ComfyUI

by comfyanonymous (open-source community)

A powerful and modular node-based UI for Stable Diffusion and open-source image/video generation — run locally, fully offline, with no internet required.

See https://github.com/comfyanonymous/ComfyUI

Features

  • Node-based visual workflow editor — chain generation steps as composable graphs
  • Text-to-image and image-to-image generation with Stable Diffusion models
  • Text-to-video and image-to-video via models like LTX 2.3, Stable Video Diffusion, AnimeDiff
  • Inpainting, outpainting, and masking workflows
  • Pose and depth map conditioning (OpenPose, ControlNet)
  • LoRA, DreamBooth, and IP-adapter support for consistent styles and subjects
  • Detailer and helper models for fixing artifacts (faces, hands, text)
  • Subgraph support (v0.3.52) for modular, reusable workflow components
  • Export workflows to Python; cloud hosting integrations
  • Community plugin ecosystem (CivitAI, custom nodes)
  • Runs fully offline on local hardware (Mac or Windows); no API key or internet needed

Superpowers

ComfyUI is the power-user’s tool for open-source generative AI. Its node graph interface lets creators compose sophisticated multi-step pipelines — combining image generation, upscaling, face restoration, video animation, and audio — in a single reusable workflow. The full offline capability means complete privacy and uncensored model access. Professional workflows go far beyond simple prompts: image-to-image conditioning, segmentation (Segment Anything), pose sequences for video, and LLM-assisted prompting can all be wired together. Shareable workflow JSON files are a community currency. Ideal for developers, researchers, and creators who need maximum control and reproducibility.

Pricing

  • Free and open-source (MIT license)
  • No subscription required; runs on your own hardware
  • Some cloud providers offer hosted ComfyUI instances (e.g., RunPod, Replicate)