Seedream 5.0 pro bbox and point tags vs region edits on Sume

Seedream 5.0 pro lets you point at an edit region inside the prompt with <point>x y</point> or <bbox>x1 y1 x2 y2</bbox> tags on a normalized 0 to 999 grid. BytePlus lists this as pro-only. Sume lists Seedream 5.0 Lite and 4.5, so the equivalent there is a normal prompt edit, or a mask: mask_url is accepted on ChatGPT Image 2.5.

BytePlus facts are from its Seedream 5.0 overview; Sume facts from the Image API docs and catalog code, read 2026-09-30.

How do the coordinate tags work?

The overview says the top-left corner of the image is 0,0 and the bottom-right is 999,999. Its example prompt reads: Replace the area in Image 1 <bbox>120 180 640 760</bbox> with a garden. It also documents up to 10 reference images for image generation, and says the interactive editing is not on 5.0 lite or the 4.x models.

What does Sume support for region edits?

The catalog has seedream-5-lite and seedream-4-5-edit, and the shared reference ceiling in the public catalog code is 10 (MAX_INPUT_REFERENCES = 10). Those ids have no documented coordinate-tag syntax. For a region, mask_url is an optional public HTTPS mask URL for ChatGPT Image 2.5 edits; the code throws an unsupported-parameter error if another model receives it.

How do I point at a region on Sume?

Either draw a mask and use ChatGPT Image 2.5, as in the mask edit how-to, or describe the region in words (left third, the sign above the door) on a Seedream id. Set aspect_ratio: "auto" on edits so the output matches the source; omitting the field is not the same.

Should I paste bbox tags into a Sume prompt?

Do not rely on it. The tags are documented for Seedream 5.0 pro only, and Sume does not document them for its ids. Compare the ids in Seedream 5.0 Pro vs Lite.