Forge HiDiffusion

Overview

Forge HiDiffusion is a port of Panchovix's JankHidiffusion from reForge to lllyasviel's Forge. HiDiffusion is a technique designed to improve the ability to quickly generate images at higher resolutions without structural collapse.

Features

• High Resolution Stability: Ensures that images generated at higher resolutions maintain structural integrity.
• Compatibility with Forge: Seamless integration with lllyasviel's Forge WebUI for enhanced user experience.
• Advanced Controls: Offers advanced settings for users who require finer control over the image generation process.

Installation

To use Forge HiDiffusion, follow these steps:

1. Install the Extension:

   • Download and install the Forge HiDiffusion extension in your WebUI.

2. Accessing the Extension:

   • Navigate to either the txt2img or img2img tab in the WebUI.
   • Locate the "Forge HiDiffusion" section.

Usage Instructions

1. Adjust Image Dimensions:

   • Set the desired image width and height. Ensure that the resolutions are divisible by 64 for optimal results.

2. Select Resolution Mode:

   • Choose a resolution mode that closely matches your desired resolution from the RAUNet tab.
   • Note: This option is disabled if Advanced mode is checked.

3. Enable Advanced Mode (Optional):

   • For finer control, you can enable advanced mode in both the RAUNet and MSW-MSA tabs.

Troubleshooting

• Common Issues:

  • If images are not generating as expected, ensure that the dimensions are set correctly and are divisible by 64.
  • Make sure to select the appropriate resolution mode to match your desired output.

• Support:

  • For additional help, visit our GitHub Issues page.

Contributing

We welcome contributions to Forge HiDiffusion. To contribute, follow these steps:

1. Fork the repository.
2. Create a new branch (git checkout -b feature-branch).
3. Make your changes and commit them (git commit -m 'Add new feature').
4. Push to the branch (git push origin feature-branch).
5. Open a Pull Request.

Credits

• reForge Implementation: Panchovix's JankHidiffusion
• ComfyUI Implementation: blepping's comfyui_jankhidiffusion
• Original Implementation: Megvii Research's HiDiffusion