Logo sketch to a transparent PNG with GPT Image 2.5

To get a transparent logo from a sketch, send the sketch as an input_references entry to GPT Image 2.5, set background to transparent and output_format to png, then check the alpha channel of the file you get back. On Sume only the two GPT Image 2.5 variants accept background.

Treat the result as a draft mark, not a finished brand. Raster output has soft edges, so a logo that must scale to a billboard still needs a vector redraw.

Why png or webp

The OpenAI image guide says transparent backgrounds need an output format that supports alpha, which means png or webp. JPEG has no alpha channel. So request png for a logo you will place on other colors.

Describe the mark flat. Ask for solid shapes, two or three colors, no gradient, no shadow and no mockup scene. A shadow turns into semi-transparent pixels that look dirty on a dark background.

Request and alpha check

Reference URLs must be public HTTPS, and Sume answers 400 input_media_unreachable when it cannot download one. A file on your laptop needs a public upload first, for example through the assets upload flow.

import os, requests
H = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}

def generate(body):
    r = requests.post("https://api.sume.com/v1/images", headers=H, json=body, timeout=60)
    if r.status_code != 200:  # 202 = still running, read data.status_url
        raise SystemExit(f"{r.status_code}: {r.text[:300]}")
    return r.json()["data"][0]["url"]
from io import BytesIO
from PIL import Image

url = generate({
    "model": "openai/gpt-image-2.5",
    "prompt": "Clean flat logo from this sketch, two colors, no shadow, no gradient",
    "input_references": [{"type": "image_url", "image_url": {"url": "https://example.com/logo-sketch.jpg"}}],
    "background": "transparent",
    "output_format": "png",
    "image_size": "1024x1024",
})
im = Image.open(BytesIO(requests.get(url, timeout=60).content))
low = im.convert("RGBA").getchannel("A").getextrema()[0]
print("has transparency" if low < 255 else "opaque: check the prompt")
im.save("logo.png")

Price

GPT Image 2.5 is billed on tokens. The fal pages list $30 per million output image tokens, $8 per million image input tokens and $5 per million text input tokens. Sume bills the provider list price times 1.25. The table is output-only, so reference and prompt tokens add a little on top of it.

Check on dark and light

Paste the PNG on a black and a white square before you trust it. Fringing, a faint box around the mark or leftover paper texture shows up on one of the two.

If the call returns 202

POST /v1/images waits up to 30 seconds and returns 200 with the images. A slow job falls back to a 202 job envelope, and you read the images from GET /v1/jobs/{id}/result. The code above exits on any non-200 so you notice, and a failed synchronous job returns 502 and is not billed.

When the file is not transparent

Three causes explain most opaque results. A short checklist finds them quickly. If the three checks pass and the file is still opaque, read the response media_type and re-run with png, then open the file in an editor that shows a checkerboard behind transparent pixels.

• Confirm the request carried background: "transparent" and an output_format of png or webp.
• Remove words such as "on a white background" or "mockup" from the prompt.
• Check the model id. Off GPT Image 2.5, background answers 400 unsupported_parameter.