https://github.com/Legorobotdude/ComfyUI-VariationLab
Helps explore different parameters quickly
https://github.com/Legorobotdude/ComfyUI-VariationLab
Last synced: 3 months ago
JSON representation
Helps explore different parameters quickly
- Host: GitHub
- URL: https://github.com/Legorobotdude/ComfyUI-VariationLab
- Owner: Legorobotdude
- License: mit
- Created: 2025-03-02T04:17:22.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2025-03-02T04:59:28.000Z (3 months ago)
- Last Synced: 2025-03-02T05:24:47.393Z (3 months ago)
- Language: Python
- Size: 7.81 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- awesome-comfyui - **ComfyUI-VariationLab**
README
# ComfyUI-VariationLab
A collection of experimental nodes for ComfyUI that enable systematic exploration of generation parameters.
## Features
This custom node collection provides specialized variation nodes for ComfyUI:
### VariationLab: CFG Explorer
- Generate a batch of images with different CFG (Classifier-Free Guidance) values
- Useful for finding the ideal CFG value for your prompt
### VariationLab: Step Explorer
- Generate a batch of images with different step counts
- Perfect for comparing how many steps are needed for good results
### VariationLab: Checkpoint Explorer
- Generate images using multiple checkpoints in a single workflow
- Supports up to 5 different checkpoints with customization for each:
- Model-specific prompts (suffixes)
- Clip skip values
- Samplers, schedulers, CFG values, and step counts
## Installation
1. Clone this repository into your ComfyUI custom_nodes folder:
```bash
cd ComfyUI/custom_nodes
git clone https://github.com/yourusername/ComfyUI-VariationLab.git
```
2. Restart ComfyUI
## Usage
### CFG Explorer
Connect model, conditioning (positive/negative), VAE, and latent inputs, then specify:
- Seed value
- Step count
- CFG range (start, end, and number of steps)
- Sampler and scheduler
### Step Explorer
Connect model, conditioning, VAE, and latent inputs, then specify:
- Seed value
- Step range (start, end, and number of steps)
- CFG value
- Sampler and scheduler
### Checkpoint Explorer
Connect a latent input and specify:
- Base prompts (positive/negative)
- Seed value
- Default values for steps, CFG, sampler, etc.
- Up to 5 checkpoints with per-checkpoint customization
## Example Workflows
Example workflows can be found in the `example_workflows` directory:
- `batch_cfg_demo.json`: Demonstrates the CFG Explorer (CFG variation)
- `batch_steps_demo.json`: Demonstrates the Step Explorer (step count variation)
- `batch_checkpoint_demo.json`: Demonstrates the Checkpoint Explorer (model comparison)
> **Note:** If you experience validation errors when loading example workflows, try creating a new workflow from scratch by adding the nodes manually. Ensure you're using valid samplers (like "euler", "euler_ancestral", etc.) and schedulers ("normal", "karras", etc.).
## Future Development
- Improved parameter handling
- Support for more parameter types (sampler exploration, prompt exploration)
- Integration with ComfyUI's queue system
- Modular design for extensible parameter types
- Visual grid layout enhancements
## License
[MIT License](LICENSE)