fix(quick-003): trim audio to match stitched video duration

After all video frames are processed (phase 1), the final video duration
is known from highestWrittenTimestamp. Audio is now trimmed to that
exact duration before being encoded into the MP4 (phase 2). Prevents
audio tracks from extending beyond the video.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

src/hooks/useStitchVideos.ts

@@ -154,6 +154,32 @@ export async function checkEncoderSupport(): Promise<boolean> {
   }
 }
 
+/**
+ * Trim an AudioBuffer to a target duration in seconds.
+ * If the buffer is shorter than the target, returns it unchanged.
+ */
+function trimAudioBuffer(buffer: AudioBuffer, targetDuration: number): AudioBuffer {
+  const targetSamples = Math.floor(targetDuration * buffer.sampleRate);
+
+  if (targetSamples >= buffer.length) {
+    return buffer;
+  }
+
+  const trimmed = new AudioBuffer({
+    length: Math.max(1, targetSamples),
+    numberOfChannels: buffer.numberOfChannels,
+    sampleRate: buffer.sampleRate,
+  });
+
+  for (let ch = 0; ch < buffer.numberOfChannels; ch++) {
+    trimmed.getChannelData(ch).set(
+      buffer.getChannelData(ch).subarray(0, targetSamples)
+    );
+  }
+
+  return trimmed;
+}
+
 /**
  * Standalone async function to stitch multiple video blobs together sequentially
  */
@@ -384,10 +410,13 @@ export async function stitchVideosAsync(
       }
     }
 
-    // Encode audio after all video frames have been queued so container metadata stays accurate
+    // Phase 2: Apply audio, trimmed to match the stitched video duration
     if (audioSource && pendingAudioBuffer) {
       updateProgress('processing', 'Encoding audio track...', 92);
-      await audioSource.add(pendingAudioBuffer);
+
+      const videoDuration = highestWrittenTimestamp + frameInterval;
+      const trimmedBuffer = trimAudioBuffer(pendingAudioBuffer, videoDuration);
+      await audioSource.add(trimmedBuffer);
       await audioSource.close();
     }