wav or mp3 for joined audio: priming padding at every edge

Keep wav. Sume timeline audio's output.format is wav by default (pcm_s16le, sample-exact) or mp3, which is smaller but adds priming padding again at every edge. The docs say to keep wav if you will join the file again or if the file drives lip-sync. Use mp3 only for a final file where size matters more than sample accuracy.

The two formats side by side

Why edges matter

A join is only gapless if the seams are exact. Timeline audio joins in the sample domain, with no re-TTS and no silence at the seams. With mp3 output, every edge gets priming padding, and joining those files again stacks the padding. Cut four times and the padding appears four times.

Lip-sync is the strict case. Avatar 1.0 image-to-video audio and Timeline 1.0 audio.url can both take the joined file, and a drift of a few milliseconds is what a face shows first. Detach also defaults to wav for the same reason; its mp3 option is 128 kbps.

Limits that apply to both

The produced audio is up to 1800 seconds. A concat takes 1 to 20 parts, and a split takes 1 to 20 ranges that may overlap. The public rate is $0.01 flat per job and uses no provider inference, only worker ffmpeg. Confirm the live rate in GET /v1/catalog. Format choice does not change the price in the docs.

• Chain wav to wav, convert once at the end.
• If you must send mp3 into a later join, expect the padding.
• Import files first with POST /v1/media-imports.

Request with mp3 output

Send output.format only when you really want mp3.

curl -X POST https://api.sume.com/v1/timeline-1.0/audio \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: timeline-audio-mp3-001" \
  -d '{
    "operation": "concat",
    "output": { "format": "mp3" },
    "parts": [
      { "url": "https://media.sume.com/artifacts/artf_demo/a.wav" },
      { "url": "https://media.sume.com/artifacts/artf_demo/b.wav" }
    ]
  }'