HeyGen speaker rules as a checklist for a Sume face swap clip
HeyGen asks for a face within 45 degrees, one speaker at a time and under 10 feet. Use that as a pre-check, then Sume's own 4 to 15 second rules.

What this post is and is not
HeyGen publishes plain-language guidance for source clips: speakers should face the camera within 45 degrees, only one person should speak at a time, and speakers should be within 10 feet of the camera. These come from HeyGen's video translation help article (read 2026-10-04) and apply to its own translation feature. Sume's face swap docs do not publish those three numbers, so this post does not claim they are Sume limits. It uses them as a sensible pre-check before you spend a Beta run.
What Sume documents for face swap
The Beta endpoint is POST /v1/models/sume/avatar-face-swap/v1.0/runs. Required fields are avatar_handle, video_url and quality, with no default quality. The video_url must be public HTTPS. Source clips are planned at about 4 to 15 seconds with usable audio, and prompts, transcripts and aspect ratio are unsupported. The result appears as resource_status and a video_url on media.sume.com.
None of this says what happens with several speakers or a face turned away, so the rules below are test-first advice, not guarantees.
Pre-check list
Run these before you submit. They are cheap to check by eye.
- One person on camera, facing the lens most of the time. HeyGen's 45 degree figure is a good ceiling to test against.
- One speaker at a time. HeyGen lists this as a rule for its own feature.
- Close enough that the face is large in frame; HeyGen says within 10 feet.
- Clip length inside Sume's planned 4 to 15 seconds, with audio you can hear clearly.
- A public HTTPS file URL, not a share page.
Both sets of rules
Where the two sources overlap, follow the stricter one.
| Check | HeyGen translation help | Sume face swap docs |
|---|---|---|
| Face angle | Within 45 degrees of the camera | Not stated |
| Speakers | One at a time | Not stated |
| Distance | Within 10 feet | Not stated |
| Length | Plan limits 30 min to 5 h | About 4 to 15 seconds planned |
| Source URL | File upload up to 5 GB | Public HTTPS video_url |
A cautious test plan
Cut one clean 6 to 8 second clip that meets every row, submit it at the quality you intend to ship, and review the result before batching. If a clip that breaks one rule works, you have learned something about your footage, not about a documented limit; if it fails, the clip requirements post lists the documented rejection reasons. The Beta label in the docs means the behaviour may change, so keep the test clip around as a regression check.
Sources
Related posts
More in Comparisons
- Ideogram Plus or Pro vs API per image: what the page shows
Ideogram sells Plus at $15 and Pro at $42 a month in credits, and its page lists no per-image API price. How to compare it with a per-image row, with a script.
- Kling 4.0 voice reference vs Sume reference_audio_urls
Kling 4.0 Omni Reference accepts voice references. Sume takes 1-3 reference_audio_urls on models such as MiniMax H3; here is what each source documents.
- Kling 4.0 vs Kling 3.0: the spec differences in one dated table
Length, resolution, HDR, references, keyframes, audio and prompt size, Kling 4.0 against 3.0 as stated on Kling's pages, plus how each maps to a Sume request.
- Kling V3 Pro vs Standard vs O3 per minute
Hedra's Artificial Analysis table puts Kling V3 Pro at $20.16 per minute and V3 Standard at $15.12. How to compare per-minute prices with Sume per-second SKUs.
Written by Sume