Everything between a sentence and a finished film
Most AI video tools give you a text box. CanvasVid gives you a workspace where the story, the shots and the characters stay connected from the first idea to the final export.
The canvas is the product
Every node is either an asset, a character or a generation step. Wire them together and the storyboard becomes executable.
Infinite canvas
Drag assets, characters and generation nodes anywhere. Connections define what feeds what, and the whole project autosaves as versions you can roll back.
One line to a film
Describe the piece once. The agent drafts the script, splits it into shots, generates each one and stitches the result, and you can intervene at any step.
Live task progress
Queue position, rendering state and completion stream over a persistent connection. Close the tab if you like; the render keeps going.
Character consistency that survives the edit
The hardest problem in AI video is the person who changes face between shots. We treat a character as four locked dimensions instead of a prompt.
Face
A reference portrait is generated once and injected every time the character appears, so facial structure stays stable.
Body
Height, build and posture are described once and carried through the shot list.
Wardrobe
Costume details are pinned so a jacket does not silently change colour between scenes.
Voice
Clone a voice, assign it to a character, and every line is dubbed in that voice across the whole project.
Two rendering paths
Different shots need different levels of control. Both run on the same cloud inference cluster and share one credit balance.
Text-to-video
Describe the shot and get a cinematic clip. Best for establishing shots, b-roll and anything where you want the model to improvise the frame.
Reference-to-video
Feed in a character reference and drive the motion from it. Best for dialogue, continuity and any shot where identity matters more than novelty.
Resolution and framing
720P or 1080P, in 16:9, 9:16 or 1:1. The same storyboard exports to whichever platform you are publishing on.
Built for series, not one-offs
Long-form work breaks most tools. Chapters, continuity and credits all have to hold over dozens of shots.
Chapter workflow
Split a long piece into chapters, generate each, then stitch the whole thing into one film.
Shot chaining
Carry the last frame of one shot into the first frame of the next for smoother continuity.
Autopilot
Let the agent advance chapter by chapter while you review, or step in and drive manually.
Per-shot retry
Only the shot that failed gets regenerated, or rerun every failed shot with a revised prompt.
Predictable cost
Generation is metered in credits, estimated before you commit and settled after the render.
Quoted up front
The estimate appears before submission, with the balance you hold shown alongside it.
Held, then settled
Credits are reserved when a job starts and reconciled against real usage when it ends.
Failures refunded
A render that fails returns its hold automatically. No support ticket needed.
Try it with starter credits
Create an account and the canvas is yours. No card required to start.
Start free