Why OpenAI is only letting some Sora users create videos of real people

OpenAI's Sora video tool limits features for real person videos pending safety testing.

: OpenAI's Sora video-generating tool, launched recently, limits its feature to generate videos of real people to a select group of users for further safety testing. This cautious approach stems from concerns over misuse, such as deepfake fraud, which has surged over tenfold from 2022 to 2023. Sora includes safeguards like underage filters and C2PA metadata while employing prompt re-writing to prevent infringement on living artists' styles. Access is available to ChatGPT Pro and Plus users in certain countries.

OpenAI introduced Sora, a video-generating tool, that currently has limited access to a feature allowing the creation of videos from real people’s images. This feature is restricted to a select user group to refine OpenAI’s approach to safety, responding to potential misuse concerns like deepfakes.

Generative video technology raises issues, as deepfake fraud increased over 10 times from 2022 to 2023. OpenAI counters misuse by implementing filters for content moderation, especially for under-18 individuals, and embeds C2PA standard metadata in Sora-generated videos for traceability.

Attempts to produce videos mimicking living artists trigger prompt re-writing to avoid copyright issues. OpenAI utilizes various datasets in training Sora, and users subscribed to ChatGPT Pro and Plus in certain regions gain access to Sora’s current capabilities, with tools for creating and editing content presented.

OpenAISoraVideo Generation