Ideogram style_type REALISTIC, DESIGN, FICTION: the Sume equivalent

Ideogram 3.0 accepts style_type with AUTO, GENERAL, REALISTIC, DESIGN and FICTION, defaulting to GENERAL. The Sume image API has no style_type; sending one returns 400 unsupported_parameter. The way to steer toward those looks on Sume is to name the look in the prompt and to pin a model that suits it.

What the vendor page says

From the Ideogram API reference.

A mapping to prompt openers

These are suggestions, not equivalences. Test them on your own subject:

• REALISTIC: begin with "Photograph, natural light, real materials".
• DESIGN: begin with "Flat graphic design, clean layout, legible type".
• FICTION: begin with "Illustrated scene, painterly, imaginative".
• GENERAL and AUTO: no opener; let the model decide.

Keep the mapping in code

A tiny table keeps the team consistent and makes the opener a reviewed value rather than a habit.

OPENERS = {
    "REALISTIC": "Photograph, natural light, real materials.",
    "DESIGN": "Flat graphic design, clean layout, legible type.",
    "FICTION": "Illustrated scene, painterly, imaginative.",
    "GENERAL": "",
    "AUTO": "",
}

def build(style, subject, model="openai/gpt-image-2.5"):
    prompt = f"{OPENERS[style]} {subject}".strip()
    return {"model": model, "prompt": prompt}

print(build("DESIGN", "poster for a ceramics sale"))

Choose the model too

Style words only go so far. The catalog lists many models and their capabilities with GET /v1/images/models, so pin one that is known to do the look you need and keep it in the same table. Re-check after catalog changes.