TikTok impersonation deepfake ban: use face swap with your own avatar

Swap in only an avatar you made for the purpose, onto footage you own, and label the result. Darkroom reports that TikTok prohibits AI-generated endorsements and impersonation deepfakes (read 2026-10-02; an agency summary, not TikTok's page). Sume's face swap does not decide whether a swap is impersonation. It runs the job you describe.

What does Face Swap (Beta) do?

Face swap (Beta) applies a ready avatar face onto a public HTTPS source video, through POST /v1/models/sume/avatar-face-swap/v1.0/runs. Required fields are avatar_handle, video_url and quality (standard, plus or max). The Beta targets source videos of about 4 to 15 seconds with usable audio. There are no prompt, transcript or duration knobs.

What counts as impersonation here?

The agency page does not define it, so use the plain test: would a viewer believe a specific real person did or said this when they did not? A swap that puts a famous face, a competitor's founder or a neighbour into a clip fails that test whatever the label says.

What should you do before you post a swap?

Label it as AI-made on the post, since the same summary family lists face swaps as a label trigger, and keep the avatar's creation record with the job id. If the avatar started from a photo of a person, keep that person's written permission with it.

What does a good record look like?

One folder per swap: the avatar's job id, the source video file, the permission document if a person is involved, the swap job id, and the final file. Add the post URL and the label setting once it is live.

If a swap is ever questioned, that folder answers who is in the clip, where the footage came from and what was changed. It takes five minutes to keep and is hard to rebuild later.