatroskaDemuxer extends Demuxer {
 			this.currentSegment = segment;
 
 			let dataSlice = this.reader.requestSlice(slice.filePos, size);
-			if (dataSlice instanceof Promise) dataSlice = await dataSlice;
+			if (isThenable(dataSlice)) dataSlice = await dataSlice;
 			if (dataSlice) {
 				this.readContiguousElements(dataSlice);
 			}
@@ -1198,7 +1203,7 @@ export class MatroskaDemuxer extends Demuxer {
 						rotation: 0,
 						codec: null,
 						codecDescription: null,
-						colorSpace: null,
+						colorSpace: { ...EMPTY_COLOR_SPACE },
 						alphaMode: false,
 						proresFormat: null,
 					};
@@ -1362,37 +1367,39 @@ export class MatroskaDemuxer extends Demuxer {
 			case EBMLId.Colour: {
 				if (this.currentTrack?.info?.type !== 'video') break;
 
-				this.currentTrack.info.colorSpace = {};
 				this.readContiguousElements(slice.slice(dataStartPos, size));
 			}; break;
 
 			case EBMLId.MatrixCoefficients: {
-				if (this.currentTrack?.info?.type !== 'video' || !this.currentTrack.info.colorSpace) break;
+				if (this.currentTrack?.info?.type !== 'video') break;
 
 				const matrixCoefficients = readUnsignedInt(slice, size);
-				const mapped = MATRIX_COEFFICIENTS_MAP_INVERSE[matrixCoefficients] ?? null;
+				const mapped = MATRIX_COEFFICIENTS_MAP_INVERSE[matrixCoefficients];
 				this.currentTrack.info.colorSpace.matrix = mapped as VideoColorSpaceInit['matrix'];
 			}; break;
 
 			case EBMLId.Range: {
-				if (this.currentTrack?.info?.type !== 'video' || !this.currentTrack.info.colorSpace) break;
+				if (this.currentTrack?.info?.type !== 'video') break;
 
-				this.currentTrack.info.colorSpace.fullRange = readUnsignedInt(slice, size) === 2;
+				const range = readUnsignedInt(slice, size);
+				this.currentTrack.info.colorSpace.fullRange = range === 1 || range === 2
+					? range === 2
+					: undefined;
 			}; break;
 
 			case EBMLId.TransferCharacteristics: {
-				if (this.currentTrack?.info?.type !== 'video' || !this.currentTrack.info.colorSpace) break;
+				if (this.currentTrack?.info?.type !== 'video') break;
 
 				const transferCharacteristics = readUnsignedInt(slice, size);
-				const mapped = TRANSFER_CHARACTERISTICS_MAP_INVERSE[transferCharacteristics] ?? null;
+				const mapped = TRANSFER_CHARACTERISTICS_MAP_INVERSE[transferCharacteristics];
 				this.currentTrack.info.colorSpace.transfer = mapped as VideoColorSpaceInit['transfer'];
 			}; break;
 
 			case EBMLId.Primaries: {
-				if (this.currentTrack?.info?.type !== 'video' || !this.currentTrack.info.colorSpace) break;
+				if (this.currentTrack?.info?.type !== 'video') break;
 
 				const primaries = readUnsignedInt(slice, size);
-				const mapped = COLOR_PRIMARIES_MAP_INVERSE[primaries] ?? null;
+				const mapped = COLOR_PRIMARIES_MAP_INVERSE[primaries];
 				this.currentTrack.info.colorSpace.primaries = mapped as VideoColorSpaceInit['primaries'];
 			}; break;
 
@@ -2347,7 +2354,7 @@ abstract class MatroskaTrackBacking implements InputTrackBacking {
 
 			// Load the header
 			let slice = demuxer.reader.requestSliceRange(currentPos, MIN_HEADER_SIZE, MAX_HEADER_SIZE);
-			if (slice instanceof Promise) slice = await slice;
+			if (isThenable(slice)) slice = await slice;
 			if (!slice) break;
 
 			const elementStartPos = currentPos;
@@ -2418,7 +2425,7 @@ abstract class MatroskaTrackBacking implements InputTrackBacking {
 				// the first segment.
 
 				let slice = demuxer.reader.requestSliceRange(endPos, MIN_HEADER_SIZE, MAX_HEADER_SIZE);
-				if (slice instanceof Promise) slice = await slice;
+				if (isThenable(slice)) slice = await slice;
 				if (!slice) break;
 
 				const elementId = readElementId(slice);
@@ -2489,11 +2496,16 @@ class MatroskaVideoTrackBacking extends MatroskaTrackBacking implements InputVid
 	}
 
 	async getColorSpace(): Promise<VideoColorSpaceInit> {
+		const decoderConfig = await this.getDecoderConfig();
+		if (!decoderConfig) {
+			return this.internalTrack.info.colorSpace;
+		}
+
 		return {
-			primaries: this.internalTrack.info.colorSpace?.primaries,
-			transfer: this.internalTrack.info.colorSpace?.transfer,
-			matrix: this.internalTrack.info.colorSpace?.matrix,
-			fullRange: this.internalTrack.info.colorSpace?.fullRange,
+			primaries: decoderConfig.colorSpace?.primaries,
+			transfer: decoderConfig.colorSpace?.transfer,
+			matrix: decoderConfig.colorSpace?.matrix,
+			fullRange: decoderConfig.colorSpace?.fullRange,
 		};
 	}
 
@@ -2516,6 +2528,7 @@ class MatroskaVideoTrackBacking extends MatroskaTrackBacking implements InputVid
 			const needsPacketForAdditionalInfo
 				= this.internalTrack.info.codec === 'vp9'
 					|| this.internalTrack.info.codec === 'av1'
+					|| this.internalTrack.info.codec === 'prores'
 					// Packets are in Annex B format:
 					|| (this.internalTrack.info.codec === 'avc' && !this.internalTrack.info.codecDescription)
 					// Packets are in Annex B format:
@@ -2525,32 +2538,47 @@ class MatroskaVideoTrackBacking extends MatroskaTrackBacking implements InputVid
 				firstPacket = await this.getFirstPacket({});
 			}
 
+			const codecInfo = {
+				width: this.internalTrack.info.width,
+				height: this.internalTrack.info.height,
+				codec: this.internalTrack.info.codec,
+				codecDescription: this.internalTrack.info.codecDescription,
+				colorSpace: this.internalTrack.info.colorSpace,
+				avcType: 1 as const, // We don't know better (or do we?) so just assume 'avc1'
+				avcCodecInfo: this.internalTrack.info.codec === 'avc' && firstPacket
+					? extractAvcDecoderConfigurationRecord(firstPacket.data)
+					: null,
+				hevcCodecInfo: this.internalTrack.info.codec === 'hevc' && firstPacket
+					? extractHevcDecoderConfigurationRecord(firstPacket.data)
+					: null,
+				vp9CodecInfo: this.internalTrack.info.codec === 'vp9' && firstPacket
+					? extractVp9CodecInfoFromPacket(firstPacket.data)
+					: null,
+				av1CodecInfo: this.internalTrack.info.codec === 'av1' && firstPacket
+					? extractAv1CodecInfoFromPacket(firstPacket.data)
+					: null,
+				proresCodecInfo: this.internalTrack.info.codec === 'prores' && firstPacket
+					? extractProresCodecInfoFromPacket(firstPacket.data)
+					: null,
+				proresFormat: this.internalTrack.info.proresFormat,
+			};
+
+			if (!colorSpaceIsComplete(this.internalTrack.info.colorSpace)) {
+				const colorSpace = extractColorSpace(codecInfo);
+
+				// Fill the missing values
+				this.internalTrack.info.colorSpace.primaries ??= colorSpace.primaries;
+				this.internalTrack.info.colorSpace.transfer ??= colorSpace.transfer;
+				this.internalTrack.info.colorSpace.matrix ??= colorSpace.matrix;
+				this.internalTrack.info.colorSpace.fullRange ??= colorSpace.fullRange;
+			}
+
 			const config: VideoDecoderConfig = {
-				codec: extractVideoCodecString({
-					width: this.internalTrack.info.width,
-					height: this.internalTrack.info.height,
-					codec: this.internalTrack.info.codec,
-					codecDescription: this.internalTrack.info.codecDescription,
-					colorSpace: this.internalTrack.info.colorSpace,
-					avcType: 1, // We don't know better (or do we?) so just assume 'avc1'
-					avcCodecInfo: this.internalTrack.info.codec === 'avc' && firstPacket
-						? extractAvcDecoderConfigurationRecord(firstPacket.data)
-						: null,
-					hevcCodecInfo: this.internalTrack.info.codec === 'hevc' && firstPacket
-						? extractHevcDecoderConfigurationRecord(firstPacket.data)
-						: null,
-					vp9CodecInfo: this.internalTrack.info.codec === 'vp9' && firstPacket
-						? extractVp9CodecInfoFromPacket(firstPacket.data)
-						: null,
-					av1CodecInfo: this.internalTrack.info.codec === 'av1' && firstPacket
-						? extractAv1CodecInfoFromPacket(firstPacket.data)
-						: null,
-					proresFormat: this.internalTrack.info.proresFormat,
-				}),
+				codec: extractVideoCodecString(codecInfo),
 				codedWidth: this.internalTrack.info.width,
 				codedHeight: this.internalTrack.info.height,
 				description: this.internalTrack.info.codecDescription ?? undefined,
-				colorSpace: this.internalTrack.info.colorSpace ?? undefined,
+				colorSpace: this.internalTrack.info.colorSpace,
 			};
 
