Face swap on someone else's Short: does YouTube call it original?

Probably not, and Sume's face swap is built in a way that makes the problem worse, not better. On October 1, 2026, YouTube said it is updating Shorts recommendations to prioritise original content. As reported by PPC Land, channels that mainly aggregate or re-upload other creators' videos without significant changes will likely see less distribution in the Shorts feed.

A face swap keeps the source's scene, camera, timing and audio and changes the person. If the source is somebody else's Short, the result is their footage with a different face on it. That is an inference about how YouTube will judge it, not a YouTube statement, and YouTube has given no start date.

What did YouTube actually say?

Those lines are from a secondary report of YouTube's post, so check YouTube's own help page before you rely on any detail.

Why is a swap on borrowed footage a weak bet?

• Same performance: the Beta clip follows the source's timing and sound, so the creative choices are still the original creator's.
• Same audio: the source audio is kept, so even the words are theirs.
• No rights: Sume asks for a public HTTPS source URL and does not check who owns the clip. That check is yours.
• Impersonation risk: swapping a real person's performance onto your avatar is the pattern platforms' deepfake rules target.

What does Sume do with YouTube links?

Sume's media import takes only public TikTok and Instagram video URLs; YouTube and other hosts are rejected with unsupported_platform. Face swap itself reads any public HTTPS video URL that is not signed or private, so the tool will not stop you. The limit is your own rights and the platform's rules.

How would a Shorts team check a clip before posting?

Ask three questions before you upload a swapped clip. Whose footage is it? Did the creator agree to AI re-casting of their performance? What did you add that was not in the source? If the answer to the first is "someone else", the second is "no" and the third is "a different face", the clip is the pattern YouTube described, and it is also the pattern most platforms' deepfake rules cover.

If you own the footage, keep a record: the original take, the avatar handle, the job id from Sume's response and the date. That record helps if a platform asks how a clip was made. Sume returns durable media.sume.com URLs for the finished video, which is the file you upload; the source stays wherever you hosted it.

What would make a swapped clip your own work?

Nothing in a face swap changes who made the footage. Sume's workflow keeps the source's camera, scene, lighting, timing and audio, and replaces the person's face, hair, outfit and hands. If the clip is someone else's Short, the performance, the words and the framing are still theirs. Swapping the face changes how it looks, not whose idea it was.

The cases that are plainly yours are footage you shot, footage a creator licensed to you in writing for AI re-casting, and clips from your own team. For those, the swap is a production tool, like a re-shoot you did not have to book.

What should you do instead of swapping borrowed footage?

• Write your own script and render a talking video: you control every word, and no one else's performance is involved.
• Re-shoot the idea: use the clip as a brief for what to show, then film or generate your own version.
• Ask the creator: a short licence that names AI re-casting and the channels you will post on turns a risk into a deal.
• Keep a record: save the licence, the original file and your edit with the date, in case a platform asks.

How should you label it?

Platform rules on synthetic media apply whether or not the clip is original, so label AI-generated people where the platform asks. Sume does not decide platform policy, and nothing in the product removes the need to read each platform's current rules before you post.

What is the original way to use face swap on Shorts?

Swap onto footage you own. A founder films three takes on a phone, you re-cast one onto a brand avatar, and the idea, script and timing are yours. Or skip the source: write a new script and render a talking video, which is new content by construction. Either way, add something of your own, such as commentary, a different cut or your own data, because YouTube's reported test is whether you added anything.

Keep the source clip 4 to 15 seconds with audible speech, the Beta's stated range, and label the result as AI-generated where the platform asks. The faceless channel checklist covers what to vary per episode.