mediabunny 1.15.2 → 1.16.1

package/README.md

@@ -39,6 +39,10 @@ Mediabunny is a JavaScript library for reading, writing, and converting media fi
   <a href="https://www.reactvideoeditor.com/" target="_blank">
     <img src="./docs/public/sponsors/rve.svg" width="40" height="40" alt="React Video Editor">
   </a>
+  &nbsp;&nbsp;&nbsp;&nbsp;
+  <a href="https://www.mux.com/" target="_blank">
+    <img src="./docs/public/sponsors/mux.jpg" width="40" height="40" alt="Mux">
+  </a>
 </div>
 
 [Sponsor Mediabunny's development](https://github.com/sponsors/Vanilagy)

package/dist/bundles/mediabunny.cjs

@@ -1610,7 +1610,7 @@ var Mediabunny = (() => {
       }
       if (timestampInSeconds < timestampInfo.maxTimestampBeforeLastKeyFrame) {
         throw new Error(
-          `Timestamps cannot be smaller than the highest timestamp of the previous run (a run begins with a key frame and ends right before the next key frame). Got ${timestampInSeconds}s, but highest timestamp is ${timestampInfo.maxTimestampBeforeLastKeyFrame}s.`
+          `Timestamps cannot be smaller than the highest 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 is ${timestampInfo.maxTimestampBeforeLastKeyFrame}s.`
         );
       }
       timestampInfo.maxTimestamp = Math.max(timestampInfo.maxTimestamp, timestampInSeconds);
@@ -1812,7 +1812,7 @@ var Mediabunny = (() => {
       } else if (lengthSize === 2) {
         nalUnitLength = dataView.getUint16(offset, false);
       } else if (lengthSize === 3) {
-        nalUnitLength = (dataView.getUint16(offset, false) << 8) + dataView.getUint8(offset + 2);
+        nalUnitLength = getUint24(dataView, offset, false);
       } else if (lengthSize === 4) {
         nalUnitLength = dataView.getUint32(offset, false);
       } else {
@@ -3077,10 +3077,11 @@ var Mediabunny = (() => {
     ]);
   };
   var mdat = (reserveLargeSize) => ({ type: "mdat", largeSize: reserveLargeSize });
-  var moov = (muxer, fragmented = false) => box("moov", void 0, [
+  var free = (size) => ({ type: "free", size });
+  var moov = (muxer) => box("moov", void 0, [
     mvhd(muxer.creationTime, muxer.trackDatas),
     ...muxer.trackDatas.map((x) => trak(x, muxer.creationTime)),
-    fragmented ? mvex(muxer.trackDatas) : null,
+    muxer.isFragmented ? mvex(muxer.trackDatas) : null,
     udta(muxer)
   ]);
   var mvhd = (creationTime, trackDatas) => {
@@ -4883,9 +4884,9 @@ var Mediabunny = (() => {
     return value;
   };
   var readU24Be = (slice) => {
-    const high = readU16Be(slice);
-    const low = readU8(slice);
-    return high * 256 + low;
+    const value = getUint24(slice.view, slice.bufferPos, false);
+    slice.bufferPos += 3;
+    return value;
   };
   var readI16Be = (slice) => {
     const value = slice.view.getInt16(slice.bufferPos, false);
@@ -5054,6 +5055,7 @@ var Mediabunny = (() => {
       this.auxWriter = this.auxTarget._createWriter();
       this.auxBoxWriter = new IsobmffBoxWriter(this.auxWriter);
       this.mdat = null;
+      this.ftypSize = null;
       this.trackDatas = [];
       this.allTracksKnown = promiseWithResolvers();
       this.creationTime = Math.floor(Date.now() / 1e3) + TIMESTAMP_OFFSET;
@@ -5090,8 +5092,16 @@ var Mediabunny = (() => {
           this.format._options.onFtyp(data, start);
         }
       }
+      this.ftypSize = this.writer.getPos();
       if (this.fastStart === "in-memory") {
-        this.mdat = mdat(false);
+      } else if (this.fastStart === "reserve") {
+        for (const track of this.output._tracks) {
+          if (track.metadata.maximumPacketCount === void 0) {
+            throw new Error(
+              "All tracks must specify maximumPacketCount in their metadata when using fastStart: 'reserve'."