 			if (

src/matroska/matroska-muxer.ts

@@ -14,7 +14,7 @@ import {
 	UNDETERMINED_LANGUAGE,
 	assert,
 	assertNever,
-	colorSpaceIsComplete,
+	colorSpaceIsEmpty,
 	imageMimeTypeToExtension,
 	keyValueIterator,
 	normalizeRotation,
@@ -405,29 +405,31 @@ export class MatroskaMuxer extends Muxer {
 			(hasNonSquarePixelAspectRatio ? { id: EBMLId.DisplayHeight, data: trackData.info.aspectRatio!.den } : null),
 			(hasNonSquarePixelAspectRatio ? { id: EBMLId.DisplayUnit, data: 3 } : null), // 3 = display aspect ratio
 			trackData.info.alphaMode ? { id: EBMLId.AlphaMode, data: 1 } : null,
-			(colorSpaceIsComplete(colorSpace)
-				? {
+			(colorSpaceIsEmpty(colorSpace)
+				? null
+				: {
 						id: EBMLId.Colour,
 						data: [
 							{
 								id: EBMLId.MatrixCoefficients,
-								data: MATRIX_COEFFICIENTS_MAP[colorSpace.matrix],
+								data: colorSpace?.matrix != null ? MATRIX_COEFFICIENTS_MAP[colorSpace.matrix] : 2,
 							},
 							{
 								id: EBMLId.TransferCharacteristics,
-								data: TRANSFER_CHARACTERISTICS_MAP[colorSpace.transfer],
+								data: colorSpace?.transfer != null
+									? TRANSFER_CHARACTERISTICS_MAP[colorSpace.transfer]
+									: 2,
 							},
 							{
 								id: EBMLId.Primaries,
-								data: COLOR_PRIMARIES_MAP[colorSpace.primaries],
+								data: colorSpace?.primaries != null ? COLOR_PRIMARIES_MAP[colorSpace.primaries] : 2,
 							},
 							{
 								id: EBMLId.Range,
-								data: colorSpace.fullRange ? 2 : 1,
+								data: colorSpace?.fullRange != null ? (colorSpace.fullRange ? 2 : 1) : 0,
 							},
 						],
-					}
-				: null),
+					}),
 			(flippedRotation
 				? {
 						id: EBMLId.Projection,
@@ -858,8 +860,8 @@ export class MatroskaMuxer extends Muxer {
 
 			decoderConfig.description = buildAacAudioSpecificConfig({
 				objectType: adtsFrame.objectType,
-				sampleRate,
-				numberOfChannels,
+				outputSampleRate: sampleRate,
+				outputNumberOfChannels: numberOfChannels,
 			});
 			requiresAdtsStripping = true;
 		}

src/media-sink.ts

@@ -29,6 +29,7 @@ import {
 	assertNever,
 	CallSerializer,
 	clamp,
+	colorSpaceIsComplete,
 	getInt24,
 	getUint24,
 	insertSorted,
@@ -38,6 +39,7 @@ import {
 	isWebKit,
 	last,
 	mapAsyncGenerator,
+	missingWebCodecsClassMessage,
 	promiseWithResolvers,
 	removeItem,
 	Rotation,
@@ -943,20 +945,42 @@ class VideoDecoderWrapper extends DecoderWrapper<VideoSample> {
 				}
 			};
 
-			if (codec === 'avc' && this.decoderConfig.description && isChromium()) {
-				// Chromium has/had a bug with playing interlaced AVC (https://issues.chromium.org/issues/456919096)
-				// which can be worked around by requesting that software decoding be used. So, here we peek into the
-				// AVC description, if present, and switch to software decoding if we find interlaced content.
-				const record = deserializeAvcDecoderConfigurationRecord(toUint8Array(this.decoderConfig.description));
-				if (record && record.sequenceParameterSets.length > 0) {
-					const sps = parseAvcSps(record.sequenceParameterSets[0]!);
-					if (sps && sps.frameMbsOnlyFlag === 0) {
-						this.decoderConfig = {
-							...this.decoderConfig,
-							hardwareAcceleration: 'prefer-software',
-						};
+			if (isChromium()) {
+				if (codec === 'avc' && this.decoderConfig.description) {
+					// Chromium has/had a bug with playing interlaced AVC (https://issues.chromium.org/issues/456919096)
+					// which can be worked around by requesting that software decoding be used. So, here we peek into
+					// the AVC description, if present, and switch to software decoding if we find interlaced content.
+					const record = deserializeAvcDecoderConfigurationRecord(
+						toUint8Array(this.decoderConfig.description),
+					);
+					if (record && record.sequenceParameterSets.length > 0) {
+						const sps = parseAvcSps(record.sequenceParameterSets[0]!);
+						if (sps && sps.frameMbsOnlyFlag === 0) {
+							this.decoderConfig = {
+								...this.decoderConfig,
+								hardwareAcceleration: 'prefer-software',
+							};
+						}
 					}
 				}
+
+				if (!colorSpaceIsComplete(this.decoderConfig.colorSpace)) {
+					// Found via https://github.com/remotion-dev/remotion/issues/10841.
+					// If the color space is incomplete (which is often that it's just partially filled), Chromium has
+					// some nasty logic where it doesn't pass that information along to the GPU at all. The result is
+					// that information is genuinely lost, like the color matrix for example. Chromium has other code
+					// paths where it just fills the missing values with a hardcoded default, so we do the exact same
+					// thing here, with the same hardcoded defaults:
+					this.decoderConfig = {
+						...this.decoderConfig,
+						colorSpace: {
+							primaries: this.decoderConfig.colorSpace?.primaries ?? 'bt709',
+							matrix: this.decoderConfig.colorSpace?.matrix ?? 'bt709',
+							transfer: this.decoderConfig.colorSpace?.transfer ?? 'bt709',
+							fullRange: this.decoderConfig.colorSpace?.fullRange ?? false,
+						},
+					};
+				}
 			}
 
 			const stack = new Error('Decoding error').stack;
@@ -1691,8 +1715,12 @@ export class VideoSampleSink extends BaseMediaSampleSink<VideoSample> {
 		onError: (error: unknown) => unknown,
 	) {
 		if (!(await this._track.canDecode())) {
+			if (typeof VideoDecoder === 'undefined') {
+				throw new Error(missingWebCodecsClassMessage('VideoDecoder'));
+			}
+
 			throw new Error(
-				'This video track cannot be decoded by this browser. Make sure to check decodability before using'
+				'This video track cannot be decoded in this environment. Make sure to check decodability before using'
 				+ ' a track.',
 			);
 		}
@@ -2090,7 +2118,7 @@ class AudioDecoderWrapper extends DecoderWrapper<AudioSample> {
 		const sampleHandler = (sample: AudioSample) => {
 			let sampleTimestamp = sample.timestamp;
 
-			if (this.expectedFirstTimestamp && this.currentTimestamp === null) {
+			if (this.expectedFirstTimestamp !== null && this.currentTimestamp === null) {
 				this.timestampOffset = this.expectedFirstTimestamp - sampleTimestamp; ;
 			}
 
@@ -2426,8 +2454,12 @@ export class AudioSampleSink extends BaseMediaSampleSink<AudioSample> {
 		onError: (error: unknown) => unknown,
 	) {
 		if (!(await this._track.canDecode())) {
+			if (typeof AudioDecoder === 'undefined') {
+				throw new Error(missingWebCodecsClassMessage('AudioDecoder'));
+			}
+
 			throw new Error(
-				'This audio track cannot be decoded by this browser. Make sure to check decodability before using'
+				'This audio track cannot be decoded in this environment. Make sure to check decodability before using'
 				+ ' a track.',
 			);
 		}

src/media-source.ts

@@ -28,7 +28,9 @@ import {
 	clamp,
 	clearIntervalUnthrottled,
 	floorToDivisor,
+	isThenable,
 	last,
+	missingWebCodecsClassMessage,
 	promiseWithResolvers,
 	roundToDivisor,
 	setInt24,
@@ -425,7 +427,7 @@ class VideoEncoderWrapper {
 		// Apply the user-defined process function, if any
 		if (config.transform?.process) {
 			let processed = config.transform.process(videoSample);
-			if (processed instanceof Promise) {
+			if (isThenable(processed)) {
 				processed = await processed;
 			}
 
@@ -718,7 +720,7 @@ class VideoEncoderWrapper {
 
 			if (!selected) {
 				if (typeof VideoEncoder === 'undefined') {
-					throw new Error('VideoEncoder is not supported by this browser.');
+					throw new Error(missingWebCodecsClassMessage('VideoEncoder'));
 				}
 
 				// The candidates only differ in their rate control, so we describe them as one config with a
@@ -731,7 +733,7 @@ class VideoEncoderWrapper {
 				throw new Error(
 					`This specific encoder configuration (${firstConfig.codec}, ${rateControls.join(' / ')},`
 					+ ` ${firstConfig.width}x${firstConfig.height}, hardware acceleration:`
-					+ ` ${firstConfig.hardwareAcceleration ?? 'no-preference'}) is not supported by this browser.`
+					+ ` ${firstConfig.hardwareAcceleration ?? 'no-preference'}) is not supported in this environment.`
 					+ ` Consider using another codec or changing your video parameters.`,
 				);
 			}
@@ -1694,8 +1696,8 @@ export class MediaStreamVideoTrackSource extends VideoSource {
 				});
 			} else {
 				throw new Error(
-					'When no explicit frame rate is set, MediaStreamTrackProcessor is required; but it\'s not supported'
-					+ ' by this browser.',
+					'When no explicit frame rate is set, MediaStreamTrackProcessor is required; but it\'s not available'
+					+ ' in this environment.',
 				);
 			}
 		}
@@ -1936,7 +1938,7 @@ class AudioEncoderWrapper {
 		if (config.transform?.process) {
 			try {
 				let processed = config.transform.process(audioSample);
-				if (processed instanceof Promise) {
+				if (isThenable(processed)) {
 					processed = await processed;
 				}
 
@@ -2210,7 +2212,7 @@ class AudioEncoderWrapper {
 				this.initPcmEncoder();
 			} else {
 				if (typeof AudioEncoder === 'undefined') {
-					throw new Error('AudioEncoder is not supported by this browser.');
+					throw new Error(missingWebCodecsClassMessage('AudioEncoder'));
 				}
 
 				let supported: boolean;
@@ -2226,7 +2228,8 @@ class AudioEncoderWrapper {
 					throw new Error(
 						`This specific encoder configuration (${encoderConfig.codec}, ${encoderConfig.bitrate} bps,`
 						+ ` ${encoderConfig.numberOfChannels} channels, ${encoderConfig.sampleRate} Hz) is not`
-						+ ` supported by this browser. Consider using another codec or changing your audio parameters.`,
+						+ ` supported in this environment. Consider using another codec or changing your`
+						+ ` audio parameters.`,
 					);
 				}
 
@@ -2254,8 +2257,8 @@ class AudioEncoderWrapper {
 
 								meta.decoderConfig.description = buildAacAudioSpecificConfig({
 									objectType,
-									numberOfChannels: meta.decoderConfig.numberOfChannels,
-									sampleRate: meta.decoderConfig.sampleRate,
+									outputNumberOfChannels: meta.decoderConfig.numberOfChannels,
+									outputSampleRate: meta.decoderConfig.sampleRate,
 								});
 							}
 						}

src/misc.ts

@@ -162,6 +162,25 @@ export const colorSpaceIsComplete = (
 	);
 };
 
