Ideogram 4.5 transparency claim vs Sume's background field
Higgsfield lists transparent backgrounds for Ideogram 4.5. On Sume, background works only on ChatGPT Image 2.5; other models need RMBG or Image 1.0.

If you need a transparent PNG on Sume, send background: "transparent" to ChatGPT Image 2.5. Do not send it to Ideogram 4.5. The Higgsfield changelog (read 2026-10-05) lists transparent backgrounds among Ideogram 4.5's features on Sep 30, but Sume's Image API accepts background only on ChatGPT Image 2.5 (both Flare and Sunburst variants). Elsewhere the request comes back as 400 unsupported_parameter.
Claim vs what Sume exposes
The left column is a distributor's feature list. The right column is what the Sume Image API docs say you can send today.
| Question | Answer | Source |
|---|---|---|
| Does Ideogram 4.5 offer transparent backgrounds? | Listed as a feature on the Sep 30 entry | Higgsfield changelog |
| Can you send background on ideogram/ideogram-v4.5 via Sume? | No. The catalog gives it no background descriptor | Sume Image API docs |
| Which Sume models take background? | openai/gpt-image-2.5 and openai/gpt-image-2.5-sunburst | Sume Image API docs |
| Values | auto, transparent, opaque | Sume Image API docs |
| Output format on Ideogram 4.5 | Chosen by the provider; output_format is not accepted | Sume Image API docs |
Three ways to get alpha on Sume
Pick the route by what you already have.
- Generate with alpha: ChatGPT Image 2.5 with
background: "transparent". The OpenAI guide pairs transparent output with PNG or WebP. - Generate on another model, then cut out: Sume's background-removal model
sume/rmbg-1.0lists $0.0225 per image. - Legacy shape: the Image API docs say that for transparent stills on other models, use Image 1.0 with
transparency: true. Image 1.0 is retiring and is an alias of Auto, so avoid it in new code.
A transparent request
This asks ChatGPT Image 2.5 for a PNG with a transparent background. The sticker prompt is only an example.
import asyncio, os, httpx
async def main():
body = {
"model": "openai/gpt-image-2.5",
"prompt": "a die-cut sticker of a smiling fox, flat colors",
"background": "transparent",
"output_format": "png",
}
headers = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}
async with httpx.AsyncClient(timeout=60) as c:
r = await c.post("https://api.sume.com/v1/images", headers=headers, json=body)
print(r.status_code, r.json())
asyncio.run(main())Which route to pick
If the look has to come from Ideogram 4.5 (text-heavy layouts, for example), generate there and cut out with RMBG, then check the edges at 100 percent on your real background. If any model will do, the single-call alpha on ChatGPT Image 2.5 avoids a second step. We have not measured edge quality for either route. A related recipe covers replacing headline text on a finished Ideogram image, and the edit chain cost post prices repeated passes.
Check the result
Whichever route you use, place the output on a dark and a light background and look at the edges at 100 percent zoom. A transparent PNG that looks clean on white can show a halo on black. Also open the file's channels to confirm there is an alpha channel at all, since a file can be PNG and still be opaque.
Sources
Related posts
More in Developers
- Image API changes in October 2026: what to do and when
October 2026 image API changes: gpt-image-1 shuts down 2026-10-23, three more OpenAI ids on 2026-12-01, GPT Image 2.5 ships, and Sume retires Image 1.0.
- Retry an image edit after a timeout: same key, same payload
After a client timeout on an image edit, resend the same payload with the same Idempotency-Key. If you change the prompt, change the key.
- image-generations_create is now generate_image: fix old MCP prompts
The Sume MCP tool image-generations_create was retired for generate_image. Update agent prompts and allow-lists, and check with tools_list. Dated facts.
- MCP: read image-models_get, then dry-run generate_image first
An agent checklist for hosted Sume MCP: read the model with image-models_get, dry-run generate_image with a spend cap, then submit with a new idempotency key.
Written by Sume