+            );
+          }
+        }
       } else if (this.isFragmented) {
       } else {
         if (this.format._options.onMdat) {
@@ -5548,6 +5558,8 @@ var Mediabunny = (() => {
       if (this.isFragmented) {
         trackData.sampleQueue.push(sample);
         await this.interleaveSamples();
+      } else if (this.fastStart === "reserve") {
+        await this.registerSampleFastStartReserve(trackData, sample);
       } else {
         await this.addSampleToTrack(trackData, sample);
       }
@@ -5555,6 +5567,15 @@ var Mediabunny = (() => {
     async addSampleToTrack(trackData, sample) {
       if (!this.isFragmented) {
         trackData.samples.push(sample);
+        if (this.fastStart === "reserve") {
+          const maximumPacketCount = trackData.track.metadata.maximumPacketCount;
+          assert(maximumPacketCount !== void 0);
+          if (trackData.samples.length > maximumPacketCount) {
+            throw new Error(
+              `Track #${trackData.track.id} has already reached the maximum packet count (${maximumPacketCount}). Either add less packets or increase the maximum packet count.`
+            );
+          }
+        }
       }
       let beginNewChunk = false;
       if (!trackData.currentChunk) {
@@ -5631,10 +5652,8 @@ var Mediabunny = (() => {
     }
     async interleaveSamples(isFinalCall = false) {
       assert(this.isFragmented);
-      if (!isFinalCall) {
-        if (!this.allTracksAreKnown()) {
-          return;
-        }
+      if (!isFinalCall && !this.allTracksAreKnown()) {
+        return;
       }
       outer:
         while (true) {
@@ -5663,7 +5682,7 @@ var Mediabunny = (() => {
         if (this.format._options.onMoov) {
           this.writer.startTrackingWrites();
         }
-        const movieBox = moov(this, true);
+        const movieBox = moov(this);
         this.boxWriter.writeBox(movieBox);
         if (this.format._options.onMoov) {
           const { data, start } = this.writer.stopTrackingWrites();
@@ -5727,6 +5746,46 @@ var Mediabunny = (() => {
         await this.writer.flush();
       }
     }
+    async registerSampleFastStartReserve(trackData, sample) {
+      if (this.allTracksAreKnown()) {
+        if (!this.mdat) {
+          const moovBox = moov(this);
+          const moovSize = this.boxWriter.measureBox(moovBox);
+          const reservedSize = moovSize + this.computeSampleTableSizeUpperBound() + 4096;
+          assert(this.ftypSize !== null);
+          this.writer.seek(this.ftypSize + reservedSize);
+          if (this.format._options.onMdat) {
+            this.writer.startTrackingWrites();
+          }
+          this.mdat = mdat(true);
+          this.boxWriter.writeBox(this.mdat);
+          for (const trackData2 of this.trackDatas) {
+            for (const sample2 of trackData2.sampleQueue) {
+              await this.addSampleToTrack(trackData2, sample2);
+            }
+            trackData2.sampleQueue.length = 0;
+          }
+        }
+        await this.addSampleToTrack(trackData, sample);
+      } else {
+        trackData.sampleQueue.push(sample);
+      }
+    }
+    computeSampleTableSizeUpperBound() {
+      assert(this.fastStart === "reserve");
+      let upperBound = 0;
+      for (const trackData of this.trackDatas) {
+        const n = trackData.track.metadata.maximumPacketCount;
+        assert(n !== void 0);
+        upperBound += (4 + 4) * Math.ceil(2 / 3 * n);
+        upperBound += 4 * n;
+        upperBound += (4 + 4) * Math.ceil(2 / 3 * n);
+        upperBound += (4 + 4 + 4) * Math.ceil(2 / 3 * n);
+        upperBound += 4 * n;
+        upperBound += 8 * n;
+      }
+      return upperBound;
+    }
     // eslint-disable-next-line @typescript-eslint/no-misused-promises
     async onTrackClose(track) {
       const release = await this.mutex.acquire();
@@ -5766,7 +5825,7 @@ var Mediabunny = (() => {
         }
       }
       if (this.fastStart === "in-memory") {
-        assert(this.mdat);
+        this.mdat = mdat(false);
         let mdatSize;
