FLUX.2 multi-reference editing API: how many images can I send?

On Sume, send up to 10 public HTTPS image URLs in input_references with model: "black-forest-labs/flux.2-pro". Black Forest Labs' own editing page says its API takes up to 8 references (10 in its playground) and outputs up to 4MP, so the two ceilings differ.

Sume's side is from the Image API docs; the vendor's from its editing page, both read 2026-09-29.

How do I send several references?

Each entry is an image_url object. Reference URLs must be public HTTPS; localhost, private-network and non-HTTPS URLs are rejected before submission. Use aspect_ratio: "auto" if you want the output to follow the reference, as the docs advise for edits.

curl -X POST https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "black-forest-labs/flux.2-pro",
    "prompt": "Put the mug from image 1 on the desk from image 2, same lighting",
    "input_references": [
      {"type": "image_url", "image_url": {"url": "https://example.com/mug.png"}},
      {"type": "image_url", "image_url": {"url": "https://example.com/desk.jpg"}}
    ],
    "aspect_ratio": "auto"
  }'

Do the ceilings match?

What if a model lists zero references?

Models whose input_references descriptor is {min: 0, max: 0} are text-to-image only and reject references. Read the catalog descriptor for the model you pick.

Does it stay consistent across edits?

Black Forest Labs' launch post markets FLUX.2 for character, product and style consistency across references. Its editing page does not detail a character-consistency technique, and Sume's docs make no consistency guarantee, so compare the output with your reference images before you publish. Consistent character AI image generator covers the workflow on Sume.