中文 ｜ English

Project source code address：

Qwen Text2Image & Image2Image Dify Plugin

📖 Project Overview

This is a comprehensive Dify plugin based on ModelScope Qwen-Image models that supports both text-to-image generation and image-to-image editing. Generate high-quality images from text descriptions or edit existing images with text prompts. The plugin uses asynchronous task processing to ensure stable and reliable image generation.

✨ Key Features

• 🎨 High-Quality Image Generation: Supports advanced Qwen-Image-2512 and other models
• ✏️ Image Editing: Edit existing images with text prompts using Qwen-Image-Edit-2511, Qwen-Image-Edit-2509 and other models
• 📐 Custom Image Size Support: Flexible image dimensions with custom size configuration (WxH format)
• 🖼️ Automatic Size Detection: Image2Image tool automatically detects input image dimensions as default
• ⚡ Asynchronous Processing: Uses task submission + polling async mode to avoid timeouts
• 🔄 Real-time Feedback: Provides detailed generation progress and status information
• 🛡️ Error Handling: Comprehensive exception handling with user-friendly error messages
• 🌐 Bilingual Support: Supports both English and Chinese interface and messages

🏗️ Project Architecture

🚀 Quick Start

1. Get ModelScope API Key

1. Visit ModelScope Official Website
2. Register and login to your account
3. Go to My Access Token page
4. Create a new API Key (format: )

2. Install Dependencies

3. Configure Environment

Copy to and configure the parameters:

4. Install Plugin in Dify

1. Upload the plugin folder to Dify plugin directory
2. Enable the plugin in Dify management interface
3. Configure ModelScope API Key

🔧 Usage

Basic Usage

1. Add "Qwen Text2Image" tool in Dify workflow

2. Configure ModelScope API Key

   
   

3. Input image description prompt

4. Select model (default: Qwen-Image)

5. Run the tool to generate image

   Workflow DSL example:

   

Prompt Suggestions

For best image generation results, we recommend:

• Detailed Description: Provide specific information about scene, objects, colors, styles, etc.
• Clear Expression: Use concise and clear language for description
• Style Specification: You can specify artistic styles like "oil painting style", "cartoon style", etc.

Example prompt:

⚙️ Technical Implementation

Core Workflow

1. Task Submission: Submit asynchronous image generation task to ModelScope API
2. Status Polling: Query task status every 5 seconds, wait up to 5 minutes
3. Image Download: Download generated image after task completion
4. Format Conversion: Use PIL to convert image to PNG format and return

API Call Pattern

🔍 Troubleshooting

Common Issues

1. Invalid API Key

   • Check if API Key format starts with
   • Confirm API Key is valid and not expired

2. Generation Timeout

   • Check if network connection is normal
   • Try simplifying prompt description
   • Retry later

3. Image Download Failed

   • Check network connection
   • Confirm firewall settings allow access to ModelScope domains

Error Codes

• : Invalid or unauthorized API Key
• : API call rate limit exceeded
• : Internal server error

📋 Development Standards

This plugin strictly follows the Dify text-to-image plugin development standards defined in CLAUDE2.md:

• ✅ Asynchronous task processing mode
• ✅ Complete error handling mechanism
• ✅ Real-time progress feedback
• ✅ Bilingual support (English/Chinese)
• ✅ Standard ModelScope API calls

🤝 Contributing

Welcome to submit Issues and Pull Requests to improve this plugin!

📄 License

This project is licensed under the MIT License.

🔗 Related Links

• ModelScope Official Website
• Qwen-Image Model
• Dify Official Documentation

📦 Release Notes

0.0.4

• Fixed Image2Image Functionality: Resolved issue where ModelScope server couldn't access Dify internal image URLs causing task failures
• Added Temporary Image Hosting: Integrated litterbox.catbox.moe temporary image hosting service to automatically upload images and obtain publicly accessible URLs
• Improved Image Processing:
  • Automatically handles RGBA, LA, P and other color modes, converting to RGB format
  • Supports image URLs from various sources (including Dify internal addresses, intranet addresses, etc.)
• Increased Network Timeout:
  • API submission request timeout: 300 seconds
  • Task status polling timeout: 120 seconds
  • Image hosting upload timeout: 120 seconds
• Updated Default Model: Image2Image default model updated to
• Improved Error Handling: Optimized error message extraction logic, providing more detailed debugging information

0.0.3

• Enhanced Custom Image Size Support: Both Text2Image and Image2Image tools now support flexible custom image dimensions
• Automatic Size Detection: Image2Image tool automatically detects and uses input image dimensions as default size
• Improved Size Validation: Added comprehensive size format validation with user-friendly error messages
• Better Error Handling: Enhanced error messages for invalid size parameters with automatic fallback
• Code Optimization: Improved parameter handling and validation logic in both tools
• Updated Documentation: Enhanced README with detailed size configuration examples and usage guidelines

0.0.2

• Added Image-to-Image tool (Image2Image) based on ModelScope Qwen-Image-Edit
• New files: ,
• Registered the tool in and imported in
• Updated description and labels to reflect both text-to-image and image-to-image
• Updated README docs (EN/ZH)
• Backward compatible; no breaking changes; existing Text2Image workflows are unaffected
• Usage: In Dify, choose the "Image to Image" tool, then provide a prompt and a public image URL

0.0.1

• Initial release with Text2Image tool based on ModelScope Qwen-Image