         for (let i = 0; i < 2; i++) {
           const movieBox2 = moov(this);
@@ -5828,11 +5887,22 @@ var Mediabunny = (() => {
           const { data, start } = this.writer.stopTrackingWrites();
           this.format._options.onMdat(data, start);
         }
-        if (this.format._options.onMoov) {
-          this.writer.startTrackingWrites();
-        }
         const movieBox = moov(this);
-        this.boxWriter.writeBox(movieBox);
+        if (this.fastStart === "reserve") {
+          assert(this.ftypSize !== null);
+          this.writer.seek(this.ftypSize);
+          if (this.format._options.onMoov) {
+            this.writer.startTrackingWrites();
+          }
+          this.boxWriter.writeBox(movieBox);
+          const remainingSpace = this.boxWriter.offsets.get(this.mdat) - this.writer.getPos();
+          this.boxWriter.writeBox(free(remainingSpace));
+        } else {
+          if (this.format._options.onMoov) {
+            this.writer.startTrackingWrites();
+          }
+          this.boxWriter.writeBox(movieBox);
+        }
         if (this.format._options.onMoov) {
           const { data, start } = this.writer.stopTrackingWrites();
           this.format._options.onMoov(data, start);
@@ -6897,10 +6967,8 @@ ${cue.notes ?? ""}`;
       }
     }
     async interleaveChunks(isFinalCall = false) {
-      if (!isFinalCall) {
-        if (!this.allTracksAreKnown()) {
-          return;
-        }
+      if (!isFinalCall && !this.allTracksAreKnown()) {
+        return;
       }
       outer:
         while (true) {
@@ -12410,8 +12478,10 @@ ${cue.notes ?? ""}`;
       if (!options || typeof options !== "object") {
         throw new TypeError("options must be an object.");
       }
-      if (options.fastStart !== void 0 && ![false, "in-memory", "fragmented"].includes(options.fastStart)) {
-        throw new TypeError('options.fastStart, when provided, must be false, "in-memory", or "fragmented".');
+      if (options.fastStart !== void 0 && ![false, "in-memory", "reserve", "fragmented"].includes(options.fastStart)) {
+        throw new TypeError(
+          "options.fastStart, when provided, must be false, 'in-memory', 'reserve', or 'fragmented'."
+        );
       }
       if (options.minimumFragmentDuration !== void 0 && (!Number.isFinite(options.minimumFragmentDuration) || options.minimumFragmentDuration < 0)) {
         throw new TypeError("options.minimumFragmentDuration, when provided, must be a non-negative number.");
@@ -14381,6 +14451,9 @@ ${cue.notes ?? ""}`;
     if (metadata.name !== void 0 && typeof metadata.name !== "string") {
       throw new TypeError("metadata.name, when provided, must be a string.");
     }
+    if (metadata.maximumPacketCount !== void 0 && (!Number.isInteger(metadata.maximumPacketCount) || metadata.maximumPacketCount < 0)) {
+      throw new TypeError("metadata.maximumPacketCount, when provided, must be a non-negative integer.");
+    }
   };
   var Output = class {
     /**
@@ -18232,6 +18305,7 @@ ${cue.notes ?? ""}`;
       this.currentCluster = null;
       this.currentBlock = null;
       this.currentCueTime = null;
+      this.currentDecodingInstruction = null;
       this.currentTagTargetIsMovie = true;
       this.currentSimpleTagName = null;
       this.currentAttachedFile = null;
@@ -18494,6 +18568,7 @@ ${cue.notes ?? ""}`;
       let dataSlice = this.reader.requestSlice(dataStartPos, size);
       if (dataSlice instanceof Promise) dataSlice = await dataSlice;
       const cluster = {
+        segment,
         elementStartPos,
         elementEndPos: dataStartPos + size,
         dataStartPos,
@@ -18506,8 +18581,8 @@ ${cue.notes ?? ""}`;
       if (dataSlice) {
         this.readContiguousElements(dataSlice);
       }
-      for (const [trackId, trackData] of cluster.trackData) {
-        const track = segment.tracks.find((x) => x.id === trackId) ?? null;
+      for (const [, trackData] of cluster.trackData) {
+        const track = trackData.track;
         assert(trackData.blocks.length > 0);
