Real footage trimmed and captioned on Sume: TikTok AI disclaimer?

A real recording that you only cut, join and caption does not become AI-generated media under TikTok's wording, but a clip that an AI model generated or significantly changed does need the disclaimer. TikTok decides on the final content, so sort your Sume jobs by whether a generative model touched the pixels, then answer per clip.

The line TikTok draws

TikTok's help page on ad disclaimers lists mandatory disclaimers for AI-generated, synthetic or manipulated media, defined as images, video or audio that are completely AI-generated or real source material significantly modified by AI. It says the AI-generated content disclaimer adds a textual label to the ad. It does not publish a list of editing tools that are exempt, so this page compares job types, not tools.

Sume jobs sorted by what they do

Sume's docs say the timeline render and video trim run only worker ffmpeg with no provider inference. Those jobs rearrange, cut or re-encode existing pictures and sound. Video captions burn text onto an existing clip; the speech step produces words for the captions, not new footage. Face swap and avatar jobs generate or replace a face, so they belong in the other column.

What to do in practice

For an ad built from real footage with only trim, timeline and captions, no disclaimer is triggered by those steps alone. If any clip in the join came from an AI video or image model, or a voice was synthesized, the ad contains AI-generated media and the toggle applies to the whole ad, so tick it. A mixed ad is still an ad with AI content in it.

Keep a note per ad of which clips were generated. A one-line record (clip id, source, generated or recorded) is enough to answer a reviewer, and it costs nothing. If you are unsure whether a change is significant, TikTok's page says to contact your sales representative if the disclaimer feature is not available to you, so ask there rather than guess.

Gray areas to settle before you upload

Two cases cause most of the doubt. The first is a real recording that has been heavily altered, for example a face replaced or a background generated. TikTok's wording covers real source material significantly modified by AI, so treat those as AI media. The second is real footage with a synthetic voice-over. TikTok's definition names audio as well as images and video, so a generated voice counts as AI-generated audio in the same sense.

Plain cuts, a re-encode to hit a size or length target, and burned captions of words that were spoken or written by a person are ordinary editing. Keep the original files for anything you ship, because they are the simplest proof of what was recorded.

Where to look in the Sume docs

Check the model page for each job you used and read whether it says the job runs only ffmpeg or calls a generative model. The trim and timeline pages state the ffmpeg-only point directly; the face swap and avatar pages describe generation. If a page does not say, do not assume the job is exempt.