Tag Prompt Builder for ComfyUI

A flexible ComfyUI custom node that simplifies prompt building through organized dropdown categories. Works with any tag-based prompting system including Danbooru-style models (Illustrious, Pony Diffusion, NovelAI), Stable Diffusion, and more.

Features

• Organized Categories: Quality, Style, Character, Pose, Camera, Lighting, Time, Expression, Background
• Negative Prompts: Dedicated node with Quality, Anatomy, Text, Artifacts, and Composition categories
• Fully Customizable: Edit prompt_tags.json to add your own tags and categories
• Clean Output: Automatic formatting and organization of your prompts
• Flexible Separators: Choose how tags are joined (commas, newlines, custom)
• Universal Compatibility: Works with Danbooru tags, natural language, or any prompting style

Installation

Method 1: Manual Installation

1. Navigate to your ComfyUI custom nodes directory:

   cd ComfyUI/custom_nodes/
   

2. Create a new folder:

   mkdir ComfyUI-PromptTagBuilder
   cd ComfyUI-PromptTagBuilder
   

3. Copy these files into the folder:

   • tag_prompt_builder.py (the main node file)
   • prompt_tags.json (configuration file)
   • __init__.py (optional)

4. Restart ComfyUI

Method 2: Git Clone

cd ComfyUI/custom_nodes/
git clone https://github.com/Cordux/ComfyUI-PromptTagBuilder.git

File Structure

ComfyUI/
└── custom_nodes/
    └── ComfyUI-PromptTagBuilder/
        ├── tag_prompt_builder.py
        ├── prompt_tags.json
        ├── __init__.py (optional)
        └── README.md

Method 3: ComfyUI Manager

1. search for ComfyUI-PromptTagBuilder and click install
2. Restart ComfyUI

Usage

Basic Workflow

1. Add "Tag Prompt Builder" node to your workflow (found in utils/prompt category)
2. Select tags from the dropdown menus
3. Add specific details in the Custom field
4. Connect output to your text encoder

<details><summary><h3>Screenshot</h3></summary> <img width="945" height="687" alt="image" src="https://github.com/user-attachments/assets/f2900ce5-6333-47cb-9b39-6e219468db9c" /> </details>

Example: Creating an Anime Portrait

Node Settings:

• NSFW: none
• Quality: masterpiece, best quality, absurdres
• Style: anime
• Character: 1girl, solo
• Pose: standing
• Camera: portrait
• Lighting: soft lighting
• Time: none
• Expression: smile
• Background: simple background
• Custom: long hair, blue eyes, school uniform, white shirt, looking at viewer

Output:

masterpiece, best quality, absurdres, anime, 1girl, solo, standing, portrait, soft lighting, smile, long hair, blue eyes, school uniform, white shirt, looking at viewer, simple background

Using the Negative Prompt Node

Add "Negative Tag Builder" node for organized negative prompts:

Settings:

• Quality: lowres, bad quality, worst quality
• Anatomy: bad anatomy, bad hands, missing fingers
• Text: signature, watermark, text, username, artist name
• Artifacts: blurry, deformed, ugly
• Composition: cropped, out of frame, bad composition
• Custom: (add any specific negatives)

Output:

lowres, bad quality, worst quality, bad anatomy, bad hands, missing fingers, signature, watermark, text, username, artist name, blurry, deformed, ugly, cropped, out of frame, bad composition

Customization

Editing Tags

Edit prompt_tags.json to customize available options:

{
  "quality": [
    "none",
    "masterpiece, best quality",
    "your custom quality tags here"
  ],
  "style": [
    "none",
    "anime",
    "photorealistic",
    "your custom styles here"
  ]
}

Available Categories (Positive Prompt)

• nsfw: NSFW rating tags
• quality: Image quality descriptors
• style: Art style (anime, realistic, sketch, etc.)
• character_count: Number of characters (1girl solo, 2girls, etc.)
• pose: Character poses and actions
• camera: Camera angles and shot types
• lighting: Lighting conditions
• time: Time of day
• expression: Facial expressions
• background: Background descriptions

Available Categories (Negative Prompt)

• negative_quality: Quality issues to avoid
• negative_anatomy: Anatomical problems to avoid
• text_issues: Unwanted text/watermarks
• negative_artifacts: Visual artifacts to avoid
• negative_composition: Composition problems to avoid

Adding New Options to Existing Categories

1. Open prompt_tags.json
2. Find the category you want to edit
3. Add your new option:

   "quality": [
     "none",
     "masterpiece, best quality",
     "your new option here"
   ]
   

4. Save and restart ComfyUI

Tag Format Guidelines

• Always include "none" as the first option in each category
• Use commas to separate multiple tags within one option
• Keep options concise and descriptive
• Group related concepts together

Node Parameters

Tag Prompt Builder (Positive)

Required Inputs:

• NSFW: NSFW rating (default: none)
• Quality: Quality tags (default: masterpiece, best quality, absurdres)
• Style: Art style (default: none)
• Character: Character count (default: 1girl, solo)
• Pose: Character pose (default: none)
• Camera: Camera angle (default: none)
• Lighting: Lighting type (default: none)
• Time: Time of day (default: none)
• Background: Background type (default: none)
• Custom: Multi-line text field for detailed descriptions
• Expression: Facial expression (default: none)

