AI Workflow Canvas

The node canvas where campaign shots
become repeatable graphs.

Chain a generate-image node into a generate-video node. Wire any output into the next node’s references. Branch one source into several looks, then save the graph that worked and run it on the next campaign.

Start free with 20 VX Browse the Workflow Marketplace

12video engines16image engines8.5 VXstandard 5s shot20 VXfree on signup

Every stage of a shot is a node

NODE PALETTE

The palette covers the whole pipeline, so a shot never has to leave the graph to get finished.

Input nodes

Bring in an image, a video, or an audio track. Cast references work in node inputs, so the brand identity is locked from the first node in the graph.

Generate nodes

Generate-image, generate-video, and generate-audio or TTS. Each one routes to any engine on the roster and is metered in VX like any composer run.

Extract nodes

Pull pose, depth, or canny edge from a source. Extraction turns a reference into a control signal the next generation can follow.

Faceswap and enhance

Swap a face onto a take that got everything else right, or run upscale and enhance before the render leaves the graph.

Clip and condition

Clip nodes cut a render down. Condition and anchor nodes set the constraints a generation must respect, like the frame a shot starts from.

Output nodes

Every branch ends in an output node that collects the finished render. What comes out is a normal VisionX generation, ready for review.

Still to shot, one wire at a time

THE GRAPH

The canvas move that matters most: approve a frame once, then let everything downstream inherit it.

A composer run gives you one prompt and one result. A graph gives you the shot as a system: sources on the left, decisions in the middle, renders on the right.

Included:Chain a generate-image node into a generate-video node.
Included:Wire any node output into the next node’s references.
Included:Branch one source into several looks.
Included:Cast references hold the same identity across every branch.

Lay down the source

Drop an input node for the image, video, or audio the shot starts from, and add a Cast reference so the identity is pinned before anything renders.

How Cast holds identity

Generate the key frame

A generate-image node renders the still that anchors the shot. Route it to whichever image engine fits the look.

See AI image generation

Chain into video

Wire the image output into a generate-video node as a reference. The motion pass starts from the exact frame you approved, not a fresh interpretation of the prompt.

See AI video generation

Branch and finish

Split the same source into several looks and keep the take that lands. Extract pose or depth where you need control, then faceswap, upscale, or clip before the output node.

Save the graph, not the renders

REUSE

The second campaign should not start from a blank canvas. Templates carry the structure forward.

Canvas templates

Save any canvas as a reusable template. A template stores structure only: the nodes, the wiring, the shape of the pipeline. Load it on the next campaign, point it at new inputs, and the graph you already trust does the rest.

The Workflow Marketplace

Import workflow templates from the public Workflow Marketplace and run them on your own sources. Proven graphs are a starting point, not a screenshot.

Browse the Workflow Marketplace

One meter, every engine

VX METERING

A graph is not a separate product with separate pricing. It is the same studio, arranged as nodes.

Metered like the composer

Every generating node bills VX exactly like a composer run. A 5-second Standard shot at 720p is 8.5 VX whether it starts from a prompt box or a node.

Any engine per node

Each generate node routes to any engine on the roster: 12 video and 16 image. Send a draft branch to a light tier and the final branch to Cinema.

Cast in the graph

Cast references work in node inputs, so the identity that holds in the composer holds across every branch. Engines change per node. The face does not.

FAQ

The workflow canvas, answered

What is an AI workflow canvas?

A node graph editor for generation pipelines. Instead of one prompt and one output, you build the shot as a graph: input nodes for images, video, and audio, generate nodes for image, video, and audio or TTS, extract nodes for pose, depth, and canny edge, plus faceswap, upscale, clip, condition, and output nodes.

Can I chain image generation into video generation?

Yes. Generate a still, wire the image output into a generate-video node as a reference, and the shot starts from the frame you approved. Branch the same source into several looks to compare takes.

How is a canvas run billed?

Every generating node is metered in VX like any composer run. A 5-second Standard shot at 720p costs 8.5 VX whether it starts from the composer or from a generate-video node. You start with 20 VX free on signup.

Can I reuse a graph on the next campaign?

Yes. Save it as a canvas template. Templates store the structure, the nodes and the wiring, not the renders. You can also import workflow templates from the Workflow Marketplace and run them on your own inputs.

Which engines can a canvas node use?

Any engine on the roster: 12 video engines and 16 image engines. Cast references work in node inputs, so the identity holds no matter which engine a node routes to.

Is the canvas a ComfyUI alternative for video work?

It covers the same node-graph territory for campaign work: chain generation steps, wire references between nodes, and branch one source into several looks. It runs inside VisionX with the engine roster and VX metering built in, so there is no local pipeline to stand up.

The rest of the platform

INDEX

Build it once. Run it again.

20 VX free on signup, no card required.

Open Studio Compare plans