AI disclosure
How UGCFact clips are generated, which models are involved, and how to label AI content on TikTok, Reels and Shorts.
How a clip is made
- Casting. Each persona (Ava, Mia, Chloe, Jake, Nova) starts from a casting prompt and one generated reference image. That image is reused in every clip so the face stays the same.
- Still. For every reaction and setting an image model (Nano Banana 2) composes a vertical 1080×1920 frame that mimics an old phone camera in hard light.
- Motion. A video model (Grok Imagine 1.5) animates that frame for 5 seconds: brows, hands, breath, a small handheld jolt. No speech.
- Publishing. The clip is transcoded, checked automatically (duration, silence, resolution) and enters the library with generated-content metadata.
What is synthetic and what is not
Faces, bodies, clothing and settings are fully generated. The examples combine a synthetic clip with a real screen recording of an app; that part is real.
How to label on each platform
| Platform | Where | What to enable |
|---|---|---|
| TikTok | Post screen → More options | "AI-generated content" |
| Instagram Reels | Advanced settings | "Add AI label" |
| YouTube Shorts | Details → Altered content | "Yes, altered or synthetic content" |
| Meta Ads | Ad level → Settings | "Digitally created" (where applicable) |
Our commitments
- We never sell or generate the likeness of real people.
- Every clip page declares
digitalSourceType: trainedAlgorithmicMediain its structured data. - We do not generate minors, sexual content or impersonations of public figures.