model nodes
pick an image, video, audio or 3D model from the catalog, fill in its inputs, wire pictures and text into it, and press run.
A model node is one AI model on your canvas. You choose the model, the node shows the inputs that model takes, and you press a button to make something.
Every model comes from ixi's catalog, so each one has a price and a known set of inputs.
add a model
- Hover the + on the dock (or click it).
- Choose Models at the bottom of the menu. Its tooltip reads "Browse and add models".
- Click a model card.
A new model node lands on the canvas with that model already selected. You can also ask a pixie to add one: see skills and tools.
The browser has a search box (Search models...) and filter chips: All, Image Gen, Video Gen, Upscalers, Vectorizers, Background Removers and 3D Modelers. Featured models list first. Each card shows a cover picture or loop, the name and a short description.
To change a node's model later, click Change next to the model name, or right-click the node and choose Swap model. A node with no model yet shows Select Model.
fill in the inputs
The node builds its form from the model's own definition. A different model gives you a different form: a prompt box, a pick-one list, a number, a switch, or a slot for pictures.
- Inputs you leave alone keep the model's defaults.
- Type a prompt or number straight into the field.
- A picture slot takes a wire from an image node, or from another model's result.
Right-click an input and choose Mark "…" as Canvas Input. That is how you make a model's field a variable of a process.
wire things in
Every input that can take text or a picture has a handle on the left edge of its row. Drag a wire from a text node, an image node, or another model node into it.
- A wired input shows Mapped from node and the typed value is ignored.
- A picture slot that accepts several pictures shows them as thumbnails. Drag the thumbnails to reorder them; the order is the order the model receives them in.
- A group or an asset node can feed all its pictures into one slot.
run it
Press Let's go!. The button pauses for under a second so a double click doesn't run twice.
A row appears in the node's Runs table with a spinner while the model works. You can close the tab: the result still lands. When it finishes, the table row fills with the result. More on that table in generations.
If a pixie sets the node up and wants you to press the button, a yellow RUN tab points at it and clears the moment you press Let's go!.
A node wired into another model, a sandbox or a process starts that next node as soon as this one's result arrives. Chain two models and the second runs on its own.
what it costs
Each run spends credits from your organisation, and the price depends on the model and what you asked for, such as image size or video length. The charge is taken when the run succeeds. A run that fails costs nothing.
You don't see a price on the node. You see it afterward, run by run, in your organisation's Usage tab: model, project, source (who or what started it), duration and Cost. The credit balance and top-ups are in Billing. See billing and credits.
When your organisation is out of credits, the run is refused and the node shows the reason in red under the button.
limits & good to know
- 500 runs per person per 24 hours by default. At the limit the run is refused with "daily run limit reached". The count is a rolling 24 hours, and runs that failed don't count.
- A model with no price in the catalog can't run. The browser only lists models that can.
related
- generations: the runs table, pinning and putting a result on the canvas.
- nodes and wires: how connections work on the canvas.
- skills and tools: a pixie that picks, runs and judges models for you.
checked against the product · 2026-10-23