Layerize Image
Break a flat image into its objects, each as its own transparent layer, plus a background plate with the holes filled.
Layerize Image
Badge: New
Node type: process:layerize_image
Category: process
Description
Break a flat image into its objects, each as its own transparent layer, plus a background plate with the holes filled.
Pricing
- Cost: ~16 credits per images
Canvas ports
These appear as port handles on the left side of the node.
| ID | Label | Details |
|---|---|---|
image |
Image | IMAGE (required) |
Sidebar config
These render as form fields in the right-side config panel when the node is selected.
| ID | Label | Details |
|---|---|---|
maxLayers |
Max layers | NUMBER |
labelLayers |
Label layers | BOOLEAN |
Outputs
| ID | Label | Type |
|---|---|---|
layers |
Layers (JSON) | JSON |
plate |
Background plate | IMAGE |
preview |
Layers preview | IMAGE |
stage |
Source size | JSON |
Wire layers into a Compositor
This node's whole point is the Layers (JSON) output. Connect it to a
Compositor's Layers (JSON) input and every object becomes a separate,
draggable layer in the compositor editor. Connect Background plate to the
Compositor's Background input in the same move, or the objects float on
nothing.
Image -> Layerize Image --Layers (JSON)--> Compositor
--Background plate-->
Leave the Compositor's own layer list empty. The layers arrive over the wire, so the canvas shows where they came from, and re-running Layerize refreshes them in place.
Objects come back whole
Layers are semantic, not just visually separated. An open laptop comes back as one laptop rather than a screen and a body. More usefully, an object that was partly hidden is reconstructed: a notebook with a plant sitting on it comes out as a complete notebook, so you can move either one and nothing has a bite taken out of it.
The background plate is generated the same way, with every object removed, so sliding a layer aside reveals a clean backdrop instead of a hole.
Layers are named
A short naming pass labels each layer, so the panel reads "laptop", "coffee mug", "spiral notebook" instead of numbers. Turn it off with Label layers to shave a few seconds off the run. It does not change the price: the node costs the same flat amount either way. Naming never fails a run: if it does not come back cleanly the layers ship numbered.
What the node itself shows
The node's own thumbnail is a grid of the objects it cut, so you can tell at a glance what it separated without opening the compositor. The Background plate output is a different image: the backdrop with everything removed. That one is nearly empty by design, which is why it is not what the card displays.
Your edits survive a re-run
Drag a layer in the compositor and that position is yours: a later Layerize run updates the layer's image but keeps where you put it. Scale, rotation, opacity, visibility and lock behave the same. Only layers you have never touched adopt the new position, so you can re-run after changing the source without losing your composition.
Max layers is a ceiling, not a target
Asking for more layers does not find more objects, it splits the ones already found and starts returning the same object twice. Near-duplicates are merged automatically and the node tells you when it merged any, but the setting is best left near its default. Raising it costs nothing extra; the price is flat per image no matter how many layers come out.
Full resolution
Layers are cut from your original image, not from the decomposition model's smaller working copy, so a 2048px source gives 2048px layers. Images larger than 2048px on the long edge are worked at 2048 to bound memory, and the node says so.
Auto-generated from the Wireflow node registry.