Sora 2 API - Are the cameo features coming to the API?

I’ve been playing around with the new preview Sora 2 API. Is there a way, through the API, to use the cameos feature? One of the greatest aids to character consistency is the cameo feature on the app. Is it possible to make use of that in the Sora 2 API? Right now the only option I see is a reference image/video, but of course if you use an image of a person in those, it’s blocked by the content moderation system.

This is one of the most powerful things about the model. Is this coming to the API?

G

2 Likes

Published today:

Guardrails and restrictions

The API enforces several content restrictions:

  • Only content suitable for audiences under 18 (a setting to bypass this restriction will be available in the future).
  • Copyrighted characters and copyrighted music will be rejected.
  • Real people—including public figures—cannot be generated.
  • Input images with faces of humans are currently rejected.

Make sure prompts, reference images, and transcripts respect these rules to avoid failed generations.

Nothing about “soon to be unlocked for you to replicate ChatGPT features”.

Another notable absence: DALL-E 3 inpainting was available in ChatGPT to edit images. Already solving what gpt-image-1 cannot do: edit based on a true mask, leaving other pixels alone. Never available on API, likely for similar reasons of the inputs that might be sent to it.

1 Like

yeah, nothing about it here:

https://platform.openai.com/docs/models/sora-2

but here it talks about consistency:

https://platform.openai.com/docs/guides/video-generation

“For more advanced prompting techniques, please refer to our dedicated Sora 2 prompting guide.”

but all they say is to just uses prompts for consistency, so no word on cameos on the api for sora 2 like J said

Yet another crippled feature on arrival from OpenAI.

For my current use cases, even the API is useless without cameo feature to capture my personal likeness.

Frack saftey OpenAI. You have 2 months max at the top before the competion, especially from China, will dominate ai video generation again.

You cripled the first version including image generation in ChatGPT to not reproduce a user’s likeness, now look where you are. Higgsfield AI is lapping circles around you by integrating the bestbin class models with a sprinkle of theirmown technologies.

Digital watermark thenhell out of the videos, especially with hidden markers. This should eliminate the need for strict safety protocols.

Or at least i tegrate into the API a robust authentication system like you have with Cameo for verification.

Don’t do this and you tender Sora 2, yet again, to a somewhat useless platform for endusers and professionals to do serious work.

OpenAI , you have the lead yet again. Don’t ruin it.

But, my confidence and gut says you will.

2 Likes

@OpenAI_Support can we please get an official update on this ? it seems that some companies are advertising that they have access to this, is it possible?