ComfyUI Node
Ideogram4 Bbox Editor
Visual bbox editor for the Ideogram-4 caption JSON. Outputs the caption string, a preview image, pixel-space bounding boxes (SAM3/crop), and the resolved size.
Ideogram4 Bbox Editor
- image
- florence_data
- prompt
- preview
- bboxes
- width
- height
◄caption_json{}►
◄width0►
◄height0►
◄import_json—►
◄florence_caption—►
◄florence_regions—►
CategoryIdeogram4
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| caption_json | STRING | {} | — |
| width | INT | 00–16384 | Target width. 0 = use the aspect ratio set in the editor. When both width and height are > 0 they override aspect_ratio (W:H) and set the pixel grid for the bboxes/preview outputs. |
| height | INT | 00–16384 | Target height. 0 = use the aspect ratio set in the editor. |
| imageopt | IMAGE | Optional reference image: dimmed behind the preview output and shown as the editor backdrop (loads on run). | |
| import_jsonopt | STRING | Optional caption JSON loaded into the editor on run. The output always reflects the editor, never this raw input. | |
| florence_captionopt | STRING | Florence2Run `caption` -> high_level_description (auto-fill on run). | |
| florence_dataopt | JSON | Florence2Run `data` (region/OCR boxes). Auto-placed as elements; needs the `image` input connected (boxes are in image pixel space). Note: kijai's `data` has no per-region labels for OD/dense — use `florence_regions` for labeled boxes. | |
| florence_regionsopt | STRING | A region task's raw `caption` (label<loc_..> string). Parsed into labeled boxes (desc = label) — recovers labels that `data` drops. Coords are normalized, so no `image` needed. |
Outputs (5)
| Name | Type | Description |
|---|---|---|
| prompt | STRING | — |
| preview | IMAGE | — |
| bboxes | BOUNDING_BOX | — |
| width | INT | — |
| height | INT | — |