# Image to Layers

> One flat image in, aligned transparent layers out (PSD / PNG / ZIP).

## What it does

- Splits a single raster image into RGBA layers that all share one canvas at the original position.
- Repaints the areas hidden behind foreground objects, so removing a subject does not leave a hole.
- Exports a layered PSD (PSD 8-bit RGB, plain raster layers, no groups), one PNG per layer, or a ZIP of both.
- Any layer can be split again, so the total layer count has no fixed ceiling.

## What it cannot do

- It cannot recover an original PSD. JPG and PNG files contain no layer data to recover.
- Text returns as a raster layer, not as editable type.
- Hair, glass, smoke and soft shadows need manual review.
- It does not produce rigging-ready Live2D or Spine assets — first-pass parts only.
- Rebuilt areas are generated, not photographed, and may not match the real scene.

## Limits (as of 2026-10-23)

- Input for a split: JPG / PNG / WEBP attached in Layer Agent, up to 10 MB. Larger files are refused, not resized, and the layers can come back smaller than the upload.
- The free single-layer PSD save runs in the browser; it also reads HEIC where the browser can decode it, and scales images over 4096 px on the long edge down first.
- Output: 2-8 layers per run on subscriptions, up to 4 on credit packs; any layer can be split again.

## Pricing

- Every run costs 10 credits, whatever the number of layers. Splitting a layer again is a new run. Failed runs are refunded; re-downloads are free.
- No free tier: trying it out is what the small credit pack is for.
- Every plan includes the Layer Agent chat, where splits run; credit packs do not include the chat.
- Starter: $14.90/mo for 30 runs, or $118.80/yr for 450 runs.
- Pro: $24.90/mo for 80 runs, or $178.80/yr for 1100 runs.
- Studio: $59.90/mo for 250 runs, or $430.80/yr for 3000 runs.
- Plan credits last one month: a monthly plan’s expire when the billing month ends; a yearly plan delivers a twelfth of its credits each month, each expiring at that month’s end.
- Credit packs (one-time, valid 30 days): $9.90 for 20 runs, $29.90 for 70 runs.
- We claim no rights in your output, on every plan.

## Data handling

- Images attached in Layer Agent are stored in our cloud storage and sent to the provider that runs the layer model. The finished layer files are hosted by that provider and kept in our storage too. Chat messages go to a language-model provider.
- The free single-layer PSD save runs in the browser and uploads nothing.

## Model

- Qwen-Image-Layered (Apache-2.0). Open weights mean the output carries no model-licence restriction.

## Pages

- [Home](https://image2layers.org/): what the product does, with an interactive layer stack.
- [Upload](https://image2layers.org/app): the upload box and the free single-layer PSD save.
- [Layer Agent](https://image2layers.org/chat): the chat where splits run, for subscribers.
- [Pricing](https://image2layers.org/pricing): plans, credit rules and the commercial licence.
- [PNG to PSD with layers](https://image2layers.org/png-to-psd): Transparent PNGs, logos and screenshots, transparency kept.
- [JPG to PSD with layers](https://image2layers.org/jpg-to-psd): Product shots and phone photos, background filled in.
- [WebP to PSD with layers](https://image2layers.org/webp-to-psd): Images saved from websites and social posts.
- [AI image to PSD with layers](https://image2layers.org/ai-generated-image-to-psd): ChatGPT and Midjourney images — fix a word, move a part.
- [Canva Magic Layers alternative](https://image2layers.org/canva-magic-layers-alternative): The same split, with the layers downloaded as PSD or PNG.
- [How to separate an image into layers](https://image2layers.org/guides/how-to-separate-an-image-into-layers): By hand in Photoshop or GIMP, or in one step with AI.
- [PSD to PNG](https://image2layers.org/psd-to-png): The whole picture or every layer as a PNG, in your browser.
- [PSD to JPG](https://image2layers.org/psd-to-jpg): A PSD as one JPG to share, email or print — no Photoshop.
- [Qwen Image Layered online](https://image2layers.org/qwen-image-layered): The open Qwen-Image-Layered model online — no GPU, no ComfyUI.