What is CatClaw Layers? Turn flat images into editable layers

A flat image is an endpoint. A layered image is raw material, and that difference is where the real creative work begins.

Published 2026-08-14 by Cat Claw, the UK's create-and-distribute AI studio. About a 8 minute read.

The short version

• CatClaw Layers splits any flat image into up to 16 transparent PNG layers, each with its own name, stacking order and description, on an interactive canvas.
• Every region hidden behind an extracted subject is AI inpainted, so each layer and the base beneath it is a complete, standalone asset.
• No source image? Prompt-to-layers mode generates a brand new image already separated into layers.
• Edit any single layer from a prompt with up to 4 reference images, or press a layer into the surface beneath it for instant mockups.
• Rearranging is free, the flattened PNG export is free, and credits reserved for outputs that never arrive are refunded automatically.

CatClaw Layers is an AI layer separation tool from Cat Claw that turns any flat image into an editable stack of up to 16 transparent PNG layers, with everything hidden behind each subject already inpainted. It launched in August 2026 and lives in the Tools section of the studio: sign in, open Tools, then CatClaw Layers.

This guide covers what the tool actually does, how AI layer separation works underneath, the full workflow from upload to export, what people are making with it, how it compares with Higgsfield Layers, and how the pricing settles.

What is CatClaw Layers?

Quick answer. CatClaw Layers is an AI layer decomposition tool from Cat Claw, the UK's create-and-distribute AI studio. Give it a single flat image and it returns a full, editable layer stack: up to 16 transparent PNG layers, each with its own stacking order, bounding box, auto-generated name and a short semantic description. The base under every extracted subject is AI inpainted, so you get a complete, usable asset at every level of the stack.

There are 2 ways to feed it. You can upload an existing image, optionally add a guidance prompt to steer how the split happens, and let the tool do the work. Or, if you have no image at all, you can type a text prompt and get a brand new image generated pre-separated into layers from the start. No source image required. One input box, 2 modes, entirely different creative starting points.

The output lands on an interactive canvas where you can click, drag, resize, hide and isolate every layer without spending a single credit. Your arrangement auto-saves and restores when you reopen the stack, and your history keeps your last 20 runs.

How does AI layer separation actually work?

AI layer separation works by detecting the subjects in an image, cutting a clean transparent PNG for each one, inpainting the pixels that were hidden behind each subject, and then naming and ordering the full stack. In CatClaw Layers the whole pipeline runs in about 1 to 2 minutes.

The detection step identifies distinct elements, whether that's a person, a product, a background object or a piece of text. Each one gets cut with a clean edge into its own transparent PNG. That part is familiar territory for anyone who has used an AI cutout tool before.

What makes CatClaw Layers different is what happens next. When a subject is lifted out of a photo, there is always something hiding behind it. Most tools leave a hole, or fill it with a guess that looks wrong. CatClaw Layers uses AI inpainting, powered by Seedream 5.0 Pro, to reconstruct the hidden region properly, so the layer beneath reads as a complete image on its own. That inpainting step is the moment most competitors quietly skip past.

Every layer is named automatically, with a label and a short semantic description. When you export the stack and drop it into Photoshop, Figma or After Effects, you're not staring at Layer 1, Layer 2 and Layer 3. You know exactly what you're looking at.

Turning a flat image into editable content, step by step

Here's the full process, from upload to export.

1. Upload an image or type a prompt. Sign in at catclaw.ai, open Tools, then CatClaw Layers. Upload a JPG or PNG, and add a short guidance prompt if you want to steer what gets separated. No source image? Skip the upload and just type what you want: the tool generates the image and separates it in the same run.
2. Wait 1 to 2 minutes. The AI detects subjects, cuts transparent PNGs, inpaints the hidden regions and builds the named, ordered stack. Pick an output resolution of 1K, 1.5K or 2K depending on what you need.
3. Arrange freely on the canvas. Your layers land on an interactive checkerboard canvas. Click to select, drag to move, pull a corner handle to resize, and hide or show any layer. Rearranging costs nothing, and the layout is saved with the stack for next time.
4. Edit one layer with AI. Select any single layer and restyle it from a prompt, with up to 4 reference images to guide the output. Useful for outfit swaps, texture changes and character variations. The rest of the image stays exactly as it is, and the edited layer drops back into the stack at the same size and position it left from.
5. Press a layer into a surface. Select a layer and press it into the surface beneath it with the Press Into function. The layer reads as part of the material at the exact spot you placed it. It's the cheapest AI action in the studio, and the fastest way to make product mockups, posters, packaging designs and clothing graphics.
6. Export what you need. Download a free full-resolution PNG flatten of your arrangement, export any single layer as a transparent PNG, or grab a ZIP of every layer plus a manifest of names, order, positions and visibility that drops straight into Photoshop, Figma or After Effects. The flattened PNG also lands in your Cat Claw gallery under the Layers filter.

