Shorts split audio from video: which Sume models let you drop AI sound

YouTube Shorts' editor now separates the audio and video tracks, as reported by Heyorca (read 2026-10-07). For people who generate clips with AI, that moves a decision upstream: do you want the model's own sound, or do you plan to lay your own music and voice over silent footage?

On Sume the answer depends on the model. Three rows always bake audio in, three let you switch it off, and one has no audio at all.

Audio behaviour by model

Which to choose for a Short you will score yourself

If you are adding music or a voiceover in the Shorts editor, buy silent footage. Kling 3 is the only row where silence is cheaper in the catalog: a 15-second silent clip is $2.10 against $3.15 with audio. For Wan 3.0 and Seedance the catalog does not price audio separately, so turning it off saves nothing, but it does keep a stray AI voice out of your mix.

If you want the model's sound, use Omni Flash. At 720p a 10-second clip is $1.25 with synced audio, and the audio cannot be separated before you download. In the Shorts editor you can then drop that track and keep the picture, because the new editor handles the two tracks apart.

A silent request

Send the flag explicitly. A model that does not accept it returns an error, which tells you the row has fixed audio.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"kling-3","prompt":"Barista pours latte art, overhead, no speech","duration":8,"resolution":"1080p","aspect_ratio":"9:16","generate_audio":false}'