SwarmUI is a modular, high-performance web UI for AI image and video generation, supporting Stable Diffusion, Flux, and various video models. It targets both beginners with an intuitive "Generate" tab and advanced users with a "Comfy Workflow" tab, aiming to be a comprehensive tool for AI media creation.

How It Works

SwarmUI is built with extensibility and performance in mind, leveraging a modular architecture. It supports a wide range of AI models, including Stable Diffusion and Flux for images, and LTX-V, Hunyuan Video, and Cosmos for video, with future plans for audio and more. The UI offers a user-friendly "Generate" tab for ease of use and a "Comfy Workflow" tab for direct access to underlying graph structures, facilitating advanced customization and experimentation.

Quick Start & Requirements

• Windows: Download Install-Windows.bat. Requires Git and .NET 8 SDK (automated on Win 11).
• Linux: Download install-linux.sh. Requires Git, Python 3.10-3.12 (with pip and venv), and .NET 8 SDK.
• macOS: Requires M-Series Apple silicon. Install via brew install dotnet python@3.11 virtualenv, then clone and run ./launch-macos.sh.
• Docker: See Docs/Docker.md for instructions.
• Colab/Cloud: Links provided for Google Colab and Runpod/Vast.ai.
• Setup: Initial setup may take several minutes due to model downloads.

Highlighted Details

• Supports a broad spectrum of AI image and video models.
• Offers both a simplified "Generate" tab and an advanced "Comfy Workflow" tab.
• Designed for extensibility with a modular architecture.
• Includes powertools like a Grid Generator and image editor.

Maintenance & Community

• Beta status, with active development and calls for PRs.
• Discord community for support and announcements.
• Feature Announcements Thread available.

Licensing & Compatibility

SwarmUI itself is licensed under the MIT License. However, it embeds or can auto-install components with various licenses, including LGPL (7-zip), GPL (ComfyUI), AGPL (AUTOMATIC1111, ultralytics), Apache2 (improved-aesthetic-predictor, ImageSharp, Material Symbols), and MIT (various Python packages, insightface, web assets). Users must be aware of potential AGPL/GPL implications when using these components, and model licenses are separate.

Limitations & Caveats

The project is in Beta, with noted areas for improvement including better mobile browser support, native LLM-assisted prompting integration, and more convenient self-contained installers (e.g., .exe, .msi). Some Linux installations may require manual .NET runtime setup.