Skip to main content
This takes you from install to a first local render and an exported project. You need Python 3.11 or newer and, for local generation, a GPU. Z-Image Turbo, the model used below, fits in about 11.5GB of VRAM at 1024×1024.
1

Install and start the app

On Windows, use .\webui.bat in place of ./webui.sh. Open http://127.0.0.1:8848. See Install for GPU-specific steps.
2

Create a project

On the home screen, name the project and click Create. A project is one portable folder holding your imported assets, every generated take and the graphs behind them, so you can move it, back it up or hand it to someone.
3

Generate your first frame

Add a Z-Image Turbo node to the canvas. If the model isn’t on disk yet, the node’s model popup downloads it with visible progress. Write a prompt and press Run.Each run adds a new take to the frame and nothing is overwritten. Star the take you want to keep.
4

Add a character

Open the Characters panel and build a character on the canvas from a photo or two, or import a .char from Omnichar Cloud or ComfyUI. Then pick it on any model node that takes one. See The .char character file for which models do.
5

Export the project

From the home screen, Export zips the project into one file with its assets, takes and graphs. Whoever opens it can re-run the same pipeline.

​
No GPU?

Add an API node, pick a hosted model such as GPT Image 2 or Seedance, and add your fal key in Settings. The result lands on the canvas as a take, next to your local renders. You can also run the whole app on a rented GPU with the RunPod template.

​
Next steps

Supported models

Every local model, with VRAM needs and file layout.

Command-line options

Port, LAN access, low-VRAM mode and more.