ComfyUI SUPIR Custom Nodes Repository

The ComfyUI SUPIR repository provides a suite of custom nodes designed to enhance the capabilities of ComfyUI through improved image processing and upscaling techniques. These nodes are specifically tailored for integrating with the SUPIR (Scaling Up to Excellent) image restoration models, allowing users to leverage advanced photo-realistic image restoration methods within the ComfyUI framework.

Installation

To install the ComfyUI SUPIR custom nodes repository, follow these steps:

1. Clone the Repository: Clone this repository into the custom_nodes directory of your ComfyUI setup.

2. Install Dependencies: Run the following command to install the required dependencies:

   pip install -r requirements.txt
   

   If you are using the portable version of ComfyUI, run:

   python_embeded\python.exe -m pip install -r ComfyUI\custom_nodes\ComfyUI-SUPIR\requirements.txt
   

3. Model Setup: Obtain the SUPIR models from the original sources and place them in the ComfyUI/models/checkpoints directory. You will also need an SDXL model, which should be stored in the same directory.

4. Optional Dependency: The xformers library can be installed to potentially improve performance:

   pip install -U xformers --no-dependencies
   

   For the portable version:

   python_embeded\python.exe -m pip install -U xformers --no-dependencies
   

Repository Purpose

This repository serves to integrate the SUPIR image restoration techniques into the ComfyUI environment. It provides custom nodes that facilitate advanced image upscaling and processing, which are essential for high-quality image restoration tasks.

Provided Nodes

The repository includes a range of custom nodes, each designed to perform specific functions within the SUPIR integration:

• SUPIR_Upscale: Manages the upscaling of images using the SUPIR model.
• SUPIR_sample: Handles sample generation using the SUPIR process.
• SUPIR_model_loader: Loads the SUPIR models for use within ComfyUI.
• SUPIR_first_stage: Deals with the initial denoising stage in the SUPIR pipeline.
• SUPIR_encode: Encodes images for processing within SUPIR.
• SUPIR_decode: Decodes processed images after SUPIR processing.
• SUPIR_conditioner: Applies conditioning to the images for improved restoration.
• SUPIR_tiles: Manages tile-based processing to handle larger images.
• SUPIR_model_loader_v2: An updated version for loading SUPIR models with additional capabilities.
• SUPIR_model_loader_v2_clip: Integrates CLIP models in the SUPIR loading process.

Special Features

• Efficient Model Loading: Optimized loading processes for SUPIR models, reducing memory usage and supporting a broader range of hardware.
• Flexible Node Design: Nodes can be used in various combinations to suit different workflow needs, allowing customization and skipping of unnecessary stages.
• Hardware Optimization: Improved hardware compatibility, enabling use on devices with varying specifications, including those with limited GPU memory.
• High-Quality Results: Utilizes state-of-the-art denoising and image enhancement methods to achieve the best possible results.

Usefulness in ComfyUI Workflows

The ComfyUI SUPIR custom nodes are instrumental in enhancing image restoration workflows by providing:

• Advanced Upscaling: Achieve higher resolution images with excellent quality, ideal for tasks requiring detailed visuals.
• Flexibility and Customization: Combine nodes in unique ways to optimize and customize image processing pipelines.
• Efficient Resource Management: Effective use of system and GPU resources, making it accessible for different hardware setups.

By integrating these nodes into ComfyUI, users can leverage sophisticated image processing techniques for improved workflow outcomes, catering to both novice and advanced users seeking high-quality image restoration solutions.