Real-person deepfake on TikTok: the reported 30-day tier and avatars

Do not build an avatar from a photo of someone who has not agreed to it. AuditSocials reports that a real-person deepfake can bring a 30-day suspension on TikTok, in the same tier as four or more violations (third-party summary, read 2026-10-02). A label does not fix that case; the summary treats it as a separate, heavier outcome.

How does an avatar start from a photo on Sume?

Create new avatar offers three inputs: a prompt, a structured profile, or a reference image. The image route is the one that can resemble a real person. The prompt and profile routes start from a description instead of a face. Sume's docs on that page do not say the API checks who is in a photo, so consent is a step you run outside the call.

What is the safe default?

Use a face you have the right to use: your own, a model who signed a release for AI use, or a prompt-made character with no real-world match. Check a prompt-made face by eye for a lookalike of a public figure before you ship it.

What should you keep on file?

Keep the signed release or the written permission, the source photo reference, and the Sume job id of the avatar you made from it. If a clip is questioned, those three answer who, what and when.

Face swap works the same way. Avatar Face Swap (Beta) applies a ready avatar face onto a public source video. Use it with your own avatar and your own footage. This is not legal advice.

Does a label help in a deepfake case?

The summary places real-person deepfakes in the suspension tier by themselves, not as a labeling mistake. So the practical rule is to avoid the case, not to label your way through it.

If a clip shows a recognizable person doing or saying something they did not, do not post it, labeled or not. If it shows someone who agreed to it, label it as AI-made and keep their permission with the clip.