FLUX 3 Image bounding boxes vs Sume masked edit: what each takes

On Sume you change one region of an image with a mask image, not boxes in the prompt. Send image_urls plus mask_image_url on the Image 1.0 route. FLUX 3 Image works differently: its release notes say to add a JSON list of boxes to the end of prompt, with no separate parameter. Sume's docs list no FLUX 3 model id.

BFL's side is from its release notes and Sume's from the Image 1.0 and Image generation docs, read 2026-10-01.

How do bounding boxes work in FLUX 3 Image?

BFL describes placing every element of a layout, such as a headline, a face in a crowd or each panel of a grid, by boxes appended to the prompt. It also lists pixel-exact local edits: mark the element to change and keep the rest. Its docs link a separate bounding-box tutorial for the exact box format, which this post does not reproduce.

What does Sume take to change one region?

A mask. The Image 1.0 table lists a masked edit as image_urls plus mask_image_url, described as a mask image URL for edit flows. The ChatGPT Image 2.5 models on the Image Router use mask_url instead, so the field name depends on the route.

How do I send a masked edit?

Host the source and the mask at public HTTPS URLs and send both. The Image 1.0 docs say new integrations should use POST /v1/images and that Image 1.0 is retiring soon, so check the mask field for your route. The prompt says what to put in the marked area.

curl -X POST https://api.sume.com/v1/image-1.0/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: image-mask-001" \
  -d '{
    "prompt": "Replace the masked sign with a plain wooden board",
    "image_urls": ["https://example.com/street.png"],
    "mask_image_url": "https://example.com/street-mask.png"
  }'

Can I convert boxes to a mask?

Sume's docs do not describe a boxes input, so you would draw the mask yourself, one filled rectangle per box, before upload. The docs do not state a mask colour convention here; check the mask field notes and test on one image first.