+export const colorSpaceIsEmpty = (colorSpace: VideoColorSpaceInit | undefined) => {
+	return (
+		!colorSpace
+		|| (
+			colorSpace.primaries == null
+			&& colorSpace.transfer == null
+			&& colorSpace.matrix == null
+			&& colorSpace.fullRange == null
+		)
+	);
+};
+
+export const EMPTY_COLOR_SPACE: VideoColorSpaceInit = {
+	primaries: undefined,
+	transfer: undefined,
+	matrix: undefined,
+	fullRange: undefined,
+};
+
 export const isAllowSharedBufferSource = (x: unknown) => {
 	return (
 		x instanceof ArrayBuffer
@@ -721,6 +740,16 @@ export const getChromiumVersion = () => {
 	return chromiumVersionCache = Number(match[1]!);
 };
 
+export const missingWebCodecsClassMessage = (className: string) => {
+	if (typeof globalThis.isSecureContext !== 'undefined' && !globalThis.isSecureContext) {
+		// WebCodecs is not exposed in insecure contexts
+		return `${className} is not available in this environment; this may be because this page is running in an`
+			+ ` insecure context. Try serving your page over HTTPS or use localhost.`;
+	}
+
+	return `${className} is not available in this environment.`;
+};
+
 /**
  * T or a promise that resolves to T.
  * @group Miscellaneous
@@ -728,6 +757,20 @@ export const getChromiumVersion = () => {
  */
 export type MaybePromise<T> = T | Promise<T>;
 
+const NativePromiseConstructor = /* #__PURE__ */ (async () => {})().constructor as PromiseConstructor;
+
+/**
+ * Needed to properly deal with custom Promise implementations and because this is closer to how the JS spec does it.
+ */
+export const isThenable = <T>(value: MaybePromise<T>): value is Promise<T> => {
+	if (value instanceof NativePromiseConstructor || value instanceof Promise) {
+		return true;
+	}
+
+	// Fall back to a crude duck typing check
+	return typeof (value as PromiseLike<T> | null | undefined)?.then === 'function';
+};
+
 /** Acts like `??` except the condition is -1 and not null/undefined. */
 export const coalesceIndex = (a: number, b: number) => {
 	return a !== -1 ? a : b;

src/mp3/mp3-demuxer.ts

@@ -17,6 +17,7 @@ import {
 	AsyncMutex,
 	binarySearchExact,
 	binarySearchLessOrEqual,
+	isThenable,
 	toDataView,
 	UNDETERMINED_LANGUAGE,
 } from '../misc';
@@ -102,7 +103,7 @@ export class Mp3Demuxer extends Demuxer {
 			// Let's skip all ID3v2 tags at the start of the file
 			while (true) {
 				let slice = this.reader.requestSlice(this.lastLoadedPos, ID3_V2_HEADER_SIZE);
-				if (slice instanceof Promise) slice = await slice;
+				if (isThenable(slice)) slice = await slice;
 
 				if (!slice) {
 					this.lastSampleLoaded = true;
@@ -136,7 +137,7 @@ export class Mp3Demuxer extends Demuxer {
 		const xingOffset = getXingOffset(header.mpegVersionId, header.channel);
 
 		let slice = this.reader.requestSlice(result.startPos + xingOffset, 4);
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 		if (slice) {
 			const word = readU32Be(slice);
 			const isXing = word === XING || word === INFO;
@@ -151,7 +152,7 @@ export class Mp3Demuxer extends Demuxer {
 
 				if (!this.xingData) {
 					let xingDataSlice = this.reader.requestSlice(result.startPos + xingOffset + 4, 12);
-					if (xingDataSlice instanceof Promise) xingDataSlice = await xingDataSlice;
+					if (isThenable(xingDataSlice)) xingDataSlice = await xingDataSlice;
 					if (xingDataSlice) {
 						const xingData = readBytes(xingDataSlice, 12);
 						const view = toDataView(xingData);
@@ -216,7 +217,7 @@ export class Mp3Demuxer extends Demuxer {
 
 			while (true) {
 				let headerSlice = this.reader.requestSlice(currentPos, ID3_V2_HEADER_SIZE);
-				if (headerSlice instanceof Promise) headerSlice = await headerSlice;
+				if (isThenable(headerSlice)) headerSlice = await headerSlice;
 				if (!headerSlice) break;
 
 				const id3V2Header = readId3V2Header(headerSlice);
@@ -227,7 +228,7 @@ export class Mp3Demuxer extends Demuxer {
 				id3V2HeaderFound = true;
 
 				let contentSlice = this.reader.requestSlice(headerSlice.filePos, id3V2Header.size);
-				if (contentSlice instanceof Promise) contentSlice = await contentSlice;
+				if (isThenable(contentSlice)) contentSlice = await contentSlice;
 				if (!contentSlice) break;
 
 				parseId3V2Tag(contentSlice, id3V2Header, this.metadataTags);
@@ -238,7 +239,7 @@ export class Mp3Demuxer extends Demuxer {
 			if (!id3V2HeaderFound && this.reader.fileSize !== null && this.reader.fileSize >= ID3_V1_TAG_SIZE) {
 				// Try reading an ID3v1 tag at the end of the file
 				let slice = this.reader.requestSlice(this.reader.fileSize - ID3_V1_TAG_SIZE, ID3_V1_TAG_SIZE);
-				if (slice instanceof Promise) slice = await slice;
+				if (isThenable(slice)) slice = await slice;
 				assert(slice);
 
 				const tag = readAscii(slice, 3);
@@ -388,7 +389,7 @@ class Mp3AudioTrackBacking implements InputAudioTrackBacking {
 			data = PLACEHOLDER_DATA;
 		} else {
 			let slice = this.demuxer.reader.requestSlice(rawSample.dataStart, rawSample.dataSize);
-			if (slice instanceof Promise) slice = await slice;
+			if (isThenable(slice)) slice = await slice;
 
 			if (!slice) {
 				return null; // Data didn't fit into the rest of the file

src/mp3/mp3-reader.ts

@@ -7,6 +7,7 @@
  */
 
 import { MP3_FRAME_HEADER_SIZE, getMp3ChannelCount, Mp3FrameHeader, readMp3FrameHeader } from '../../shared/mp3-misc';
+import { isThenable } from '../misc';
 import { Reader, readU32Be } from '../reader';
 
 export const readNextMp3FrameHeader = async (
@@ -27,7 +28,7 @@ export const readNextMp3FrameHeader = async (
 			: CHUNK_SIZE;
 
 		let slice = reader.requestSliceRange(currentPos, MP3_FRAME_HEADER_SIZE, maxLength);
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 		if (!slice || slice.length < MP3_FRAME_HEADER_SIZE) break;
 
 		while (slice.remainingLength >= MP3_FRAME_HEADER_SIZE) {

src/mpeg-ts/mpeg-ts-demuxer.ts

@@ -63,6 +63,7 @@ import {
 	COLOR_PRIMARIES_MAP_INVERSE,
 	findLastIndex,
 	floorToMultiple,
+	isThenable,
 	last,
 	MATRIX_COEFFICIENTS_MAP_INVERSE,
 	readExpGolomb,
@@ -190,7 +191,7 @@ export class MpegTsDemuxer extends Demuxer {
 		return this.metadataPromise ??= (async () => {
 			const lengthToCheck = TS_PACKET_SIZE + 16 + 1;
 			let startingSlice = this.reader.requestSlice(0, lengthToCheck);
-			if (startingSlice instanceof Promise) startingSlice = await startingSlice;
+			if (isThenable(startingSlice)) startingSlice = await startingSlice;
 			assert(startingSlice);
 
 			const startingBytes = readBytes(startingSlice, lengthToCheck);
@@ -951,7 +952,7 @@ export class MpegTsDemuxer extends Demuxer {
 
 	async readPacketHeader(pos: number): Promise<TsPacketHeader | null> {
 		let slice = this.reader.requestSlice(pos, 4);
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 
 		if (!slice) {
 			return null;
@@ -987,7 +988,7 @@ export class MpegTsDemuxer extends Demuxer {
 	async readPacket(pos: number): Promise<TsPacket | null> {
 		// Code in here is duplicated from readPacketHeader for performance reasons
 		let slice = this.reader.requestSlice(pos, TS_PACKET_SIZE);
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 
 		if (!slice) {
 			return null;
@@ -2099,7 +2100,7 @@ class PacketReadingContext {
 
 			while (true) {
 				let remaining = this.ensureBuffered(CHUNK_SIZE);
-				if (remaining instanceof Promise) remaining = await remaining;
+				if (isThenable(remaining)) remaining = await remaining;
 
 				if (remaining === 0) {
 					break;
@@ -2253,7 +2254,7 @@ class PacketReadingContext {
 
 			while (true) {
 				let remaining = this.ensureBuffered(CHUNK_SIZE);
-				if (remaining instanceof Promise) remaining = await remaining;
+				if (isThenable(remaining)) remaining = await remaining;
 
 				const startPos = this.currentPos;
 
@@ -2269,7 +2270,7 @@ class PacketReadingContext {
 						const possibleHeaderStartPos = this.currentPos;
 
 						let remaining = this.ensureBuffered(MAX_ADTS_FRAME_HEADER_SIZE);
-						if (remaining instanceof Promise) remaining = await remaining;
+						if (isThenable(remaining)) remaining = await remaining;
 
 						if (remaining < MAX_ADTS_FRAME_HEADER_SIZE) {
 							return;
@@ -2282,7 +2283,7 @@ class PacketReadingContext {
 							this.seekTo(possibleHeaderStartPos);
 
 							let remaining = this.ensureBuffered(header.frameLength);
-							if (remaining instanceof Promise) remaining = await remaining;
+							if (isThenable(remaining)) remaining = await remaining;
 
 							return this.supplyPacket(
 								remaining,
@@ -2300,7 +2301,7 @@ class PacketReadingContext {
 						const possibleHeaderStartPos = this.currentPos;
 
 						let remaining = this.ensureBuffered(MP3_FRAME_HEADER_SIZE);
-						if (remaining instanceof Promise) remaining = await remaining;
+						if (isThenable(remaining)) remaining = await remaining;
 
 						if (remaining < MP3_FRAME_HEADER_SIZE) {
 							return;
@@ -2314,7 +2315,7 @@ class PacketReadingContext {
 							this.seekTo(possibleHeaderStartPos);
 
 							let remaining = this.ensureBuffered(result.header.totalSize);
-							if (remaining instanceof Promise) remaining = await remaining;
+							if (isThenable(remaining)) remaining = await remaining;
 
 							const duration = result.header.audioSamplesInFrame * TIMESCALE
 								/ elementaryStream.info.sampleRate;
@@ -2332,7 +2333,7 @@ class PacketReadingContext {
 
