ComfyUI Pack Of ThatAIGod

A custom node pack for ComfyUI providing LLM integration, image utilities, resolution management, and wildcard-based prompt generation.

Features

• LLM Chat - Connect to OpenRouter API or local LLM servers (LM Studio) with streaming responses
• Image Saver Plus - Save images with format selection, quality control, template variables, and text sidecars
• Dynamic Resolution Picker - Calculate optimal width/height for any aspect ratio
• Dynamic Resolution Selector - Multi-select aspect ratios with max/min side constraint and custom W:H ratio support
• Wildcard Reader - Resolve __wildcard__ placeholders from curated text files with {A|B|C} choice syntax
• Sequential Image Loader - Iterate through directory images with natural sorting
• Upscale By Max Side - Aspect-ratio-preserving upscaler
• Truncate LLM Thinking - Strip thinking/reasoning blocks from LLM output

Installation

Option 1: Git Clone

Navigate to your ComfyUI custom_nodes directory and clone:

cd ComfyUI/custom_nodes
git clone https://github.com/thataigod/ComfyUI_Pack_Of_ThatAIGod.git

Option 2: ComfyUI Manager

Search for "Pack Of ThatAIGod" in ComfyUI Manager and install directly.

Option 3: Manual Download

1. Download the ZIP from GitHub
2. Extract to ComfyUI/custom_nodes/ComfyUI_Pack_Of_ThatAIGod

API Key Setup

OpenRouter (Recommended)

Set your API key as an environment variable:

# Windows (PowerShell)
[System.Environment]::SetEnvironmentVariable('OPENROUTER_API_KEY', 'sk-or-v1-your-key-here', 'User')

# Linux/macOS
export OPENROUTER_API_KEY='sk-or-v1-your-key-here'

The node supports three environment variable names:

• OPENROUTER_API_KEY (default)
• OPENROUTER_API_KEY_BACKUP
• OPENROUTER_API_KEY_EXTRA

This allows switching between multiple API keys without modifying system variables.

Local LLM (LM Studio)

1. Start LM Studio with a loaded model
2. Enable the local server (default: http://localhost:1234/v1)
3. In ComfyUI, set Mode to Local

Node Reference

LLM Chat (OpenRouter/Local)

Category: ThatAIGod/LLM

Connects to OpenRouter or a local LLM server for text generation with streaming responses.

Input	Type	Description
Mode	Dropdown	OpenRouter or Local
Model	Dropdown	Auto-fetched model list from OpenRouter, or local model
System Prompt	STRING	System instructions for the LLM
User Prompt	STRING	Your input text (supports dynamic connections)
Temperature	FLOAT	Randomness (0.0 = deterministic, 2.0 = max creative)
Max Tokens	INT	Maximum response length (1-128000)
seed	INT	For deterministic generation. Note: seed 0 is treated as "no seed" and is omitted from the request payload; use any non-zero value for reproducible outputs
Timeout (Seconds)	INT	Request timeout (1-300)
API Key Env Var	Dropdown	Which environment variable holds your API key
Local URL	STRING	Local LLM server endpoint
Image(s)	IMAGE	Optional image input for vision models. Images larger than 8192×8192 pixels will raise an error

Output	Type	Description
Generated Text	STRING	The LLM response
Status (Boolean)	BOOLEAN	True if generation succeeded
Information	STRING	Latency and credit info
Reasoning Content	STRING	Model's reasoning/thinking text (if provided)

Features:

• Streaming responses displayed in real-time via WebSocket
• Response caching (last 10 requests) for faster re-runs
• Vision support - connect images for multimodal models
• Credit balance display for OpenRouter
• Reasoning content extracted separately from the clean response

LLM Fallback Switch

Category: ThatAIGod/LLM

Passes generated text through if the LLM succeeded, otherwise falls back to the original input.

