An Azure service that provides access to OpenAI’s GPT-3 models with enterprise capabilities.
Solved — the video generation API has been migrated to a new endpoint.
The old endpoint POST /openai/v1/video/generations/jobs is deprecated and has the model.name == "sora" validation bug described above. The new endpoint is:
POST {endpoint}/openai/v1/videos?api-version=preview
Key differences from the old API:
| Parameter | Old API | New API |
|-----------|---------|---------|
| Endpoint | /openai/v1/video/generations/jobs | /openai/v1/videos |
| seconds | integer (5) | string ("4", "8", or "12") |
| size | width/height integers | string ("720x1280", "1280x720", "1024x1792", "1792x1024") |
| Poll status | GET .../video/generations/jobs/{id} | GET .../videos/{id} |
| Download | GET .../video/generations/{gen_id}/content/video | GET .../videos/{id}/content |
| Status values | succeeded / failed | completed / failed |
Working example:
POST {endpoint}/openai/v1/videos?api-version=preview
{
"prompt": "A golden retriever running on a beach at sunset",
"model": "<your-deployment-name>",
"size": "720x1280",
"seconds": "4"
}
This works with a sora-2 model deployed under any deployment name. The new endpoint is documented in the v1 API reference (search for VideoResource), though the quickstart docs have not been updated yet.
Hope this saves others the debugging time!