Skip to content

SwarmUI Guide (2026): Install, First Run, Models and the Comfy Workflow Tab

6–9 minutes
Amazed woman with wavy brown hair in a lime and grey bomber jacket by a night city window holding a card reading SWARMUI, with the text Easy AI Art with SwarmUI

SwarmUI is a free, open-source web interface for AI image and video generation. It gives you a clean “type a prompt and press Generate” screen like the classic WebUIs, but runs ComfyUI underneath, so you also get ComfyUI’s speed, model support and full node editor when you need it.

It was formerly called StableSwarmUI, and it supports image models like Stable Diffusion, Flux and Krea 2, video models such as Wan, LTX-2 and MiniMax H3, and some audio models like ACE-Step.

This guide shows how to install SwarmUI on Windows, Linux and Mac, walk through the first-run installer, add your models (including ones you already have in ComfyUI or Forge), use the Comfy Workflow tab, open it to your network, and fix common problems. It follows the official SwarmUI README and docs as of November 2026.

TL;DR

  • What: a simple-to-use generation UI with ComfyUI as its backend. MIT licensed, currently v0.9.8 Beta.
  • Windows: download install-windows.bat from the GitHub release, save it outside Program Files, run it.
  • Linux: Python 3.11 or 3.12, then install-linux.sh or clone and run launch-linux.sh.
  • Mac: Apple Silicon only; brew install dotnet python@3.11, clone, ./launch-macos.sh.
  • Open: http://localhost:7801.
  • Reuse models: Server > Server Configuration > ModelRoot.

What SwarmUI Is

The README describes SwarmUI as a modular AI image generation web UI with an emphasis on making power tools easy to access. Key points:

  • Backend: it can install ComfyUI for you and drive it. It can also use AUTOMATIC1111’s WebUI as a backend.
  • The name: “Swarm” refers to using a swarm of GPUs for one person’s generations, for example big grids across several machines.
  • Status: v0.9.8 Beta, which the README calls “Almost-Release”.
  • License: MIT for SwarmUI itself. Bundled parts such as ComfyUI have their own licenses, and models have theirs.
  • Cost: free; the developer accepts donations on Patreon.

In practice, SwarmUI suits people who find ComfyUI’s node graph intimidating but still want its model support and speed. You get a familiar prompt box, sliders and a model browser, and when a tutorial gives you a ComfyUI workflow, you can load it in the Comfy tab and turn it into simple parameters on the Generate screen.

SwarmUI vs ComfyUI vs Forge

SwarmUI ComfyUI Forge
Interface Simple Generate tab plus node editor Node graph Classic tabs and sliders
Engine ComfyUI backend Itself Its own (A1111-based)
Video models Yes Yes Limited
Multi-GPU / multi-machine Built-in backends Manual No
Best for Easy UI with ComfyUI power Custom workflows A1111-style users

More on the others: our ComfyUI guide and Forge WebUI guide.

Install on Windows

  1. Download install-windows.bat from the latest SwarmUI release on GitHub.
  2. Save it in a folder where you want SwarmUI installed. Don’t use Program Files or another protected folder.
  3. Double-click it. On Windows 11, git and the .NET SDK install automatically. On Windows 10 you may need to install git and the .NET 8 SDK yourself first.
  4. If the window closes early, run it again. The README says it sometimes needs two runs.
  5. Wait for the browser to open the installer at http://localhost:7801/Install.

The installer adds a desktop icon you can use to launch SwarmUI later. Don’t run it as Administrator; SwarmUI warns against that.

Manual route: install git and the .NET 8 SDK (the README also suggests the .NET 10 SDK for future versions), then:

git clone https://github.com/mcmonkeyprojects/SwarmUI
cd SwarmUI
launch-windows.bat

Install on Linux

  1. Install git and Python (3.11 or 3.12 recommended; avoid 3.14+):
    sudo apt install git python3-pip python3-venv
  2. Easy way: download install-linux.sh from the GitHub release, then:
    chmod +x install-linux.sh
    ./install-linux.sh
  3. Manual way:
    git clone https://github.com/mcmonkeyprojects/SwarmUI
    cd SwarmUI
    ./launchtools/linux-dotnet-install.sh
    ./launch-linux.sh
  4. Open http://localhost:7801.

On a headless server, start it with ./launch-linux.sh --launch_mode none --host 0.0.0.0.

Install on Mac

SwarmUI supports Apple Silicon (M1 and later) only, not Intel Macs.

brew update
brew doctor
brew install dotnet
brew install python@3.11
brew install virtualenv
git clone https://github.com/mcmonkeyprojects/SwarmUI
cd SwarmUI
./launch-macos.sh

Make sure brew doctor shows no errors before continuing. If you’d rather use ComfyUI directly on a Mac, see how to install ComfyUI on Mac.

The First-Run Installer

  1. Pick a theme.
  2. Choose who it’s for: just yourself, or yourself plus your local network.
  3. Choose a backend. The docs recommend ComfyUI for local use; SwarmUI installs it for you.
  4. Pick starter models to download, or skip if you already have models.
  5. Confirm and install. Model downloads can take several minutes.

