Skip to main content

Building workflows

Every workflow is a graph: input nodes bring in your media and text, generator nodes transform it with models, and Output nodes mark what the run should deliver. This guide covers the canvas itself — for a full list of node types, see the Node reference.

Add nodes​

Create a workflow from the Workflows page (New workflow) and the canvas opens with a node palette. Click a palette entry to drop that node onto the canvas, then drag it anywhere. Nodes are roomy on purpose — give the graph space and it stays readable at a glance.

Select a node to configure it: type a Prompt node's text, pick a Generate node's model and aspect ratio, or attach media to an Image or Video input.

Wire the ports​

Drag from a node's output port to another node's input port to connect them. Ports are typed — text, image, video, or audio — and only matching types connect: a Prompt's text output fits a Generate Image node's Prompt input, an Image input's output fits its References input. The canvas won't let you wire an image into a text port, so a saved graph is always structurally sound.

Some inputs accept multiple connections (a Generate Image node can take several reference images; Combine Text merges any number of texts), and optional ports can be left unwired.

A worked example​

Here's the classic product pipeline — photo in, video out:

app.corte.so/workflows interactive
⌁ Prompt
“Alpine trail-running capsule, cold dawn light”
▣ Image input
✦ Generate image
Nano Banana · 4:5
⌕ Generate video
Kling 3.0 · 5s
◎ Output
Waiting…
Product photo + prompt → Generate Image → Generate Video → Output.
  1. Image input — attach your product photo.
  2. Prompt — describe the scene: "Alpine trail-running capsule, cold dawn light."
  3. Generate Image — wire the prompt to Prompt and the photo to References. The model restages your real product in the described scene.
  4. Generate Video — wire the prompt in again and the generated image to Frames to use it as the first frame. Pick a video model and duration.
  5. Output — wire the video in. Outputs mark the run's deliverables, so results are easy to find later.

Run it, swap the photo next season, run it again.

Saving​

The top bar shows the workflow's name (click to rename) and its save state — Unsaved changes, Saving…, Saved. Press Save after editing; running always uses the saved graph. Building and saving are free.

What costs credits​

Only generator nodes spend credits, and only when a run actually executes them — an unwired branch costs nothing. Each Generate Image, Generate Video, Generate Text, Generate Audio, Edit Video, and Upscale Video node bills like its Studio equivalent. Model choice drives price; see Choosing a model.

Keep going​