Layer Composer
Add Node → Compose → Layer Composer
Wire in up to ten pictures, arrange them by hand on one canvas, and the node publishes the arrangement as a single JPG. Move things, scale them, rotate them, set their opacity and blend mode, draw and write on top.
It’s the one node in the Generate/Compose reference that calls nothing at all: no provider, no model, no cost. The picture is rendered on your own machine the moment the arrangement changes, so there’s no strip of results and no Generate button — there is exactly one output and it is always current.
Downstream, it’s an ordinary image. Wire it into an Image node as a reference, into a Video node as a first frame, into an Upscale — anything that eats a picture takes this with nothing to explain.
What’s on the node
| Area | What goes in it |
|---|---|
| The frame | The composition as it stands. Double-click it to open the editor. |
| Aspect ratio | The canvas’s shape: 16:9, 4:3, 3:2, 1:1, 2:3, 3:4 or 9:16. It sits on the card because it decides what the node outputs. |
| Resolution | 2K or 4K, shown here and set in the editor. New compositions start at 4K. |
| Edit | Opens the compositor full-screen. |
| 123 | The layers, up to ten. Ring 1 is the back of the stack, the last one is the front. Fill one and a fresh empty ring appears below. |
There is no “add layer” button, because a layer is an input ring — layers appear by wiring a picture in. And because the rings are the stack, dragging a ring to another place in the column re-stacks the composition: the layer travels with its picture, keeping its frame, rotation and blend wherever it lands.
A picture arriving on an empty canvas starts full-bleed, covering the whole frame. There’s more on handles in How nodes connect.
The editor
Edit — or a double-click on the frame — opens the compositor as a full-screen window: the canvas in the middle, the layers list down one side, the properties of the selected layer down the other. There’s no draft and no Apply. It edits the node directly, the node re-renders as you go, and closing the window is just closing a window.
Drag a layer on the canvas to move it, grab a corner to scale it, and use the properties panel for anything you want to be exact about:
| Control | What it does |
|---|---|
| Position and Dimensions | X, Y, W, H in canvas pixels, with a link toggle to keep width and height in proportion. |
| Alignment | Snap the layer to the canvas’s left, centre, right, top, middle or bottom. |
| Rotation | In degrees, turning about the layer’s own centre — plus flip horizontally, flip vertically, and a reset back to centred, upright and unflipped. |
| Opacity | 0 to 100%. |
| Blend mode | The sixteen usual ones — multiply, screen, overlay, soft light, difference and the rest. |
| Scale | How the picture sits inside its frame: fill covers it, fit letterboxes inside it, stretch distorts to match. |
The canvas itself has its own three settings at the top of the panel: aspect ratio, resolution (2K or 4K, naming the long edge whichever way the canvas is turned) and background colour behind every layer.
Switching between 2K and 4K scales the whole canvas, so nothing in your arrangement moves — it just renders sharper. Do it whenever you like.
Drawing and writing on it
The compositor carries the same annotation tools as the image editor, under Add Text, Paint or Forms: text, freehand paint with an eraser, and rectangles, circles and lines — each on its own layer above the pictures. They’re stored on the node and rendered into the output like everything else, so a caption or a callout you draw here comes out in the file.
3D layers
A 3D object wired into a Layer Composer arrives as a layer you can still turn. Select it and the layer’s still is replaced in place by the live object, with the Texture toggle, the colour swatches and Reset on the selection box — so you aim it where it stands, at the size and rotation it already has in the composition.
The layer’s view is its own. It’s seeded from however the source is facing the moment the wire lands, so the composition opens showing exactly what you plugged in; from then on the two are independent. Turning the layer doesn’t move the source, and turning the source doesn’t move the layer — which is what lets two compositions show the same object from two different angles.
What it’s for
Three jobs it does better than anything else here:
- Building a reference by hand. Put the product on the left, the character on the right, the background behind both, and wire the result into an Image node as a single reference — instead of hoping a model reads three separate images the way you meant.
- Framing a first frame. Compose the exact opening shot, then send it to a Video node.
- Finishing. Titles, watermarks, a caption, a logo in the corner — the last thing a picture passes through before it’s exported.
Because it’s always current and costs nothing to re-render, it’s also the cheapest thing on the canvas to fiddle with.