Fix Matroska files where no audio packets are key packets (fixes #192)

dev/demux.html

@@ -14,8 +14,8 @@
 			source: new Mediabunny.BlobSource(file),
 		});
 
-		const videoTrack = await input.getPrimaryVideoTrack();
+		const audioTrack = await input.getPrimaryAudioTrack();
-		const sink = new Mediabunny.EncodedPacketSink(videoTrack);
+		const sink = new Mediabunny.EncodedPacketSink(audioTrack);
 
 		for await (const packet of sink.packets()) {
 			console.log(packet.timestamp, packet.duration, packet.type)

src/matroska/matroska-demuxer.ts

@@ -1374,9 +1374,17 @@ export class MatroskaDemuxer extends Demuxer {
 				const relativeTimestamp = readI16Be(slice);
 
 				const flags = readU8(slice);
-				const isKeyFrame = !!(flags & 0x80);
 				const lacing = (flags >> 1) & 0x3 as BlockLacing; // If the block is laced, we'll expand it later
 
+				let isKeyFrame = !!(flags & 0x80);
+				if (trackData.track.info?.type === 'audio' && trackData.track.info.codec) {
+					// Some files don't mark their audio packets as key packets (I'm looking at you, Firefox). But, we
+					// can fix this in most cases: if we recognize the codec of the track, then we know every packet is
+					// necessarily a key packet, no matter what the container says.
+					// https://github.com/Vanilagy/mediabunny/issues/192
+					isKeyFrame = true;
+				}
+
 				const blockData = readBytes(slice, size - (slice.filePos - dataStartPos));
 				const hasDecodingInstructions = trackData.track.decodingInstructions.length > 0;
 

src/muxer.ts

@@ -46,21 +46,21 @@ export abstract class Muxer {
 
 	private trackTimestampInfo = new WeakMap<OutputTrack, {
 		maxTimestamp: number;
-		maxTimestampBeforeLastKeyFrame: number;
+		maxTimestampBeforeLastKeyPacket: number;
 	}>();
 
-	protected validateAndNormalizeTimestamp(track: OutputTrack, timestampInSeconds: number, isKeyFrame: boolean) {
+	protected validateAndNormalizeTimestamp(track: OutputTrack, timestampInSeconds: number, isKeyPacket: boolean) {
 		timestampInSeconds += track.source._timestampOffset;
 
 		let timestampInfo = this.trackTimestampInfo.get(track);
 		if (!timestampInfo) {
-			if (!isKeyFrame) {
+			if (!isKeyPacket) {
-				throw new Error('First frame must be a key frame.');
+				throw new Error('First packet must be a key packet.');
 			}
 
 			timestampInfo = {
 				maxTimestamp: timestampInSeconds,
-				maxTimestampBeforeLastKeyFrame: timestampInSeconds,
+				maxTimestampBeforeLastKeyPacket: timestampInSeconds,
 			};
 			this.trackTimestampInfo.set(track, timestampInfo);
 		}
@@ -69,15 +69,15 @@ export abstract class Muxer {
 			throw new Error(`Timestamps must be non-negative (got ${timestampInSeconds}s).`);
 		}
 
-		if (isKeyFrame) {
+		if (isKeyPacket) {
-			timestampInfo.maxTimestampBeforeLastKeyFrame = timestampInfo.maxTimestamp;
+			timestampInfo.maxTimestampBeforeLastKeyPacket = timestampInfo.maxTimestamp;
 		}
 
-		if (timestampInSeconds < timestampInfo.maxTimestampBeforeLastKeyFrame) {
+		if (timestampInSeconds < timestampInfo.maxTimestampBeforeLastKeyPacket) {
 			throw new Error(
-				`Timestamps cannot be smaller than the highest timestamp of the previous GOP (a GOP begins with a key`
+				`Timestamps cannot be smaller than the largest timestamp of the previous GOP (a GOP begins with a key`
-				+ ` frame and ends right before the next key frame). Got ${timestampInSeconds}s, but highest timestamp`
+				+ ` packet and ends right before the next key packet). Got ${timestampInSeconds}s, but largest`
-				+ ` is ${timestampInfo.maxTimestampBeforeLastKeyFrame}s.`,
+				+ ` timestamp is ${timestampInfo.maxTimestampBeforeLastKeyPacket}s.`,
 			);
 		}