You can also skip the installer and accept the defaults.

Add Models and Reuse Existing Folders

Already have models in ComfyUI, Forge or AUTOMATIC1111? Point SwarmUI at them instead of downloading twice:

  1. Open Server > Server Configuration.
  2. Set ModelRoot to your existing models folder. You can list several folders separated by ;; the first is used for downloads.
  3. Set SDModelFolder to checkpoints for ComfyUI or Stable-diffusion for A1111/Forge. You can enter both, like Stable-diffusion;checkpoints.
  4. Click Save and refresh the Models list on the Generate tab.

Need models? See how to download models from Hugging Face.

Generate Your First Image

  1. Open the Generate tab.
  2. Pick a model from the Models browser at the bottom.
  3. Type a prompt and, optionally, a negative prompt.
  4. Adjust parameters on the left, such as resolution, steps and CFG, or leave the model’s defaults.
  5. Click Generate.

The Grid Generator tool is one of SwarmUI’s strengths: it renders the same prompt across different settings so you can compare them side by side.

The Comfy Workflow Tab

  1. Open Comfy Workflow Editor (needs a ComfyUI backend).
  2. Build or load a workflow. Use Primitive nodes with clear titles for anything you want as a parameter.
  3. End with one Save Image node.
  4. Click Use This Workflow. Back on Generate, your parameter list now matches the workflow.
  5. To go back, scroll up in the parameters and click Disable Custom Comfy Workflow.

Looking for graphs to import? Try our ComfyUI workflows.

Use SwarmUI From Other Devices

  • Local network: Server > Server Configuration, change Host from localhost to 0.0.0.0, save and restart. Allow port 7801 in your firewall if needed.
  • Over the internet: SwarmUI supports Cloudflared (--cloudflared-path) and ngrok (--ngrok-path) so you don’t need port forwarding.
  • Security: the docs say public hosting is for experienced admins only. Put authentication and HTTPS in front, as some routes pass through raw ComfyUI access.

Launch Options and Docker

Option Meaning
--host Address to listen on (default localhost)
--port Port (default 7801)
--launch_mode web opens a browser, none doesn’t

Launch scripts: launch-windows.bat, launch-linux.sh and launch-macos.sh. Docker setups are documented in the repo’s Docker doc, and there are templates for Runpod, Vast.ai and Colab if you don’t have a local GPU.

Tips for New SwarmUI Users

  • Start on the Generate tab. You don’t need to touch nodes until you want a custom workflow.
  • Use the model’s recommended settings. Resolution, steps and CFG differ a lot between SDXL, Flux and video models.
  • Compare settings with grids. Rendering one prompt across several samplers or CFG values is the fastest way to learn what each does.
  • Keep one model folder. Point ModelRoot at the folder your other tools use so you never download twice.
  • Check Server > Logs first when something fails; the backend error is usually spelled out there.
  • Add more GPUs later. SwarmUI can use multiple backends, including other machines, for big batches.
  • Back up your settings before big updates by copying SwarmUI’s Data folder.

Troubleshooting

Problem Likely cause Fix
Installer window closes instantly First run setting up dependencies Run install-windows.bat again
Browser doesn’t open Launch mode or slow start Go to http://localhost:7801 yourself
Install fails on Windows 10 Missing git or .NET 8 SDK Install both, then rerun
Python errors on Linux or Mac Unsupported Python version Use Python 3.11 or 3.12
Models don’t appear Wrong ModelRoot or folder name Fix ModelRoot and SDModelFolder, save, refresh
Can’t reach it from phone or other PC Host still localhost Set Host to 0.0.0.0 and restart
Generation errors Backend problem Check Server > Logs
Black images VAE or precision issue in ComfyUI See our ComfyUI black image fix

FAQ

Is SwarmUI free?

Yes. It’s free and open source under the MIT license.

Is SwarmUI the same as StableSwarmUI?

Yes. SwarmUI was formerly known as StableSwarmUI.

Does SwarmUI need ComfyUI?

It uses ComfyUI as its main backend and can install it for you during setup.

What port does SwarmUI use?

7801 by default, at http://localhost:7801.

Can I use my existing ComfyUI or Forge models?

Yes. Set ModelRoot and SDModelFolder in Server Configuration to point at your existing folders.

Does SwarmUI work on Mac?

Yes, on Apple Silicon Macs. Intel Macs aren’t supported.

Which Python version does SwarmUI need?

The README recommends Python 3.11 or 3.12 on Linux and Mac, and says not to use 3.14 or later.

Sources

Free weekly email

Get new AI guides every week

One short weekly email with our newest local AI, ComfyUI and AI tool guides. No spam, no pop-ups, just the good stuff.

Free. Unsubscribe anytime in one click. We only use your email to send this newsletter. See our privacy policy.

Amazed woman with wavy brown hair in a lime and grey bomber jacket by a night city window holding a card reading SWARMUI, with the text Easy AI Art with SwarmUI