This project automates the creation of large batches of AI-generated artwork locally, targeting users who want to generate art at scale using their own hardware. It offers flexibility in combining subjects and styles, providing a powerful tool for artists and hobbyists.

How It Works

The generator supports Stable Diffusion, CLIP-guided Diffusion (Disco Diffusion), and VQGAN+CLIP. It processes user-defined prompt files, allowing for batch generation by combining multiple subjects and styles. Advanced directives within prompt files enable fine-grained control over generation parameters like image size, model choice, and upscaling.

Quick Start & Requirements

• Install: Clone the repository, create a conda environment (conda create --name ai-art python=3.9), activate it (conda activate ai-art), install PyTorch with CUDA 11.3, and then install other required Python packages via pip and conda. Clone additional required repositories (CLIP, taming-transformers, guided-diffusion, ResizeRight, latent-diffusion). Download pre-trained model checkpoints.
• Prerequisites: Nvidia GPU (8GB VRAM minimum, 12GB recommended for 512x512, 24GB+ for 1024x1024), Anaconda, CUDA 11.3. AMD GPU support is possible on Linux via ROCm but untested.
• Setup Time: Significant, involving multiple Git clones, package installations, and model downloads.
• Docs: Example prompts

Highlighted Details

• Supports batch generation of up to 100 images per prompt file (e.g., 5 subjects x 20 styles).
• Allows pausing/resuming generation and reloading prompt files via keyboard shortcuts (Windows only).
• Output images embed generation settings as EXIF metadata.
• Offers optional upscaling via ESRGAN/GFPGAN.

Maintenance & Community

The project's primary developer has launched a successor, "Dream Factory," and will only provide bug fixes for this repository, not new features.

Licensing & Compatibility

The repository itself is licensed under the MIT License. However, it utilizes and requires models from other projects, which may have different licenses. Compatibility for commercial use depends on the licenses of the underlying models used.

Limitations & Caveats

The project is in maintenance mode, with new features directed to "Dream Factory." Keyboard shortcuts for pausing and reloading are Windows-specific. The README notes that AMD GPU support is untested.