Optional Inputs:

• separator: String to separate tags (default: ", ")
• newlines: Add line breaks between tag groups (default: false)

Output:

• Positive Prompt: Combined text string

Negative Tag Builder

Required Inputs:

• Quality: Quality issues (default: none)
• Anatomy: Anatomical issues (default: none)
• Text: Text/watermark issues (default: none)
• Artifacts: Visual artifacts (default: none)
• Composition: Composition issues (default: none)
• Custom: Multi-line text field for custom negatives

Optional Inputs:

• separator: String to separate tags (default: ", ")
• newlines: Add line breaks between tag groups (default: false)

Output:

• Negative Prompt: Combined text string

Tips & Best Practices

1. Organize Your Workflow

Use the Custom field for specific details about your subject. Let the dropdowns handle the technical and stylistic aspects.

2. Tag Order

The node outputs tags in this order:

1. NSFW → Quality → Style → Character → Pose → Camera → Lighting → Time → Expression
2. Custom details
3. Background

This follows common prompting best practices where technical tags come first.

3. Use "none" Liberally

Set categories to "none" if you don't need them. This keeps your prompts clean and focused.

4. Custom Separator

Try different separators for different models:

• Standard: , (comma space)
• Weighted: :1.2, for emphasis systems
• Natural: (just space) for natural language models

5. Enable Newlines for Readability

Turn on newlines when building complex prompts to make them easier to read and debug.

6. Model-Specific Presets

Create different JSON configurations for different models by editing prompt_tags.json before use.

7. Combine with Other Nodes

• Chain multiple builders for complex scenes
• Use with wildcard nodes for variation
• Connect to prompt styling/weighting nodes

Examples

Example 1: Fantasy Character

Settings:
- Style: anime
- Character: 1girl, solo
- Pose: standing
- Camera: full body
- Lighting: dramatic lighting
- Expression: serious
- Custom: elf ears, long silver hair, green eyes, leather armor, holding sword
- Background: outdoors, forest

Output: anime, 1girl, solo, standing, full body, dramatic lighting, serious, elf ears, long silver hair, green eyes, leather armor, holding sword, outdoors, forest

Example 2: Realistic Portrait

Settings:
- Quality: best quality, high quality
- Style: realistic
- Character: 1boy, solo
- Camera: close-up
- Lighting: natural lighting
- Time: golden hour
- Expression: smile
- Custom: stubble, brown eyes, casual clothing
- Background: detailed background

Output: best quality, high quality, realistic, 1boy, solo, close-up, natural lighting, golden hour, smile, stubble, brown eyes, casual clothing, detailed background

Example 3: Chibi Art

Settings:
- Style: chibi
- Character: 2girls
- Pose: waving
- Camera: full body
- Expression: laughing
- Custom: colorful outfits, cat ears, holding hands
- Background: simple background, white background

Output: chibi, 2girls, waving, full body, laughing, colorful outfits, cat ears, holding hands, simple background, white background

Troubleshooting

Node doesn't appear in ComfyUI:

• Verify files are in the correct directory
• Restart ComfyUI completely
• Check console for error messages

Dropdowns are empty:

• Ensure prompt_tags.json is in the same folder as the Python file
• Check JSON syntax is valid (use jsonlint.com)
• Look for error messages in the console

Tags not updating after JSON edit:

• Restart ComfyUI (the JSON is loaded on startup)
• Verify JSON syntax is correct
• Check file permissions

Output has unexpected formatting:

• Check your separator settings
• Try toggling the newlines option
• Verify no trailing spaces in your JSON tags

Model Compatibility

Fully Compatible:

• ✅ Pony Diffusion (all versions)
• ✅ Illustrious XL
• ✅ NovelAI (Danbooru tags)
• ✅ Stable Diffusion (all versions)
• ✅ SDXL
• ✅ Most anime/manga models
• ✅ Any tag-based prompting system

Adaptable:

• ⚠️ Natural language models (customize JSON for natural descriptions)
• ⚠️ Midjourney-style prompts (adjust separator and tag format)

FAQ

Q: Can I add my own categories? A: Yes, but it requires editing the Python code to add new inputs. Adding new options to existing categories only requires editing the JSON.

Q: Will this work with SDXL? A: Yes! Just edit the JSON to use natural language descriptions instead of Danbooru tags.

Q: Can I have multiple JSON files for different models? A: Currently the node uses one JSON file. You can manually swap files or edit the filename in the code.

Q: Does this support weighted prompts like (tag:1.2)? A: The node passes through whatever you put in the JSON, so yes! Just include the weights in your tag options.

Q: Can I use this for negative prompts only? A: Yes, use the "Negative Tag Builder" node independently.

License

Free to use and modify for personal and commercial projects.

Credits

Created for the ComfyUI community to streamline tag-based prompt creation across different models and prompting systems.

Changelog

Version 1.0

• Initial release
• Positive and negative prompt builders
• 11 customizable positive categories
• 5 customizable negative categories
• JSON-based configuration
• Flexible separator options
• Newline formatting support

---

Need help? Check the ComfyUI documentation or ask in the community forums!