Stitch four voice takes into one 3-minute Short with audio.parts

Use audio.parts[] when narration comes as separate takes and you want one 180-second Short. Each part is a url plus an optional source_in and duration; up to 20 parts are joined gaplessly in the sample domain, with no re-synthesis. audio.url and parts are mutually exclusive. YouTube's Shorts page, read on 2026-10-06, caps a Short at three minutes. Timeline details are from the Sume docs of the same date.

What does the body look like?

Declare audio.duration_seconds as 180 and make the part lengths add up to at least that, or the job is refused with audio_parts_shorter_than_duration. Place the video slots on the joined timeline.

{
  "audio": {
    "parts": [
      {"url": "https://media.sume.com/artifacts/artf_demo/t1.wav", "duration": 45},
      {"url": "https://media.sume.com/artifacts/artf_demo/t2.wav", "duration": 45},
      {"url": "https://media.sume.com/artifacts/artf_demo/t3.wav", "duration": 45},
      {"url": "https://media.sume.com/artifacts/artf_demo/t4.wav", "duration": 45}
    ],
    "duration_seconds": 180
  },
  "video": [{"source_url": "https://media.sume.com/artifacts/artf_demo/b.mp4",
             "start": 0, "duration": 180}]
}

When should I use Timeline audio instead?

If the merged narration is reused in several renders, mint it once with Timeline audio concat at $0.01 per job. It returns offsets in segments[] that you can use to re-base each video[].start. For a one-off render, audio.parts[] avoids the extra job.

What is the cost?

A 180-second render is three billable minutes at $0.10, so $0.30. Run the unbilled plan call to confirm billable_minutes is 3.