Tongyi Qianwen Image Editing Model Interface Document

1. Interface Basic Information

Model Name: qwen-image-edit-plus / qwen-image-edit-plus-2025-10-30 (Qwen Image Editing Model)

Base URL: https://llm.ai-nebula.com/v1/images/generations

Authentication Method: Bearer Token

Auth Token: Bearer sk-xxxxxxxxxx

Core Capabilities:

✅ Single Image Editing (1 input image)

✅ Multi-Image Fusion (2-3 input images)

✅ Modify Text in Image

✅ Add/Remove/Move Objects

✅ Change Subject Action

✅ Transfer Image Style

✅ Enhance Image Details

❌ Pure Text-to-Image Not Supported (Input image required)

Image Input: 1-3 images (Supports URL or Base64)

Image Output: 1-6 images

Billing Method: Per image billing (¥0.18/image, Official price ¥0.2/image)

Official Documentation: Alibaba Cloud Bailian Model Interface Documentation (qwen-image-edit-plus)

1.1 Core Parameters Description

Parameter Name	Type	Required	Description	Example Value
model	string	Yes	Model name	qwen-image-edit-plus
prompt	string	Yes	Image editing instruction (text prompt)	"The girl in image 1 sits down in the pose of image 3 wearing the black dress from image 2"
parameters	object	Yes	Generation parameters	See description below

1.2 Request Parameter Structure

qwen-image-edit-plus uses a simplified format:

{
 "model": "qwen-image-edit-plus",
 "prompt": "Image editing instruction",
 "parameters": {
  "n": 2,             // Number of output images (1-6)
  "negative_prompt": "",      // Negative prompt (Optional)
  "watermark": false,       // Watermark (Optional, default false)
  "seed": 12345          // Random seed (Optional)
 },
 "contents": [
  {
   "role": "user",
   "parts": [
    { "image": "https://..." }, // 1st image (Required)
    { "image": "https://..." }, // 2nd image (Optional)
    { "image": "https://..." }, // 3rd image (Optional)
    { "text": "Editing instruction" }    // Text instruction (Required, placed at the end)
   ]
  }
 ]
}

1.3 Image Input Requirements

Image Format

Supported Formats: JPG, JPEG, PNG, BMP, TIFF, WEBP

Number of Images: 1-3 images

Input Method: URL or Base64 encoding

Image Specifications

Resolution: Both width and height must be within the range of 384-3072 pixels

File Size: Not exceeding 10MB

Aspect Ratio: No limit

Base64 Encoding Format

data:{mime_type};base64,{base64_data}

MIME Type Mapping:

JPEG/JPG: image/jpeg

PNG: image/png

BMP: image/bmp

TIFF: image/tiff

WEBP: image/webp

Example:

data:image/jpeg;base64,/9j/4AAQSkZJRgABAQEAYABgAAD...

2. Single Image Editing

2.1 Modify Text in Image

Use a single image to precisely modify the text content within the image:

2.2 Add/Remove Objects

Add or delete objects in the image:

2.3 Style Transfer

Change the artistic style of the image:

2.4 Change Subject Action

Modify the action/pose of a person or animal:

3. Multi-Image Fusion

3.1 Clothing Transfer (2-Image Fusion)

Apply clothing from one image to a person in another image:

3.2 Pose Transfer (3-Image Fusion)

Combine elements from multiple images:

3.3 Scene Fusion

Fuse elements from multiple images into one scene:

4. Advanced Features

4.1 Using Negative Prompts

Exclude unwanted effects via negative prompts:

Common Negative Prompt Scenarios:

Person Editing: "distortion, deformation, extra limbs, wrong proportions, uncoordinated, distorted face"

Style Transfer: "over-rendered, distortion, mismatched colors, unnatural, inconsistent style"

Text Editing: "blurred text, typos, distorted font, unclear strokes, overlapping text"

Object Editing: "unnatural, sense of violation, obvious seams, wrong shadows, incorrect perspective"

Multi-Image Fusion: "uncoordinated, obvious seams, mismatched lighting, inconsistent tone, stitching marks"

4.2 Random Seed (Seed)

Using the same seed parameter keeps the generation content relatively stable:

Note:

Seed range: [0, 2147483647]

Using the same seed produces similar (but not guaranteed identical) results.

