AI avatar of a real person: release checklist before the photo
Before you turn an employee's or creator's photo into a Sume avatar, get a written release. A checklist for scope, term and revocation, matched to the API.
Yes, get a written release before you upload a real person's photo to make an avatar. Sume's avatar request has no consent step: the photo input takes a public HTTPS image URL and the avatar handle, and the schema has no field for a release, a consent video or a signer. The permission lives in your paperwork, not in the API.
That is a design fact you can verify in the avatar docs and the OpenAPI reference, both read 2026-10-03. It is also the reason to settle scope, term and revocation up front, because a likeness is hard to take back once clips are published. This page is a practical checklist, not legal advice; for the law, read the linked posts on likeness statutes and have counsel review your release.
What does the Sume API do, and not do, with a photo?
| Question | What the docs and schema show |
|---|---|
| How is a real face supplied? | input.type photo with a public HTTPS image_url |
| Is there a consent or release field? | None in the request schema |
| Who owns the avatar record? | Catalog search returns public system avatars plus avatars owned by your workspace |
| Can I delete an avatar through the API? | The Avatar 1.0 routes in the OpenAPI are create, list and read; no delete route is listed |
| What does creation cost? | $0.95 per avatar, flat |
| Who can render it? | Any request that names the handle with a valid key |
What should the release cover?
- Subject and source: name the person and the exact photo or photos used to build the avatar.
- Uses: ads, site, social, internal training, paid media. List channels so nothing is implied.
- Synthetic output: say plainly that the avatar will speak words the person never said, and who writes the script.
- Face swap: if you may re-cast their recorded footage onto another avatar, or their avatar onto others' footage, authorise that separately.
- Term and territory: how long, and where the clips may run.
- Revocation: what you will do, such as stop new renders and pull published clips, and how fast.
- Compensation and credit, if any.
- Approval: who signs off on each script, especially anything that reads as a testimonial.
What should you do in the workspace after signing?
Pick a handle that names the purpose and the person's role, not a vague label, so a teammate can tell whose face it is. The sume_ prefix is reserved for Sume's own avatars, so yours cannot use it. Store the release next to the handle in your records, and note the date and the photo used.
Because the API lists no delete route for avatars, plan revocation as a process: stop rendering with the handle, take down published clips, and ask Sume support about removal of the record. Do not assume a call exists that erases it. If you need the avatar to be a person who does not exist, a prompt or profile avatar skips the release question for the face.
What about voice?
A photo gives an avatar a face. The voice in an Avatar 1.0 video is chosen separately, and the release should say whose voice, if any, the video uses and whether a cloned or stock voice is allowed. Do not let a release for a face quietly stand in for a release for a voice.
When should I ask a lawyer?
Ask before you run paid media, before you use the likeness of a minor, a public figure or a person outside your own team, and before you use it across borders. State laws on digital replicas differ, which is why the linked posts exist, but they summarise rather than advise.
Does a release replace disclosure?
No. A signed release covers the person's rights; labels cover the audience. An avatar speaking for a founder or an employee can still trigger the FTC's rules on insider reviews, and platform labels still apply. Keep both: the release for the likeness, and a caption line on the clip for the viewer. The FTC fake reviews post covers the testimonial side.
Sources
Related posts
- California SB 1111 digital replicas: avatar and face-swap checklist
- Revised NO FAKES Act: counter-notice and what it means for APIs
- Creatify custom avatar lipsync_input and consent video, explained
- Create an AI avatar from a reference image: URL rules and cost
- FTC fake reviews rule: founder avatar testimonial and insider reviews
More in Use cases
- AI avatar sales agent: live SDR or personalised clips?
Tavus builds live SDR avatars on a per-minute plan. Sume renders one 4-60 second avatar clip per lead from a script. How to pick, plus a Python loop.
- AI backing track generator: an instrumental to sing or play over
Generate an AI backing track by prompt: tempo, key, form, no vocals. Sume returns one mixed MP3, no stems or click track, so trim and check the take by ear.
- AI character series on Shorts: avoid the same situation each time
YouTube's inauthentic content policy flags characters in identical situations with the same outcomes. How to keep an AI character and vary the story in Sume.
- AI classroom background music for lesson videos, under narration
Make calm instrumental music for a lesson video: a prompt that keeps vocals out, a Python script, and a Timeline bed that ducks under the teacher's voice.
Written by Sume