Qwen Image Layered: what Sume returns, flat or transparent

Sume's image API hands back each result as one flat image at a Sume-hosted URL, so it does not return a stack of separate layers. It lists two Qwen models, qwen/qwen-image and qwen/qwen-image-max, and its docs describe neither as producing layers. What Sume does document is a transparent background on a single image, which covers most of what people want a layered file for.

Everything here comes from the Image API docs, the Image 1.0 page and the catalog code, read 2026-09-29. It makes no claim about how any outside model works.

What does Sume's image API return?

One data[].url per image, Sume-hosted and signed, with a media_type beside it. There is no field for layers, masks per element or an alpha-only channel. Transparency, when you ask for it, lives inside the single image you get back.

Is a transparent image the same as layers?

No. A transparent image is one canvas whose empty parts are see-through, so the subject and everything painted around it are still fused in a single picture. A layered result would keep each element separately editable. Sume's transparent option removes the backdrop, which is enough to drop a product or a character onto another design, but you cannot later move one part of the subject without editing pixels.

Which Qwen models does Sume list?

Two ids: qwen/qwen-image and qwen/qwen-image-max. In the catalog code qwen/qwen-image can take reference images for edits, while qwen/qwen-image-max is text-to-image only. Neither advertises a background parameter, and a parameter a model does not list returns 400 unsupported_parameter. For the Qwen side of the catalog in more detail, read Qwen Image 2.1 API: open weights, and what Sume lists.

How do I get a see-through subject today?

ChatGPT Image 2.5 accepts background: auto|transparent|opaque. Ask for transparent and describe only the subject, so nothing in the prompt paints a backdrop.

curl -X POST https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-image-2.5",
    "prompt": "A ceramic coffee mug, studio lighting, no backdrop",
    "background": "transparent",
    "output_format": "png"
  }'

How do I build layers myself?

Generate each element as its own request, each with background: transparent, and keep one aspect_ratio across them so they line up. Stack the returned files in your own editor or renderer. Sume's docs do not list a step that merges stills into layers: Timeline compose puts one still and one video on screen, not several stills. Copy the files you keep into your own storage, since Sume's image URLs are signed.

Will Sume return layers later?

The docs make no promise either way, so this post makes none. If the catalog changes, GET /v1/images/models shows each model's parameters.