MNeMoNiCuZandClaude Opus 5.5
92a58d6774
- Migrate all nodes to the V3 node API; old workflows migrate automatically - Add help pages for every node, opened from a ? icon in the title bar - Minor fixes and tooltip, placeholder and default improvements - Bump version to 3.0.0 Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2.5 KiB
2.5 KiB
🧩 Ideogram 4 Prompt Builder w. String Inputs
A visual editor for Ideogram 4's structured JSON caption format. Draw regions on a canvas, describe each one, and the node assembles the caption JSON.
Adapted from KJNodes' Ideogram4PromptBuilderKJ by kijai, with per-region
string inputs added.
Canvas controls
- Drag on empty canvas — draw a new region
- + Region — add a region without drawing
- Click a region — select it and show its fields
- Drag inside / on an edge or corner — move / resize
- Delete or Backspace — delete the selected region
- Ctrl+C / Ctrl+V — copy / paste a region
- Ctrl+D — duplicate
- Escape — cancel the current draw or edit
Inputs
- width / height — Canvas aspect and the pixel grid the boxes are measured against. Ideogram 4 wants multiples of 16.
- high_level_description (optional) — One line describing the whole image. Omitted from the JSON when blank.
- background (optional) — The scene background description.
- style (optional) —
noneomits the style block;photoandart_stylechoose which style key is written. - photo / art_style (optional) — The descriptor for the chosen style.
- aesthetics / lighting / medium (optional) — Extra style descriptors. Blank ones are omitted.
- image (optional) — A reference image shown behind the editor canvas.
- import_json (optional) — A full caption JSON. When connected it loads into the editor on the next run; the output always reflects the editor, never the raw input.
- style_palette_data / elements_data — Serialized editor state, managed by the node's UI. Do not edit these by hand.
- region_0, region_1, … — One string input per region, added automatically. A connected, non-empty string overrides that region's description, which is how you drive a region from a wildcard or an LLM node.
Outputs
- prompt — The caption as JSON text.
- preview — A rendered preview of the regions, their text and their palettes.
- bboxes — Region boxes in pixels as
{x, y, width, height}, nested one list per frame. Freeform regions are excluded. - width / height — Passed through.
Notes
In the JSON, bbox is normalised to a 0-1000 grid as
[ymin, xmin, ymax, xmax]; width / height only set the aspect ratio.
Regions marked freeform carry no box, which lets Ideogram blend them into the
scene instead of pinning them to a rectangle — they show as dashed in the
preview.