Face swap source clip: 4-15 seconds with usable audio on Sume

Sume's Avatar Face Swap Beta targets source videos of about 4-15 seconds with usable audio. What the docs say on length, silent clips, and unsupported fields.

5 min readSume
All posts

Sume's Avatar Face Swap Beta is planned for source videos of about 4 to 15 seconds with usable audio. The docs word it as what the beta worker validation targets, not as a hard schema check, so a clip outside that range may be refused or may fail the job. If your video is silent or longer than 15 seconds, prepare it before you submit.

This post summarizes the Face swap (Beta) page and the Media inputs page, and it flags where the docs stop.

What does the beta expect from the source clip?

The face swap page lists three requirements for the source: a fetchable public HTTPS video URL, a length of about 4 to 15 seconds, and usable audio. Separately, the endpoint needs a ready Avatar 1.0 identity as avatar_handle and a quality of standard, plus, or max. It does not take a prompt, transcript, duration, aspect ratio, or provider field.

Source clip expectations for Avatar Face Swap Beta, from the Sume docs (read 2026-10-02)
PropertyWhat the docs say
LengthAbout 4 to 15 seconds (planned)
AudioUsable audio expected
URLFetchable public HTTPS
QualityRequired: standard, plus, or max
Prompt, transcript, duration, aspect ratioNot supported

What happens with a silent clip?

The docs do not say. They say the beta targets clips with usable audio and nothing more about the failure mode, so this post cannot tell you whether a silent clip is refused at submit, fails the job, or produces output. Because the endpoint is a Beta, do not build a pipeline on a guess.

Video inspect and video trim read clips hosted on media.sume.com, while face swap takes a fetchable public HTTPS video_url. Keep the trimmed output URL in your own records.

How do I check my clip first?

Use video inspect on a clip hosted with Sume to read the probe: it returns a probe and sampled stills, plus an optional transcript. If the clip is too long, video trim returns a new clip for a range you choose, and you can host that at a public HTTPS URL for the face swap.

Video inspect and video trim work on clips hosted on media.sume.com, while face swap takes a public HTTPS URL, so the two steps do not share an input. Keep the trimmed output URL in your own records.

curl -X POST https://api.sume.com/v1/models/sume/avatar-face-swap/v1.0/runs \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: face-swap-short-001" \
  -d '{
    "avatar_handle": "sume_clawra",
    "video_url": "https://example.com/inputs/10-second-clip.mp4",
    "quality": "standard",
    "mode": "async"
  }'

Why is there a 15-second ceiling?

The docs say the range is the target of the beta worker validation and do not give a reason. A practical reading is that the beta is built for short social clips. If your source is a 40-second video, cut it into pieces under 15 seconds, swap each, and join them. Each piece is its own job and its own charge, and the face swap page does not promise that identity will match across pieces, so compare the cuts.

Do not treat 15 seconds as a guaranteed limit either: the word the docs use is "planned", so check the page again before you rely on it in production.

Sources

Related posts

More in Developers

All Developers posts

Written by Sume