Change caption render logic

utils/CaptionRender.js

@@ -23,19 +23,17 @@ export class CaptionRenderer {
     const outDir = path.join(process.cwd(), 'out');
     if (!fs.existsSync(outDir)) fs.mkdirSync(outDir, { recursive: true });
 
-    const outFile = path.join(outDir, `${jobId}_final.mp4`);
+    const segmentFiles = [];
 
-    let inputs = [];
-    let filter = [];
-    let inputIndex = 0;
-    let videoLabels = [];
-    let audioLabels = [];
+    // Render each transcript section separately
+    for (const [i, section] of originalManuscript.transcript.entries()) {
+      const segmentFile = path.join(outDir, `${jobId}_segment_${i}.mp4`);
+      const inputs = [];
+      const filter = [];
+      let inputIndex = 0;
 
-    originalManuscript.transcript.forEach((section, i) => {
       const media = section.mediaAbsPaths || [];
-      if (!media.length) return;
-
-      const audioIndex = inputIndex + media.length;
+      if (!media.length) continue;
 
       media.forEach(m => {
         inputs.push('-i', m.path);
@@ -70,32 +68,39 @@ export class CaptionRenderer {
       const aout = `aout${i}`;
 
       filter.push(`[${sub}]null[${vout}]`);
-      filter.push(`[${audioIndex}:a]anull[${aout}]`);
-
-      videoLabels.push(`[${vout}]`);
-      audioLabels.push(`[${aout}]`);
-    });
+      filter.push(`[${inputIndex - 1}:a]anull[${aout}]`);
+
+      const args = [
+        ...inputs,
+        '-filter_complex', filter.join(';'),
+        '-map', `[${vout}]`,
+        '-map', `[${aout}]`,
+        '-c:v', 'libx264',
+        '-c:a', 'aac',
+        '-y', segmentFile
+      ];
+
+      await this.runFFmpeg(args, controller, onLog);
+      segmentFiles.push(segmentFile);
+    }
 
-    filter.push(`${videoLabels.join('')}concat=n=${videoLabels.length}:v=1:a=0[finalv]`);
-    filter.push(`${audioLabels.join('')}concat=n=${audioLabels.length}:v=0:a=1[finala]`);
+    // Concatenate all segments together
+    const concatFile = path.join(outDir, `${jobId}_final.mp4`);
+    const concatList = path.join(outDir, `${jobId}_concat_list.txt`);
 
-    if (originalManuscript.bgMusic) {
-      inputs.push('-i', originalManuscript.bgMusic);
-      filter.push(`[${inputIndex}:a]volume=${originalManuscript.bgMusicVolume || 0.25}[bgm]`);
-      filter.push(`[finala][bgm]amix=inputs=2:duration=first:dropout_transition=2[mixeda]`);
-    }
+    fs.writeFileSync(concatList, segmentFiles.map(f => `file '${f}'`).join('\n'));
 
-    const args = [
-      ...inputs,
-      '-filter_complex', filter.join(';'),
-      '-map', '[finalv]',
-      '-map', originalManuscript.bgMusic ? '[mixeda]' : '[finala]',
-      '-c:v', 'libx264',
-      '-c:a', 'aac',
-      '-y', outFile
+    const concatArgs = [
+      '-f', 'concat',
+      '-safe', '0',
+      '-i', concatList,
+      '-c', 'copy',
+      '-y', concatFile
     ];
 
-    return this.runFFmpeg(args, controller, onLog);
+    await this.runFFmpeg(concatArgs, controller, onLog);
+
+    return concatFile;
   }
 
   runFFmpeg(args, controller, onLog) {