Titan color-guided generation: 1 to 10 hex codes vs a Sume prompt

Brand teams like Titan Image Generator v2 for one feature. Per Amazon's model page (read 2026-10-06), its color-guided content takes a list of 1 to 10 hex codes plus a prompt, and the image incorporates that palette. There is no such field on Sume, but the need is real, so here is what to do instead.

The Sume version

Write the hex codes into the prompt, name where each one goes, and check the output. Sume's API rejects parameters a model does not list, so a colors field would return 400 unsupported_parameter. FLUX.2 models are known for reading hex in the prompt; see FLUX.2 hex color prompt for wording.

• Name each colour and its job: "background #0B3D2E, product label #F4E04D".
• Limit it to three or four colours. Ten is hard to hold in a prompt.
• Check the result instead of trusting it.

Request and palette check

The call asks for a palette in the prompt and then reports the dominant colours from the returned image.

import io, os, requests
from PIL import Image

r = requests.post("https://api.sume.com/v1/images", timeout=90,
    headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"},
    json={"model": "black-forest-labs/flux.2-pro",
          "prompt": "Flat product banner, background #0B3D2E, accent #F4E04D, white text area."})
r.raise_for_status()
img = Image.open(io.BytesIO(requests.get(r.json()["data"][0]["url"]).content)).convert("RGB")
small = img.resize((64, 64)).quantize(4).convert("RGB")
colors = sorted(small.getcolors(), reverse=True)
for count, rgb in colors:
    print("#%02X%02X%02X" % rgb, count)

For a strict brand match, measure distance in Lab space; the post on brand colour checks with Delta E shows how.