# ComfyUi-Untwisting-RoPE (Training-Free Style Transfer)

<video controls width="100%" max-width="600">
  <source src="https://huggingface.co/9r4n4y/Untwisting-RoPE-Style-tranfer-for-img-models-with-tutorial/resolve/main/vid.mp4" type="video/mp4">
  Your browser does not support the video tag.
</video>

### Official Repository
This is the official GitHub repository: [BigStationW/ComfyUi-Untwisting-RoPE](https://github.com/BigStationW/ComfyUi-Untwisting-RoPE)

### Repository Clone
The GitHub repository clone is already included inside this Hugging Face repository as a `.7z` archive.

This is a ComfyUi implementation of [Untwisting RoPE: Frequency Control for Shared Attention in DiTs](https://arxiv.org/abs/2602.05013)

https://untwisting-rope.github.io/

<img width="1280" alt="Image 4" src="https://github.com/user-attachments/assets/ea4b7f49-45b0-4812-845b-0bd58824f27c" />


You can see more examples [here](https://github.com/BigStationW/ComfyUi-Untwisting-RoPE/tree/main/Examples).

## Supported models
- Z-image and Z-image turbo
- Anima
- The whole Flux.2 family
- The whole Qwen-Image/Edit family

## Installation

Navigate to the **ComfyUI\custom_nodes** folder, [open cmd](https://www.youtube.com/watch?v=bgSSJQolR0E&t=47s) and run:

```bash
git clone https://github.com/BigStationW/ComfyUi-Untwisting-RoPE
```
Restart ComfyUI after installation.

## Usage

- Here's [some workflows](https://github.com/BigStationW/ComfyUi-Untwisting-RoPE/tree/main/workflows) for those interested.
- You also need these custom nodes to make those workflows work.

https://github.com/BigStationW/ComfyUi-Scale-Image-to-Total-Pixels-Advanced

https://github.com/BigStationW/ComfyUi-TextEncodeEditAdvanced

Don't hesitate to consult [the documentation](https://github.com/BigStationW/ComfyUi-Untwisting-RoPE/tree/main/documentation) to better understand this repository.