To download and set up the Flux AI models for use with ComfyUI, follow these steps based on the available information:
Available Flux Checkpoints
1. Flux FP8 Checkpoints
These models can be used directly with minimal setup:
| Model Name |
File Size |
Download Link |
| flux1-dev-fp8 |
17.2 GB |
Download |
| flux1-schnell-fp8 |
17.2 GB |
Download |
2. NF4 Version
This version is optimized for performance and requires the installation of a specific plugin:
| Model Name |
Download Link |
| flux1-dev-bnb-nf4 |
Download |
Installation Steps
- Download the Model:
- Choose the appropriate model (e.g.,
flux1-dev-fp8 or flux1-schnell-fp8) and download it from the provided links.
- Place in Correct Directory:
- Move the downloaded model file to the directory
- Execute Command：ComfyUI/models/checkpoints/
- Update ComfyUI:
- Ensure that your ComfyUI installation is up to date. You can do this through the ComfyUI Manager by selecting:
- Execute Command：Manager > Update All
- After updating, reload the ComfyUI page.
- Load the Checkpoint:
- In ComfyUI, use the "Load Checkpoint" node to select your downloaded model. Set the CFG (Classifier-Free Guidance) to 1.0 when using FP8 models, as this may affect output quality.
- Install Necessary Plugins (for NF4 models):
- Install the NF4 plugin by navigating to:
- Execute Command：ComfyUI manager > Install via Git URL
- Enter the following URL:
- Execute Command：https://github.com/comfyanonymous/ComfyUI_bitsandbytes_NF4.git
- Additionally, run
pip install bitsandbytes in your command prompt or terminal to ensure all dependencies are met.
Additional Notes
- Performance Considerations: FP8 models may degrade quality slightly compared to full 16-bit versions, so consider your hardware capabilities when choosing a model.
- Documentation and Support: For detailed workflows and troubleshooting, refer to resources like the ComfyUI Wiki or community forums.
By following these steps, you should be able to successfully download and utilize Flux AI models within ComfyUI for your projects.
