Splitting an image into layers

LUVI can split an image into layers: you get a base image, and each element in it becomes a separate, transparent layer.

Search the model list for "layer" and you'll find Seedream's Pro and Flash layer decomposition models. We're using Flash, the cheaper of the two.

These models take a single image, which goes in the slot in the Reference section. A prompt is optional.

Leave the prompt empty and the model separates every main element; if you only want some of them, name them in the prompt.

The image must be between five hundred and twelve by five hundred and twelve and six thousand by six thousand pixels in total, no longer than sixteen to one, and no bigger than thirty megabytes. If it isn't, it's turned away before the model runs, and nothing is charged.

The estimate says "up to six hundred and twelve credits", since no one knows in advance how many layers the model will find. The exact price is set at the end, by the number of files it returns, and never goes over that.

Press Generate. The result comes back as a base image plus up to sixteen transparent layers.

The first time a result like this opens, the Layers panel opens too, and the Layers button on the right-hand toolbar shows or hides it. The list starts with the top layer and ends with the base image.

The eye icon hides a layer or shows it again. Drag a layer on the canvas to move it; while it's selected, the arrow keys nudge it one pixel, or ten with Shift, and Esc deselects it.

These moves stay on screen; the files themselves never change. Reset layout puts every layer back where it started.

The download button on each row saves that layer on its own, while Download visible saves what's on screen as a single PNG at the base image's resolution.

In short: add the image, press Generate, then hide, move and download the layers you need.