Alibaba: Qwen-Image-Layered
qwen-image-layered- Price
- From $0.0275
- Modalities
- Image
About Qwen-Image-Layered
Qwen-Image-Layered is an Alibaba Qwen model that decomposes a single image into several RGBA layers, so each element sits on its own transparent layer. Because layers are physically separate, you can move, resize, recolour, or delete one element without touching the others. This is a different approach from prompt-based editors, which repaint the whole picture. The weights are released under Apache 2.0.
Where it works well
- Returns separate RGBA layers with transparency, so a background, a subject, and a text block can each be edited on their own.
- The number of layers is flexible; the maker's examples use 3, 4, or 8 layers depending on how finely you want to split the image.
- A single layer can be decomposed again, which helps when one layer still holds several objects.
- Basic operations such as resizing, repositioning, recolouring, and deletion leave other layers untouched instead of risking drift.
When to choose another model
- It splits images; it does not follow free-form edit instructions, so use an instruction-based editor to change what an object looks like.
- The model card recommends a working resolution of 640 pixels for this version, so very large artwork may need to be scaled first.
- Output is a set of layers, not a finished picture, so you need your own tool to compose or export them.
Getting started
Create API key
Create a key in Console, then use it with every model on the platform.
Send your first request
Copy the example for your language and run it against the endpoint.
Image editingTokenLab endpointPOST/v1/images/generationscurl -X POST "https://api.tokenlab.sh/v1/images/generations" \ -H "Authorization: Bearer sk-xxx" \ -H "Content-Type: application/json" \ -d '{ "operation": "image-to-image", "model": "qwen-image-layered", "prompt": "A minimalist product photo of matte black headphones on a soft blue background.", "image_url": "https://example.com/source.png" }'
Pricing
Official price is the model maker's public baseline. TokenLab price is what you pay for this model on TokenLab.
Image editing
per request- Official price
- $0.0275
- Official
- $0.0275
- Discount
- —
| Official priceper request | Officialper request | Discount | |
|---|---|---|---|
| Image editing | $0.0275 | $0.0275 | — |
Usage & activity
Success rate is the share of requests that completed. Latency is how long a full response takes; P95 means 95% of requests finished within that time.
Usage & availability
Last 24 hours- Requests
- Success rate
- P95 latency
- Total tokens
Data is based on aggregate user requests, excluding status checks.
Open in Console
Open Qwen-Image-Layered in Console with a prompt ready to edit or send.
Help me create an image with qwen-image-layered using image-edit at /v1/images/edits. Show the result, cost, and any limits I should know.
Use cases
Rearranging a poster
Split a flat poster into layers, then move the headline, shift the product, and change the background colour without regenerating the whole picture.
Cutting out objects with transparency
Separate a subject from its background and keep soft edges in the alpha channel, ready for use in a design file.
Localisation of marketing images
Isolate a text layer so a designer can swap or reposition the copy for another market while the artwork stays identical.
Animation prep
Break an illustration into parallax-ready layers for motion graphics, one per foreground, midground, and background element.
Prompt examples
Decompose this product banner into 4 layers: background, product, headline text, and decorative shapes.
Split this illustration into 8 layers so I can move the character and the lantern separately.
Separate the subject from the scenery in this photo and keep the transparency around the hair.
FAQ
What does Qwen-Image-Layered output?
A set of RGBA images, one per layer, which stack back into the original picture. Each layer has its own transparency, so an object on one layer can be edited or removed without leaving a hole or a smear on the others.
How many layers can it produce?
The number is adjustable. The model card shows decompositions into 3, 4, and 8 layers, and it supports recursive decomposition, where you take one layer and split it again into finer parts.
Is it the same as an image editing model?
No. Editing models such as Qwen-Image-Edit-2511 repaint pixels from a written instruction. Layered does not repaint; it separates what is already in the image, and you make the edits yourself on the resulting layers.
What image size works best?
The Hugging Face model card recommends a resolution of 640 pixels for the current version. Larger source images can be scaled down before decomposition and the layers upscaled afterwards if you need print size.
Can I use the layers commercially?
The model weights are published under Apache 2.0, which permits commercial use of the software. Rights to the source image itself remain with whoever owns it, so only decompose artwork you are allowed to edit.
How much does Qwen-Image-Layered cost?
On TokenLab, Qwen-Image-Layered costs $0.0275 per request. The pricing table above shows the full breakdown. Rates depend on the billing unit, specification, and usage. Compare matching conditions in the model's detailed pricing; a single rate does not determine the total cost.
Which endpoint should Qwen-Image-Layered use?
Use https://api.tokenlab.sh/v1/images/edits for Qwen-Image-Layered. The request example below shows the matching code shape.
Which operations does Qwen-Image-Layered support?
Qwen-Image-Layered supports Image editing. Select an operation above to see its endpoint and request example.
Compare Qwen-Image-Layered
Sources
Reviewed Oct 2, 2026
