How to Use Sora 2 Cameo for Consistent On-Screen People
Sora 2’s Cameo feature verifies a real person once, then lets that person appear in generated videos with their consent, solving the consistent-identity problem for talking-head content.
Every other approach to a recurring on-screen person involves either a real shoot or a character that subtly changes between clips. Cameo ties the identity to a verified recording, so the same face shows up in generation after generation.
How does Cameo verification work?
You record a short video of yourself, once. OpenAI uses it to confirm you are a real person and to build the likeness reference. Verification is a one-time step per person; after that, you can generate clips featuring your verified likeness through the Sora app. This runs on ChatGPT Plus at $20 per month or Pro at $200 per month.
How do you generate a clip with your cameo?
- Select your verified cameo as the character in the Sora app.
- Write the scene and the spoken line in quotes.
- Keep the line under about twelve words for reliable lip sync.
- Describe the setting and ambient sound you want generated.
Because audio is generated in the same pass as the picture, the mouth movement matches the words. That beats the old workflow of generating a silent clip and dubbing it, which never looked right.
What are the consent rules?
Nobody else can use your cameo unless you allow it. That consent model is the reason Sora 2 can offer this at all; earlier likeness tools operated in a grey area. Treat it seriously: if you are generating content featuring someone else, you need their verified cameo, not a lookalike prompt.
What breaks with cameo clips?
| Issue | Behaviour | Workaround |
|---|---|---|
| Long dialogue | Loses sync past a sentence | Split into short takes |
| Fast head movement | Frame drift | Keep motion calm |
| Extreme expressions | Distortion | Reroll for a cleaner take |
| Watermarks | Always present | Not removable |
The watermark applies to every output, cameo or not, and there is no removal path. If your deliverable needs clean footage, this is the wrong tool.
Where does Cameo make sense?
Training videos and explainers where a presenter is needed but a shoot is not, social series that need the same face episode to episode, and concept ads that require lip-synced lines for review. It is not suited to anything requiring precise, physical performance or long monologues.
Key Takeaways
- Cameo verifies a person once and reuses the likeness with consent.
- Requires the Sora app with ChatGPT Plus from $20/mo.
- Dialogue and sound effects generate in the same pass as video.
- Clips are short, watermarked, and lose sync on long lines.
