Get it running
Paste this into the AI that runs on your computer. It does the install, checks it works, and tells you what to type first. If anything fails, paste the error back to it.
Paste into Claude Code
Install ComfyUI (https://github.com/comfyanonymous/ComfyUI) on this computer and get it working for me. I am not a developer; explain each step in one line as you go and never skip one. You are Claude Code, running on my machine. 1. Check what this machine already has (git, Python or Node as the project needs, Docker if the README says so). Tell me anything missing and install it, asking me before anything that needs my password. 2. Clone https://github.com/comfyanonymous/ComfyUI into ~/tools/comfyui and follow the README's install exactly. 3. Configure it. If it needs an API key or a login, stop and ask me for it; never guess one and never store it anywhere except where the README says. Tool-specific notes: Match the pytorch install command to your hardware — NVIDIA, AMD/ROCm, Intel XPU, and Apple Silicon each need a different command from the README; getting this step wrong is the most common failure. Put any model checkpoint files you already have into models/checkpoints, and any VAE files into models/vae, before the first run. Start it with python main.py from the project folder and open the local URL it prints in the terminal. If you see "Torch not compiled with CUDA enabled", uninstall torch (pip uninstall torch) and reinstall it with the exact command that matches your GPU. 4. Run the smallest test the README gives, and show me the output. 5. When it works, tell me what to type first, in one line, for this job: Run an example workflow. If anything fails, show me the exact error and fix it before going on.
Paste into Kimi
Install ComfyUI (https://github.com/comfyanonymous/ComfyUI) on this computer and get it working for me. I am not a developer; explain each step in one line as you go and never skip one. You are Kimi, running on my machine. 1. Check what this machine already has (git, Python or Node as the project needs, Docker if the README says so). Tell me anything missing and install it, asking me before anything that needs my password. 2. Clone https://github.com/comfyanonymous/ComfyUI into ~/tools/comfyui and follow the README's install exactly. 3. Configure it. If it needs an API key or a login, stop and ask me for it; never guess one and never store it anywhere except where the README says. Tool-specific notes: Match the pytorch install command to your hardware — NVIDIA, AMD/ROCm, Intel XPU, and Apple Silicon each need a different command from the README; getting this step wrong is the most common failure. Put any model checkpoint files you already have into models/checkpoints, and any VAE files into models/vae, before the first run. Start it with python main.py from the project folder and open the local URL it prints in the terminal. If you see "Torch not compiled with CUDA enabled", uninstall torch (pip uninstall torch) and reinstall it with the exact command that matches your GPU. 4. Run the smallest test the README gives, and show me the output. 5. When it works, tell me what to type first, in one line, for this job: Run an example workflow. If anything fails, show me the exact error and fix it before going on.
Paste into Gemini CLI
Install ComfyUI (https://github.com/comfyanonymous/ComfyUI) on this computer and get it working for me. I am not a developer; explain each step in one line as you go and never skip one. You are Gemini CLI, running on my machine. 1. Check what this machine already has (git, Python or Node as the project needs, Docker if the README says so). Tell me anything missing and install it, asking me before anything that needs my password. 2. Clone https://github.com/comfyanonymous/ComfyUI into ~/tools/comfyui and follow the README's install exactly. 3. Configure it. If it needs an API key or a login, stop and ask me for it; never guess one and never store it anywhere except where the README says. Tool-specific notes: Match the pytorch install command to your hardware — NVIDIA, AMD/ROCm, Intel XPU, and Apple Silicon each need a different command from the README; getting this step wrong is the most common failure. Put any model checkpoint files you already have into models/checkpoints, and any VAE files into models/vae, before the first run. Start it with python main.py from the project folder and open the local URL it prints in the terminal. If you see "Torch not compiled with CUDA enabled", uninstall torch (pip uninstall torch) and reinstall it with the exact command that matches your GPU. 4. Run the smallest test the README gives, and show me the output. 5. When it works, tell me what to type first, in one line, for this job: Run an example workflow. If anything fails, show me the exact error and fix it before going on.
Paste into Codex
Install ComfyUI (https://github.com/comfyanonymous/ComfyUI) on this computer and get it working for me. I am not a developer; explain each step in one line as you go and never skip one. You are Codex, running on my machine. 1. Check what this machine already has (git, Python or Node as the project needs, Docker if the README says so). Tell me anything missing and install it, asking me before anything that needs my password. 2. Clone https://github.com/comfyanonymous/ComfyUI into ~/tools/comfyui and follow the README's install exactly. 3. Configure it. If it needs an API key or a login, stop and ask me for it; never guess one and never store it anywhere except where the README says. Tool-specific notes: Match the pytorch install command to your hardware — NVIDIA, AMD/ROCm, Intel XPU, and Apple Silicon each need a different command from the README; getting this step wrong is the most common failure. Put any model checkpoint files you already have into models/checkpoints, and any VAE files into models/vae, before the first run. Start it with python main.py from the project folder and open the local URL it prints in the terminal. If you see "Torch not compiled with CUDA enabled", uninstall torch (pip uninstall torch) and reinstall it with the exact command that matches your GPU. 4. Run the smallest test the README gives, and show me the output. 5. When it works, tell me what to type first, in one line, for this job: Run an example workflow. If anything fails, show me the exact error and fix it before going on.
What to point it at first
1
Run an example workflow
Load one of the official example workflows and generate a single image end to end — this proves your GPU, Python environment, and models are wired together correctly before you build anything custom.
2
Rebuild one manual task as a workflow
Take something you currently do by hand across several image tools and lay it out as one wired workflow, so it becomes a single click next time.
3
Save and reopen a workflow
Save a working graph as JSON, close ComfyUI, and reopen that file to confirm it reproduces the exact same image and seed — that repeatability is the whole point of using it.
What it must never do unattended
- Never publish a cut you have not watched to the end.
- Check the licence of every model and every clip before it goes out under your name.
- Never let it install or enable a third-party custom node without you reviewing what it adds — custom nodes run as ordinary code on your machine.
Who made it
comfyanonymous/ComfyUI on GitHub, under the GPL-3.0 licence. 132,121 stars, checked 9 September 2026. Last change 9 September 2026. We did not write it; we checked it, and wrote this page so you can use it.
Also in Content
Tool · Content · autoclip
One long video into a week of short clips
AutoClip watches a long video for you and pulls out the interesting parts.
20 min · your AI on your computer · docker, an API key for one AI provider (Qwen, an OpenAI-compatible API, Gemini, or SiliconFlow) — or a local Ollama/LM Studio install instead, with no key neededOpen
Tool · Content · MoneyPrinterTurbo
Type a topic, get a finished short video
MoneyPrinterTurbo turns a topic into a finished short video without you touching an editing timeline.
20 min · your AI on your computer · python 3.11 or later, an API key for a script-writing AI model (or a local Ollama setup instead), ffmpeg (usually auto-downloaded, see brief)Open
Tool · Content · Pixelle-Video
Make short videos without touching an editor
Pixelle-Video turns a single topic into a finished short video, end to end, through a web page rather than a code editor.
25 min · your AI on your computer · uv (Python package manager), ffmpeg, an API key for an LLM to write the script, plus an image/video provider key or a local ComfyUI/RunningHub setupOpen