ComfyUI Installation Guide for use with Pixart Sigma

This guide will help you install ComfyUI, a powerful and customizable user interface, along with several popular modules. For questions, I can frequently be found on the civitai discord in general chat @eface60.

update: Seems the sigmas_tools_and_the_golden_scheduler has had an update, giving us better sigma values -> better image quality for the same amount of compute!

Be sure to grab the v1.1 of the workflow, using these updates Sigma values.

Currently using: https://civitai.com/models/462854 as Sigma Model, anyMODE did an awesome job tuning this

Prerequisites

Before starting, ensure you have the following:

• Python 3.8 or above installed. You can download it from python.org.

• Git installed. You can download it from git-scm.com.

Installation Steps

1. Clone the ComfyUI Repository

First, clone the ComfyUI repository to your local machine using Git.

git clone https://github.com/comfyanonymous/ComfyUI

2. Navigate to the ComfyUI Directory

Change your current working directory to the newly cloned ComfyUI directory.

cd ComfyUI

3. Set Up a Virtual Environment

Create a new virtual environment to isolate the project's dependencies.

python -m venv venv

4. Activate the Virtual Environment

Activate the virtual environment to ensure all dependencies are installed inside it.

venv\Scripts\activate

5. Install Required Dependencies

Install the primary dependencies required for ComfyUI.

pip install -r requirements.txt

6. Add Custom Modules

Navigate to the custom_nodes directory to add additional modules.

cd custom_nodes

Only ExtraModels is needed to get Pixart working, But I highly recommend all the other modules so pixart will work properly with a lot more fidelity
Clone these repositories:

git clone https://github.com/city96/ComfyUI_ExtraModels
git clone https://github.com/ltdrdata/ComfyUI-Manager
git clone https://github.com/Extraltodeus/ComfyUI-AutomaticCFG
git clone https://github.com/Extraltodeus/sigmas_tools_and_the_golden_scheduler
git clone https://github.com/Clybius/ComfyUI-Extra-samplers

7. Install Additional Requirements for Modules

Install additional requirements for the newly added modules.

pip install -r ComfyUI_ExtraModels/requirements.txt
pip install accelerate xformers==0.0.26.post1 matplotlib

8. Install PyTorch with CUDA Support

Install PyTorch with CUDA support for GPU acceleration.

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

9. Add PixArt-Sigma-XL Model

Clone the full Pixart repository, and move the files to the right places afterwards.

1. Clone the PixArt-Sigma-XL model from Hugging Face:

git clone https://huggingface.co/PixArt-alpha/PixArt-Sigma-XL-2-1024-MS

1. Move the model files to appropriate directories:

move PixArt-Sigma-XL-2-1024-MS\vae\diffusion_pytorch_model.safetensors models\vae\pixart_sigma_sdxl.safetensors
move PixArt-Sigma-XL-2-1024-MS\transformer\diffusion_pytorch_model.safetensors models\checkpoints\pixart_sigma.safetensors

1. Create a directory for the T5 model and move the text encoder files:

mkdir models\t5
move PixArt-Sigma-XL-2-1024-MS\text_encoder models\t5\pixart_t5_v11

1. Remove the nearly empty cloned PixArt-Sigma-XL model directory:

rmdir PixArt-Sigma-XL-2-1024-MS /s /q

10. Create a Startup Batch File

Create a start.bat file to easily activate the virtual environment and start ComfyUI.

echo call venv\scripts\activate.bat > start.bat && echo python main.py >> start.bat

11. Run ComfyUI

Execute the start.bat file to launch ComfyUI.

start.bat

12. Import workflow

After opening ComfyUI on http://localhost:8188 you can drag the attached workflow, or load it into comfy ui. If you changed some directory settings, make sure to update it in the workflow.

Have fun generating images!