Optimize caption render into single command

utils/CaptionRender.js

@@ -23,9 +23,15 @@ export class CaptionRenderer {
       fs.mkdirSync(outDir, { recursive: true });
     }
 
-    const chunkFiles = [];
+    const outFile = path.join(outDir, `${jobId}_final.mp4`);
+
+    // Build a single complex filter chain for all sections
+    let filterComplex = '';
+    let inputs = [];
+    let sectionOutputs = [];
+    let inputIndex = 0;
 
-    // Process each transcript section
+    // Process each transcript section into the filter graph
     for (let i = 0; i < originalManuscript.transcript.length; i++) {
       const section = originalManuscript.transcript[i];
       const audioDuration = section.durationInSeconds;
@@ -33,27 +39,24 @@ export class CaptionRenderer {
       const captionFile = section.audioCaptionFile;
       const mediaFiles = section.mediaAbsPaths || [];
 
-      const chunkOutput = path.join(outDir, `chunk_${i}.mp4`);
-
       if (mediaFiles.length === 0) {
         console.log(`Skipping section ${i}: no media files`);
         continue;
       }
 
-      // Build filter_complex for this section
-      let filterComplex = '';
-      let inputs = [];
-
-      // Add all media files as inputs
+      // Add all media files as inputs for this section
+      const sectionVideoStart = inputIndex;
       for (let j = 0; j < mediaFiles.length; j++) {
         inputs.push('-i', mediaFiles[j].path);
+        inputIndex++;
       }
 
-      // Add audio input
+      // Add audio input for this section
       inputs.push('-i', audioPath);
-      const audioIndex = mediaFiles.length;
+      const audioIndex = inputIndex;
+      inputIndex++;
 
-      // Trim and concatenate video clips to match audio duration
+      // Build video filter chain for this section
       let videoFilters = [];
       let totalVideoDuration = 0;
 
@@ -61,73 +64,53 @@ export class CaptionRenderer {
         const clipDuration = Math.min(mediaFiles[j].durationSec, audioDuration - totalVideoDuration);
         if (clipDuration <= 0) break;
 
-        videoFilters.push(`[${j}:v]trim=duration=${clipDuration},setpts=PTS-STARTPTS[v${j}]`);
+        const idx = sectionVideoStart + j;
+        videoFilters.push(`[${idx}:v]trim=duration=${clipDuration},setpts=PTS-STARTPTS[v${i}_${j}]`);
         totalVideoDuration += clipDuration;
 
         if (totalVideoDuration >= audioDuration) break;
       }
 
-      // Concatenate trimmed videos
+      // Concatenate trimmed videos for this section
       const videoCount = videoFilters.length;
-      filterComplex = videoFilters.join(';') + ';';
-      filterComplex += videoFilters.map((_, idx) => `[v${idx}]`).join('') + `concat=n=${videoCount}:v=1:a=0[vconcat]`;
-
-      // Apply subtitles
-      filterComplex += `;[vconcat]subtitles=${captionFile.replace(/\\/g, '/')}[vout]`;
-
-      // Execute ffmpeg for this chunk
-      await this.runFFmpeg([
-        ...inputs,
-        '-filter_complex', filterComplex,
-        '-map', '[vout]',
-        '-map', `${audioIndex}:a`,
-        '-t', audioDuration.toString(),
-        '-c:v', 'libx264',
-        '-c:a', 'aac',
-        '-y',
-        chunkOutput
-      ], controller, onLog);
-
-      chunkFiles.push(chunkOutput);
-      console.log(`Completed chunk ${i}: ${chunkOutput}`);
+      if (filterComplex) filterComplex += ';';
+      filterComplex += videoFilters.join(';');
+
+      if (videoCount > 1) {
+        filterComplex += `;${videoFilters.map((_, idx) => `[v${i}_${idx}]`).join('')}concat=n=${videoCount}:v=1:a=0[vconcat${i}]`;
+      } else {
+        filterComplex += `;[v${i}_0]copy[vconcat${i}]`;
+      }
+
+      // Apply subtitles for this section
+      const escapedCaptionFile = captionFile.replace(/\\/g, '/').replace(/:/g, '\\:');
+      filterComplex += `;[vconcat${i}]subtitles='${escapedCaptionFile}'[vsub${i}]`;
+
+      // Combine video with its audio and trim to audio duration
+      filterComplex += `;[vsub${i}][${audioIndex}:a]concat=n=1:v=1:a=1[vout${i}][aout${i}]`;
+
+      sectionOutputs.push(`[vout${i}][aout${i}]`);
     }
 
-    // Concatenate all chunks
-    const outFile = path.join(outDir, `${jobId}_final.mp4`);
-    const concatListPath = path.join(outDir, 'concat_list.txt');
-    fs.writeFileSync(concatListPath, chunkFiles.map(f => `file '${path.basename(f)}'`).join('\n'));
+    // Concatenate all sections
+    if (sectionOutputs.length > 1) {
+      filterComplex += `;${sectionOutputs.join('')}concat=n=${sectionOutputs.length}:v=1:a=1[finalv][finala]`;
+    } else {
+      filterComplex += `;[vout0]copy[finalv];[aout0]copy[finala]`;
+    }
 
+    // Execute single ffmpeg command
     await this.runFFmpeg([
-      '-f', 'concat',
-      '-safe', '0',
-      '-i', concatListPath,
-      '-c', 'copy',
+      ...inputs,
+      '-filter_complex', filterComplex,
+      '-map', '[finalv]',
+      '-map', '[finala]',
+      '-c:v', 'libx264',
+      '-c:a', 'aac',
       '-y',
       outFile
     ], controller, onLog);
 
-    // Clean up intermediate files
-    console.log('Cleaning up intermediate files...');
-    for (const chunkFile of chunkFiles) {
-      try {
-        if (fs.existsSync(chunkFile)) {
-          fs.unlinkSync(chunkFile);
-          console.log(`Deleted: ${chunkFile}`);
-        }
-      } catch (err) {
-        console.error(`Failed to delete ${chunkFile}:`, err);
-      }
-    }
-
-    try {
-      if (fs.existsSync(concatListPath)) {
-        fs.unlinkSync(concatListPath);
-        console.log(`Deleted: ${concatListPath}`);
-      }
-    } catch (err) {
-      console.error(`Failed to delete ${concatListPath}:`, err);
-    }
-
     console.log(`Final output: ${outFile}`);
     return outFile;
   }