Qwen Image API on MuAPI

Compare Qwen image generation and editing endpoints, then run a request with the same REST API shape across the Qwen model family. MuAPI provides unified access to these media endpoints; this repository is an independent integration guide and is not an official Alibaba Cloud or Qwen project.

• MuAPI: https://muapi.ai
• Qwen API comparison and examples: https://muapi.ai/qwen-api
• Qwen Image playground: https://muapi.ai/qwen-image
• Qwen 2.1 playground: https://muapi.ai/qwen-2.1
• Create/manage an API key: https://muapi.ai/access-keys
• API introduction: https://muapi.ai/docs/introduction
• Pricing: https://muapi.ai/pricing

Related Projects

• Open Generative AI — practical generative-media API examples and model references.
• OpenAI API examples — compare another unified API covering image-generation and image-editing workflows.
• Google Gemini Media API — image and video generation API examples.
• MiniMax Media API — video, speech, music, and image editing examples.
• Kling Video API — examples for another model-family API exposed through MuAPI.

Qwen API scope

This is a Qwen image API guide. The live MuAPI catalog currently exposes Qwen image generation, image-to-image editing, LoRA-assisted image generation/editing, and LoRA training endpoints. It does not expose Qwen text/chat endpoints through this Qwen media catalog. Use Alibaba Cloud's official service for Qwen chat, embeddings, or other text-model APIs.

All examples use MuAPI's unified task API:

1. Submit POST https://api.muapi.ai/api/v1/{endpoint} with an x-api-key header and JSON body.
2. Read the returned request_id.
3. Poll GET https://api.muapi.ai/api/v1/predictions/{request_id}/result until the task is complete.

Choose an endpoint

Workflow	Endpoint	What it does	Required input
Text to image, Qwen 2.1	qwen2-1-text-to-image	Qwen 2.1 1K/2K image generation; supports aspect ratio, transparent background, format, prompt enhancement, and seed	prompt
Image editing, Qwen 2.1	qwen2-1-image-to-image	Edit or combine reference images; optional mask enables local edits	prompt, images_list
Text to image, Qwen 3	qwen3-text-to-image	Qwen 3 image generation with 1K/2K resolution and prompt extension controls	prompt
Image editing, Qwen 3	qwen3-image-to-image	Edit up to three reference images	prompt, images_list
Qwen Image 2.0 / Pro	qwen-image-2.0, qwen-image-2.0-pro	Generation endpoints with aspect-ratio control	prompt
Qwen Image 2.0 / Pro editing	qwen-image-2.0-edit, qwen-image-2.0-pro-edit	Image editing from reference image URLs	prompt, images_list
LoRA image workflows	qwen-image-*-lora	Generate or edit using supplied LoRA weights	Depends on endpoint; see schema below
Train a LoRA	qwen-image-lora-trainer, qwen-image-2512-lora-trainer	Train Qwen image LoRA weights from a dataset	data

The comparison above is based on MuAPI's live model catalog and request schemas. Endpoint availability and inputs can change; check the model playground for the current form and availability before production use.

Quick start: generate an image

Set your API key in the environment:

export MUAPI_KEY="your_muapi_api_key"

Submit a Qwen 2.1 text-to-image request:

curl -X POST "https://api.muapi.ai/api/v1/qwen2-1-text-to-image" \
  -H "x-api-key: $MUAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "A corgi wearing a yellow rain hat sitting on stone steps after the rain, shallow depth of field",
    "aspect_ratio": "16:9",
    "resolution": "1K",
    "output_format": "png"
  }'

The response includes a request_id and status. Use the result endpoint to retrieve the generated image URL after completion:

curl "https://api.muapi.ai/api/v1/predictions/REQUEST_ID/result" \
  -H "x-api-key: $MUAPI_KEY"

Image editing example

images_list is a list of publicly reachable image URLs. Qwen 2.1 accepts 1–10 reference URLs; image order matters when the prompt refers to the first or second image. mask_url is optional and enables local editing when used with exactly one input image.

curl -X POST "https://api.muapi.ai/api/v1/qwen2-1-image-to-image" \
  -H "x-api-key: $MUAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Replace the backpack in the first image with the colour scheme from the second image, and leave everything else unchanged",
    "images_list": [
      "https://example.com/backpack.jpg",
      "https://example.com/color-reference.jpg"
    ],
    "resolution": "1K",
    "aspect_ratio": "auto"
  }'

Replace the example URLs with image URLs that the API can access. The API receives URLs; this endpoint does not upload local files directly.

Python example

Requires Python 3.9+ and requests. Run python examples/qwen_image.py --help for options. The script submits either a Qwen 2.1 generation or edit request and polls for the result.

python -m pip install -r requirements.txt
export MUAPI_KEY="your_muapi_api_key"
python examples/qwen_image.py --prompt "A tiny greenhouse on a rainy city rooftop at dusk"

For editing, provide one or more accessible reference URLs:

python examples/qwen_image.py \
  --mode edit \
  --prompt "Turn the product into a brushed aluminum desk lamp; keep the composition" \
  --image-url "https://example.com/product.jpg"

Qwen 2.1 request schema

qwen2-1-text-to-image

Field	Type	Required	Accepted values / behavior
prompt	string	Yes	Up to 5,000 characters; any language is supported
aspect_ratio	string	No	1:1, 4:3, 3:4, 3:2, 2:3, 16:9, 9:16, 21:9, 9:21; default 1:1
resolution	string	No	1K or 2K; default 1K
background	string	No	opaque or transparent; default opaque
output_format	string	No	png, webp, or jpeg; default png
enhance_prompt	boolean	No	Prompt rewrite; default true
seed	integer	No	Omit for a random seed

Transparent background is incompatible with JPEG.

qwen2-1-image-to-image

Field	Type	Required	Accepted values / behavior
prompt	string	Yes	Describe the desired result; for masked edits, describe the white mask region
images_list	string[]	Yes	1–10 publicly reachable reference image URLs; order is meaningful
mask_url	string	No	Inpainting mask; requires exactly one reference image
aspect_ratio	string	No	auto, 1:1, 4:3, 3:4, 3:2, 2:3, 16:9, 9:16, 21:9, 9:21
resolution	string	No	1K or 2K; default 1K
background	string	No	opaque or transparent; default opaque
output_format	string	No	png, webp, or jpeg; default png
enhance_prompt	boolean	No	Ignored in local-edit mode
seed	integer	No	Optional seed

Response and polling

Submission returns a request_id and task status. Poll the predictions result endpoint with that ID. The Qwen 2.1 result schema contains request_id (identifier), status (current task state), and output (array of generated image URL strings). On completion, read the image URLs from output. Handle pending/processing statuses, terminal errors, and network timeouts in your client. Output URLs should be downloaded or copied to your own storage if you need long-term availability.

Pricing and limits

MuAPI uses account credits and per-model billing. Exact costs can vary by model and settings such as output count or resolution. Check the live MuAPI pricing page and each Qwen model's playground before submitting production workloads. The request examples do not promise a fixed cost, latency, or output availability period.

Local verification

The Python example can be syntax-checked without an API key:

python -m py_compile examples/qwen_image.py

License

MIT. MuAPI, Qwen, and Alibaba Cloud are trademarks of their respective owners. This repository is an independent community example and is not endorsed by Alibaba Cloud.