sora

Run the bundled CLI scripts/sora.py with a prompt, model, size, and seconds parameters, with OPENAI_API_KEY set in your environment. Video creation is asynchronous, so poll the job until complete, then download the video, thumbnail, or spritesheet assets.

Provide the existing video ID and a remix prompt that explicitly states invariants, such as keeping the same shot and camera move while changing only one element. The CLI submits the remix job, which you then poll and download like a normal generation.

The API supports only sora-2 and sora-2-pro models, with durations limited to 4, 8, or 12 seconds passed as string values via the seconds parameter. Output sizes are limited by model and must be set through API parameters, not prompt prose.

Sora API access requires an organization-verified OpenAI account and a valid OPENAI_API_KEY environment variable. Create a key at platform.openai.com/api-keys, set it locally in your shell, and never paste the full key into chat.

Generated content must be suitable for audiences under 18, with no copyrighted characters or music and no real people including public figures. Input reference images containing human faces are rejected, and reference images must be jpg, png, or webp matching the target size.