@pictx/gemini-veo-watermark-remover 0.2.3 → 0.2.4

package/dist/browser.js

@@ -272,6 +272,79 @@ var init_frameProcessor = __esm({
   }
 });
 
+// src/video/codecDescription.js
+function buildAvcDecoderConfigurationRecord(avcC) {
+  const spsList = avcC.SPS ?? [];
+  const ppsList = avcC.PPS ?? [];
+  let size = 7;
+  for (const sps of spsList) {
+    size += 2 + getNaluByteLength(sps);
+  }
+  size += 1;
+  for (const pps of ppsList) {
+    size += 2 + getNaluByteLength(pps);
+  }
+  if (avcC.ext?.length) {
+    size += avcC.ext.length;
+  }
+  const buffer = new Uint8Array(size);
+  const view = new DataView(buffer.buffer, buffer.byteOffset, buffer.byteLength);
+  let offset = 0;
+  buffer[offset++] = avcC.configurationVersion;
+  buffer[offset++] = avcC.AVCProfileIndication;
+  buffer[offset++] = avcC.profile_compatibility;
+  buffer[offset++] = avcC.AVCLevelIndication;
+  buffer[offset++] = (avcC.lengthSizeMinusOne & 3) + (63 << 2);
+  buffer[offset++] = spsList.length + (7 << 5);
+  for (const sps of spsList) {
+    const nalu = getNaluBytes(sps);
+    view.setUint16(offset, nalu.length);
+    offset += 2;
+    buffer.set(nalu, offset);
+    offset += nalu.length;
+  }
+  buffer[offset++] = ppsList.length;
+  for (const pps of ppsList) {
+    const nalu = getNaluBytes(pps);
+    view.setUint16(offset, nalu.length);
+    offset += 2;
+    buffer.set(nalu, offset);
+    offset += nalu.length;
+  }
+  if (avcC.ext?.length) {
+    buffer.set(avcC.ext, offset);
+  }
+  return buffer;
+}
+function extractCodecDescription(mp4, trackId) {
+  const trak = mp4.getTrackById(trackId);
+  const entries = trak?.mdia?.minf?.stbl?.stsd?.entries ?? [];
+  for (const entry of entries) {
+    if (entry.avcC) {
+      return buildAvcDecoderConfigurationRecord(entry.avcC);
+    }
+    if (entry.hvcC) {
+      throw new Error("HEVC (hvcC) WebCodecs decoding is not implemented yet.");
+    }
+  }
+  throw new Error("Codec description box (avcC or hvcC) not found for video track.");
+}
+function getNaluBytes(entry) {
+  if (entry?.nalu instanceof Uint8Array) return entry.nalu;
+  if (entry instanceof Uint8Array) return entry;
+  throw new Error("Invalid SPS/PPS entry in avcC box.");
+}
+function getNaluByteLength(entry) {
+  if (entry?.nalu instanceof Uint8Array) return entry.nalu.length;
+  if (typeof entry?.length === "number") return entry.length;
+  if (entry instanceof Uint8Array) return entry.length;
+  throw new Error("Invalid SPS/PPS entry in avcC box.");
+}
+var init_codecDescription = __esm({
+  "src/video/codecDescription.js"() {
+  }
+});
+
 // src/video/videoDecoder.base.js
 var VideoDecoderBase;
 var init_videoDecoder_base = __esm({
@@ -296,6 +369,7 @@ var init_videoDecoder_base = __esm({
 var WebCodecsDecoder;
 var init_videoDecoder_webcodecs = __esm({
   "src/video/videoDecoder.webcodecs.js"() {
+    init_codecDescription();
     init_videoDecoder_base();
     WebCodecsDecoder = class extends VideoDecoderBase {
       #videoInfo = null;
@@ -321,6 +395,15 @@ var init_videoDecoder_webcodecs = __esm({
         if (!videoTrack) {
           throw new Error("No video track found");
         }
+        this.#videoMp4 = createFile();
+        await new Promise((resolve, reject) => {
+          this.#videoMp4.onReady = resolve;
+          this.#videoMp4.onError = reject;
+          const buffer = this.#sourceBuffer.slice(0);
+          buffer.fileStart = 0;
+          this.#videoMp4.appendBuffer(buffer);
+          this.#videoMp4.flush();
+        });
         this.#videoInfo = {
           width: videoTrack.video.width,
           height: videoTrack.video.height,
@@ -328,21 +411,13 @@ var init_videoDecoder_webcodecs = __esm({
           duration: videoTrack.duration / videoTrack.timescale,
           fps: videoTrack.nb_samples / (videoTrack.duration / videoTrack.timescale),
           codec: videoTrack.codec,
-          videoTrackId: videoTrack.id
+          videoTrackId: videoTrack.id,
+          description: extractCodecDescription(this.#videoMp4, videoTrack.id)
         };
         const audioTrack = info.audioTracks?.[0] ?? null;
         if (audioTrack) {
           this.#audioData = await this.#readTrackSamples(createFile, audioTrack);
         }
-        this.#videoMp4 = createFile();
-        await new Promise((resolve, reject) => {
-          this.#videoMp4.onReady = resolve;
-          this.#videoMp4.onError = reject;
-          const buffer = this.#sourceBuffer.slice(0);
-          buffer.fileStart = 0;
-          this.#videoMp4.appendBuffer(buffer);
-          this.#videoMp4.flush();
-        });
         return this.#videoInfo;
       }
       /**
