MOSS-Image Prompt Builder KJ

Visual layout builder for MOSS-Image: the prompt becomes a JSON layout with the scene caption and one item per box (label object / character / text, caption, text to render, bbox in pixels from the canvas centre). Without boxes the plain prompt is output. Draw boxes on the canvas; select one to set its label, caption, (for text) the exact words to render, and optional tags and palette. meta / canvas / the summary palette row are the optional summary fields. width/height must match the generation size: the boxes are absolute pixels. Prompt weighting: write (phrase:weight) inside any caption and wire the prompt output into the text input of MOSS-Image Prompt Weight, which applies the weights and strips the markup. Canvas controls are the same as in the Ideogram 4 builder; right-click lists the boxes. Turn the prompt upsampler off when using layouts, it rewrites them.