The canvas workspace

The canvas gets the whole screen, and the two side columns float over it:

• Left · Canvas elements — every node on this canvas as one list. Search it, filter by type, click a row and the canvas moves to that node.
• Middle · Canvas — where the work happens. Nodes sit here, and a wire between two of them says who feeds whom.
• Right · AI Agent — a chat assistant. Ask it to generate images, change a frame or build a workflow, all from here.

The top of the left column shows the project name and the current canvas — the name is also a dropdown: open it to switch projects, create a new one, or delete the current one — every row carries the project's cover thumbnail, since a column of "Untitled" projects is impossible to tell apart by name alone. The strip below it has two tabs, Canvas and Asset. The Asset tab splits into two further tabs, Personal and Agent: Personal holds the material kept in your asset library (it spans canvases and follows your account) — click a row and it drops onto the canvas; Agent holds the nodes the Agent built on this canvas. Under that is the node list, and the small buttons to its right handle sorting, filtering by type and search. The button at the right end of the tab row opens Assets, where folders, uploads and batch management live; the row in that dialog's title switches between Assets and Subject library (see 10.7).

Both columns start collapsed, leaving the whole screen to the canvas; their open / collapsed state is remembered per project, so collapsing one project does not affect another. Collapse the left column from the button at its bottom-left — a pill then floats at the top-left that keeps showing the project and the canvas, and you bring the column back with Assets, which joins the group of controls at the bottom-left. Collapse Agent from the rightmost button in its header — an AI Agent button then stays at the top-right to call it back. Drag either column's edge to resize it — the width you set survives a refresh.

On a phone the canvas takes the whole screen and both columns become drawers: pull the left one out with Assets at the bottom-left, and Agent with the chat icon on the bottom toolbar.

A node is one piece of media or one instruction: image, video, audio, text, script, video editor, 3D previz, frame study, plus the style / effect samples you pick out of the library. Run a wire between two nodes and the first one's output is the second one's input — which is what makes the recipe for a film something you can see.

• The name sits above the frame; double-click to change it. The small text at the top-right is the resolution or the duration.
• Double-click empty canvas, press Tab, or hit the + at the far left of the bottom toolbar — any of the three makes a new node.
• Drag from the dot on a node's edge out to empty space and it asks you what to create there.
• Select a node and an input bar appears under it: write the prompt, pick the model, hit generate.

The strip at the bottom center of the canvas holds this page's main actions:

On a phone the toolbar keeps only New node, Pointer tool, My toolbox and AI Agent; the rest move into the ⋯ at its right end — which also holds Account (credits, your account page, switching spaces), the style and effect libraries, and the two share actions that sit at the top right on desktop.

They only change how you look at this canvas, never its content:

• Tidy up — packs the scattered nodes into a near-square grid at their real sizes, keeping the order they were already in and clearing the overlaps (⌥⇧F).
• Minimap — opens a thumbnail at the bottom-right; use it to jump around once there are a lot of nodes.
• Hide connections — enough wires and the canvas turns into a net; click this to take them all away and look at the pictures only, click again to bring them back. The connections themselves are untouched, just not drawn.
• Snap to grid — with it on, a dragged node snaps to the grid lines and everything lines up.
• Zoom — click the percentage at the right end for a menu: Zoom in, Zoom out, Fit to screen, and 50% / 100% / 200%. Pinching on a trackpad and ⌘ plus scroll work too.

Two or three pills float at the top right of the canvas. Left to right:

• Share — opens Publish and share: Publish on DOTV puts the finished film on the home page, and Share a link hands out a read-only link that whoever gets it can copy into a project of their own.
• Collaborator avatars — whoever is on this canvas right now shows up here. Click one to follow their view: when they move the canvas, yours moves with them.
• Your avatar — the one at the right end of the second pill is you.

One last button sits outside any pill and changes with the face you are on: with Agent collapsed it is an AI Agent button that brings the panel back; on the chat face it is Split with canvas, which pulls the canvas out. The storyboard face adds a search box next to the avatars.

While anything is generating, one more pill appears to the left of Share: "2 running · ~3 min" — how many tasks are going, and how long until the whole batch finishes.

Open it for a list: which node each job is, what percentage it is at, and how much longer (estimated from average runtimes, ticking by the second). Click an entry to jump straight to that node — even on another canvas: the canvas switches, then the camera and the selection land on it. So you don't have to sit on this canvas waiting for a render: keep working on another one, and click back when it's nearly done. When everything finishes, the pill turns into a quiet "all done" (if some failed, they're marked red) — read it and carry on; it takes no space.

The right column is a conversational way in, good for anything you would rather just ask for — a set of storyboard shots, a frame reworked to follow a reference image, the nodes you selected wired into a workflow.

• Type @ in the box to reference a node, an asset, a model or a Skill on the canvas.
• It can create nodes on the canvas, submit generations and run a whole group. It spends credits by default, and the first time you use it you are asked to confirm that setting.
• Conversations are kept as threads; starting one at the top-right opens a new one.