Hobbyists, artists and developers building custom Stable Diffusion, video and audio generation pipelines.

A node-based graphical interface for building and running Stable Diffusion, video and audio diffusion model workflows.

Self-hosted desktop or server application, runs locally or on a rented GPU cloud instance.

First released in 2023 by comfyanonymous, now maintained by Comfy-Org with continuous updates through 2026.

Gives fine-grained visual control over complex multi-model diffusion pipelines that simple text-box UIs cannot express.

Requires Python, PyTorch and a CUDA-capable GPU (or CPU/MPS fallback) plus downloaded model checkpoints.

Open source under GPL-3.0, self-hosted, no credentials required, actively maintained by Comfy-Org, over 120k GitHub stars.