ComfyUI
A node-based visual interface for building, running, and sharing AI workflows that generate images, video, 3D assets, and audio
It takes care to curate a safe software space. Help us do more.
&
| + | Node-Based Workflow Editor free | Connects processing steps as nodes on a canvas so each parameter and intermediate step stays visible and adjustable |
|---|---|---|
| + | Custom Node Ecosystem free | Extends the base engine with thousands of community-built node packs covering samplers, preprocessors, and niche model integrations |
| + | Multi-Modal Generation free | Runs image, video, 3D, and audio generation models within the same workflow graph |
| + | App Mode free | Exposes a chosen workflow as a simplified form-style interface without hiding access to the underlying node graph |
| + | Low-VRAM Memory Management free | Offloads model weights automatically so large diffusion models can run on GPUs with as little as 1GB of VRAM |
| + | Offline Operation free | Runs entirely on local hardware without transmitting workflow data once models are downloaded |
| + | Workflow Portability free | Saves workflows as JSON or embeds them directly in generated PNG, WebP, and FLAC files for sharing and reuse |
| + | API Endpoints api | Exposes a workflow as a callable HTTP endpoint for integration into external applications and pipelines |
| + | Cloud GPU Access cloud | Runs the same workflow files on hosted GPUs without requiring local hardware setup |
...6 more features/limitations. Contact us to get a complete list of features and system requirements.
System Requirements
| # | Minimum | Recommended |
|---|---|---|
| 1 | Windows 10 or 11 (64-bit) | |
| 2 | modern multi-core CPU (Intel i5 / Ryzen 5 or higher) | |
| 3 | 8GB (Comfy Desktop minimum) | 16GB-32GB (Comfy Desktop recommends 16GB; heavier FLUX/video workloads commonly need up to 32GB) |
| 4 | 4.85GB (Comfy Desktop base install, excludes model files) | |
| 5 | NVIDIA GPU with CUDA 12.8+ (RTX 50-series/Blackwell requires 12.8+ minimum); smart offloading can run models on as little as 1GB VRAM | NVIDIA GPU with 8GB+ VRAM for SDXL/FLUX-class workloads |
| 6 | Python 3.12-3.13 with PyTorch 2.5+ (Windows Portable bundles Python 3.13 + CUDA 13.0) | |
| 7 | Google Chrome 143 or later recommended for the UI |
| # | Minimum | Recommended |
|---|---|---|
| 1 | 64-bit Linux | |
| 2 | modern multi-core CPU | |
| 3 | 8GB | 16GB-32GB for heavier workloads |
| 4 | ~5GB (base install, excludes model files) | |
| 5 | NVIDIA GPU with CUDA 12.8+, or AMD GPU with ROCm 7.2 (stable or nightly) | NVIDIA GPU with 8GB+ VRAM for SDXL/FLUX-class workloads |
| 6 | Python 3.12–3.13 with PyTorch 2.5+ | |
| 7 | ComfyUI Desktop not officially supported for Linux. |
| # | Minimum | Recommended |
|---|---|---|
| 1 | macOS 13+ (Apple Silicon, Comfy Desktop); manual install works on ‘any recent macOS version’ per official README with no stated minimum | |
| 2 | Apple Silicon (M1, M2, M3, or M4) | |
| 3 | 8GB (Comfy Desktop minimum); | 16GB+ unified memory recommended for heavier workloads |
| 4 | 4.85GB (Comfy Desktop base install, excludes model files) | |
| 5 | Apple Silicon GPU via Metal acceleration (PyTorch nightly install) |
Ratings
Not available, but we appreciate help! You can help us improve this page by contacting us.
Repository
License
Categories
Alternatives
Notes
- Comfy Local (the engine documented here) is free and open source; Comfy Cloud, Comfy API, and Comfy Enterprise are separate paid products built on the same engine.