Image Layered: an AI model that splits a picture into layers

The NeuralBox catalog has a model with a unique ability: Image Layered (wavespeed qwen-image-layered) takes a finished image and splits it into 2–8 RGBA layers — like layers in Photoshop: objects separately, background separately. Each layer is a PNG with transparency, ready to use in any editor.

It's the only model in the catalog that does the reverse of generation: instead of building a picture from a prompt, it takes an existing one apart into editable pieces. No prompt needed — just upload the image and choose the number of layers. Prices start at $0.093.

What the model does

The input is any raster image: a photo, a generated picture, a screenshot of a layout. The output is a set of PNG layers with transparency (RGBA): the model decides what in the image is a separate object and separates it from the background and neighboring elements.

You set the number of layers — from 2 (object + background) to 8 (a detailed breakdown of a complex scene). The more layers, the finer the split and the higher the price.

Why it's useful

Before, tasks like these meant cutting out objects by hand or settling for a simple background remover, which gives you one object and nothing more. Image Layered hands you the whole scene in pieces.

Pricing

The price depends on the number of layers the image is split into. There is no subscription: you top up your balance by card (via Stripe) or with crypto. Prices are on the Basic plan ($5 = 150,000 tokens); on higher plans tokens are up to 40% cheaper.

How to try it

Sign in with Telegram, Google or email, top up your balance (from $5 by card, from $2 in crypto) and open the Images tab → Image Layered. Upload an image and choose the number of layers — no prompt needed. A minute later you can download the finished transparent PNG layers.

Tip: start with 2–3 layers for $0.093–0.14 — for most object-plus-background tasks that's enough. Use 6–8 layers for busy scenes with several objects you need separately.