 						// Need at least 5 bytes for sync word + CRC + fscod/frmsizecod
 						let remaining = this.ensureBuffered(5);
-						if (remaining instanceof Promise) remaining = await remaining;
+						if (isThenable(remaining)) remaining = await remaining;
 
 						if (remaining < 5) {
 							return;
@@ -2361,7 +2362,7 @@ class PacketReadingContext {
 						this.seekTo(possibleSyncPos);
 
 						remaining = this.ensureBuffered(frameSize);
-						if (remaining instanceof Promise) remaining = await remaining;
+						if (isThenable(remaining)) remaining = await remaining;
 
 						const duration = Math.round(
 							AC3_SAMPLES_PER_FRAME * TIMESCALE / elementaryStream.info.sampleRate,
@@ -2377,7 +2378,7 @@ class PacketReadingContext {
 
 						// Need at least 5 bytes for E-AC-3 header parsing (sync word + frmsiz + fscod/numblkscod)
 						let remaining = this.ensureBuffered(5);
-						if (remaining instanceof Promise) remaining = await remaining;
+						if (isThenable(remaining)) remaining = await remaining;
 
 						if (remaining < 5) {
 							return;
@@ -2399,7 +2400,7 @@ class PacketReadingContext {
 						this.seekTo(possibleSyncPos);
 
 						remaining = this.ensureBuffered(frameSize);
-						if (remaining instanceof Promise) remaining = await remaining;
+						if (isThenable(remaining)) remaining = await remaining;
 
 						// Duration = numblks * 256 samples per block
 						const samplesPerFrame = numblks * 256;
@@ -2416,7 +2417,7 @@ class PacketReadingContext {
 						const possibleSyncPos = this.currentPos;
 
 						let remaining = this.ensureBuffered(DTS_CORE_FRAME_HEADER_SIZE);
-						if (remaining instanceof Promise) remaining = await remaining;
+						if (isThenable(remaining)) remaining = await remaining;
 
 						if (remaining < DTS_CORE_FRAME_HEADER_SIZE) {
 							return;
@@ -2438,7 +2439,7 @@ class PacketReadingContext {
 
 							const headerBound = Math.min(leadingExss.frameSize, DTS_EXSS_MAX_HEADER_SIZE);
 							let remaining = this.ensureBuffered(headerBound);
-							if (remaining instanceof Promise) remaining = await remaining;
+							if (isThenable(remaining)) remaining = await remaining;
 
 							leadingExss = parseDtsExssHeader(this.readBytes(remaining)) ?? leadingExss;
 						}
@@ -2456,7 +2457,7 @@ class PacketReadingContext {
 
 								const neededBytes = nextSubstreamPos + DTS_EXSS_HEADER_PREFIX_SIZE;
 								let remaining = this.ensureBuffered(neededBytes);
-								if (remaining instanceof Promise) remaining = await remaining;
+								if (isThenable(remaining)) remaining = await remaining;
 
 								if (remaining < neededBytes) {
 									break;
@@ -2484,7 +2485,7 @@ class PacketReadingContext {
 						this.seekTo(possibleSyncPos);
 
 						remaining = this.ensureBuffered(frameSize);
-						if (remaining instanceof Promise) remaining = await remaining;
+						if (isThenable(remaining)) remaining = await remaining;
 
 						const duration = Math.round(
 							sampleCount * TIMESCALE / elementaryStream.info.sampleRate,

src/ogg/ogg-demuxer.ts

@@ -18,6 +18,7 @@ import {
 	AsyncMutex,
 	binarySearchLessOrEqual,
 	findLast,
+	isThenable,
 	last,
 	roundIfAlmostInteger,
 	toDataView,
@@ -72,7 +73,7 @@ export class OggDemuxer extends Demuxer {
 
 			while (true) {
 				let slice = this.reader.requestSliceRange(currentPos, MIN_PAGE_HEADER_SIZE, MAX_PAGE_HEADER_SIZE);
-				if (slice instanceof Promise) slice = await slice;
+				if (isThenable(slice)) slice = await slice;
 				if (!slice) break;
 
 				const page = readPageHeader(slice);
@@ -274,7 +275,7 @@ export class OggDemuxer extends Demuxer {
 		while (true) {
 			// Load the entire page data
 			let pageSlice = this.reader.requestSlice(currentPage.dataStartPos, currentPage.dataSize);
-			if (pageSlice instanceof Promise) pageSlice = await pageSlice;
+			if (isThenable(pageSlice)) pageSlice = await pageSlice;
 			assert(pageSlice);
 			const pageData = readBytes(pageSlice, currentPage.dataSize);
 
@@ -299,7 +300,7 @@ export class OggDemuxer extends Demuxer {
 			let currentPos = currentPage.headerStartPos + currentPage.totalSize;
 			while (true) {
 				let headerSlice = this.reader.requestSliceRange(currentPos, MIN_PAGE_HEADER_SIZE, MAX_PAGE_HEADER_SIZE);
-				if (headerSlice instanceof Promise) headerSlice = await headerSlice;
+				if (isThenable(headerSlice)) headerSlice = await headerSlice;
 				if (!headerSlice) {
 					return null;
 				}
@@ -358,7 +359,7 @@ export class OggDemuxer extends Demuxer {
 		let currentPos = lastPacket.endPage.headerStartPos + lastPacket.endPage.totalSize;
 		while (true) {
 			let slice = this.reader.requestSliceRange(currentPos, MIN_PAGE_HEADER_SIZE, MAX_PAGE_HEADER_SIZE);
-			if (slice instanceof Promise) slice = await slice;
+			if (isThenable(slice)) slice = await slice;
 			if (!slice) {
 				return null;
 			}
@@ -658,7 +659,7 @@ class OggAudioTrackBacking implements InputAudioTrackBacking {
 				);
 
 				let searchSlice = this.demuxer.reader.requestSlice(searchStartPos, until - searchStartPos);
-				if (searchSlice instanceof Promise) searchSlice = await searchSlice;
+				if (isThenable(searchSlice)) searchSlice = await searchSlice;
 				assert(searchSlice);
 
 				const found = findNextPageHeader(searchSlice, until);
@@ -672,7 +673,7 @@ class OggAudioTrackBacking implements InputAudioTrackBacking {
 					MIN_PAGE_HEADER_SIZE,
 					MAX_PAGE_HEADER_SIZE,
 				);
-				if (headerSlice instanceof Promise) headerSlice = await headerSlice;
+				if (isThenable(headerSlice)) headerSlice = await headerSlice;
 				assert(headerSlice);
 
 				const page = readPageHeader(headerSlice);
@@ -685,7 +686,7 @@ class OggAudioTrackBacking implements InputAudioTrackBacking {
 					pageValid = true;
 				} else {
 					let pageSlice = this.demuxer.reader.requestSlice(page.headerStartPos, page.totalSize);
-					if (pageSlice instanceof Promise) pageSlice = await pageSlice;
+					if (isThenable(pageSlice)) pageSlice = await pageSlice;
 					assert(pageSlice);
 
 					// Validate the page by checking checksum
@@ -759,7 +760,7 @@ class OggAudioTrackBacking implements InputAudioTrackBacking {
 
 			const nextPos = currentPage.headerStartPos + currentPage.totalSize;
 			let slice = this.demuxer.reader.requestSliceRange(nextPos, MIN_PAGE_HEADER_SIZE, MAX_PAGE_HEADER_SIZE);
-			if (slice instanceof Promise) slice = await slice;
+			if (isThenable(slice)) slice = await slice;
 			assert(slice);
 
 			const nextPage = readPageHeader(slice);

src/output-format.ts

@@ -30,7 +30,7 @@ import { MpegTsMuxer } from './mpeg-ts/mpeg-ts-muxer';
 import { WaveMuxer } from './wave/wave-muxer';
 import { HlsMuxer } from './hls/hls-muxer';
 import { HLS_MIME_TYPE } from './hls/hls-misc';
-import { MaybePromise, FilePath, toArray } from './misc';
+import { MaybePromise, FilePath, isNumber, toArray } from './misc';
 import { Target } from './target';
 
 /**
@@ -229,7 +229,7 @@ export abstract class IsobmffOutputFormat extends OutputFormat {
 		}
 		if (
 			options.minimumFragmentDuration !== undefined
-			&& (!Number.isFinite(options.minimumFragmentDuration) || options.minimumFragmentDuration < 0)
+			&& (!isNumber(options.minimumFragmentDuration) || options.minimumFragmentDuration < 0)
 		) {
 			throw new TypeError('options.minimumFragmentDuration, when provided, must be a non-negative number.');
 		}
@@ -514,7 +514,7 @@ export class MkvOutputFormat extends OutputFormat {
 		}
 		if (
 			options.minimumClusterDuration !== undefined
-			&& (!Number.isFinite(options.minimumClusterDuration) || options.minimumClusterDuration < 0)
+			&& (!isNumber(options.minimumClusterDuration) || options.minimumClusterDuration < 0)
 		) {
 			throw new TypeError('options.minimumClusterDuration, when provided, must be a non-negative number.');
 		}
@@ -861,7 +861,7 @@ export class OggOutputFormat extends OutputFormat {
 		}
 		if (
 			options.maximumPageDuration !== undefined
-			&& (!Number.isFinite(options.maximumPageDuration) || options.maximumPageDuration <= 0)
+			&& (!isNumber(options.maximumPageDuration) || options.maximumPageDuration <= 0)
 		) {
 			throw new TypeError('options.maximumPageDuration, when provided, must be a positive number.');
 		}

src/output.ts

@@ -6,7 +6,16 @@
  * file, You can obtain one at https://mozilla.org/MPL/2.0/.
  */
 
