Realistic AI b-roll in a Short: is the Altered content flag needed?
A realistic generated scene that did not occur needs YouTube's disclosure; a fantasy scene does not. Sort Sume clips before upload using the page's own list.
B-roll is where AI footage most often slips into a Short unmarked. YouTube's disclosure help page gives the test: does the content meaningfully alter or generate photorealistic content? It lists generating a realistic scene that did not actually occur.
Sort each clip
Run the test per clip, not per video. A Short can be mostly filmed and carry one generated shot.
| Clip | Disclose? |
|---|---|
| Generated street scene that looks filmed | Yes, realistic scene that did not occur |
| Generated product in a fantasy world | No: unrealistic, per the page's unicorn example |
| Your own footage with a color grade | No: color adjustment is exempt |
| Edited footage of a real event or place | Yes: alters footage of a real event or place |
Label placement
The page says a label shows in the video player for photorealistic content and in the expanded description for non-photorealistic or animated content.
Keep the source list
When you assemble in Timeline 1.0, each video[] slot has a source_url. Keep that array with the finished Short. It is a ready list of which shots came from where, so the Altered content answer at upload is a lookup, not a guess.
Rule of thumb
If a viewer could take any one shot for real camera footage of something that did not happen, disclose the whole upload. The cost is a description label.
Sources
Related posts
More in Use cases
- Recipe blog post to video: scrape text, send step photos to Wan 3.0
Wan 3.0 can take a webpage as a reference at Alibaba, but Sume does not expose web_url. Scrape the recipe, then send step photos as references. Steps and curl.
- Recolor a product with a swatch reference and mask on GPT Image 2.5
BFL's FLUX 3 Image lists recolor, replace and move edits by bounding box. On Sume, a recolor job takes a swatch reference and a mask_url on GPT Image 2.5.
- Redact names from a transcript and the audio: STT words, then split
Sume STT returns word timings with a type field. Find the names you list, blank them in text, and cut the same spans from the audio for $0.01.
- Reels ad disclaimer: keep the bottom 40% clear at 1080x1920
Meta says to leave the bottom 40% of a Reels ad free of text and logos when it carries a disclaimer. Pixel math and where to burn captions with Sume.
Written by Sume