Explore the canvas
Learn how nodes and connections work, add source material, select content, and hand it to Agent for the next step.
The canvas is where you keep and make content. Uploaded images, written notes, reference videos, and anything Agent creates all stay on the canvas.
A canvas is made of nodes and connections. Nodes hold content. Connections show how that content is related.
What is a node?
Every piece of content on the canvas is a node. For example:
- An uploaded product image;
- A brief or script;
- An image created by Agent;
- A video or audio result;
- Research organized by Agent.
Each node remains independent. You can select one node to view it or select several nodes and give them to Agent together. When you create a new result, the source nodes remain on the canvas.
What is a connection?
A connection shows a relationship between two nodes. If you use a product image to create an ad visual, the product image and the new image are connected. If you then turn that image into a video, the video is connected after it.
A simple chain might look like this:
Product image node + instruction node → image result node → video result node
Follow the connections to see which material led to a result. Connections help you understand the relationship, but you should still check which nodes Agent is referencing before generation begins.
The main parts of the interface
Canvas
| What you want to do | Mouse or trackpad | Mac shortcut |
|---|---|---|
| Pan across the canvas | Scroll the mouse wheel or swipe with two fingers on a trackpad. You can also hold the middle or right mouse button and drag | — |
| Zoom in or out | Pinch with two fingers on a trackpad. With a mouse, hold ⌘ while scrolling. You can also drag the zoom control in the lower-left corner | ⌘ + = or ⌘ + + to zoom in; ⌘ + - to zoom out |
| Select one node | Click the node | — |
| Select several nodes | Hold Shift and click each node, or drag a selection box from an empty area of the canvas | Shift + click |
| Move a node | Place the pointer over the node, then hold the left mouse button and drag | — |
| View every node | Select Fit view in the lower-left corner. The canvas adjusts until all nodes are visible | — |
| Navigate with the minimap | Open the minimap from the lower-left corner. Click a destination to jump there, or drag the white viewport box to keep moving | — |
Dragging from an empty area with the left mouse button starts a selection box; it does not pan the canvas. To pan, use the mouse wheel, a two-finger trackpad gesture, or hold the middle or right mouse button while dragging.
When the cursor is inside the Agent input, a text node, or another text editor, ⌘ + and ⌘ - do not zoom the canvas. Click an empty area first, then use the shortcut.
Agent
Describe the task in Agent and give it the nodes it should use. Agent reads the content you explicitly select and adds new results back to the canvas.
When you reference a node, state what it is for:
Use @Front product reference to preserve the bottle shape and label placement. Use @Warm lighting reference only for the lighting, not for the product shown in that image. Give me three composition directions first. Do not generate video yet.
This tells Agent which image defines the product and which one is only a style reference.
Add menu
Use the Add menu to create text, image, video, and audio nodes, or to upload local files. Once added, each item becomes a node on the canvas.
| Entry point | Mouse | Mac trackpad |
|---|---|---|
| Add button on the left | Select + at the top of the left toolbar, then choose the content type | Move the pointer to +, tap the trackpad, then tap the content type |
| Add at a specific position | Double-click an empty area. The Add menu opens where you clicked | Move the pointer to the destination and tap or press the trackpad twice |
| Context menu | Right-click an empty area, then choose the content or tool you want to add | Two-finger click an empty area, then choose an item from the menu |
To upload an image, video, audio file, or document from your computer, select +, choose Upload, and select the file. Once the upload finishes, the file appears as a new node on the current canvas.
The items shown in the menu may vary with your account and canvas permissions. If an option is unavailable or missing, follow the message shown in the interface.
Asset library and templates
- The asset library stores characters, settings, styles, sounds, and objects you expect to use again;
- Elements group references for the same person, product, or character so supported video models can use them together;
- Templates store reusable arrangements of nodes and connections.
A good first canvas workflow
- Add one image or source file to the canvas. It becomes a node;
- Select the node, open Agent, and describe what you want to make from it;
- If you selected several nodes, explain the purpose of each one;
- When Agent finishes, review the new node on the canvas;
- Follow the connections to check which inputs the result used;
- Move the nodes into a clear arrangement so the current material is easy to find.
Your first canvas does not need a complex structure. Complete one select a node → give it to Agent → receive a new node flow to understand how the canvas works.
Save status
The canvas shows its save or sync status. Before closing the page, wait if you still see saving, offline, or sync failed, and check your network connection.
Completion checklist
- You know that content is stored as nodes on the canvas;
- You can give one or more selected nodes to Agent;
- You can follow a connection back to the inputs used by a result;
- New content created by Agent appears on the canvas;
- The page shows that your latest changes have been saved.