-import { assert, AsyncMutex, EventEmitter, isIso639Dash2LanguageCode, MaybePromise, Rotation, toArray } from './misc';
+import {
+	assert,
+	AsyncMutex,
+	EventEmitter,
+	isIso639Dash2LanguageCode,
+	isThenable,
+	MaybePromise,
+	Rotation,
+	toArray,
+} from './misc';
 import { MetadataTags, TrackDisposition, validateMetadataTags, validateTrackDisposition } from './metadata';
 import { Muxer } from './muxer';
 import { OutputFormat } from './output-format';
@@ -451,7 +460,7 @@ export class Output<
 		}
 
 		const rootTargetResult = this._getRootTarget();
-		if (rootTargetResult instanceof Promise) {
+		if (isThenable(rootTargetResult)) {
 			throw new TypeError(errorMessage);
 		}
 
@@ -516,7 +525,7 @@ export class Output<
 			return result;
 		};
 
-		if (result instanceof Promise) {
+		if (isThenable(result)) {
 			return result.then(handleResult);
 		} else {
 			return handleResult(result);
@@ -604,7 +613,7 @@ export class Output<
 			return target;
 		};
 
-		if (result instanceof Promise) {
+		if (isThenable(result)) {
 			return this._rootTargetPromise = result.then(handleResult);
 		} else {
 			return handleResult(result);

src/packet.ts

@@ -155,7 +155,7 @@ export class EncodedPacket {
 			throw new TypeError('Metadata-only packets cannot be converted to a video chunk.');
 		}
 		if (typeof EncodedVideoChunk === 'undefined') {
-			throw new Error('Your browser does not support EncodedVideoChunk.');
+			throw new Error('EncodedVideoChunk is not available in this environment.');
 		}
 
 		return new EncodedVideoChunk({
@@ -179,7 +179,7 @@ export class EncodedPacket {
 			throw new TypeError('Metadata-only packets cannot be converted to a video chunk.');
 		}
 		if (typeof EncodedVideoChunk === 'undefined') {
-			throw new Error('Your browser does not support EncodedVideoChunk.');
+			throw new Error('EncodedVideoChunk is not available in this environment.');
 		}
 
 		return new EncodedVideoChunk({
@@ -198,7 +198,7 @@ export class EncodedPacket {
 			throw new TypeError('Metadata-only packets cannot be converted to an audio chunk.');
 		}
 		if (typeof EncodedAudioChunk === 'undefined') {
-			throw new Error('Your browser does not support EncodedAudioChunk.');
+			throw new Error('EncodedAudioChunk is not available in this environment.');
 		}
 
 		return new EncodedAudioChunk({

src/reader.ts

@@ -7,7 +7,7 @@
  */
 
 import { InputDisposedError } from './input';
-import { assert, clamp, getUint24, MaybePromise, textDecoder, toDataView } from './misc';
+import { assert, clamp, getUint24, isThenable, MaybePromise, textDecoder, toDataView } from './misc';
 import { DEFAULT_MAX_READ_POSITION, DEFAULT_MIN_READ_POSITION, Source } from './source';
 
 export class Reader {
@@ -47,7 +47,7 @@ export class Reader {
 		const end = start + length;
 		const result = this.source._read(start, end, DEFAULT_MIN_READ_POSITION, DEFAULT_MAX_READ_POSITION);
 
-		if (result instanceof Promise) {
+		if (isThenable(result)) {
 			return result.then((x) => {
 				if (!x) {
 					return null;
@@ -95,7 +95,7 @@ export class Reader {
 				);
 			};
 
-			if (promisedAttempt instanceof Promise) {
+			if (isThenable(promisedAttempt)) {
 				return promisedAttempt.then(handleAttempt);
 			} else {
 				return handleAttempt(promisedAttempt);
@@ -121,7 +121,7 @@ export class Reader {
 				}
 
 				let slice = this.requestSliceRange(currentSize, 0, CHUNK_SIZE);
-				if (slice instanceof Promise) slice = await slice;
+				if (isThenable(slice)) slice = await slice;
 
 				if (!slice || slice.length === 0) {
 					break;

src/sample.ts

@@ -11,6 +11,7 @@ import {
 	clamp,
 	COLOR_PRIMARIES_MAP,
 	isAllowSharedBufferSource,
+	isThenable,
 	MATRIX_COEFFICIENTS_MAP,
 	Rotation,
 	SECOND_TO_MICROSECOND_FACTOR,
@@ -864,7 +865,7 @@ export class VideoSample implements Disposable {
 
 		if (this._data instanceof VideoSampleResource) {
 			let result = this._data.getDataPlanes();
-			if (result instanceof Promise) result = await result;
+			if (isThenable(result)) result = await result;
 
 			if (
 				!Array.isArray(result)
@@ -1022,7 +1023,7 @@ export class VideoSample implements Disposable {
 			}
 
 			const planes = this._data.getDataPlanes();
-			if (planes instanceof Promise) {
+			if (isThenable(planes)) {
 				throw new Error(
 					'Cannot convert a VideoSampleResource-backed VideoSample to VideoFrame if getDataPlanes() returns'
 					+ ' a promise.',
@@ -1586,7 +1587,7 @@ export class VideoSample implements Disposable {
 		// Description's finalized; let's see if a registered transformer wants to handle it
 		for (const transformer of registeredVideoSampleTransformers) {
 			let result = transformer(this, description);
-			if (result instanceof Promise) result = await result;
+			if (isThenable(result)) result = await result;
 
 			if (result !== null) {
 				return result;
@@ -2657,7 +2658,7 @@ export class AudioSample implements Disposable {
 		const { format, frameCount: optFrameCount, frameOffset: optFrameOffset } = options;
 		let { planeIndex } = options;
 
-		const srcFormat = this.format;
+		let srcFormat = this.format;
 		const destFormat = format ?? this.format;
 		if (!destFormat) throw new Error('Destination format not determined');
 
@@ -2704,88 +2705,133 @@ export class AudioSample implements Disposable {
 					frameOffset,
 					copyFrameCount,
 				);
+				return;
 			} else {
-				// Per spec, only f32-planar conversion must be supported, but in practice, all browsers support all
-				// destination formats, so let's just delegate here:
-				this._data.copyTo(destination, {
-					planeIndex,
-					frameOffset,
-					frameCount: copyFrameCount,
-					format: destFormat,
-				});
+				try {
+					// The spec only requires f32-planar to work here, but most of the time, all formats work. Sometimes
+					// they don't, and that's why we try-catch it.
+					this._data.copyTo(destination, {
+						planeIndex,
+						frameOffset,
+						frameCount: copyFrameCount,
+						format: destFormat,
+					});
+					return;
+				} catch (error) {
+					if (destFormat === 'f32-planar') {
+						throw error;
+					}
+
+					// Format conversion likely wasn't supported, so let's fall back to the manual conversion path
+					srcFormat = 'f32-planar';
+				}
 			}
-		} else {
-			const readFn = getReadFunction(srcFormat);
-			const srcBytesPerSample = getBytesPerSample(srcFormat);
-			const srcIsPlanar = formatIsPlanar(srcFormat);
+		}
 
-			let uint8Data: Uint8Array;
-			if (this._data instanceof AudioSampleResource) {
-				const getDataPlaneValidated = (index: number) => {
-					const result = (this._data as AudioSampleResource).getDataPlane(index);
-					if (!(result instanceof Uint8Array)) {
-						throw new TypeError('getDataPlane() must return a Uint8Array.');
-					}
+		const readFn = getReadFunction(srcFormat);
+		const srcBytesPerSample = getBytesPerSample(srcFormat);
+		const srcIsPlanar = formatIsPlanar(srcFormat);
 
-					const expectedSize = numFrames * srcBytesPerSample * (srcIsPlanar ? 1 : numChannels);
-					if (result.byteLength !== expectedSize) {
-						throw new TypeError(
-							`Data plane ${index} has invalid size. Expected exactly ${expectedSize} bytes, got`
-							+ ` ${result.byteLength} bytes.`,
-						);
-					}
+		let uint8Data: Uint8Array;
+		if (this._data instanceof AudioSampleResource) {
+			const getDataPlaneValidated = (index: number) => {
+				const result = (this._data as AudioSampleResource).getDataPlane(index);
+				if (!(result instanceof Uint8Array)) {
+					throw new TypeError('getDataPlane() must return a Uint8Array.');
+				}
 
-					return result;
-				};
+				const expectedSize = numFrames * srcBytesPerSample * (srcIsPlanar ? 1 : numChannels);
+				if (result.byteLength !== expectedSize) {
+					throw new TypeError(
+						`Data plane ${index} has invalid size. Expected exactly ${expectedSize} bytes, got`
+						+ ` ${result.byteLength} bytes.`,
+					);
+				}
 
-				if (srcIsPlanar) {
-					if (destIsPlanar) {
-						// Only one source plane will be extracted, so let's fetch only that one
-						uint8Data = getDataPlaneValidated(planeIndex);
-						planeIndex = 0; // To fix the subsequent access
-					} else {
-						// Pack all planes tightly together
-						uint8Data = new Uint8Array(numFrames * srcBytesPerSample * numChannels);
-						for (let ch = 0; ch < numChannels; ch++) {
-							const planeData = getDataPlaneValidated(ch);
-							uint8Data.set(planeData, ch * numFrames * srcBytesPerSample);
-						}
-					}
+				return result;
+			};
+
+			if (srcIsPlanar) {
+				if (destIsPlanar) {
+					// Only one source plane will be extracted, so let's fetch only that one
+					uint8Data = getDataPlaneValidated(planeIndex);
+					planeIndex = 0; // To fix the subsequent access
 				} else {
-					uint8Data = getDataPlaneValidated(0); // That's the only plane there is
+					// Pack all planes tightly together
+					uint8Data = new Uint8Array(numFrames * srcBytesPerSample * numChannels);
+					for (let ch = 0; ch < numChannels; ch++) {
+						const planeData = getDataPlaneValidated(ch);
+						uint8Data.set(planeData, ch * numFrames * srcBytesPerSample);
+					}
 				}
 			} else {
-				uint8Data = this._data;
+				uint8Data = getDataPlaneValidated(0); // That's the only plane there is
 			}
+		} else if (this._data instanceof Uint8Array) {
+			uint8Data = this._data;
+		} else {
+			assert(srcFormat === 'f32-planar');
 
