Image & Vision MIT

SwarmUI

Polished, multi-user Stable Diffusion UI. Runs ComfyUI under the hood with a classic interface on top. Best UX out of the box in 2026.

Website
Platforms: windowsmacoslinux

SwarmUI (formerly StableSwarmUI) is a polished frontend for local image generation that uses ComfyUI as its inference engine but presents a familiar form-based interface instead of a node graph. The thesis is simple: most people want the power and model support of ComfyUI without learning to wire nodes. SwarmUI gives you that. You type a prompt, pick a checkpoint, set resolution and sampler, and hit generate. Behind the scenes it builds a ComfyUI graph for you, and if you want to see or edit that graph, a click opens it in ComfyUI’s canvas. It also ships real multi-user support with login, per-user model access, and a shared backend — something neither A1111 nor ComfyUI offer natively. By 2026 it is the easiest way to put a Stable Diffusion server in front of a small team or household.

Key Features

  • ComfyUI backend, form-based frontend. You get Flux, SDXL, SD3.5, ControlNet, LoRAs, and every model ComfyUI supports, but through a tabbed UI with dropdowns instead of nodes. Power users can pop into the underlying graph at any time.
  • Real multi-user with permissions. Logins, per-user history, per-user default models, and admin controls over who can install extensions or change backend settings. Deploy one SwarmUI instance and let several people use it without stepping on each other.
  • Backends API and TTY backend pooling. SwarmUI can drive multiple ComfyUI instances simultaneously — useful for a multi-GPU box or for queuing work across a small cluster. It also exposes a clean REST API and a built-in OpenAI-compatible endpoint.
  • Polish out of the box. Image history with thumbnails, prompt templates, batch generation with grid preview, one-click Civitai model download, and a dark/light theme. It feels like a finished product, not a developer tool.

When to Use SwarmUI

Choose SwarmUI when you want ComfyUI’s model coverage without the node-graph learning curve, or when more than one person will use the server. It is the best default for a home server shared by a family or a small studio. Skip it if you specifically want the A1111 tab layout (use Forge) or if you live in node graphs and want nothing in between (use ComfyUI directly).

Setup

# Windows: download the release zip from GitHub and run the launcher -- it
# installs ComfyUI and Python for you.
# Cross-platform from source:
git clone https://github.com/mcmonkeyprojects/SwarmUI
cd SwarmUI
./make.sh run            # or: make run

# Docker:
docker run -d --gpus all -p 7801:7801 \
  -v $PWD/data:/data -v $PWD/models:/models \
  --name swarmui mcmy/swarmui:latest

The first run downloads and configures a ComfyUI backend in src/ComfyUI/. Models can be shared with an existing ComfyUI install via a symlink or the --externals flag. The UI is at http://localhost:7801.

See also