All docs

ComfyUI

Run ComfyUI on one of our GPUs, inside this project.

ComfyUI opens a real ComfyUI in the page, running on a GPU we rent for you. Pick a machine, pick the shot your saves should land on, and press Start. Install any custom nodes you like with the Manager. Your trained models are already in the LoRA list.

You pay for every second the machine is running, including the time you spend thinking. The meter in the bar shows the running total. Stop the session when you are done.

  1. Choose a machine. Bigger machines cost more per hour and fit bigger models.
  2. Choose an environment: Fresh, or one you saved before. Under Run for, choose how long it may run.
  3. Under Save outputs to shot, choose the shot that saved images and videos should go to.
  4. Press Start ComfyUI and wait for it to appear. You are not billed while it boots. The first 30 minutes are reserved from your credits when you start.
  5. In ComfyUI, open the Root tab in the left sidebar. It shows this project's images, videos and audio; click one to add it to your graph.
  6. Every result a Save node writes (images, videos, audio) lands by itself as a new take on the shot you chose, and everyone on the project can see it. To keep a Preview Image result, click its node, then press Save selected output to my shot in the Root tab.
  7. Press Save environment to keep your installed nodes, open canvas and saved workflows for next time. Save as new… keeps them as a separate environment.
  8. Full screen gives ComfyUI the whole window, and Exit full screen brings the page back. While it runs, +30 min and +1 h give it more time, up to your plan's limit.
  9. Press Stop when you are finished.

A session with no activity for 5 minutes is stopped for you, so a forgotten tab does not run up a bill. Anything you install is your responsibility, including its licence.

Still stuck? Support has the common problems and a way to reach us.