Guide
English
Search the docs⌘K

Release notes

This page covers everything in this DOTV update. Every entry comes with a demo screen — follow the "See" link to the guide section that explains it.

Connect your AI assistant to the DOTV canvas

The new connection page walks Codex, Claude Code, Cursor, and other assistants through linking to DOTV. Their actions and results stay in the Agent timeline, and credit-consuming steps still ask you first.

Two ways to installInstall via an AI agent: hand that one prompt to your AI assistant and it installs the thing itself. This is the default tab.
1 / 2
Connect your AI assistant to the DOTV canvas

Agent understands your canvas better — and stays under your control

The composer accepts images and videos and can @-mention nodes or models. Save reusable prompts in your own library, and each conversation keeps its own running progress and results.

Agent understands your canvas better — and stays under your control

The previz stage can now block real scenes

The previz stage now covers character actions, hand props, keyframe performance, follow shots, and segmented camera moves. One sentence can block a scene, and the preview can return to the canvas as a reference.

Animate from one sentence"She walks to the window, camera pushes in with her" — one sentence becomes a timeline: motion paths, poses and camera follow, ready to fine-tune track by track.
1 / 3
The previz stage can now block real scenes

Run a scene check before you shoot

Before export, DOTV checks seated poses, props against walls, cameras that are too close, and other easy-to-miss issues, then summarizes each shot so you can fix it first.

3D sceneScale / move / rotate the whole scene. People and cameras move together, so you never have to adjust them one by one just to change how big a person reads against a camera.
1 / 4
Run a scene check before you shoot

Point at “exactly here” on the image

Circle, box, or annotate the exact area on a generated image. Later generations can reference it instead of describing the location again.

The image node toolbarIt only shows when you select an image node that has already generated. Everything in this chapter starts from this bar.
Point at “exactly here” on the image

The compose desk understands speech and subtitles better

Subtitles can be generated from speech, splits and protected ranges are more accurate, and DOTV checks for missing assets before export.

Selecting a clipClick once to select it: a white outline, with white handles at both ends. Click empty track space to deselect.
1 / 3
The compose desk understands speech and subtitles better

More models, safer switching

Stronger text-model tiers are available. Switching keeps valid settings and warns early when a model does not support image or audio references.

Pick the model hereEach kind of node has its own list of models, and this is where you switch. Change the model and the resolution and duration options on the right change with it — vendors support different ones.
1 / 2
More models, safer switching

Smoother large canvases, more reliable jobs

Large canvases zoom and switch more smoothly, card text is selectable, new nodes avoid overlaps, and local jobs do not rerun after a refresh.

The cluster in the bottom leftTidy up, Minimap, Snap to grid, and the current zoom. They only change how you look at this canvas, never what's on it.
Smoother large canvases, more reliable jobs
Try it in DOTVReading only goes so far: open DOTV and make an AI video with the feature this section covers.
Open DOTVDownload the desktop app
Next
1.1 Home and all projects
On this page
Connect your AI assistant to the DOTV canvas
Agent understands your canvas better — and stays under your control
The previz stage can now block real scenes
Run a scene check before you shoot
Point at “exactly here” on the image
The compose desk understands speech and subtitles better
More models, safer switching
Smoother large canvases, more reliable jobs