Skip to main content
This page builds a two-step workflow: a text description goes into an image model, which makes a promo image for a yoga class.
The workflow editor for Class promo image: a #1 Text Input node describing a bright yoga studio at dawn, connected to a #2 Text to Image node with Model Nano Banana Pro, Size 1:1, the prompt @#1 Text Input, Number of images 1, and Run (7 credits), with Add Node and zoom controls

​
Build a workflow

1

Add your first node

Click Add Node at the top right and pick one, such as Text Input. It appears in the middle of the canvas.
The Add Node list: Text Input, Upload Media, Select Media, Text to Image, Image to Image, Upscale Image, Image Splitter, Text to Video, and Image to Video
2

Connect the next node

Drag from the round handle on the right of a node to the left handle of another. Or drop the connection on an empty spot: a list of nodes that can take that output opens, with a search box.
The Add text node list that opens after dropping a connection, with Search nodes and the options Text to Image, Text Generation, Text to Video, Text to Speech, Output, and First-Last Frame to Video
Handles are colored by type: blue for text, purple for images, cyan for video, and indigo for audio. You can only connect matching types.
3

Use the result in the prompt

Connecting into a prompt adds a reference like @#1 Text Input. When the workflow runs, the reference is replaced by that node’s result. Type @ in any prompt to reference another connected node, and add your own words around it, for example: “@#1 Text Input, in the style of a watercolor painting”.
4

Set up each node

Pick the model and settings in each node. The Run button shows what a node costs before you run it.

​
Run a workflow

Click Run in a node, or the ▶ in its header. Before it runs, any node it depends on that doesn’t have a result yet runs first, so running the last node runs the whole chain. Nodes that don’t depend on each other run at the same time.
The same workflow after running: the #2 Text to Image node shows the generated image of a small group walking into a bright yoga studio with plants and wooden floors, with a green check
  • A running node has a blue border. A finished one gets a green check, and its settings fold away; click the node’s settings icon to show them again.
  • A node that fails turns red and shows why, for example a missing prompt or a model that isn’t on your plan.
  • Results stay in the node. Open or download them from the node’s menu. Images and videos you make also appear in AI Images and AI Videos, and in History.
To make a new version, change the input and click Regenerate on the node you want to redo. Nodes that already have a result are reused, not run again, so regenerate each node you want to change.
Workflows run in your browser. Keep the page open until your nodes finish; if you leave, the page warns you, and nodes that were still running stop updating.

​
Nodes

Video nodes have the same models, lengths, aspect ratios, and resolutions as AI Videos. Their outputs include the video’s first and last frames, to chain clips together.
Select several nodes and click Group (or press Cmd/Ctrl + G) to put them in a group you can color, lock, and move together.
In Text Generation, Text to Speech, and Talking Video, choose the model or voice yourself before running, rather than relying on the one shown.

​
Work on the canvas

The top bar shows All changes saved once your edits are saved. The download button exports the workflow as a file.

​
What each node costs

Each node uses the same allowance, at the same price, as the feature it runs. Whoever runs a node pays for it, including in a shared workflow. On the Free plan, a node set to a premium model fails with a message to upgrade. Pick a model that’s available on Free; the default image model, Nano Banana Pro, isn’t. See Plans and credits.

Share and collaborate

Invite people and edit together.

Workflows

Templates, import, and export.