         let blockReferencesExist = false;
         let hasLacedBlocks = false;
@@ -18531,7 +18606,7 @@ ${cue.notes ?? ""}`;
             const nextEntry = trackData.presentationTimestamps[i + 1];
             currentBlock.duration = nextEntry.timestamp - currentBlock.timestamp;
           } else if (currentBlock.duration === 0) {
-            if (track?.defaultDuration != null) {
+            if (track.defaultDuration != null) {
               if (currentBlock.lacing === 0 /* None */) {
                 currentBlock.duration = track.defaultDuration;
               } else {
@@ -18547,12 +18622,10 @@ ${cue.notes ?? ""}`;
         const lastBlock = trackData.blocks[last(trackData.presentationTimestamps).blockIndex];
         trackData.startTimestamp = firstBlock.timestamp;
         trackData.endTimestamp = lastBlock.timestamp + lastBlock.duration;
-        if (track) {
-          insertSorted(track.clusters, cluster, (x) => x.elementStartPos);
-          const hasKeyFrame = trackData.firstKeyFrameTimestamp !== null;
-          if (hasKeyFrame) {
-            insertSorted(track.clustersWithKeyFrame, cluster, (x) => x.elementStartPos);
-          }
+        insertSorted(track.clusters, cluster, (x) => x.elementStartPos);
+        const hasKeyFrame = trackData.firstKeyFrameTimestamp !== null;
+        if (hasKeyFrame) {
+          insertSorted(track.clustersWithKeyFrame, cluster, (x) => x.elementStartPos);
         }
       }
       insertSorted(segment.clusters, cluster, (x) => x.elementStartPos);
@@ -18562,7 +18635,12 @@ ${cue.notes ?? ""}`;
     getTrackDataInCluster(cluster, trackNumber) {
       let trackData = cluster.trackData.get(trackNumber);
       if (!trackData) {
+        const track = cluster.segment.tracks.find((x) => x.id === trackNumber);
+        if (!track) {
+          return null;
+        }
         trackData = {
+          track,
           startTimestamp: 0,
           endTimestamp: 0,
           firstKeyFrameTimestamp: null,
@@ -18579,6 +18657,10 @@ ${cue.notes ?? ""}`;
         if (originalBlock.lacing === 0 /* None */) {
           continue;
         }
+        if (!originalBlock.decoded) {
+          originalBlock.data = this.decodeBlockData(track, originalBlock.data);
+          originalBlock.decoded = true;
+        }
         const slice = FileSlice.tempFromBytes(originalBlock.data);
         const frameSizes = [];
         const frameCount = readU8(slice) + 1;
@@ -18643,7 +18725,7 @@ ${cue.notes ?? ""}`;
         for (let i = 0; i < frameCount; i++) {
           const frameSize = frameSizes[i];
           const frameData = readBytes(slice, frameSize);
-          const blockDuration = originalBlock.duration || frameCount * (track?.defaultDuration ?? 0);
+          const blockDuration = originalBlock.duration || frameCount * (track.defaultDuration ?? 0);
           const frameTimestamp = originalBlock.timestamp + blockDuration * i / frameCount;
           const frameDuration = blockDuration / frameCount;
           blocks.splice(blockIndex + i, 0, {
@@ -18652,7 +18734,8 @@ ${cue.notes ?? ""}`;
             isKeyFrame: originalBlock.isKeyFrame,
             referencedTimestamps: originalBlock.referencedTimestamps,
             data: frameData,
-            lacing: 0 /* None */
+            lacing: 0 /* None */,
+            decoded: true
           });
         }
         blockIndex += frameCount;
@@ -18776,9 +18859,16 @@ ${cue.notes ?? ""}`;
               defaultDuration: null,
               name: null,
               languageCode: UNDETERMINED_LANGUAGE,
+              decodingInstructions: [],
               info: null
             };
             this.readContiguousElements(slice.slice(dataStartPos, size));
+            if (this.currentTrack.decodingInstructions.some((instruction) => {
+              return instruction.data?.type !== "decompress" || instruction.data.algorithm !== 3 /* HeaderStripping */;
+            })) {
+              console.warn(`Track #${this.currentTrack.id} has an unsupported content encoding; dropping.`);
