Ideogram negative_prompt: what to send to the Sume image API instead

Ideogram's reference describes negative_prompt as a description of what to exclude, and adds that descriptions in the prompt take precedence. The Sume image API has no negative_prompt field, so a request that carries one will be rejected as 400 unsupported_parameter. The workable substitute is to state what you want in positive terms and to add a short "keep out" sentence at the end of the prompt.

What the vendor page says

From the Ideogram API reference.

Rewrite exclusions

Models follow what you name, including things you name to avoid. Prefer the positive form:

• Instead of "no text", write "clean surface with no lettering" or simply describe the empty surface.
• Instead of "no people", write "empty street at dawn".
• Instead of "not blurry", ask for "sharp focus, crisp edges".

Convert a body automatically

If you port requests from another API, move the negative text into the prompt in one place rather than by hand.

def to_sume(body):
    body = dict(body)
    neg = body.pop("negative_prompt", "")
    if neg:
        body["prompt"] += f" Keep out: {neg}."
    return body

print(to_sume({
    "model": "openai/gpt-image-2.5",
    "prompt": "A quiet cafe interior",
    "negative_prompt": "people, text",
}))

Verify

Review outputs for the excluded items. If one keeps appearing, rephrase positively rather than repeating the word, since naming it may keep it in play.