Omni edit plus a reference image returns 400: send one or the other
Sume rejects video_url together with image or reference fields. Choose edit or reference-to-video on gemini-omni-flash-1.1, and see each field list.
Sume's validator returns 400 when a request carries both video_url and any frame or reference field: use either video_url (edit) or frame/reference fields, not both. If you wanted to edit a clip and also show the model the replacement object, put the object in the prompt, or run a reference-to-video call to make a new clip instead.
Two modes, two field sets
| Mode | Send | Do not send |
|---|---|---|
| Edit | prompt, video_url, optional resolution | image_url, end_image_url, reference_*_urls, aspect_ratio |
| Reference to video | prompt, reference_image_urls (up to 10) and/or reference_video_urls (up to 3, each up to 3 s) | video_url |
| Image to video | prompt, image_url, optional end_image_url | video_url |
A workaround that stays inside the rules
Describe the replacement in the edit prompt, as in the docs example: replace the bottle with an apple and keep everything else the same. If a photo is essential, generate a new clip with the photo as a reference and accept that it is a new clip, not your original footage. In reference mode the docs address media as <IMAGE_REF_0> and <VIDEO_REF_0>, zero-based in list order.
Omni takes no reference_audio_urls.
Sources
Related posts
More in Developers
- GPT Image 2.5 curl command: generate and download in a shell
A copy-paste curl call to Sume's POST /v1/images for GPT Image 2.5, with jq to pull the URL, download the file, and a check for the 202 job response.
- Use a local photo as a GPT Image 2.5 reference: it needs a URL
Sume's input_references take public HTTPS image URLs only; localhost and private URLs are rejected. Three ways to turn a file on disk into a usable reference.
- GPT Image 2.5 negative prompt: no field, so write exclusions
Sume's /v1/images has no negative_prompt for GPT Image 2.5. Put exclusions in the prompt as positive rules and a preserve list; examples and a curl call.
- Batch GPT Image 2.5 from a CSV in Python: async jobs, safe retries
Render one GPT Image 2.5 packshot per CSV row on Sume: mode async, an Idempotency-Key per SKU, a small worker pool, and the 429 queue_full rule.
Written by Sume