explaingit

chrisjohnson89/comfyui-neuralbooru

Analysis updated 2026-05-18

★ 26PythonAudience · vibe coderComplexity · 3/5Setup · hard

TLDR

A ComfyUI node that converts plain-English scene descriptions into validated Danbooru tags using a local LLM, so anime image models get prompts they were actually trained on.

Mindmap

mindmap
  root((NeuralBooru))
    What it does
      Plain English to Danbooru tags
      Tag validation against 140k list
      Alias remapping
      Tag ordering by category
    Tech Stack
      Python
      ComfyUI node
      OpenAI-compatible API
      Local LLM server
    Use Cases
      Anime image generation
      Tag discovery and cleanup
      Custom template wrapping
    Setup
      LM Studio or Ollama
      Qwen3-1.7B recommended
      ComfyUI Manager install
Click or tap to explore — scroll the page freely

Code map

Detail Auto

An interactive map of this repo's files and how they connect — its source is parsed live in your browser. Click Visualize to build it.

filefunction / class

What do people build with it?

USE CASE 1

Describe a scene in plain English and get a validated Danbooru tag set ready to feed into an anime image generation checkpoint in ComfyUI.

USE CASE 2

Clean up messy LLM-generated tag lists by running them through NeuralBooru's alias remapping and word-form correction against the real Danbooru vocabulary.

USE CASE 3

Swap in different local LLMs (Qwen, Llama, Mistral) to improve tag quality without changing the rest of your ComfyUI workflow.

What is it built with?

PythonComfyUIOpenAI APILM StudioOllama

How does it compare?

chrisjohnson89/comfyui-neuralboorualicankiraz1/gemma-4-31b-mtp-vllm-serverparadigmxyz/centaur
Stars262626
LanguagePythonPythonPython
Setup difficultyhardhardhard
Complexity3/54/55/5
Audiencevibe coderops devopsops devops

Figures from each repo's GitHub metadata at analysis time.

How do you get it running?

Difficulty · hard Time to first run · 1h+

Requires ComfyUI and a local LLM server (LM Studio or Ollama) with a model loaded before the node can function.

In plain English

NeuralBooru is a custom node for ComfyUI, a popular open-source tool for generating images with AI, that converts plain-English scene descriptions into the specific tag vocabulary that anime-style image generation models expect. Instead of memorizing hundreds of Danbooru tags like "1girl", "crossed arms", or "denim shorts", you type a normal sentence describing your scene and the node translates it for you. The translation happens through a local language model running on your own computer via tools like LM Studio or Ollama. NeuralBooru sends your description to that local server, gets back a rough list of tag-like phrases, then validates every candidate against a bundled list of roughly 140,000 real Danbooru tags. Tags that do not exist in that list are either remapped to their correct alias ("jean shorts" becomes "denim shorts"), adjusted for word form ("smirking" becomes "smirk"), extracted from multi-word phrases ("black crop top" yields "crop top"), or dropped entirely. The result is a clean set of tags your image model was actually trained on. Beyond validation, the node orders the surviving tags the way booru-trained models expect, from character counts to general descriptors. It then wraps them in a template suited to the specific checkpoint you are using: there are presets for popular models like Illustrious, Pony, Animagine XL, and NovaAnimeXL, or you can write your own. A second output shows exactly which tags were filtered out so you can see what was changed. The node connects to any server that speaks the OpenAI-compatible API format, including LM Studio, Ollama, llama.cpp, and vLLM. There is no cloud dependency and no API key required. The recommended starting model is Qwen3-1.7B, which is fast and works well for tag generation. Installation is available through ComfyUI Manager or by cloning the repository into the custom_nodes folder. The node has no Python dependencies beyond the standard library.

Copy-paste prompts

Prompt 1
I have ComfyUI and LM Studio running locally with Qwen3-1.7B. Walk me through installing NeuralBooru and connecting it so I can describe a scene in plain English and get validated Danbooru tags.
Prompt 2
How does NeuralBooru validate tags against the Danbooru vocabulary? Explain the alias remapping, word-form fixing, and sub-phrase recovery steps in plain terms.
Prompt 3
Show me how to write a custom prompt_template in NeuralBooru for a Pony-based checkpoint, using the {prompt} placeholder and my own quality tag prefix.
Prompt 4
I want to see which tags NeuralBooru dropped from my last prompt. Where does that information appear and how do I use the dropped_tags output in my ComfyUI workflow?

Frequently asked questions

What is comfyui-neuralbooru?

A ComfyUI node that converts plain-English scene descriptions into validated Danbooru tags using a local LLM, so anime image models get prompts they were actually trained on.

What language is comfyui-neuralbooru written in?

Mainly Python. The stack also includes Python, ComfyUI, OpenAI API.

How hard is comfyui-neuralbooru to set up?

Setup difficulty is rated hard, with roughly 1h+ to a first successful run.

Who is comfyui-neuralbooru for?

Mainly vibe coder.

Open on GitHub → Explain another repo

This repo across BitVibe Labs

Scan in gitsafehub Deploy in gitdeployhub chrisjohnson89 on gitmyhub

Verify against the repo before relying on details.

Alternatives & similar repos