This project provides a web-based playground for generating images from text prompts, currently utilizing Stable Diffusion V2. It's designed for enthusiasts and developers interested in exploring text-to-image generation without complex setup.

How It Works

The project consists of a Python backend serving a React frontend. The backend handles image generation requests using Stable Diffusion V2, while the frontend provides a user interface for inputting prompts and viewing results. The architecture allows for flexible deployment, with options for cloud-hosted backends (like Google Colab) or local execution.

Quick Start & Requirements

• Colab Backend: Run the provided Google Colab notebook, then access the frontend at https://saharmor.github.io/dalle-playground/?backendUrl=<your_colab_url>.
• Local Backend: Requires Python 3.x, PyTorch with CUDA support. Install dependencies via pip install -r requirements.txt and run python3 app.py --port 8080. The frontend requires npm install and npm start.
• Docker: Requires Docker and NVIDIA Container Toolkit. Run with docker-compose up.
• WSL2: Requires specific setup for GPU support, including installing CUDA/CuDNN in Linux and compiling JAX from source with CUDA enabled.

Highlighted Details

• Supports Stable Diffusion V2 for image generation.
• Offers a GitHub-hosted frontend for quick testing with a remote backend.
• Provides detailed instructions for local development and Docker deployment.
• Includes specific guidance for WSL2 GPU setup challenges.

Maintenance & Community

The project was originally based on DALL-E Mini by @borisdayma. Further community or maintenance details are not specified in the README.

Licensing & Compatibility

The repository's license is not explicitly stated in the README.

Limitations & Caveats

Running the backend on Google Colab's free tier may lead to timeouts for generating more than a few images. The WSL2 setup for GPU acceleration requires significant manual compilation and configuration of JAX.