If seed is not provided, the algorithm automatically uses a random seed.

4.3 Batch Generation (n Parameter)

Generate multiple variants in a single request:

n Parameter Range: 1-6 images

Billing Note: Billed by the number of generated images, generating 6 images = 6 images × ¥0.18 = ¥1.08

5. Response Handling

5.1 Response Format

A successful response (Status Code 200) returns JSON containing image data:

{
 "code": 200,
 "msg": "Operation successful",
 "data": {
  "data": [
   {
    "url": "https://dashscope-result-sz.oss-cn-shenzhen.aliyuncs.com/xxx.png?Expires=xxx",
    "b64_json": "",
    "revised_prompt": ""
   },
   {
    "url": "https://dashscope-result-sz.oss-cn-shenzhen.aliyuncs.com/yyy.png?Expires=xxx",
    "b64_json": "",
    "revised_prompt": ""
   }
  ],
  "created": 1762415000,
  "metadata": {
   "output": {
    "choices": [
     {
      "finish_reason": "stop",
      "message": {
       "content": [
        { "image": "https://..." },
        { "image": "https://..." }
       ],
       "role": "assistant"
      }
     }
    ]
   },
   "usage": {
    "width": 1248,
    "image_count": 2,
    "height": 832
   },
   "request_id": "bf37ca26-0abe-98e4-8065-xxxxxx"
  }
 }
}

5.2 Response Field Description

Field Path	Type	Description
code	int	Response status code, 200 indicates success
msg	string	Response message
data.data[]	array	Generated image array
data.data[].url	string	Image URL address (valid for approx. 24 hours)
data.created	int	Generation timestamp
data.metadata	object	Manufacturer's original response data
data.metadata.usage.image_count	int	Number of images generated
data.metadata.usage.width	int	Image width
data.metadata.usage.height	int	Image height

5.3 Error Handling

If the request fails, an error message will be returned:

{
 "code": 400,
 "msg": "Parameter Error: Image quantity exceeds limit",
 "data": null
}

Common Error Codes:

400: Parameter error (e.g., incorrect image format, size over limit, etc.)

401: Authentication failed (Invalid or expired API key)

403: Insufficient permissions or balance

429: Request too frequent, rate limit exceeded

500: Internal Server Error

6. Best Practices

6.1 Multi-Image Editing Prompt Tips

Explicit Image Marking

When processing multiple images, you must use "Image 1", "Image 2", "Image 3" to refer explicitly:

✅ Good Prompt (Explicit marking):

"The girl in image 1 sits down in the pose of image 3 wearing the black dress from image 2"

❌ Bad Prompt (Missing marking):

"Girl sits down wearing a black dress"

Describe Desired Effect in Detail

Providing detailed descriptions leads to better editing results:

✅ Good Prompt:

"The girl in image 1 sits down in the pose of image 3 wearing the black dress from image 2, keep her clothing, hairstyle and expression unchanged, movement is natural and smooth, lighting and shadows remain consistent"

❌ Simple Prompt:

"Girl in image 1 changes clothes"

6.2 Image Input Suggestions

Image Quality

✅ Use high-resolution, clear images

✅ Ensure sufficient lighting and moderate contrast

❌ Avoid blurred, too dark, or too bright images

Image Content

✅ Clear subject, simple background

✅ Reasonable composition, correct proportions

❌ Avoid overly complex scenes

Image Size

Minimum: 384×384 pixels

Maximum: 3072×3072 pixels

Recommended: 1024×1024 to 2048×2048 pixels

6.3 Negative Prompt Strategy

Use appropriate negative prompts based on different editing types:

Edit Type	Recommended Negative Prompts
Person Editing	"distortion, deformation, extra limbs, wrong proportions, uncoordinated, face distortion"
Style Transfer	"over-rendered, distortion, mismatched colors, unnatural, inconsistent style"
Text Editing	"blurred text, typos, distorted font, unclear strokes, overlapping text"
Object Editing	"unnatural, sense of violation, obvious seams, wrong shadows, incorrect perspective"
Multi-Image Fusion	"uncoordinated, obvious seams, mismatched lighting, inconsistent tone, stitching marks"