+              this.currentTrack = null;
+            }
             if (this.currentTrack && this.currentTrack.id !== -1 && this.currentTrack.codecId && this.currentTrack.info) {
               const slashIndex = this.currentTrack.codecId.indexOf("/");
               const codecIdWithoutSuffix = slashIndex === -1 ? this.currentTrack.codecId : this.currentTrack.codecId.slice(0, slashIndex);
@@ -19129,11 +19219,14 @@ ${cue.notes ?? ""}`;
             if (!this.currentCluster) break;
             const trackNumber = readVarInt(slice);
             if (trackNumber === null) break;
+            const trackData = this.getTrackDataInCluster(this.currentCluster, trackNumber);
+            if (!trackData) break;
             const relativeTimestamp = readI16Be(slice);
             const flags = readU8(slice);
             const isKeyFrame = !!(flags & 128);
             const lacing = flags >> 1 & 3;
-            const trackData = this.getTrackDataInCluster(this.currentCluster, trackNumber);
+            const blockData = readBytes(slice, size - (slice.filePos - dataStartPos));
+            const hasDecodingInstructions = trackData.track.decodingInstructions.length > 0;
             trackData.blocks.push({
               timestamp: relativeTimestamp,
               // We'll add the cluster's timestamp to this later
@@ -19141,8 +19234,9 @@ ${cue.notes ?? ""}`;
               // Will set later
               isKeyFrame,
               referencedTimestamps: [],
-              data: readBytes(slice, size - (slice.filePos - dataStartPos)),
-              lacing
+              data: blockData,
+              lacing,
+              decoded: !hasDecodingInstructions
             });
           }
           ;
@@ -19165,10 +19259,13 @@ ${cue.notes ?? ""}`;
             if (!this.currentCluster) break;
             const trackNumber = readVarInt(slice);
             if (trackNumber === null) break;
+            const trackData = this.getTrackDataInCluster(this.currentCluster, trackNumber);
+            if (!trackData) break;
             const relativeTimestamp = readI16Be(slice);
             const flags = readU8(slice);
             const lacing = flags >> 1 & 3;
-            const trackData = this.getTrackDataInCluster(this.currentCluster, trackNumber);
+            const blockData = readBytes(slice, size - (slice.filePos - dataStartPos));
+            const hasDecodingInstructions = trackData.track.decodingInstructions.length > 0;
             this.currentBlock = {
               timestamp: relativeTimestamp,
               // We'll add the cluster's timestamp to this later
@@ -19176,8 +19273,9 @@ ${cue.notes ?? ""}`;
               // Will set later
               isKeyFrame: true,
               referencedTimestamps: [],
-              data: readBytes(slice, size - (slice.filePos - dataStartPos)),
-              lacing
+              data: blockData,
+              lacing,
+              decoded: !hasDecodingInstructions
             };
             trackData.blocks.push(this.currentBlock);
           }
@@ -19322,10 +19420,119 @@ ${cue.notes ?? ""}`;
           }
           ;
           break;
+        case 28032 /* ContentEncodings */:
+          {
+            if (!this.currentTrack) break;
+            this.readContiguousElements(slice.slice(dataStartPos, size));
+            this.currentTrack.decodingInstructions.sort((a, b) => b.order - a.order);
+          }
+          ;
+          break;
+        case 25152 /* ContentEncoding */:
+          {
+            this.currentDecodingInstruction = {
+              order: 0,
+              scope: 1 /* Block */,
+              data: null
+            };
+            this.readContiguousElements(slice.slice(dataStartPos, size));
+            if (this.currentDecodingInstruction.data) {
+              this.currentTrack.decodingInstructions.push(this.currentDecodingInstruction);
+            }
+            this.currentDecodingInstruction = null;
+          }
+          ;
+          break;
+        case 20529 /* ContentEncodingOrder */:
+          {
