YouTube auto dubbing skips videos over 120 minutes: dub in chunks

YouTube's help page says auto dubbing generates translated audio tracks, that videos over 120 minutes are ineligible, and that lip sync is experimental and in early access for select channels. If your video is longer, or you want to dub a file that is not on YouTube, the Sume route is a pipeline you drive: cut into pieces with video-trim, extract speech with audio-detach, then transcribe, translate and synthesize each piece. Sume's trim output is capped at 900 seconds per job, so chunks of 600 seconds fit.

Cutting a long file

Trim needs Sume-hosted media, so import the source first with the media-imports route. Each call takes start plus duration, and an Idempotency-Key. The loop below creates three 600-second pieces; give each its own key. Source files are limited to 1800 seconds, so a longer program must be split in your own tools first.

for start in 0 600 1200; do
  curl -X POST https://api.sume.com/v1/video-trim \
    -H "Authorization: Bearer $SUME_API_KEY" \
    -H "Content-Type: application/json" \
    -H "Idempotency-Key: dub-chunk-$start" \
    -d "{\"video_url\":\"https://media.sume.com/artifacts/artf_demo/long.mp4\",\"start\":$start,\"duration\":600}"
done

What each system covers

Putting it back together

After synthesis, a Timeline 1.0 render can lay your dubbed audio over the video slots. Its audio spine is limited to 1800 seconds. Check each piece for drift at the cut points, since speech lengths differ across languages.

Honest limits

• This is more work than turning on auto dubbing; use YouTube's feature for videos it accepts.
• Sume has no lip sync in this pipeline.
• Source files over 1800 seconds need splitting before upload.