6.4 Parameter Combination Suggestions

Usage Scenario	n	negative_prompt	watermark	seed
Quick Test	1-2	""	false	Not set
High Quality Output	3-4	Detailed description	false	Not set
Batch Generation	6	Brief description	false	Not set
Stable Reproduction	1	Detailed description	false	Fixed value

7. Billing Instructions

7.1 Billing Method

qwen-image-edit-plus uses a per-image billing model:

Official Price: ¥0.20/image

System Price: ¥0.18/image (10% off)

Billing Unit: Billed based on the number of successfully generated images

Unified Price: Same price for all output sizes

7.2 Billing Example

Generate 2 images: 2 images × ¥0.18 = ¥0.36
Generate 6 images: 6 images × ¥0.18 = ¥1.08

7.3 Billing Log Format

After successful generation, the system records billing info:

{
 "content": "Image generation: 2 images × ¥0.18 = ¥0.36",
 "modelName": "qwen-image-edit-plus",
 "quota": 12329,
 "quotaDollar": "0.049316"
}

8. Frequently Asked Questions (FAQ)

Q1: Does qwen-image-edit-plus support pure Text-to-Image?

A: No. qwen-image-edit-plus is an image editing model and must be provided with at least 1 input image. For pure Text-to-Image functionality, please use qwen-image-plus or doubao-seedream-4-0-250828.

Q2: How many images can be uploaded at most?

A: Supports up to 3 images input simultaneously. Use 1 image for single image editing, and 2-3 images for multi-image fusion.

Q3: How many images can be generated at once?

A: You can control the output quantity via the n parameter, ranging from 1-6 images.

Q4: How long is the generated image valid?

A: The image URL is valid for approximately 24 hours. It is recommended to download and save it immediately after receiving the response, or upload it to your own storage service.

Q5: How to refer to different images during multi-image editing?

A: You must use "Image 1", "Image 2", "Image 3" in the prompt to explicitly refer to the corresponding images, for example:

"The girl in image 1 sits down in the pose of image 3 wearing the black dress from image 2"

If not explicitly marked, unexpected editing results may occur.

Q6: What image formats are supported for input?

A: Supports URL and Base64 encoding:

URL: Directly provide the image's HTTP/HTTPS link

Base64: Format data:image/{type};base64,{data}

Q7: How to improve editing quality?

A: Suggestions:

Use high-resolution, clear input images.

Provide detailed, specific editing instructions.

Explicitly use "Image 1", "Image 2" markers during multi-image editing.

Use negative prompts reasonably to exclude unwanted effects.

Appropriately adjust the n parameter to generate multiple variants for selection.

Q8: What is the use of the seed parameter?

A: The seed (random seed) can keep the generation result relatively stable. Using the same seed, same input, and parameters can produce similar (but not guaranteed identical) outputs. Suitable for:

Reproducing specific effects

A/B testing

Batch production of images with similar styles

Q9: Can watermarks be removed?

A: Yes, set watermark: false. Note: Removing watermarks may have usage restrictions according to Alibaba Cloud's terms of service; please comply with relevant regulations.

Q10: What image output formats are supported?

A: The image URL returned by qwen-image-edit-plus points to a PNG format image, ensuring high-quality output. The aspect ratio of the output image matches the input image.

9. Complete Request Examples

Example 1: Single Image Editing - Modify Text (Python)

Example 2: Multi-Image Fusion - Clothing Transfer (Node.js)

Example 3: Base64 Image Input (Python)

10. Performance Optimization Suggestions

10.1 Image Preprocessing

Preprocessing images before uploading can improve editing results:

Compress Image: Compress large images to a reasonable size (recommended 1024-2048 pixels).

Adjust Brightness: Ensure the image has sufficient lighting and moderate contrast.

Crop Irrelevant Content: Remove unnecessary background or edges.

10.2 Prompt Optimization

Be Specific: Avoid vague descriptions; give specific editing requirements.

Step-by-Step Description: Complex edits can be broken down into multiple simple steps.

Preserve Original Features: Explicitly state which original features need to be preserved.

10.3 Batch Generation Strategy

If you need to generate a large number of images:

Concurrent Requests: Initiate multiple independent requests simultaneously.

Use Queue: Implement request queue management to avoid exceeding rate limits.

