KinoGenie
Start for free
All toolsQwen-Image API

Qwen-Image API: Text to Image with Legible Text

Call Qwen-Image, the twenty-billion-parameter text-to-image model from Alibaba, over a hosted API. It is the one to reach for when the image itself has to contain readable words. Run the preset directly on this page or call the same agent-ready endpoint from your product.

Run this tool View API recipe
LIVE TOOL
Run qwen image
Describe the image, pick its shape and how many to render, and download the results here.
0 / 2000
Try:

Describe the subject, the light and the framing. Qwen-Image slower and dearer, renders legible text, honours negative prompts, up to 2k.

1024 × 1024 px

Both tiers cost the same per megapixel; 2k takes longer.

Up to 4. Each one is billed.

Optional. What to steer away from.

1 image at 1k, Qwen-Image (Apache-2.0). Billed per image, reserved when the job is queued and refunded if the render fails.

Already have an account? Sign in

BEFORE / AFTER

See what the tool changes.

A real example: the after side is the unedited output of a KinoGenie job. Run the live form above to generate the same result from your own media.

BeforeSource video
WebM · VP9
mountain.webm
AfterCompatible MP4
Ready
MP4 · H.264
mountain.mp4

Demo footage: public-domain timelapses (Bureau of Land Management, Oregon · NASA SVS). Processed by the same pipeline the API and MCP tools call.

Useful defaults, typed options.

The tool slug stays stable while your agent supplies named media inputs and a narrow set of documented options.

  • Renders readable words inside the image
  • Negative prompts work, unlike on distilled models
  • Apache-2.0 weights, 1k and 2k tiers
What this preset does
  1. 01Send a prompt, optionally a negative prompt, an aspect ratio and a size tier
  2. 02Render with Qwen-Image on a GPU worker
  3. 03Mark each file as AI-generated in its metadata
  4. 04Return one downloadable file per image in result.outputs
Stable endpointPOST /api/v1/tools/qwen-image
Live API request
{
  "inputs": [],
  "options": { "model": "qwen-image", "prompt": "A vintage enamel shop sign reading OPEN DAILY 7-19 in cream lettering on deep green, weathered paint, morning light.", "aspect_ratio": "3:2", "size": "2k", "negative_prompt": "blurry text, warped letters" }
}
Successful output example
{
  "result": {
    "output": {
      "filename": "qwen-image-output.mp4",
      "contentType": "video/mp4",
      "byteSize": 437021,
      "downloadUrl": "https://cdn.kinopipe.com/…"
    }
  }
}

Know the boundaries before you run.

Twenty billion parameters at FP8, so a render is slower and dearer than the distilled model
Both the 1k and 2k tiers are available
The model was trained around 1328 px square and 1664 by 928 for 16:9; our tiers are 1024 and 2048 square equivalents, so renders are close to but not exactly its showcase geometry
Weights: Qwen/Qwen-Image, Apache-2.0
Every image carries a synthetic-content marker in its metadata (EU AI Act art. 50)

About qwen-image api

What is Qwen-Image known for?

Rendering legible text inside the image. Signs, labels, posters and packaging come out readable far more often than with most open-weight models, which is the main reason to pay for it over a distilled one.

Do negative prompts work?

Yes. The model runs with a guidance pass, so a negative prompt genuinely steers the result. That is the other thing it offers over Z-Image Turbo, which ignores the field.

When should I use Z-Image Turbo instead?

When you are iterating, rendering batches, or generating images with no text in them. It is several times cheaper per image and fast enough to try four variations at once.

What resolutions does it render at?

The 1k and 2k tiers, in any of the seven aspect ratios. Its own training resolutions are near 1328 pixels square, so our tiers sit slightly either side of that; renders hold together at both.

What licence are the weights under?

Apache-2.0, on the Qwen/Qwen-Image repository. We host the model; the images you generate are yours to use, subject to our terms and to the AI-generated marking we write into every file.

Related media tools

Z-Image Turbo API: Fast Text to Image

Call Z-Image Turbo, the distilled six-billion-parameter text-to-image model from Alibaba Tongyi, over a hosted API: a prompt in, up to four images out, no GPU of your own.

Explore

Generate Images from Text with an API

Turn a prompt into one to four images with an open-weight model running on our GPUs, delivered as PNG or WebP files you can download.

Explore

Generate AI Video from Text with an API

Turn a text prompt, or an image to animate, into a short video clip with native stereo audio, delivered as an MP4.

Explore