-			const srcView = toDataView(uint8Data);
+			if (destIsPlanar) {
+				// Only one source plane will be read, so let's copy only that one
+				uint8Data = new Uint8Array(this._data.allocationSize({
+					format: 'f32-planar',
+					planeIndex,
+				}));
+				this._data.copyTo(uint8Data, {
+					format: 'f32-planar',
+					planeIndex,
+				});
+				planeIndex = 0; // To fix the subsequent access
+			} else {
+				// All planes will be read, so fetch 'em all
+				uint8Data = new Uint8Array(this._data.allocationSize({
+					format: 'f32-planar',
+					planeIndex: 0,
+				}) * numChannels);
 
-			for (let i = 0; i < copyFrameCount; i++) {
-				if (destIsPlanar) {
-					const destOffset = i * destBytesPerSample;
+				for (let ch = 0; ch < numChannels; ch++) {
+					this._data.copyTo(
+						uint8Data.subarray(
+							ch * numFrames * srcBytesPerSample,
+							(ch + 1) * numFrames * srcBytesPerSample,
+						),
+						{
+							format: 'f32-planar',
+							planeIndex: ch,
+						},
+					);
+				}
+			}
+		}
+
+		const srcView = toDataView(uint8Data);
+
+		for (let i = 0; i < copyFrameCount; i++) {
+			if (destIsPlanar) {
+				const destOffset = i * destBytesPerSample;
+				let srcOffset: number;
+				if (srcIsPlanar) {
+					srcOffset = (planeIndex * numFrames + (i + frameOffset)) * srcBytesPerSample;
+				} else {
+					srcOffset = (((i + frameOffset) * numChannels) + planeIndex) * srcBytesPerSample;
+				}
+
+				const normalized = readFn(srcView, srcOffset);
+				writeFn(destView, destOffset, normalized);
+			} else {
+				for (let ch = 0; ch < numChannels; ch++) {
+					const destIndex = i * numChannels + ch;
+					const destOffset = destIndex * destBytesPerSample;
 					let srcOffset: number;
 					if (srcIsPlanar) {
-						srcOffset = (planeIndex * numFrames + (i + frameOffset)) * srcBytesPerSample;
+						srcOffset = (ch * numFrames + (i + frameOffset)) * srcBytesPerSample;
 					} else {
-						srcOffset = (((i + frameOffset) * numChannels) + planeIndex) * srcBytesPerSample;
+						srcOffset = (((i + frameOffset) * numChannels) + ch) * srcBytesPerSample;
 					}
 
 					const normalized = readFn(srcView, srcOffset);
 					writeFn(destView, destOffset, normalized);
-				} else {
-					for (let ch = 0; ch < numChannels; ch++) {
-						const destIndex = i * numChannels + ch;
-						const destOffset = destIndex * destBytesPerSample;
-						let srcOffset: number;
-						if (srcIsPlanar) {
-							srcOffset = (ch * numFrames + (i + frameOffset)) * srcBytesPerSample;
-						} else {
-							srcOffset = (((i + frameOffset) * numChannels) + ch) * srcBytesPerSample;
-						}
-
-						const normalized = readFn(srcView, srcOffset);
-						writeFn(destView, destOffset, normalized);
-					}
 				}
 			}
 		}
@@ -3200,9 +3246,9 @@ export const toInterleavedAudioFormat = (format: AudioSampleFormat): 'u8' | 's16
 };
 
 /**
- * WebKit has a bug where calling AudioData.copyTo with a format different from the source format
- * crashes the tab when there are more than 2 channels. This function works around that by always
- * copying with the source format and then manually converting to the destination format.
+ * WebKit has a bug where calling AudioData.copyTo with a format different from the source format crashes the tab when
+ * there are more than 2 channels. This function works around that by always copying with the source format and then
+ * manually converting to the destination format.
  *
  * See https://bugs.webkit.org/show_bug.cgi?id=302521.
  */

src/source.ts

@@ -14,6 +14,7 @@ import {
 	closedIntervalsOverlap,
 	FilePath,
 	isNumber,
+	isThenable,
 	isWebKit,
 	MaybePromise,
 	mergeRequestInit,
@@ -328,7 +329,7 @@ export abstract class PathedSource extends Source {
 			return ref;
 		};
 
-		if (result instanceof Promise) {
+		if (isThenable(result)) {
 			return result.then(handle);
 		} else {
 			return handle(result);
@@ -387,7 +388,7 @@ export class CustomPathedSource extends PathedSource {
 					return ref;
 				};
 
-				if (result instanceof Promise) {
+				if (isThenable(result)) {
 					this._rootRequest = result.then(handle);
 				} else {
 					handle(result);
@@ -495,6 +496,14 @@ export type BlobSourceOptions = {
 	useStreamReader?: boolean;
 };
 
+const blobReaderRegistry = typeof FinalizationRegistry !== 'undefined'
+	? new FinalizationRegistry<ReadableStreamDefaultReader<Uint8Array>>((reader) => {
+		// Browsers don't GC readers that aren't "done", which creates indefinite memory leaks,
+		// see https://github.com/Vanilagy/mediabunny/issues/144#issuecomment-5465467062. So, we need to do it instead.
+		void reader.cancel().catch(() => {});
+	})
+	: null;
+
 /**
  * A source backed by a [`Blob`](https://developer.mozilla.org/en-US/docs/Web/API/Blob). Since a
  * [`File`](https://developer.mozilla.org/en-US/docs/Web/API/File) is also a `Blob`, this is the source to use when
@@ -509,6 +518,8 @@ export class BlobSource extends Source {
 	_options: BlobSourceOptions;
 	/** @internal */
 	_orchestrator: ReadOrchestrator;
+	/** @internal */
+	_readers = new WeakMap<ReadWorker, ReadableStreamDefaultReader<Uint8Array> | null>();
 
 	/**
 	 * Creates a new {@link BlobSource} backed by the specified
@@ -540,6 +551,17 @@ export class BlobSource extends Source {
 			maxCacheSize: options.maxCacheSize ?? (8 * 2 ** 20 /* 8 MiB */),
 			maxWorkerCount: 4,
 			runWorker: this._runWorker.bind(this),
+			onIdleWorkerRemoved: (worker) => {
+				const reader = this._readers.get(worker);
+
+				if (reader) {
+					this._readers.delete(worker);
+					blobReaderRegistry?.unregister(worker);
+
+					// If we don't do this, memory leaks indefinitely
+					void reader.cancel().catch(() => {});
+				}
+			},
 			prefetchProfile: PREFETCH_PROFILES.fileSystem,
 		});
 
@@ -561,9 +583,6 @@ export class BlobSource extends Source {
 		return this._orchestrator.read(start, end, minReadPosition, maxReadPosition);
 	}
 
-	/** @internal */
-	_readers = new WeakMap<ReadWorker, ReadableStreamDefaultReader<Uint8Array> | null>();
-
 	/** @internal */
 	private async _runWorker(worker: ReadWorker) {
 		assert(worker.strictTarget);
@@ -577,10 +596,15 @@ export class BlobSource extends Source {
 			// - ReadableStream stalls under backpressure (especially video)
 			// Affects Safari and all iOS browsers (Chrome, Firefox, etc.).
 			// Use arrayBuffer() fallback for WebKit browsers.
-			if ('stream' in this._blob && !isWebKit() && this._options.useStreamReader !== false) {
+			if (
+				'stream' in this._blob && !isWebKit()
+				&& this._options.useStreamReader !== false
+			) {
 				// Get a reader of the blob starting at the required offset, and then keep it around
 				const slice = this._blob.slice(worker.currentPos);
 				reader = slice.stream().getReader();
+
+				blobReaderRegistry?.register(worker, reader, worker);
 			} else {
 				// We'll need to use more primitive ways
 				reader = null;
@@ -894,7 +918,7 @@ export class UrlSource extends PathedSource {
 			return result;
 		};
 
-		if (result instanceof Promise) {
+		if (isThenable(result)) {
 			return result.then(processResult);
 		} else {
 			return processResult(result);
@@ -1198,7 +1222,7 @@ export class UrlSource extends PathedSource {
 		for (const slice of uniqueSlices) {
 			const result = backing._read(slice.start, slice.start + slice.bytes.length);
 
-			if (result instanceof Promise) {
+			if (isThenable(result)) {
 				result.then((readResult) => {
 					if (readResult) {
 						// The backing's cache is empty at this point, so the read is guaranteed to produce
@@ -1459,7 +1483,7 @@ export class CustomSource extends Source {
 
 		const result = this._options.getSize();
 
-		if (result instanceof Promise) {
+		if (isThenable(result)) {
 			return result.then((size) => {
 				if (!Number.isInteger(size) || size < 0) {
 					throw new TypeError('options.getSize must return or resolve to a non-negative integer.');
@@ -1485,7 +1509,7 @@ export class CustomSource extends Source {
 			const originalTargetPos = worker.targetPos;
 
 			let data = this._options.read(worker.currentPos, originalTargetPos);
-			if (data instanceof Promise) data = await data;
+			if (isThenable(data)) data = await data;
 
 			if (worker.aborted) {
 				break;
@@ -1969,6 +1993,7 @@ class ReadOrchestrator {
 		runWorker: (worker: ReadWorker) => Promise<void>;
 		prefetchProfile: PrefetchProfile;
 		maxWorkerCount: number;
+		onIdleWorkerRemoved?: (worker: ReadWorker) => void;
 	}) {}
 
 	read(
@@ -2262,6 +2287,7 @@ class ReadOrchestrator {
 				assert(oldestIndex !== null);
 				assert(oldestWorker.pendingSlices.length === 0);
 				this.workers.splice(oldestIndex, 1);
+				this.options.onIdleWorkerRemoved?.(oldestWorker);
 			} else {
 				return null; // All workers are still running, we can't create a new one
 			}
@@ -2409,6 +2435,7 @@ class ReadOrchestrator {
 				otherWorker.currentPos, otherWorker.targetPos, // These should typically be equal when the worker's idle
 			)) {
 				this.workers.splice(i, 1);
+				this.options.onIdleWorkerRemoved?.(otherWorker);
 				i--;
 			}
 		}
@@ -2485,6 +2512,7 @@ class ReadOrchestrator {
 
 		worker.running = false;
 		this.workers.splice(index, 1);
+		this.options.onIdleWorkerRemoved?.(worker);
 
 		if (this.fileSize === null) {
 			// We can now deduce the file size!
@@ -2596,6 +2624,11 @@ class ReadOrchestrator {
 
 			worker.pendingSlices.length = 0;
 			worker.aborted = true;
+
+			if (!worker.running) {
+				// Running workers clean up after themselves when they notice the abort
+				this.options.onIdleWorkerRemoved?.(worker);
+			}
 		}
 
 		for (const queuedRead of this.queuedReads) {
@@ -2706,7 +2739,7 @@ export class RangedSource extends Source {
 			return result;
 		};
 
