Expert Text Prompt: AST-Based Prompt Parser Node for ComfyUI
A new ComfyUI custom node parses prompts as an AST: weighted wildcards, prompt groups, inline mute and solo, automatic negative routing and a built-in syntax checker.
The Expert Text Prompt node parses the whole prompt as a syntax tree instead of chaining regex nodes or wordlist files.
One node instead of a regex chain
Expert Text Prompt is a new custom node by v3rm1ll1on that replaces prompt regex nodes and wordlist files with a single AST-based parser. Instead of building a cascade of randomizer nodes, you write one prompt string and the node parses it into a syntax tree, which also enables its most practical feature: a built-in syntax check that pinpoints the exact character of a missed bracket instead of failing silently.
Feature tour
| Feature | Syntax | What it does |
|---|---|---|
| Weighted wildcards | {70% blue | 30% green} | Custom probability per option |
| Skip chance | {20%? sunglasses} | 20% chance to omit the tag entirely |
| Prompt groups | [GRP:NAME] | Organize long prompts into semantic sections |
| Inline mute | // leather jacket | Comment out a tag or group without deleting it |
| Inline solo | ! tag | Run only the soloed tags or groups |
| Negative extraction | -sunglasses | Routes the tag into the negative output automatically |
| Number ranges | {18-50} / {0-10:2} | Random or stepped numeric wildcards |
$negative placeholder | anywhere in negative field | Controls exactly where extracted negatives are injected |
SDXL-style weights (tag:1.2) and inline LoRA tags <lora:my_lora:1.1> survive parsing untouched, and a deterministic seed makes every wildcard roll reproducible. The node also reports the total number of possible prompt combinations as an integer output, useful for estimating how much of a variation space a batch actually samples.
Availability
The node is published on the ComfyUI Registry as Expert Text Prompt (Wildcards & AST) by v3rm1ll1on (MIT). Install through the ComfyUI Manager, or clone the repository into custom_nodes. The examples folder includes seven progressively harder prompt templates, ending in a masterclass character generator combining groups, weights and nested wildcards.
Comments
Sign in with GitHub to join the discussion.