Extract audio from MP4 to WAV or MP3: the detach defaults

Audio detach gives you WAV by default: pcm_s16le, sample-exact. Set format: "mp3" for a 128 kbps MP3 instead. WAV is the shape the docs say timeline_create audio.url, POST /v1/timeline-1.0/audio and speech-to-text want; MP3 is the smaller file.

Details are from the audio detach docs, read 2026-10-01. HeyGen's Speech Cleanup post, read the same day, says it accepts MP4, MOV or WEBM uploads; Sume's detach takes a video already on media.sume.com, so import first with POST /v1/media-imports.

What are the format and audio options?

All fields are optional except video_url.

How big is each format?

Arithmetic on the stated encodings, assuming a stereo 48 kHz source: 16-bit PCM is 48,000 samples x 2 channels x 2 bytes, about 192 KB per second or 11.5 MB per minute. MP3 at 128 kbps is 16 KB per second, about 0.96 MB per minute. At the 900-second output cap that is roughly 173 MB for WAV and 14 MB for MP3. These are calculations, not measured files; your source's channels and rate change the WAV figure.

Which one should I pick?

Choose WAV when the audio goes on to speech-to-text or a Timeline spine; 16000 with channels: "mono" is the speech-to-text shape the docs name. Choose MP3 for a listening copy. If the request returns unsupported_media_type, the HEAD of the source was not a video; see also extract audio from a video URL and WAV or MP3 for general trade-offs.

curl -X POST https://api.sume.com/v1/audio-detach \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: detach-mp3-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/talk.mp4",
    "format": "mp3"
  }'