-		if (result instanceof Promise) {
+		if (isThenable(result)) {
 			return result.then(processResult);
 		} else {
 			return processResult(result);

src/wave/wave-demuxer.ts

@@ -12,7 +12,7 @@ import { Input } from '../input';
 import { InputAudioTrackBacking } from '../input-track';
 import { PacketRetrievalOptions } from '../media-sink';
 import { DEFAULT_TRACK_DISPOSITION, MetadataTags } from '../metadata';
-import { assert, UNDETERMINED_LANGUAGE } from '../misc';
+import { assert, isThenable, UNDETERMINED_LANGUAGE } from '../misc';
 import { EncodedPacket, PLACEHOLDER_DATA } from '../packet';
 import { readAscii, readBytes, Reader, readU16, readU32, readU64 } from '../reader';
 import { ID3_V2_HEADER_SIZE, parseId3V2Tag, readId3V2Header } from '../id3';
@@ -52,7 +52,7 @@ export class WaveDemuxer extends Demuxer {
 	async readMetadata() {
 		return this.metadataPromise ??= (async () => {
 			let slice = this.reader.requestSlice(0, 12);
-			if (slice instanceof Promise) slice = await slice;
+			if (isThenable(slice)) slice = await slice;
 			assert(slice);
 
 			const riffType = readAscii(slice, 4);
@@ -77,7 +77,7 @@ export class WaveDemuxer extends Demuxer {
 
 			while (totalFileSize === null || currentPos < totalFileSize) {
 				let slice = this.reader.requestSlice(currentPos, 8);
-				if (slice instanceof Promise) slice = await slice;
+				if (isThenable(slice)) slice = await slice;
 				if (!slice) break;
 
 				const chunkId = readAscii(slice, 4);
@@ -103,7 +103,7 @@ export class WaveDemuxer extends Demuxer {
 					// File and data chunk sizes are defined in here instead
 
 					let ds64Slice = this.reader.requestSlice(startPos, chunkSize);
-					if (ds64Slice instanceof Promise) ds64Slice = await ds64Slice;
+					if (isThenable(ds64Slice)) ds64Slice = await ds64Slice;
 					if (!ds64Slice) break;
 
 					const riffChunkSize = readU64(ds64Slice, littleEndian);
@@ -136,7 +136,7 @@ export class WaveDemuxer extends Demuxer {
 
 	private async parseFmtChunk(startPos: number, size: number, littleEndian: boolean) {
 		let slice = this.reader.requestSlice(startPos, size);
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 		if (!slice) return; // File too short
 
 		let formatTag = readU16(slice, littleEndian);
@@ -206,7 +206,7 @@ export class WaveDemuxer extends Demuxer {
 
 	private async parseListChunk(startPos: number, size: number, littleEndian: boolean) {
 		let slice = this.reader.requestSlice(startPos, size);
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 		if (!slice) return; // File too short
 
 		const infoType = readAscii(slice, 4);
@@ -303,7 +303,7 @@ export class WaveDemuxer extends Demuxer {
 	private async parseId3Chunk(startPos: number, size: number) {
 		// Parse ID3 tag embedded in WAV file (non-default, but used a lot in practice anyway)
 		let slice = this.reader.requestSlice(startPos, size);
-		if (slice instanceof Promise) slice = await slice;
+		if (isThenable(slice)) slice = await slice;
 		if (!slice) return; // File too short
 
 		const id3V2Header = readId3V2Header(slice);
@@ -489,7 +489,7 @@ class WaveAudioTrackBacking implements InputAudioTrackBacking {
 			// requested slice actually exists.
 
 			let slice = this.demuxer.reader.requestSlice(this.demuxer.dataStart + startOffset, sizeInBytes);
-			if (slice instanceof Promise) slice = await slice;
+			if (isThenable(slice)) slice = await slice;
 
 			if (!slice) {
 				return null;
@@ -501,7 +501,7 @@ class WaveAudioTrackBacking implements InputAudioTrackBacking {
 			data = PLACEHOLDER_DATA;
 		} else {
 			let slice = this.demuxer.reader.requestSlice(this.demuxer.dataStart + startOffset, sizeInBytes);
-			if (slice instanceof Promise) slice = await slice;
+			if (isThenable(slice)) slice = await slice;
 			assert(slice);
 
 			data = readBytes(slice, sizeInBytes);

test/browser/audio-samples.test.ts

@@ -137,3 +137,108 @@ test('trim, throws on a closed sample', () => {
 	sample.close();
 	expect(() => sample.trim(0, 5)).toThrow('AudioSample is closed.');
 });
+
+test('copyTo, AudioData-backed, planar destination falls back to manual conversion', () => {
+	withMockedNativeCopyTo((calls) => {
+		using sample = makeAudioDataBackedSample();
+
+		for (let c = 0; c < NUM_CHANNELS; c++) {
+			const out = new Int16Array(NUM_FRAMES);
+			sample.copyTo(out, { planeIndex: c, format: 's16-planar' });
+
+			for (let f = 0; f < NUM_FRAMES; f++) {
+				expect(out[f]).toBe(cellValue(f, c));
+			}
+		}
+
+		// With a frame offset
+		const out = new Int16Array(NUM_FRAMES - 3);
+		sample.copyTo(out, { planeIndex: 1, format: 's16-planar', frameOffset: 3 });
+		for (let f = 0; f < NUM_FRAMES - 3; f++) {
+			expect(out[f]).toBe(cellValue(f + 3, 1));
+		}
+
+		// The native conversion must have been attempted (and thrown)
+		expect(calls.some(options => options.format === 's16-planar')).toBe(true);
+	});
+});
+
+test('copyTo, AudioData-backed, interleaved destination falls back to manual conversion', () => {
+	withMockedNativeCopyTo((calls) => {
+		using sample = makeAudioDataBackedSample();
+
+		const out = new Float32Array(NUM_FRAMES * NUM_CHANNELS);
+		sample.copyTo(out, { planeIndex: 0, format: 'f32' });
+
+		for (let f = 0; f < NUM_FRAMES; f++) {
+			for (let c = 0; c < NUM_CHANNELS; c++) {
+				expect(out[f * NUM_CHANNELS + c]).toBe(Math.fround(cellValue(f, c) / 32767));
+			}
+		}
+
+		// The native conversion must have been attempted (and thrown)
+		expect(calls.some(options => options.format === 'f32')).toBe(true);
+	});
+});
+
+test('copyTo, AudioData-backed, control case with native format conversion', () => {
+	using sample = makeAudioDataBackedSample();
+
+	for (let c = 0; c < NUM_CHANNELS; c++) {
+		const out = new Int16Array(NUM_FRAMES);
+		sample.copyTo(out, { planeIndex: c, format: 's16-planar' });
+
+		for (let f = 0; f < NUM_FRAMES; f++) {
+			// Allow off-by-one
+			expect(Math.abs(out[f]! - cellValue(f, c))).toBeLessThanOrEqual(1);
+		}
+	}
+
+	const out = new Float32Array(NUM_FRAMES * NUM_CHANNELS);
+	sample.copyTo(out, { planeIndex: 0, format: 'f32' });
+
+	for (let f = 0; f < NUM_FRAMES; f++) {
+		for (let c = 0; c < NUM_CHANNELS; c++) {
+			expect(out[f * NUM_CHANNELS + c]).toBe(Math.fround(cellValue(f, c) / 32767));
+		}
+	}
+});
+
+const withMockedNativeCopyTo = (fn: (calls: AudioDataCopyToOptions[]) => void) => {
+	// eslint-disable-next-line @typescript-eslint/unbound-method
+	const original = AudioData.prototype.copyTo;
+	const calls: AudioDataCopyToOptions[] = [];
+
+	AudioData.prototype.copyTo = function (this: AudioData, destination, options) {
+		calls.push(options);
+		if (options.format !== undefined && options.format !== 'f32-planar') {
+			throw new DOMException('Format conversion not supported.', 'NotSupportedError');
+		}
+
+		return original.call(this, destination, options);
+	};
+
+	try {
+		fn(calls);
+	} finally {
+		AudioData.prototype.copyTo = original;
+	}
+};
+
+const makeAudioDataBackedSample = () => {
+	const data = new Float32Array(NUM_FRAMES * NUM_CHANNELS);
+	for (let f = 0; f < NUM_FRAMES; f++) {
+		for (let c = 0; c < NUM_CHANNELS; c++) {
+			data[f * NUM_CHANNELS + c] = cellValue(f, c) / 32767;
+		}
+	}
+
+	return new AudioSample(new AudioData({
+		format: 'f32',
+		sampleRate: SAMPLE_RATE,
+		numberOfFrames: NUM_FRAMES,
+		numberOfChannels: NUM_CHANNELS,
+		timestamp: TIMESTAMP * 1e6,
+		data,
+	}));
+};