Input	Type	Description
Original Input	STRING	The fallback text (required)
Generated Text	STRING	LLM output (optional)
Status (Boolean)	BOOLEAN	LLM success status (optional)

Output	Type	Description
Final Text	STRING	Generated text if successful, otherwise original input

Dynamic Resolution Picker

Category: ThatAIGod/Image Utils

Calculates optimal image dimensions based on aspect ratio and scale factor.

Input	Type	Description
Max Side Pixels	INT	Target resolution (256-16384)
Aspect Ratio	Dropdown	15 presets including random portrait/landscape
Scale Factor	FLOAT	Multiplier for upscale dimensions
seed	INT	For random ratio selection

Output	Type	Description
Width	INT	Base width (rounded to 8px)
Height	INT	Base height (rounded to 8px)
Scaled Width	INT	Width after scale factor
Scaled Height	INT	Height after scale factor
Scale Factor	FLOAT	The applied multiplier
Keywords	STRING	Aspect ratio description for prompts
Guide Size	INT	Min(width, height)
Max Size	INT	Max(width, height)

Aspect Ratio Presets:

• Square 1:1
• Portrait: 2:3, 3:4, 4:5, 9:16
• Landscape: 3:2, 4:3, 5:4, 16:9, 16:10, 21:9, 1.85:1
• Random (Any, Portrait, or Landscape)

Image Saver Plus

Category: ThatAIGod/Image Utils

Saves images with format selection, quality/compression control, filename template variables, and optional text sidecar files.

Input	Type	Description
images	IMAGE	The images to save
filename_prefix	STRING	Prefix with template variables. Supports %width%, %height%, %date:FORMAT% (e.g. %date:yyyy_MM_dd%), %year%, %month%, %day%, %hour%, %minute%, %second% (shorthand for the most common date parts), %counter% (sequential number, place anywhere in the filename), and %batch_num%
file_format	Dropdown	png, jpeg, or webp
quality	INT	Quality for JPEG/WebP (1-100, default 95)
compress_level	INT	PNG compression (0-9, default 4)
save_text	STRING	Optional text to save as a .txt sidecar alongside each image

Examples:

• ComfyUI/Chroma/%date:yyyy_MM_dd%/PiX_%date:yyyyMMdd%_%counter%_PreFaceFix produces ComfyUI/Chroma/2026_05_30/PiX_20260530_00001_PreFaceFix.png
• Using %counter% in the filename prevents overwrites — the counter scans existing files and increments automatically

Wildcard Reader

Category: ThatAIGod/Text Utils

Resolves __wildcard__ placeholders by randomly selecting lines from .txt files in the wildcards/ directory.

Input	Type	Description
text	STRING	Text containing __wildcard__ tags
Select to add Wildcard	Dropdown	Quick-insert wildcard tags
mode	Dropdown	Deterministic (Seed), Full Random, or Random (No Repeat)
seed	INT	For deterministic mode
delimiter	STRING	Separator when combining prepend/main/append text
Prependable Text	STRING	Text added before the result
Appendable Text	STRING	Text added after the result

Output	Type	Description
Text	STRING	Resolved text with wildcards replaced

Modes:

• Deterministic (Seed) - Same seed = same output every time
• Full Random - Completely random selection each run
• Random (No Repeat) - Shuffles and draws without replacement until depleted

Inline Choice Syntax: Use {option1|option2|option3} in your text to randomly pick one option (separated by |). Works alongside __wildcard__ tags and respects the same mode/seed settings.

Nested Wildcards: Wildcard files can reference other wildcards. For example, all_colors_male.txt contains:

__neutral_colors_male__
__earthtone_colors_male__
__jeweltone_colors_male__

These are recursively resolved up to 50 iterations.

Built-in Wildcard Files:

• Color palettes (8 categories x 2 genders)
• Outfit components (traditional, formal, casual, athletic, costume, swimwear, lingerie)
• Subject descriptions (male/female)
• Complete scene assemblers

Sequential Image Loader

Category: ThatAIGod/Image Utils

Iterates through images in a directory using a seed-based index.

