YouTube Create audio tools mapped to Sume: split, volume, fade

Which Sume parameter does what a Create audio tool does?

YouTube's help page on adding and editing audio in Create lists these audio-layer tools: split, volume, fade, find beats, audio cleanup and delete. Four of the six have a direct parameter in Sume's media tools, one needs a manual workaround and one has no named equivalent in the docs I read. The table below puts each next to its Sume counterpart so you can script a Short instead of tapping through it.

The mapping is about controls, not sound quality. A fade made by a different tool will not sound identical, and neither side promises that it does.

Side by side

The left column is the tool name and wording from YouTube's page; the right column is what the Sume docs state.

Split and join without re-recording

Timeline audio cuts one Sume-hosted audio file into ranges (start, optional end) and returns a durable audio_url for each. Concat goes the other way, with up to 20 ordered parts joined in the sample domain, so there is no re-synthesis and no silence at the seam. The default output is wav; mp3 is smaller but the docs say it re-adds priming padding at every edge, so keep wav when the file will be joined again.

To pull audio out of an existing Short first, use audio detach, which returns a wav of the track for $0.01 per job in the docs, then split it. For one render only, you do not need a reusable file: put the slices on audio.parts[] in the render instead.

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: split-voice-001" \
  -d '{
    "operation": "split",
    "url": "https://media.sume.com/artifacts/artf_demo/spine.wav",
    "ranges": [{ "start": 0, "end": 12.4 }, { "start": 12.4 }]
  }'

Volume and fades in a render

In Timeline 1.0, audio.gain_db sets the level of the spine between -60 and 12 decibels. A soundtrack has its own gain_db, plus loop, fade_out_seconds up to 10 and duck_db from 0 to 20, which lowers the bed under speech and needs a real spine. Whole-video fades are output.fade_in_seconds and output.fade_out_seconds, each 0 to 5 seconds, and the render refuses fades longer than the output.

Those are numbers you can set from a spreadsheet, which is the point of mapping them. A batch of 20 Shorts can share one gain_db of -6 for the bed and a 1.5-second fade-out, and any change is one edit.

Beats and cleanup: where Sume stops

Sume's docs do not describe beat detection. If you want cuts on the beat, work out the times yourself and set video[].start accordingly; the post on cutting clips to the beat shows arithmetic for that, and the one on YouTube's sync to beat covers the in-app version.

I found no noise-reduction or voice-enhancement control among the media tool docs, so this post does not claim one. If your source has hiss, fix it at the recording or generation step rather than expecting a render parameter to remove it.