test/browser/color-space.test.ts

@@ -0,0 +1,232 @@
+import { expect, test } from 'vitest';
+import { VideoCodec } from '../../src/codec.js';
+import { ALL_FORMATS } from '../../src/input-format.js';
+import { Input } from '../../src/input.js';
+import { EncodedPacketSink } from '../../src/media-sink.js';
+import { EncodedVideoPacketSource } from '../../src/media-source.js';
+import { assert, colorSpaceIsComplete } from '../../src/misc.js';
+import { Output } from '../../src/output.js';
+import { MkvOutputFormat, Mp4OutputFormat } from '../../src/output-format.js';
+import { EncodedPacket } from '../../src/packet.js';
+import { BufferSource, Source, UrlSource } from '../../src/source.js';
+import { BufferTarget } from '../../src/target.js';
+
+test('Color space extraction, AVC in MP4', async () => {
+	const source = await readPackets('/video.mp4');
+	const buffer = await remuxWithoutColorSpace(new Mp4OutputFormat(), source);
+
+	await expectColorSpace(new BufferSource(buffer), {
+		primaries: undefined,
+		transfer: undefined,
+		matrix: 'bt470bg',
+		fullRange: true,
+	});
+});
+
+test('Color space extraction, HEVC in MP4', async () => {
+	const source = await readPackets('/video-h265.mp4');
+	const buffer = await remuxWithoutColorSpace(new Mp4OutputFormat(), source);
+
+	await expectColorSpace(new BufferSource(buffer), {
+		primaries: undefined,
+		transfer: undefined,
+		matrix: 'bt470bg',
+		fullRange: true,
+	});
+});
+
+test('Color space extraction, VP9 in MP4', async () => {
+	const source = await readPackets('/vp9-smpte170m.webm');
+	const buffer = await remuxWithoutColorSpace(new Mp4OutputFormat(), source);
+
+	await expectColorSpace(new BufferSource(buffer), {
+		primaries: 'smpte170m',
+		transfer: 'smpte170m',
+		matrix: 'smpte170m',
+		fullRange: false,
+	});
+});
+
+test('Color space extraction, AV1 in MP4', async () => {
+	const source = await encodePackets('av1', 'av01.0.04M.08');
+	const buffer = await remuxWithoutColorSpace(new Mp4OutputFormat(), source);
+
+	await expectCompleteColorSpace(new BufferSource(buffer));
+});
+
+test('Color space extraction, AVC in Matroska', async () => {
+	const source = await readPackets('/video.mp4');
+	const buffer = await remuxWithoutColorSpace(new MkvOutputFormat(), source);
+
+	await expectColorSpace(new BufferSource(buffer), {
+		primaries: undefined,
+		transfer: undefined,
+		matrix: 'bt470bg',
+		fullRange: true,
+	});
+});
+
+test('Color space extraction, HEVC in Matroska', async () => {
+	const source = await readPackets('/video-h265.mp4');
+	const buffer = await remuxWithoutColorSpace(new MkvOutputFormat(), source);
+
+	await expectColorSpace(new BufferSource(buffer), {
+		// The SPS only signals the matrix and the range, leaving the rest unspecified
+		primaries: undefined,
+		transfer: undefined,
+		matrix: 'bt470bg',
+		fullRange: true,
+	});
+});
+
+test('Color space extraction, VP9 in Matroska', async () => {
+	const source = await readPackets('/vp9-smpte170m.webm');
+	const buffer = await remuxWithoutColorSpace(new MkvOutputFormat(), source);
+
+	await expectColorSpace(new BufferSource(buffer), {
+		primaries: 'smpte170m',
+		transfer: 'smpte170m',
+		matrix: 'smpte170m',
+		fullRange: false,
+	});
+});
+
+test('Color space extraction, AV1 in Matroska', async () => {
+	const source = await encodePackets('av1', 'av01.0.04M.08');
+	const buffer = await remuxWithoutColorSpace(new MkvOutputFormat(), source);
+
+	await expectCompleteColorSpace(new BufferSource(buffer));
+});
+
+const readPackets = async (path: string) => {
+	using input = new Input({
+		source: new UrlSource(path),
+		formats: ALL_FORMATS,
+	});
+
+	const track = await input.getPrimaryVideoTrack();
+	assert(track);
+
+	const codec = await track.getCodec();
+	assert(codec);
+
+	const decoderConfig = await track.getDecoderConfig();
+	assert(decoderConfig);
+
+	const sink = new EncodedPacketSink(track);
+	const packets: EncodedPacket[] = [];
+
+	for await (const packet of sink.packets()) {
+		packets.push(packet);
+
+		if (packets.length === 10) {
+			break;
+		}
+	}
+
+	return { codec, decoderConfig, packets };
+};
+
+const encodePackets = async (codec: VideoCodec, codecString: string) => {
+	const width = 320;
+	const height = 240;
+
+	const canvas = new OffscreenCanvas(width, height);
+	const context = canvas.getContext('2d')!;
+
+	let decoderConfig: VideoDecoderConfig | null = null;
+	const packets: EncodedPacket[] = [];
+
+	// We manually go through WebCodecs here so we can intercept the decoder config
+	const encoder = new VideoEncoder({
+		output: (chunk, metadata) => {
+			decoderConfig ??= metadata?.decoderConfig ?? null;
+			packets.push(EncodedPacket.fromEncodedChunk(chunk));
+		},
+		error: (error) => {
+			throw error;
+		},
+	});
+	encoder.configure({ codec: codecString, width, height });
+
+	for (let i = 0; i < 10; i++) {
+		context.fillStyle = `hsl(${36 * i}, 100%, 50%)`;
+		context.fillRect(0, 0, width, height);
+		context.fillStyle = 'black';
+		context.fillRect(20 * i, 10 * i, 80, 80);
+
+		const frame = new VideoFrame(canvas, { timestamp: i * 1e6 / 30, duration: 1e6 / 30 });
+		encoder.encode(frame, { keyFrame: i === 0 });
+		frame.close();
+	}
+
+	await encoder.flush();
+	encoder.close();
+
+	assert(decoderConfig);
+
+	return { codec, decoderConfig, packets };
+};
+
+const remuxWithoutColorSpace = async (
+	format: Mp4OutputFormat | MkvOutputFormat,
+	source: { codec: VideoCodec; decoderConfig: VideoDecoderConfig; packets: EncodedPacket[] },
+) => {
+	const output = new Output({
+		format,
+		target: new BufferTarget(),
+	});
+
+	const videoSource = new EncodedVideoPacketSource(source.codec);
+	output.addVideoTrack(videoSource);
+
+	await output.start();
+
+	const decoderConfig = { ...source.decoderConfig };
+	delete decoderConfig.colorSpace;
+
+	let isFirstPacket = true;
+	for (const packet of source.packets) {
+		await videoSource.add(packet, isFirstPacket ? { decoderConfig } : undefined);
+		isFirstPacket = false;
+	}
+
+	await output.finalize();
+
+	const buffer = output.target.buffer;
+	assert(buffer);
+
+	return buffer;
+};
+
+const expectColorSpace = async (source: Source, expected: VideoColorSpaceInit) => {
+	using input = new Input({
+		source,
+		formats: ALL_FORMATS,
+	});
+
+	const track = await input.getPrimaryVideoTrack();
+	assert(track);
+
+	const decoderConfig = await track.getDecoderConfig();
+	assert(decoderConfig);
+
+	expect(await track.getColorSpace()).toEqual(expected);
+	expect({ ...decoderConfig.colorSpace }).toEqual(expected);
+};
+
+const expectCompleteColorSpace = async (source: Source) => {
+	using input = new Input({
+		source,
+		formats: ALL_FORMATS,
+	});
+
+	const track = await input.getPrimaryVideoTrack();
+	assert(track);
+
+	const decoderConfig = await track.getDecoderConfig();
+	assert(decoderConfig);
+
+	expect(colorSpaceIsComplete(await track.getColorSpace())).toBe(true);
+	expect(colorSpaceIsComplete(decoderConfig.colorSpace)).toBe(true);
+};

test/browser/quality.test.ts

@@ -76,6 +76,7 @@ const convertVideo = async (quality: Quality, codec: VideoCodec, allowMissingQua
 		trim: { start: 0, end: 2 },
 		video: { codec, quality },
 		audio: { discard: true },
+		showWarnings: false,
 	});
 
 	if (allowMissingQuantizerSupport && !conversion.isValid) {

test/browser/read-mp4.test.ts

@@ -0,0 +1,27 @@
+import { expect, test } from 'vitest';
+import { Input } from '../../src/input.js';
+import { UrlSource } from '../../src/source.js';
+import { ALL_FORMATS } from '../../src/input-format.js';
+import { AudioSampleSink } from '../../src/media-sink.js';
+import { assert } from '../../src/misc.js';
+
+test('Encrypted MP4 without senc', async () => {
+	using input = new Input({
+		source: new UrlSource('/639955605-b752f7ad-7ce8-43c2-8ae9-88e74c6ae696.mp4'),
+		formats: ALL_FORMATS,
+		formatOptions: {
+			isobmff: {
+				resolveKeyId: () => 'aabbccddeeff00112233445566778899',
+			},
+		},
+	});
+
+	const audioTrack = await input.getPrimaryAudioTrack();
+	assert(audioTrack);
+
+	// Test that it can decode
+	const sink = new AudioSampleSink(audioTrack);
+	using firstSample = await sink.getSample(0);
+	assert(firstSample);
+	expect(firstSample.timestamp).toBe(0);
+});

test/node/read-mp4.test.ts

@@ -14,6 +14,7 @@ import {
 	Mp4OutputFormat,
 	Output,
 } from '../../src/index.js';
+import { assert, toUint8Array } from '../../src/misc.js';
 
 const __dirname = new URL('.', import.meta.url).pathname;
 
@@ -140,3 +141,43 @@ test('QuickTime nclc color information', async () => {
 	});
 	expect(await track.hasHighDynamicRange()).toBe(true);
 });
+
+// Annex B isn't supposed to exist in MP4, but some files have it anyway, with an empty avcC box
+test('Annex B', async () => {
+	const filePath = path.join(__dirname, '..', 'public/annex-b.mp4');
+	using input = new Input({
+		source: new FilePathSource(filePath),
+		formats: ALL_FORMATS,
+	});
+
+	const track = await input.getPrimaryVideoTrack();
+	if (!track) throw new Error('No video track found');
+
+	const decoderConfig = (await track.getDecoderConfig())!;
+	expect(decoderConfig.codec).toBe('avc1.424028');
+	expect(decoderConfig.description).toBeUndefined();
+
+	const sink = new EncodedPacketSink(track);
+	const firstPacket = (await sink.getFirstPacket())!;
+	expect([...firstPacket.data.slice(0, 4)]).toEqual([0, 0, 0, 1]);
+});
+
+test('HE-AAC v2 audio config is parsed correctly', async () => {
+	const filePath = path.join(__dirname, '..', 'public/he-aac-v2.mp4');
+	using input = new Input({
+		source: new FilePathSource(filePath),
+		formats: ALL_FORMATS,
+	});
+
+	const track = await input.getPrimaryAudioTrack();
+	assert(track);
+
+	expect(await track.getSampleRate()).toBe(44100);
+	expect(await track.getNumberOfChannels()).toBe(2);
+
+	const decoderConfig = (await track.getDecoderConfig())!;
+	expect(decoderConfig.codec).toBe('mp4a.40.29');
+	expect(decoderConfig.sampleRate).toBe(44100);
+	expect(decoderConfig.numberOfChannels).toBe(2);
+	expect([...toUint8Array(decoderConfig.description!)]).toEqual([0xeb, 0x8a, 0x08, 0x00]);
+});