ComfyUI Extension: Wizdroid Character Nodes for ComfyUI

Authored by wizdroid

Created

Updated

9 stars

Run ComfyUI workflows without the setup

No installs, no CUDA version roulette, no GPU sitting idle on your bill. Bring a workflow and run it in the browser.

A comprehensive collection of custom nodes for ComfyUI that provide AI-powered prompt generation tools using Ollama. These nodes help create detailed, professional prompts for text-to-image generation across various domains including character creation, fashion, fantasy scenes, and image analysis.

Looking for a different extension?

Custom Nodes (0)

    README

    πŸ§™ Wizdroid Character Nodes for ComfyUI

    Custom nodes for ComfyUI that generate and edit character-focused prompts using Ollama LLM. Optimized for consistent character generation across genders with gender-aware filtering.

    ✨ Features

    Prompt Generation (πŸ§™ Wizdroid/Prompts)

    • πŸ§™ Wizdroid: Character Prompt - Comprehensive character prompt builder with gender-aware options (body, hair, makeup, fashion, poses, backgrounds, etc.)
    • πŸ§™ Wizdroid: Character Edit - Multi-reference image editing (face, clothing, pose, background, style transfers) with gender-appropriate beauty defaults
    • πŸ§™ Wizdroid: Scene Generator - Generate vivid scene prompts for any imaginable scenario
    • πŸ§™ Wizdroid: Background - Create surreal background prompts without human figures
    • πŸ§™ Wizdroid: Meta Prompt - Expand loose keywords into detailed image prompts
    • πŸ§™ Wizdroid: Prompt Combiner - Merge multiple prompts into one coherent description
    • πŸ§™ Wizdroid: Image Edit - Generate multi-image editing instructions (face swap, style transfer)

    Analysis (πŸ§™ Wizdroid/Analysis)

    • πŸ§™ Wizdroid: Photo Aspect Extractor - Extract clothes, pose, style from images using vision models

    Training (πŸ§™ Wizdroid/Training)

    • πŸ§™ Wizdroid: LoRA Dataset Export - Export image datasets for LoRA training with built-in validation

    πŸ”„ v2025.12.01 β€” Architecture Refactor

    What Changed

    • Removed Contest Prompt, LoRA Trainer, LoRA Validate, LoRA Dataset Validator nodes (14 β†’ 9 nodes)
    • Merged Character Edit + Multi-Angle into a single unified Character Edit node
    • Consolidated data into data/shared/ β€” single source of truth for body types, emotions, hair, eyes, makeup, poses, fashion, backgrounds, camera/lighting
    • Added gender-aware filtering β€” body types, poses, makeup, fashion outfits and styles are now tagged with gender and filtered based on selected gender
    • Added male-specific options β€” 6 male body types, 12 male hairstyles, 15 male poses, 4 male makeup styles, 17 male-specific fashion items
    • Data-driven architecture β€” Background node moved from hardcoded tuples to JSON data files
    • Built-in dataset validation β€” LoRA Dataset Export now validates automatically (no separate node needed)

    πŸ“¦ Install

    1. Clone into your ComfyUI custom nodes folder:

      cd /path/to/ComfyUI/custom_nodes
      git clone https://github.com/wizdroid/wizdroid-character.git
      
    2. Install Python dependencies:

      pip install -r requirements.txt
      
    3. Make sure Ollama is running with at least one text model (and optional vision models for analysis nodes).

    βš™οΈ Configure

    Data Architecture

    data/
    β”œβ”€β”€ shared/                      # Single source of truth
    β”‚   β”œβ”€β”€ body_types.json          # Gender-tagged body types
    β”‚   β”œβ”€β”€ emotions.json            # 100 comprehensive emotions
    β”‚   β”œβ”€β”€ eye_colors.json          # 40 eye colors
    β”‚   β”œβ”€β”€ hair.json                # Colors + gender-specific styles
    β”‚   β”œβ”€β”€ skin_tones.json          # 40 skin tones
    β”‚   β”œβ”€β”€ makeup.json              # Gender-tagged makeup styles
    β”‚   β”œβ”€β”€ poses.json               # Gender-specific poses
    β”‚   β”œβ”€β”€ backgrounds.json         # Studio/real/imaginative
    β”‚   β”œβ”€β”€ camera_lighting.json     # Lighting, camera, color palettes
    β”‚   β”œβ”€β”€ fashion.json             # Outfits, styles, footwear (gender-tagged)
    β”‚   └── background_edit.json     # Background node themes/moods
    β”œβ”€β”€ character_options.json       # Character identity (gender, race, age, image category)
    β”œβ”€β”€ scene_data.json              # Scene categories and moods
    β”œβ”€β”€ meta_prompt_options.json     # Meta prompt worlds and visual styles
    β”œβ”€β”€ photo_aspect_modes.json      # Photo aspect extraction modes
    β”œβ”€β”€ prompt_styles.json           # Prompt templates per model family
    └── system_prompts/              # LLM system prompts
    

    Gender-Aware Filtering

    When a gender is selected, all gender-tagged options are automatically filtered:

    • Male selected β†’ Shows male + any-gender items (e.g., "V-taper athletic", "crew cut", "groomed masculine")
    • Female selected β†’ Shows female + any-gender items (e.g., "hourglass", "hime cut", "coquette feminine")
    • None/Random β†’ Shows all items from all genders

    Restart ComfyUI after editing data files.

    πŸš€ Use

    Once installed, look for nodes under these categories in ComfyUI:

    • πŸ§™ Wizdroid/Prompts - All prompt generation nodes
    • πŸ§™ Wizdroid/Analysis - Image analysis nodes
    • πŸ§™ Wizdroid/Training - LoRA training nodes

    Wire them into your normal image-generation workflows.

    License

    Licensed under the Apache License, Version 2.0.

    Run ComfyUI workflows without the setup

    No installs, no CUDA version roulette, no GPU sitting idle on your bill. Bring a workflow and run it in the browser.

    Learn more