Does a face-swapped video need a YouTube AI disclosure?
YouTube asks for disclosure when content makes a real person appear to say or do something they didn't. What that means for a Sume Beta face-swap output.
Most likely yes, if the result looks realistic and shows a real person doing or saying something they did not. YouTube's help page lists "Makes a real person appear to say or do something they didn't do" among the cases that need a disclosure. Sume's face swap docs say nothing about YouTube labels, so the decision is yours.
What does YouTube's page say?
Creators must disclose when content "appears realistic" and was "generated or meaningfully altered with AI". The page names three examples: a real person made to appear to say or do something they didn't, altered footage of a real event or place, and a realistic scene that didn't occur. Labels sit in the video player for photorealistic content and in the expanded description otherwise. Wording is from YouTube Help, read 2026-09-30.
What does Sume's face swap produce?
The Beta endpoint applies a ready avatar's face onto a public HTTPS source video and returns a video under media.sume.com. The face comes from an avatar you created; whose likeness that avatar carries decides the answer below.
| Whose face is on the result? | Realistic? | YouTube's page suggests |
|---|---|---|
| A real person, who did not say the words | Yes | Matches "Makes a real person appear to say or do something they didn't do"; disclose |
| A real person, in their own words | Yes | Realistic and AI-altered; the page's test is realism and meaningful alteration, so disclose to be safe |
| An avatar from a text prompt, no real person | Yes | Realistic synthetic footage; check the realistic-scene example |
What should I do before publishing?
Get the person's agreement to their face being used, keep the record, and disclose when your video meets YouTube's test. Sume's avatar docs describe prompt, traits and photo inputs and no consent step, so do not treat creating an avatar as proof of permission.
For the wider policy, see YouTube's AI content policy and YouTube likeness detection.
Is that the final word?
No. Platform policy can change and this post read only one YouTube page. Re-check it when you publish.
Sources
Related posts
More in Sume Avatar 1.0
- Prime Video AI lip sync vs Sume: human voice on a talking image
Amazon uses AI and VFX to lip-sync shows to human-dubbed audio. Sume cannot re-sync existing footage, but Fabric can animate a still to a human recording.
- Synthesia Roleplay Sessions vs Sume scripted scenario clips
Synthesia launched Roleplay and Survey Sessions on Oct 1. Sume has no live roleplay, but can render the scenario prompt as a short avatar clip.
- Synthesia Survey Sessions vs Sume: not interactive, but scripted
Synthesia's Survey Sessions avatar asks questions live and reports themes. Sume renders avatar question clips from a script, with no live conversation.
- VEED Fabric Emotions audio tags vs Sume Fabric audio_url
VEED's Fabric Emotions adds inline tags like [excited] in the script. On Sume, Fabric takes a finished audio_url; emotion lives in the TTS step.
Written by Sume