Getting started

Your first workflow

Text prompt to finished image in five steps, then how to re-run just one node.

The shortest useful graph is a prompt feeding an image model. Here is that, then one step further.

Five steps

·Press Tab on the canvas, type text, and place a Text Input. Type your prompt into it.
·Press Tab again and place an image model. Any of them will do to start.
·Drag from the Text Input's output dot to the model's prompt input. The connection takes the colour of the port type.
·Select the model node and press 1 so the viewer is watching it.
·Press Ctrl Enter to run. The node shows its queue position, then its result.

Drag a noodle into empty canvas instead of onto a node and the add-node menu opens filtered to things that can actually accept what you are dragging. It is usually faster than Tab.

One step further

Add a Merge / Composite and feed it two images to stack them, or a Grade to push the result around. Both run in your browser, so you can iterate on the look without generating again.

Re-running one node

You rarely want to re-run everything. Select a node and use Run in the properties panel to execute just that one with its current inputs. Browser nodes also re-run on their own when something upstream of them changes; AI nodes never re-run unless you ask them to.