Video starts on the canvas

Ask someone what a video production tool looks like, and they’ll picture a timeline: clips laid end to end, a playhead scrubbing left to right. But a lot of the work that actually decides what goes into the video never touches a timeline. It happens in a notepad, a shot list, a folder of screenshots with arrows drawn on them, a brief that gets passed around until someone loses it.

More and more, that work is moving onto a canvas. A shared, spatial, annotatable surface where you can put a sketch next to a script, next to a generated clip, next to a colleague’s comment.

Here are six examples of places it’s already happening throughout the video production process.

Planning: the storyboard comes first

Previz, mapping out the scenes, angles, and shot locations, is step zero of physical production. Up until now, it’s lived in a slide deck, on paper, or in storyboard software that has nothing to do with wherever the video eventually gets made.

AI video generation looked different: instead of scouting a physical location, you collected reference images. In some cases, you skipped it altogether and went straight from a text box to a generated clip. That meant iterations also happened in a chat; you had to explain every change in as much detail as possible. If the background generated for clip 38 didn’t match up to the first 37, you had to explain how.

A canvas's spatial layout makes it a better place to decide what you’re making than a text prompt field. You can see every shot at once, compare them, reorder them, and notice that scene four doesn’t match scene two before anything generates.

Lightricks’ LTX Studio puts that into practice. It starts a project from a script or a one-line idea and builds a board before it generates any video. According to its website, it extracts characters, objects, and locations from the script as reusable “Elements” so they stay consistent across shots, then breaks the script into scenes and shots you can review before generating anything.

The thing you show a client is no longer a deck you built separately from the video. It’s the board itself: approved shot by shot, with the characters and locations locked in, before anyone presses generate.

Sketching: drawing is faster than describing

Some of the best ideas start life as a scribble on the back of an envelope. A rough composition, a camera angle, a shot that’s much faster to draw than to explain in a sentence. A canvas already has the tools for that (pens, shapes, arrows), so it can treat the drawing as an input rather than as decoration around the real work.

In September 2024, Runway was reported to be testing two Storyboard prototypes where you draw on a canvas, convert the sketch to an image, and animate that image with a video model, sketch-to-image-to-motion, without leaving the surface. While these were experimental, they show how canvases keep showing up as the base layer on which video production is built.

The node pipeline: every generation step becomes visible

The first wave of AI video tools was a prompt box and a black box behind it. Think Sora and shy kids’ short Air Head (2024), which reportedly took hundreds of prompts and serious VFX work to hold together.

A newer category, which Alignify calls “AI video canvases”, lays each step out as a node on a canvas. A prompt, a reference image, a model, an upscaler, and an audio track are all connected by lines. Alignify believes this offers better step-by-step visibility, plus the freedom to mix models in one workflow, branch and compare outputs, and to build and export reusable workflows. The canvas does the same job for a generation pipeline that a storyboard does for a shoot: it makes the structure visible so people can argue about it.

The field is already crowded. Krea Nodes pairs the graph with an agent. You describe the outcome in plain language; it builds the pipeline, and then you can see and rewire what it made. Flora starts from a storyboard and branches from there. Mosaic exports its work as XML for Premiere Pro and DaVinci Resolve, so the canvas can hand off to a traditional editor instead of trying to replace one. Figma has its own entry, Weave.

Motion: the canvas gets a timeline

So far, we’ve covered how canvas-based workflows are changing previz and production. But they’re also becoming part of editing and post-production. For that to happen, the canvas needs to be able to handle time.

Traditionally, animation has been a hand-off point. A designer makes the static frames, hands them to someone with motion tools, and the two files drift apart. Figma Motion puts a timeline next to the design on the same canvas: switch a frame to Motion mode, and you can keyframe position, scale, rotation, and opacity, attach comments to a specific moment, and export to MP4, GIF, SVG, or WebM.

This is design-system motion (illustrations, banners, UI), not a film edit, and at launch Motion was in open beta. But it shows the direction of travel: the timeline is moving to the canvas, rather than the canvas to the timeline.

Annotation: drawing directly on the footage

Most feedback on video starts with someone recording their screen or exporting a cut, then trying to describe in words what’s wrong at frame 14. A canvas already has the drawing tools for this. Pointing them at a video instead of a still image removes the translation step.

The bug-reporting tool Jam already had screenshot annotation built on tldraw. They then extended it to screen recordings so you can draw directly over a video, not just a still. Instead of a recording plus three paragraphs of description, a reviewer circles the exact thing that’s wrong, on the exact frame where it’s wrong, and sends a link.

Read Jam’s case study

The whole workspace: one canvas, brief to creator

A video ad campaign has more hands on it than the credits suggest: a brand team, an agency, creators, each with their own references, feedback, and version of the brief, scattered across email threads and PDFs. What if all of that took place on a canvas, with the plan, references, generated concepts, feedback, and budget all sharing one surface?

Vidsy makes video ads for enterprise brands through a global network of creators, and it has built its creative workflow on the tldraw canvas. The team runs client strategy sessions live on a shared board instead of writing a brief, emailing it and booking a meeting to pick it apart. Each client project gets its own canvas, and before a brief goes to a real creator, the team can generate AI concept videos from the chosen strategy narrative to see whether the idea holds up.

The strategy session replaces the document. The client sits in the room (or on a call) while the idea goes from brief to storyboard to rough concept video, gets annotated by the people who’ll make it, and arrives at the creator with the context attached instead of retyped into a form.

And cut.

Video is a visual medium. It’s a sprawling, collaborative process. Expecting it to fit into a series of documents or a chatbot causes friction. Canvas-based workflows enable better visibility, collaboration, and space for creative feedback.

A canvas isn’t a redesign of the video tool. It fills in the step that was missing before the video tool even starts, and after it ends. The generation models will keep getting better. What’s still scarce is someone who knows what good looks like and a place for them to say so.

We’d love your feedback. Get in touch with us on our Discord or follow us on X. If you'd like to put a canvas like this in your own product, check out our docs or starter kits.