@@ -389,10 +464,13 @@ var init_videoDecoder_webcodecs = __esm({
         if (!this.#videoMp4 || !this.#videoInfo) {
           throw new Error("Call open() first");
         }
-        const { width, height, videoTrackId } = this.#videoInfo;
+        const { width, height, videoTrackId, frameCount } = this.#videoInfo;
         const pendingFrames = [];
         let resolveNext = null;
         let done = false;
+        let extractionDone = false;
+        let flushed = false;
+        let samplesSubmitted = 0;
         const decoder = new VideoDecoder({
           output: (frame) => {
             const canvas = new OffscreenCanvas(width, height);
@@ -416,12 +494,18 @@ var init_videoDecoder_webcodecs = __esm({
           error: (e) => {
             console.error("Decoder error:", e);
             done = true;
+            if (resolveNext) {
+              const r = resolveNext;
+              resolveNext = null;
+              r(null);
+            }
           }
         });
         decoder.configure({
           codec: this.#videoInfo.codec,
           codedWidth: width,
-          codedHeight: height
+          codedHeight: height,
+          description: this.#videoInfo.description
         });
         this.#videoMp4.onSamples = (_trackId, _user, trackSamples) => {
           for (const sample of trackSamples) {
@@ -432,6 +516,15 @@ var init_videoDecoder_webcodecs = __esm({
               data: sample.data
             });
             decoder.decode(chunk);
+            samplesSubmitted++;
+          }
+          if (samplesSubmitted >= frameCount) {
+            extractionDone = true;
+            if (resolveNext && pendingFrames.length === 0 && decoder.decodeQueueSize === 0) {
+              const r = resolveNext;
+              resolveNext = null;
+              r(null);
+            }
           }
         };
         this.#videoMp4.setExtractionOptions(videoTrackId);
@@ -439,18 +532,24 @@ var init_videoDecoder_webcodecs = __esm({
         while (!done) {
           if (pendingFrames.length > 0) {
             yield pendingFrames.shift();
-          } else {
-            const frame = await new Promise((resolve) => {
-              resolveNext = resolve;
-              if (decoder.decodeQueueSize === 0) {
-                decoder.flush().then(() => {
-                  done = true;
-                  resolve(null);
-                });
+            continue;
+          }
+          if (extractionDone && decoder.decodeQueueSize === 0) {
+            if (!flushed) {
+              await decoder.flush();
+              flushed = true;
+              if (pendingFrames.length > 0) {
+                continue;
               }
-            });
-            if (frame) yield frame;
-            else break;
+            }
+            done = true;
+            break;
+          }
+          const frame = await new Promise((resolve) => {
+            resolveNext = resolve;
+          });
+          if (frame) {
+            yield frame;
           }
         }
         decoder.close();
@@ -511,7 +610,8 @@ var init_videoEncoder_webcodecs = __esm({
             width: config.width,
             height: config.height
           },
-          fastStart: "in-memory"
+          fastStart: "in-memory",
+          firstTimestampBehavior: "offset"
         };
         if (config.audio) {
           muxerConfig.audio = {
@@ -534,7 +634,14 @@ var init_videoEncoder_webcodecs = __esm({
           width: config.width,
           height: config.height,
           bitrate: config.bitrate || 5e6,
-          framerate: config.fps
+          framerate: config.fps,
+          avc: { format: "avc" },
+          colorSpace: {
+            primaries: "bt709",
+            transfer: "bt709",
+            matrix: "bt709",
+            fullRange: false
+          }
         });
         this._target = target;
       }
@@ -543,7 +650,13 @@ var init_videoEncoder_webcodecs = __esm({
           format: "RGBA",
           codedWidth: imageData.width,
           codedHeight: imageData.height,
-          timestamp
+          timestamp,
+          colorSpace: {
+            primaries: "bt709",
+            transfer: "bt709",
+            matrix: "bt709",
+            fullRange: true
+          }
         });
         const isKey = this.#frameCount % 30 === 0;
         this.#encoder.encode(frame, { keyFrame: isKey });
@@ -594,6 +707,7 @@ async function processVideo(input, options = {}) {
     height: videoInfo.height,
     fps: videoInfo.fps,
     bitrate: bitrate || 5e6,
+    codec: videoInfo.codec,
     audio: audioData ? {
       codec: audioData.codec,
       sampleRate: audioData.sampleRate,