Workflow Automation Basics: Chaining Steps on the Canvas
Stop clicking the same buttons 40 times. Learn how the VAR2 workflow canvas chains generation steps into nodes that run once — or a hundred times — on their own.
Doing pro-level AI work by hand looks like this: generate an image, download it, upload it into the background remover, download that, feed it to the upscaler, rename the file, repeat. It works for one asset. For forty, it is a full workday of clicking — and every manual step is a chance to grab the wrong file. Workflow automation replaces that whole ritual with a diagram you build once and run forever.
What Is the Workflow Canvas?
The VAR2 workflow canvas is a visual board where each box — a node — is one operation: generate an image, animate it, remove a background, upscale, add music. You draw a line from one node to the next, and the output of the first becomes the input of the second. The canvas is a recipe: *make this, then do that to it, then finish it like so.* Hit run once and VAR2 executes the whole chain start to finish, no babysitting.
- Nodes — each one is a single generation or edit step (create-image, animate, upscale, remove background, add audio)
- Connections — arrows that pass one node's output into the next node's input
- Inputs — a prompt, a reference image, or the result flowing in from an upstream node
- A run — one full execution of the chain; the canvas can also fan a run out across many inputs at once
A Real Three-Node Chain
Say you sell products online and need clean packshots. A minimal workflow is three nodes: Generate the product on a plain studio background, Remove Background so it drops onto any layout, then Upscale to a crisp 4K asset ready for print. You build those three boxes, wire them in order, and now every future product is one run away from a finished, print-ready cutout. The skills from Background Removal & Compositing and Upscaling & Enhancement become a single automated move.
| Node | Job | Feeds into |
|---|---|---|
| Create Image | Generate product on white seamless | Remove Background |
| Remove Background | Cut out to transparent PNG | Upscale |
| Upscale | Push to 4K, sharpen edges | Final asset |
The prompt inside the first (Create Image) node
a matte-black wireless earbuds case, centered product shot, clean white seamless studio background, soft even lighting, sharp focus, no props
The white seamless background is deliberate — it makes the next node's background removal near-perfect. Design each node's prompt to feed the node after it, not just to look good on its own.
Where Automation Really Pays Off: Batch Runs
A single run saves clicks. A batch run changes your business. Instead of one prompt, you feed the chain a *list* — ten product names, twenty scene descriptions — and the entire workflow executes for every item, in parallel, while you get coffee. This is the moment AI stops being a toy and starts being a factory. Because every generation costs tokens, a batch of forty is forty times the spend of one — so test the chain on a single input first, confirm it is perfect, *then* unleash the batch. We go deep on scaling this in Batch Generation at Scale.
Your first automated chain is three boxes and two arrows away. Go build it. Open the workflow canvas