Prompt
Text
Executed
8s cafe hero — warm wood, slow pour-in.
Hand-build it. Describe it. Or call it.
The same media job — on a canvas, from a sentence, or from your app.
Place the models. Draw the wires.
Drop image, video, voice, and utilities onto a graph. Connect them in the order the job actually needs. Run once. The outputs stay attached to the nodes that made them.
Outputs never leave the graph until you want them to.
A still is not a download you re-import. It is already the next node’s input.
The frame is a starting point.
Generate with Flux, lock a reference, or upscale. That still feeds video and voice on the same canvas — no round-trip through another app.
Motion, voice, and the mix stay one job.
Animate the still, lay speech or score, combine the cut. The combiner is a node. The output is still on the graph.
Edit like a graph. Not a render queue.
Generative media is expensive when you redo work you already liked. The canvas is built so you don’t.
Re-run the node that changed.
A new prompt on one still does not rebuild the film. Only the path below that node updates.
Batch the variables. Keep the wiring.
Swap products, hooks, or voices across the same graph. Hundreds of versions. One pipeline.
Combine without exporting.
Stitch clips and lay voice on the canvas. You leave for a timeline editor only when you need frame-level finishing.
A working graph is an asset.
Once the pipeline is right, you should not rebuild it for the next brief, teammate, or release.
Save the wiring.
Keep the nodes and connections. Next time, change the brief or the assets and run again.
Share the same steps.
The team generates from one graph instead of a folder of tool links.
Publish the endpoint.
The visual workflow is the API. Keys and credits cover the runs.
Studio first. API when it repeats.
Nothing to install. Build in the browser. When the job is a system, trigger the same graph from code.
Work in the browser.
Open a canvas, drop models, run the job. No local setup.
Open studioSame graph, from your app.
Publish once. Call it with a key. Batch from your product if you need volume.
POST /workflows/{id}/run
Authorization: Bearer <api_key>
{
"inputs": { "prompt": "8 second product hero with voiceover" }
}Get started
Put the job on a canvas.
Wire the first pipeline yourself, or let Architect lay it out from a sentence.