Enclosure Visualizer

Prompt weight syntax is the sneakiest source of "why is my image weird." A misplaced parenthesis quietly changes how a whole phrase is weighted, and when you're three levels deep into (((golden hour)), (cinematic:1.4), [soft light]) nobody can tell what's nested where just by reading it. Enclosure Visualizer turns that text into a picture: it renders your prompt with each nesting level color-coded, mismatched or hanging parentheses flagged in red, and warnings when you've stacked enclosures deep enough to get silly.

This is a pure diagnostic node - it fixes nothing, and it's honest about that. What it gives you is a second set of eyes on the exact structure your tokenizer is going to parse. If you've ever spent a run cursing a generation that ignored half your prompt, this is the node that shows you the parenthetical train wreck before you spend the compute.

How it works

It parses your text and walks the parenthesis stack the way the prompt parser will, tracking depth. Each depth level gets a muted color - the nesting becomes a visual staircase instead of a sea of brackets. Unmatched opening or closing parentheses get flagged red, because that's the classic failure: one ( too few and a whole chunk of your prompt silently parses differently than you intended. It also warns on deep multi-layer enclosures, which is the "layer it a thousand times" habit that reads as a weights mistake to any experienced eye.

The settings that matter:

• text - paste or wire your prompt.
• font_size (default 24), line_height (1.5), padding (20) - just layout; bump font size for long prompts so it stays readable.
• background_color - dark or light, cosmetic only.

Outputs: visualization (an IMAGE you can preview or save), analysis_report (a string describing the structure), and issues_text (a string listing the mismatches and warnings). Because it's an output node, it always runs - good, since you want it live while you iterate on a prompt.

Installing it

Part of CorvaeOboro's ComfyUI_illumorae (CC0, pure code, no model files). ComfyUI Manager → search illumorae, or:

cd ComfyUI/custom_nodes
git clone https://github.com/CorvaeOboro/ComfyUI_illumorae
# restart ComfyUI

Common issues

• It shows red but the image was fine - some tokenizers tolerate a stray bracket; the visualizer is stricter than your model. Use the report to understand, not as a final verdict.
• Font looks ugly or tiny - it picks a system font; if you're on Linux and it falls back to something pixelated, raise font_size and keep prompts short.
• The image output seems blank - check background_color against your theme; on a light ComfyUI theme, dark mode renders black-on-black text visually until you flip it.