Canvas

How the canvas works#

1. Left — the project menu. “Canvas” is the first item of the “Production” group.
2. Top — the canvas picker. «Холст 1» (Canvas 1) is a dropdown: other canvases, “All” — a feed from every canvas in the project, “Create new”.
3. Center — the generation feed. Every image or video is numbered; videos show their duration.
4. Right — the generation panel. The Image, Video, Assistant and Errors tabs sit at the bottom of the panel: above them are the prompt and references, below them the model and settings. On the “Assistant” tab the same panel holds the chat with this canvas’s assistant.

01Generate images and video#

The canvas works like a model aggregator: one panel, dozens of image and video models. The “Image” tab is for pictures, “Video” for clips; each has its own prompt and settings.

Below the prompt:

• model — a searchable list: Seedream, Nano Banana, GPT Image for images; Kling, Seedance, WAN, Veo and others for video;
• aspect ratio and quality — 16:9, 9:16, 1:1, 2K and so on, depending on the model;
• duration and sound — video only;
• the project styles toggle — adds the style set in the project settings to the prompt; the eye icon shows which one.

Click Generate — the result appears in the feed with a number. The cost shows in the button’s tooltip before you run it.

02Put project assets into the prompt#

What sets the canvas apart from a regular aggregator: it knows your project. Insert a character, location, prop or camera into the prompt as a mention — and the model gets their references.

With @ or a double space#

Type @ — or just press space twice — and a menu appears: “References”, “Characters”, “Locations”, “Props”, “Cameras”. Keep typing a name to narrow the list and pick the one you need.

In the prompt, a mention becomes a colored chip: @Viktor Rakhmanin, @Camp Director’s Office, @Dolly In. A character’s or location’s image is added to the references above the prompt automatically, and a camera inserts its movement description.

Add your own references#

Above the prompt is the references strip. Add reference opens a picker right in the panel — the “Characters”, “Locations”, “Props” tabs — or an upload of your own file; you can also just drag an image in. For video, references can also be video and audio files.

03Ask the agent to write the prompt#

On the Assistant tab the panel turns into a chat — the Project assistant of this canvas. It knows the project’s script and assets and writes the prompt with mentions itself, switches between image and video, picks the model and settings.

The agent puts the prompt into the generation panel — you can edit it by hand. It runs generation only when you ask and after you confirm. The canvas agent does not create characters, locations or shots — the section assistants do that.

04Save the result to the project#

Click a generation in the feed to open the viewer: the prompt with mentions, the model, parameters and references it was made with.

At the bottom:

• Copy prompt;
• Use as reference — the result goes to the references of the next generation;
• Regenerate — returns the prompt, model and settings to the panel;
• Save to project — turns the result into an asset or a shot.

An image — to an asset or a shot#

For an image, the “Save to project” window offers “Character”, “Location”, “Prop”, “Shot reference frame”, “Shot storyboard” and “Create new shot”. Pick an existing character or “Create new character”, then where to put it — the portrait or the reference sheet.

A video — to a shot#

A video is saved to a shot: pick an existing one or create a new one — scene, position in the scene and title. The canvas prompt moves into the shot prompt.

What’s next#

Saved generations wait in their shots on the Storyboard, and portraits and sheets in the “Characters”, “Locations” and “Props” sections.