Z-Image Turbo Image Generation

Alibaba Z-Image Turbo — a fast, cost-effective image generation model excelling at high-fidelity portraits and product photos.

Try it Online

Test in Playground ↗ · View Models ↗

Quick Start

from openai import OpenAI

client = OpenAI(
    base_url="https://open.dieyuyun.com/v1",
    api_key="YOUR_API_KEY"
)

response = client.images.generate(
    model="z-image-turbo",
    prompt="A cute orange tabby cat basking in sunlight on a windowsill",
    size="1024x1536",
    n=1
)

print(response.data[0].url)

Endpoint

Item	Value
Method	POST
Path	/v1/images/generations
Base URL	https://open.dieyuyun.com
Protocol	OpenAI Images API

Authentication

Bearer Token: Authorization: Bearer YOUR_API_KEY

Standard Request Fields

Field	Type	Required	Default	Description
model	string	Yes	—	Model identifier: z-image-turbo
prompt	string	Yes	—	Image description text
size	string	No	1024x1536	Image size, total pixels must be within [512², 2048²]
n	integer	No	1	Number of images — fixed at 1, batch generation not supported
response_format	string	No	url	Response format: url or b64_json

Important Notes

• n is fixed at 1: Z-Image Turbo does not support batch generation. Each request produces exactly 1 image.
• No negative_prompt support: This model does not accept negative prompt parameters.
• No watermark parameter: This model does not have a watermark control parameter.

Supported Sizes

Size	Aspect Ratio	Description
1024x1024	1:1	Square
1280x1280	1:1	Large square
1536x1536	1:1	Extra-large square
1024x1536	2:3	Portrait (default)
1104x1472	3:4	Portrait
896x1184	3:4	Small portrait
768x1360	9:16	Vertical screen
960x1704	9:16	Large vertical screen
1152x2048	9:16	Extra-large vertical
1328x1760	3:4	Large portrait
1704x960	16:9	Large horizontal screen
1760x1328	4:3	Large landscape
2048x1152	16:9	Extra-wide screen
1360x768	16:9	Horizontal screen
1184x896	4:3	Landscape
1472x1104	4:3	Large landscape

Size Guidelines

Total pixel count must be within [512², 2048²]. The sizes listed above are recommended; you may also pass custom sizes as long as they satisfy the total pixel constraint.

Extended Parameters

Z-Image Turbo supports additional parameters via the parameters object:

Field	Type	Required	Default	Description
parameters.prompt_extend	boolean	No	false	Whether to optimize the prompt. When enabled, returns a revised prompt and thinking process. Costs more.
parameters.seed	integer	No	—	Random seed in range [0, 2147483647] for reproducible results

{
  "model": "z-image-turbo",
  "prompt": "A cute orange tabby cat basking in sunlight on a windowsill",
  "size": "1024x1536",
  "parameters": {
    "prompt_extend": true,
    "seed": 42
  }
}

About prompt_extend

Unlike Qwen Image, Z-Image Turbo defaults prompt_extend to false. When enabled:

• The response includes revised_prompt (optimized prompt) and reasoning_content (thinking process)
• Generation quality may improve, but token usage and cost will increase

Request Examples

cURL

curl https://open.dieyuyun.com/v1/images/generations \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "z-image-turbo",
    "prompt": "A cute orange tabby cat basking in sunlight on a windowsill",
    "size": "1024x1536",
    "n": 1
  }'

Python

from openai import OpenAI

client = OpenAI(
    base_url="https://open.dieyuyun.com/v1",
    api_key="YOUR_API_KEY"
)

response = client.images.generate(
    model="z-image-turbo",
    prompt="A cute orange tabby cat basking in sunlight on a windowsill",
    size="1024x1536",
    n=1
)

print(response.data[0].url)

Node.js

import OpenAI from 'openai'

const client = new OpenAI({
  baseURL: 'https://open.dieyuyun.com/v1',
  apiKey: 'YOUR_API_KEY',
})

const response = await client.images.generate({
  model: 'z-image-turbo',
  prompt: 'A cute orange tabby cat basking in sunlight on a windowsill',
  size: '1024x1536',
  n: 1,
})

console.log(response.data[0].url)

Enable Prompt Optimization Example

curl https://open.dieyuyun.com/v1/images/generations \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "z-image-turbo",
    "prompt": "A cute orange tabby cat basking in sunlight on a windowsill",
    "size": "1024x1536",
    "n": 1,
    "parameters": {
      "prompt_extend": true,
      "seed": 42
    }
  }'

Response Format

Success Response (prompt_extend=false)

{
  "created": 1717000000,
  "data": [
    {
      "url": "https://..."
    }
  ]
}

Success Response (prompt_extend=true)

{
  "created": 1717000000,
  "data": [
    {
      "url": "https://...",
      "revised_prompt": "A vibrant orange tabby cat lazily lying on a sun-drenched windowsill, warm afternoon light streaming through the glass, detailed fur texture visible, cozy home interior in the background",
      "reasoning_content": "The user wants an image of an orange cat sunbathing on a windowsill. I should enhance this prompt with more detail about lighting, fur texture, and ambient atmosphere to produce a more realistic image..."
    }
  ]
}

Field	Type	Description
created	integer	Unix timestamp
data	array	Array of generated results
data[].url	string	Image URL
data[].revised_prompt	string	Optimized prompt (only returned when prompt_extend=true)
data[].reasoning_content	string	Model's thinking process (only returned when prompt_extend=true)

Error Response

See Error Codes for details.

Best Practices

• Speed first: Z-Image Turbo is built for fast generation, making it ideal for high-volume scenarios like e-commerce product images and social media assets.
• Portraits & product photos: This model excels at high-fidelity portraits and product photography, well-suited for e-commerce and advertising use cases.
• Prompt optimization: For simple prompts, use the default setting (prompt_extend=false) to save costs. For complex scenes, enable prompt_extend for better results.
• Seed for reproducibility: Use the seed parameter to reproduce previous results, making it easy to iterate on images you like.
• Size selection: The default size 1024x1536 works well for portrait shots; for product images, choose square (1024x1024) or landscape sizes as needed.
• Synchronous response: This model returns results synchronously. No task_id polling is needed.

Rate Limits

See Rate Limits for details.

Related Documentation

• Wan 2.7 - Alibaba Wan video generation model
• Qwen Image - Tongyi Wanxiang image generation model
• Model List & Pricing - Browse available models and pricing
• Playground - Test image generation interactively