What can you make with CatClaw Layers?

Layer separation sounds abstract until you see the jobs it quietly does. These are the workflows it was built for.

• E-commerce product mockups: place your product layer on a different background, press a graphic into the packaging surface, or generate a new product image pre-layered and ready for variant testing. No reshoots, no manual masking.
• Ad variants and thumbnails: split one hero image into layers, swap the background, change the subject's outfit, and export 5 creative versions in the time it takes to brief a designer on the first one.
• Character and outfit swaps: extract a character layer and use the per-layer AI edit with reference images to change their look, style or setting without touching anything else in the composition.
• Parallax and motion assets: video editors need clean, separated layers for depth animations and parallax effects. The ZIP export opens in After Effects with the layers already named and ordered.
• Meme surgery: replace one element, keep everything else. No explanation needed.

CatClaw Layers vs Higgsfield Layers

Both CatClaw Layers and Higgsfield Layers offer AI-driven image-to-layers separation, and if straight splitting is all you need, both are worth a look. The table reflects CatClaw Layers' verified capabilities and publicly available information about Higgsfield Layers at the time of writing. For anything Higgsfield-specific, check their current documentation.

If you only need to split an image into layers, both tools will serve you. If you need the stack to keep working after the split, through generation, editing, blending and clean export, that's the job CatClaw Layers is built for.

Pricing that settles honestly

CatClaw Layers charges per output image actually returned, not per attempt. When you start a split, the tool reserves an estimated credit amount. Once the job completes, it settles against what was actually delivered and refunds anything left over. If the render fails entirely, you get a full refund. Pricing is size-tiered by your chosen output resolution, and the current rates are on the Cat Claw pricing page.

This matters because AI image generation can be unpredictable. Jobs sometimes fail, and outputs sometimes come back at a different count than expected. Settle-and-refund metering means you pay for what you get, not for what was attempted.

The short rule. Reserve, settle, refund. You're charged for the images that actually arrive, and nothing else.

Frequently asked questions

What is CatClaw Layers?

CatClaw Layers is an AI tool from Cat Claw that splits any flat image into up to 16 editable transparent PNG layers. Every hidden region behind an extracted subject is inpainted, so each layer and the base beneath it is a complete, standalone asset that's ready to use immediately.

How do I split an image into layers with AI?

Sign in to Cat Claw, open Tools, then CatClaw Layers. Upload a JPG or PNG, add an optional guidance prompt to steer the split, and in about 1 to 2 minutes you get a named, movable, resizable layer stack on an interactive canvas. Output resolution options are 1K, 1.5K and 2K.

Can AI recover what's behind the subject in a photo?

Yes. CatClaw Layers inpaints the hidden regions during separation, so lifting a subject out of the image reveals a complete background underneath rather than a hole or a rough fill. Every layer in the stack is a usable standalone asset.

Can I edit one part of an image without affecting the rest?

Yes. Select any single layer in CatClaw Layers, write a prompt to restyle it, and attach up to 4 reference images to guide the output. The edited layer returns to the stack at the exact same size and position, and nothing else in the image is touched.

Can AI generate an image that's already separated into layers?

Yes. CatClaw Layers includes a prompt-to-layers mode. Type a description and the tool generates a brand new image born pre-separated into named, ordered layers. No source image is needed.

Is there a Higgsfield Layers alternative?

CatClaw Layers is a Higgsfield Layers alternative that covers image-to-layers separation and adds prompt-to-layers generation, per-layer AI editing with up to 4 reference images, and Press Into material blending. Unused render credits are refunded automatically through settle-and-refund metering.

What formats does CatClaw Layers export?

3 options: a free full-resolution flattened PNG of your arrangement, per-layer transparent PNG downloads, or a ZIP of every layer with a manifest listing names, stacking order, positions and visibility. The ZIP drops directly into Photoshop, Figma or After Effects.

How many layers can CatClaw Layers produce from one image?

Up to 16 independent transparent PNG layers from a single flat image, each with its own stacking order, bounding box, auto-generated name and a short semantic description, plus the inpainted base image underneath them all.

Every image is hiding something. See what yours is hiding.

A flat image has one version of the truth. CatClaw Layers gives you up to 16. The split takes 1 to 2 minutes, the stack stays editable, and the export drops into whatever tool you already use. The cat, as always, reveals nothing. He has 16 layers and intends to keep all of them.

• Start using it now
• More Cat Claw guides
• Prompt Marketplace · Legal & policies

CatClaw Layers turns any flat image into up to 16 editable transparent-PNG layers, backgrounds inpainted. How it works, pricing, and vs Higgsfield Layers.