Azure OpenAI Sora: Video-to-Video Inpainting Validation Errors

Su Myat Hlaing 250 Reputation points
2025-10-30T07:45:44.87+00:00

I've encountered what appears to be contradictory validation requirements in the Azure OpenAI Sora Video Inpainting endpoint when submitting video-to-video requests via multipart payload.

Issue Description:

When configuring inpaint_items for a video input, the API appears to simultaneously require two mutually exclusive fields:

clip object (with start_frame_index and end_frame_index) — Required to pass initial validation and avoid:

400 Bad Request
{"detail":"Video inpaint items must have clip start frame less than clip end frame."}

frame_index field — Required to pass subsequent validation and avoid:

422 Unprocessable Entity
{"msg": "Field required", "loc": "inpaint_items.0.frame_index"}

Could the Azure OpenAI team clarify the correct and complete payload schema for video-to-video inpainting?

Reference: Quickstart: Generate video with Sora - Azure OpenAI | Microsoft Learn
Any guidance or documentation updates would be greatly appreciated.


Azure OpenAI in Foundry Models

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.