Stable Diffusion PAG Image Generator

This project leverages Stable Diffusion XL (SDXL) with Perturbed-Attention Guidance (PAG) to generate stunning AI-powered visuals. You can create two variations of an image: one using PAG and another without PAG, and compare the results side-by-side.

Features

• Generate high-quality images with and without PAG.
• Customizable guidance scales, PAG scales, and layers.
• Ability to use custom LoRA models for fine-tuning.
• Randomized or fixed seed for unique and reproducible outputs.
• Clean and user-friendly UI built with Gradio.

Requirements

Ensure you have the following installed:

• Python 3.8+
• CUDA-enabled GPU for faster inference.

Installation

Clone the repository and install the required dependencies:

git clone https://github.com/shahram8708/stable-diffusion-pag-image-gen.git
cd stable-diffusion-pag-image-gen
pip install -r requirements.txt