Error Retry: Implement automatic retry mechanism to handle temporary failures.

11. Quick Reference

Model Parameters Cheat Sheet

Parameter	Value
Model Name	qwen-image-edit-plus, qwen-image-edit-plus-2025-10-30
Supported Function	Image Editing (Image-to-Image)
Image Input	1-3 images (URL or Base64)
Image Output	1-6 images
Input Format	JPG, JPEG, PNG, BMP, TIFF, WEBP
Output Format	PNG
Resolution Limit	384-3072 pixels
File Size Limit	≤10MB
Response Format	URL (Valid for approx. 24 hours)
Billing Method	Per image billing (¥0.18/image)

Core Functions Cheat Sheet

Function	Supported	Description
Single Image Editing	✅	1 input image
Multi-Image Fusion	✅	2-3 input images
Modify Text	✅	Precise modification of text in image
Add/Remove Objects	✅	Add or delete objects
Change Action	✅	Modify person/animal pose
Style Transfer	✅	Change artistic style
Negative Prompt	✅	Exclude unwanted elements
Random Seed	✅	Control generation stability
Pure Text-to-Image	❌	Must have input image

Parameter Combination Cheat Sheet

Scenario	Image Count	n	negative_prompt	watermark	seed
Single Image Edit	1	2-3	Brief description	false	Not set
Multi-Image Fusion	2-3	2-4	Detailed description	false	Not set
High Quality Output	1-3	3-4	Detailed description	false	Not set
Stable Reproduction	1-3	1	Detailed description	false	Fixed value
Batch Test	1-3	6	Brief description	false	Not set

Document Version: v1.0
Update Time: 2025-11-06
Model: Qwen Image Editing Model (qwen-image-edit-plus)
Technical Support: https://llm.ai-nebula.com

Technical Support

If you have any questions, please contact:

Documentation: https://llm.ai-nebula.com/docs

Technical Support: support@ai-nebula.com

Tongyi Qianwen Image Editing Model Interface Document

1. Interface Basic Information#

1.1 Core Parameters Description#

1.2 Request Parameter Structure#

1.3 Image Input Requirements#

Image Format#

Image Specifications#

Base64 Encoding Format#

2. Single Image Editing#

2.1 Modify Text in Image#

2.2 Add/Remove Objects#

2.3 Style Transfer#

2.4 Change Subject Action#

3. Multi-Image Fusion#

3.1 Clothing Transfer (2-Image Fusion)#

3.2 Pose Transfer (3-Image Fusion)#

3.3 Scene Fusion#

4. Advanced Features#

4.1 Using Negative Prompts#

4.2 Random Seed (Seed)#

4.3 Batch Generation (n Parameter)#

5. Response Handling#

5.1 Response Format#

5.2 Response Field Description#

5.3 Error Handling#

6. Best Practices#

6.1 Multi-Image Editing Prompt Tips#

Explicit Image Marking#

Describe Desired Effect in Detail#

6.2 Image Input Suggestions#

Image Quality#

Image Content#

Image Size#

6.3 Negative Prompt Strategy#

6.4 Parameter Combination Suggestions#

7. Billing Instructions#

7.1 Billing Method#

7.2 Billing Example#

7.3 Billing Log Format#

8. Frequently Asked Questions (FAQ)#

Q1: Does qwen-image-edit-plus support pure Text-to-Image?#

Q2: How many images can be uploaded at most?#

Q3: How many images can be generated at once?#

Q4: How long is the generated image valid?#

Q5: How to refer to different images during multi-image editing?#

Q6: What image formats are supported for input?#

Q7: How to improve editing quality?#

Q8: What is the use of the seed parameter?#

Q9: Can watermarks be removed?#

Q10: What image output formats are supported?#

9. Complete Request Examples#

Example 1: Single Image Editing - Modify Text (Python)#

Example 2: Multi-Image Fusion - Clothing Transfer (Node.js)#

Example 3: Base64 Image Input (Python)#

10. Performance Optimization Suggestions#

10.1 Image Preprocessing#

10.2 Prompt Optimization#

10.3 Batch Generation Strategy#

11. Quick Reference#

Model Parameters Cheat Sheet#

Core Functions Cheat Sheet#

Parameter Combination Cheat Sheet#

Technical Support#