gpt-image-2 transparent background: OpenAI preview vs Sume
OpenAI put transparent backgrounds for gpt-image-2 in preview (png or webp). On Sume, the background parameter is listed for ChatGPT Image 2.5, not gpt-image-2.
OpenAI's changelog (Aug 20) says transparent backgrounds are in preview for gpt-image-2: set background to transparent and use png or webp output. On Sume, the background parameter (auto, transparent, opaque) is listed for ChatGPT Image 2.5, and the code does not send a background field for gpt-image-2.
OpenAI facts are from its changelog; Sume facts from the Image API docs and catalog code, read 2026-09-30.
What does the OpenAI preview cover?
The entry names gpt-image-2 and the dated snapshot gpt-image-2-2026-04-21, in the Images API and the Responses API image generation tool. It adds that jpeg does not support transparent backgrounds.
Which Sume model takes background?
| Sume id | background | Source |
|---|---|---|
| openai/gpt-image-2.5 | auto, transparent, opaque | Docs and catalog descriptor |
| openai/gpt-image-2.5-sunburst | auto, transparent, opaque | Docs and catalog descriptor |
| openai/gpt-image-2 | Not listed | Provider adapter omits the field for this id |
What happens if I send background to a model that lacks it?
The docs say a parameter the selected model does not list is rejected with 400 unsupported_parameter rather than silently dropped. Read supported_parameters for the model first; a request that works on 2.5 may be refused on another id. Output format choices are in PNG, JPEG or WebP output.
What should I send for a transparent request?
Use openai/gpt-image-2.5 with background: "transparent" and output_format: "png" or "webp", following OpenAI's note that jpeg cannot carry transparency. Check the returned file rather than assuming it, since this post states only what the two pages say.
Sources
Related posts
More in Developers
- GPT Image output_compression: OpenAI has it, Sume returns 400
OpenAI lets jpeg and webp take output_compression 0-100. Sume's docs say it is not served in v1 and sending it returns 400 unsupported_parameter.
- Rotating a webhook secret during a gradual deployment
Sume signs with both the new and old secret for 24 hours after rotation, so two receiver versions can run side by side. Upgrade the verifier first, then rotate.
- Grok Imagine first and last frame: keyframes on xAI only
xAI takes up to 4 timestamped keyframes for Grok Imagine video. Sume's grok-imagine-video-1.5 row takes one start image and rejects an end frame.
- Grok Imagine: how many images per request, xAI vs Sume
xAI's image API takes n from 1 to 10 per request. Sume's catalog code gives Grok a maximum of 1, so send one image per call and loop for more.
Written by Sume