64 lines
2.1 KiB
Markdown
64 lines
2.1 KiB
Markdown
# ComfyUI-MixMod
|
|
|
|
|
|
ComfyUI-MixMod provides a powerful way to combine multiple models during sampling.
|
|
|
|
VRAM requirement: Around 8gb vram is minimum. You might need to replace some models with gguf, I got 1sdxl, 1sdxlgguf and pixart sigma with gguf te running on 8gb vram.
|
|
|
|
## Features
|
|
|
|
- Mix multiple models during sampling
|
|
- SD1.5+SDXL (use https://huggingface.co/ostris/sdxl-sd1-vae-lora to align the latents)
|
|
- SDXL+Pixart sigma for increased prompt adherence, needs [GGUF](https://github.com/city96/ComfyUI-GGUF) and [ExtraModels](https://github.com/city96/ComfyUI_ExtraModels) for Comfyui
|
|
- Scheduling everything
|
|
- Multiple experimental modes.
|
|
|
|
## Please share cool workflows you find under discussion, this stuff is really experimental and still needs discoveries.
|
|
|
|
## Installation
|
|
|
|
1. Clone this repository into your ComfyUI's `custom_nodes` directory:
|
|
```bash
|
|
cd ComfyUI/custom_nodes
|
|
git clone https://github.com/Kantsche/comfyui-mixmod.git
|
|
```
|
|
|
|
2. Restart ComfyUI if it's already running.
|
|
|
|
|
|
## Tips
|
|
|
|
- Different models excel at different aspects - try mixing a detail-focused model with a composition-focused one
|
|
- For FFT modes, base models work well for low frequencies, style models for high frequencies
|
|
- Experiment with weights and CFG values to find the best balance
|
|
- Schedule models to activate at different sampling steps for creative control
|
|
|
|
## Compatibility
|
|
|
|
- Tested with SD 1.5, SDXL and Pixart Sigma
|
|
- Works with different model architectures (base, inpainting, etc.)
|
|
- Only tested on Windows
|
|
|
|
|
|
Example workflow with pixart sigma:
|
|
|
|
|
|
It improves the prompt adherence of sdxl in general prompts:
|
|
|
|
|
|
|
|
|
|
|
|
Example with Ponyv6 and NoobAI:
|
|
|
|
Only Pony, Pony+Noob, Only Noob image
|
|
|
|
|
|
|
|
|
|
|
|
## Credits
|
|
|
|
Created by Kantsche
|
|
|