Should I use sume/auto for brand image work? Pin a model id instead

For brand work, pin a model id. Sume's Image API docs describe sume/auto as a Sume-only value: Sume selects the family and never discloses which one ran. The response echoes sume/auto as the model, and GET /v1/images/models does not list it. If you cannot see which model made the image, you cannot reproduce its look next week.

Use auto for quick drafts and one-off needs, where the best-fitting model matters more than a repeatable one.

What auto gives you and takes away

Why a brand should not leave it to auto

A brand look is a set of choices: the rendering style, how the model draws type, how it handles a logo reference. Those differ by family. Auto trades that control for convenience, and the docs say outright that the family stays hidden.

The docs promise no particular family for a given prompt, so the look can change without notice. Pinning removes the question.

How to pin well

• Pick one row for text-heavy layouts, such as ideogram/ideogram-v4.5, and one for photographic product shots, such as openai/gpt-image-2.5. Test your own brief on each at low quality.
• Record the model id with every approved image, next to the job id.
• Re-read supported_parameters when you upgrade, since a row's descriptors are what the API enforces.
• Keep auto for exploration. Switch to the id that won once you have an image you like.

A short test

Send the same prompt to sume/auto and to two pinned ids at low quality, and compare the three side by side. The point is not to find a winner for all work, only to learn which style your brief tends to produce. Then write the winning id into your template.

import os, requests

key = os.environ.get("SUME_API_KEY")
if not key:
    raise SystemExit("set SUME_API_KEY")
for model in ["sume/auto", "ideogram/ideogram-v4.5", "openai/gpt-image-2.5"]:
    r = requests.post(
        "https://api.sume.com/v1/images",
        headers={"Authorization": f"Bearer {key}"},
        json={"model": model, "quality": "low",
              "prompt": "Minimal ad: a ceramic mug on a pale table, soft light"},
        timeout=90,
    )
    print(model, r.status_code, r.json()["data"][0]["url"] if r.status_code == 200 else r.text[:100])