Input	Type	Description
Directory Path	STRING	Absolute path to image directory
seed	INT	Index of the image to load

Output	Type	Description
Image	IMAGE	The loaded image tensor
Filename	STRING	Filename without extension
Stats	STRING	Current index / total files (e.g., "3/50")

Tip: Set "Control After Generate" to "Increment" on the node to auto-advance through images.

Dynamic Resolution Selector

Category: ThatAIGod/Image Utils

Multi-select aspect ratios with constraint mode (max side or min side) and optional custom W:H ratio with scaling.

Input	Type	Description
Limit By	Dropdown	Max Side or Min Side constraint for the Pixels value
Pixels	INT	Pixel budget for the constrained side (1-16384)
Scale Factor	FLOAT	Multiplier for Scaled Width/Height outputs
Aspect Ratio Config	STRING	JSON config managed by the frontend UI with toggle buttons for each ratio, batch selectors (All/Portraits/Landscapes), and a custom W:H ratio input
seed	INT	Seed for random ratio selection from the active set

Output	Type	Description
Width	INT	Base width (rounded to 8px)
Height	INT	Base height (rounded to 8px)
Scaled Width	INT	Width after scale factor
Scaled Height	INT	Height after scale factor
Scale Factor	FLOAT	The applied multiplier
Keywords	STRING	Aspect ratio description for prompts
Guide Size	INT	Min(width, height)
Max Size	INT	Max(width, height)

Upscale By Max Side

Category: ThatAIGod/Image Utils

Upscales an image so its longest side matches the target, preserving aspect ratio.

Input	Type	Description
Image	IMAGE	Input image
Max Side	INT	Target pixel count for longest side
Divisibility	INT	Round dimensions to this value
Method	Dropdown	lanczos, bicubic, bilinear, nearest-exact, area

Output	Type	Description
Image	IMAGE	Upscaled image
Width	INT	Final width
Height	INT	Final height

Truncate LLM Thinking

Category: ThatAIGod/Text Utils

Removes content between configurable start/end tokens from LLM output.

Input	Type	Description
Text	STRING	LLM output containing thinking blocks
Start Token	STRING	Beginning marker (default: <think>)
End Token	STRING	Ending marker (default: </think>)

Output	Type	Description
Cleaned Text	STRING	Text with thinking blocks removed
Thinking Content	STRING	Extracted thinking content

Wildcard System

Directory Structure

wildcards/
  autowildcards/          # Generated wildcard files
    neutral_colors_male.txt
    neutral_colors_female.txt
    earthtone_colors_male.txt
    ... (36 files total)
  wildcard_generator.bat  # Regenerate all wildcard files
  colors.bat              # Simplified color list generator
  find.bat                # Search utility for wildcards

Adding Custom Wildcards

1. Create a .txt file in wildcards/ or wildcards/autowildcards/
2. Add one entry per line
3. Use # at the start of a line for comments
4. Reference it in your prompt as __filename__ (without .txt)

Example Prompt

A woman wearing a __female_casual_outfit__ in __bright_colors_female__ colors,
standing in a __subject_female__ pose, {smiling|looking serious|laughing}

UI Features

LLM Node

• Streaming Preview - Watch tokens appear in real-time
• Refresh Models - Fetch latest available models from OpenRouter
• Live Credit Display - See remaining OpenRouter credits

Wildcard Reader

• Dropdown Selector - Browse and insert wildcard tags directly
• Auto-insert - Selected wildcards append to the text field

Troubleshooting

"No API Key found"

Set the OPENROUTER_API_KEY environment variable or select a different one from the dropdown.

"Local URL must be localhost"

The Local URL field only accepts localhost or 127.0.0.1 for security. Ensure your LLM server is running locally.

Wildcards not resolving

• Check that your .txt files are in the wildcards/ directory
• Ensure wildcard tags use double underscores: __tag__
• Verify the filename matches (case-insensitive on Windows)

Images not loading

• Use absolute paths for the Sequential Image Loader
• Supported formats: .png, .jpg, .jpeg, .webp, .bmp, .tiff

License

MIT