+            if (!this.currentDecodingInstruction) break;
+            this.currentDecodingInstruction.order = readUnsignedInt(slice, size);
+          }
+          ;
+          break;
+        case 20530 /* ContentEncodingScope */:
+          {
+            if (!this.currentDecodingInstruction) break;
+            this.currentDecodingInstruction.scope = readUnsignedInt(slice, size);
+          }
+          ;
+          break;
+        case 20532 /* ContentCompression */:
+          {
+            if (!this.currentDecodingInstruction) break;
+            this.currentDecodingInstruction.data = {
+              type: "decompress",
+              algorithm: 0 /* Zlib */,
+              settings: null
+            };
+            this.readContiguousElements(slice.slice(dataStartPos, size));
+          }
+          ;
+          break;
+        case 16980 /* ContentCompAlgo */:
+          {
+            if (this.currentDecodingInstruction?.data?.type !== "decompress") break;
+            this.currentDecodingInstruction.data.algorithm = readUnsignedInt(slice, size);
+          }
+          ;
+          break;
+        case 16981 /* ContentCompSettings */:
+          {
+            if (this.currentDecodingInstruction?.data?.type !== "decompress") break;
+            this.currentDecodingInstruction.data.settings = readBytes(slice, size);
+          }
+          ;
+          break;
+        case 20533 /* ContentEncryption */:
+          {
+            if (!this.currentDecodingInstruction) break;
+            this.currentDecodingInstruction.data = {
+              type: "decrypt"
+            };
+          }
+          ;
+          break;
       }
       slice.filePos = dataStartPos + size;
       return true;
     }
+    decodeBlockData(track, rawData) {
+      assert(track.decodingInstructions.length > 0);
+      let currentData = rawData;
+      for (const instruction of track.decodingInstructions) {
+        assert(instruction.data);
+        switch (instruction.data.type) {
+          case "decompress":
+            {
+              switch (instruction.data.algorithm) {
+                case 3 /* HeaderStripping */:
+                  {
+                    if (instruction.data.settings && instruction.data.settings.length > 0) {
+                      const prefix = instruction.data.settings;
+                      const newData = new Uint8Array(prefix.length + currentData.length);
+                      newData.set(prefix, 0);
+                      newData.set(currentData, prefix.length);
+                      currentData = newData;
+                    }
+                  }
+                  ;
+                  break;
+                default:
+                  {
+                  }
+                  ;
+              }
+            }
+            ;
+            break;
+          default:
+            {
+            }
+            ;
+        }
+      }
+      return currentData;
+    }
     processTagValue(name, value) {
       if (!this.currentSegment?.metadataTags) return;
       const metadataTags = this.currentSegment.metadataTags;
@@ -19626,6 +19833,10 @@ ${cue.notes ?? ""}`;
       const trackData = cluster.trackData.get(this.internalTrack.id);
       const block = trackData.blocks[blockIndex];
       assert(block);
+      if (!block.decoded) {
+        block.data = this.internalTrack.demuxer.decodeBlockData(this.internalTrack, block.data);
+        block.decoded = true;
+      }
       const data = options.metadataOnly ? PLACEHOLDER_DATA : block.data;
       const timestamp = block.timestamp / this.internalTrack.segment.timestampFactor;
       const duration = block.duration / this.internalTrack.segment.timestampFactor;
@@ -20297,7 +20508,7 @@ ${cue.notes ?? ""}`;
       const description = new Uint8Array(
         1 + lacingValues.length + firstPacket.data.length + secondPacket.data.length + thirdPacket.data.length
       );
-      description[0] = lacingValues.length;
+      description[0] = 2;
       description.set(
         lacingValues,
         1