Skip to main content
Version: 1.11.0-beta

Qwen Image Edit

Advanced text-based image editing using the Qwen Image Edit model through the ImagineoAI API. This model enables powerful AI-driven image transformations using natural language prompts.

Overview​

The Qwen Image Edit model is a ComfyDeploy-powered AI model that specializes in text-based image editing. It allows you to describe changes to images using natural language, with support for both positive and negative prompts for precise control over the output.

Method​

The Qwen Image Edit model is accessed through the standard generate method with model_type: "qwen-image-edit":

Signature:

client.images.generate(input: GenerateImageInput): Promise<GenerateImageOutput | null>

Parameters​

Required Parameters​

  • prompt (string): The text prompt describing the desired edit or transformation
  • model_type (string): Must be set to "qwen-image-edit"
  • width (number): Output image width in pixels (64-2048)
  • height (number): Output image height in pixels (64-2048)

Optional Parameters​

  • lora_1 (string): Optional LoRA model name for style enhancement
  • strength_model (number): LoRA model strength (0-2, default: 1)
  • negative_prompt (string): Text describing what to avoid in the generation
  • webhook_url (string): URL for receiving completion notifications
  • webhook_secret (string): Secret for webhook authentication

Examples​

Basic Image Edit​

Transform an image with a simple text prompt:

const client = new ImagineoAIClient(apiUrl, { apiKey: 'sk-...' });

const result = await client.images.generate({
  prompt: 'Make the scene more vibrant and colorful with a sunset atmosphere',
  model_type: 'qwen-image-edit',
  width: 1024,
  height: 1024,
});

if (result) {
  console.log('Edited image URL:', result.image_url);
  console.log('Run ID:', result.run_id);
}

Advanced Edit with Negative Prompt​

Use negative prompts to avoid specific elements:

const result = await client.images.generate({
  prompt: 'Transform into a fantasy landscape with magical floating islands',
  negative_prompt: 'modern buildings, cars, technology, people',
  model_type: 'qwen-image-edit',
  width: 1920,
  height: 1080,
});

Style Enhancement with LoRA​

Apply a specific style using a LoRA model:

const result = await client.images.generate({
  prompt: 'Apply an artistic watercolor painting style to the image',
  model_type: 'qwen-image-edit',
  lora_1: 'watercolor-style-v2',
  strength_model: 1.5,  // Increase LoRA influence
  width: 1024,
  height: 1024,
});

Complex Scene Transformation​

Perform complex transformations with detailed prompts:

const result = await client.images.generate({
  prompt: 'Transform the urban street into a cyberpunk scene with neon lights, holographic advertisements, flying vehicles, and rain-slicked streets reflecting the lights',
  negative_prompt: 'daylight, traditional signs, clear sky',
  model_type: 'qwen-image-edit',
  lora_1: 'cyberpunk-aesthetic-v3',
  strength_model: 1.2,
  width: 1920,
  height: 1080,
});

Webhook Integration​

Set up webhooks for asynchronous processing:

const result = await client.images.generate({
  prompt: 'Convert the photograph into a detailed pencil sketch',
  model_type: 'qwen-image-edit',
  width: 1024,
  height: 1024,
  webhook_url: 'https://your-server.com/webhook/qwen-complete',
  webhook_secret: 'your-webhook-secret'
});

// The webhook will receive updates when the edit is complete

Use Cases​

1. Style Transfer​

Transform images into different artistic styles:

  • Convert photos to paintings, sketches, or digital art
  • Apply specific art movement styles (impressionist, cubist, etc.)
  • Create themed variations (fantasy, sci-fi, vintage)

2. Scene Modification​

Change environmental aspects of images:

  • Time of day transformations (day to night)
  • Season changes (summer to winter)
  • Weather effects (add rain, snow, fog)

3. Object Manipulation​

Modify or enhance specific elements:

  • Change colors or textures
  • Add or remove objects (using negative prompts)
  • Transform objects into different materials

4. Mood and Atmosphere​

Adjust the emotional tone of images:

  • Make scenes more dramatic or peaceful
  • Add cinematic effects
  • Enhance lighting and shadows

Best Practices​

Prompt Engineering​

  1. Be Specific: Provide detailed descriptions of desired changes

    // Good
    prompt: 'Transform the modern office into a Victorian-era library with wooden bookshelves, leather-bound books, and warm gas lamp lighting'

    // Less effective
    prompt: 'Make it old-fashioned'

  2. Use Negative Prompts: Explicitly exclude unwanted elements

    negative_prompt: 'modern technology, plastic, neon lights, contemporary furniture'

  3. Layer Your Instructions: Break complex edits into clear components

    prompt: 'First, change the time to golden hour sunset. Second, add warm orange and pink tones to the sky. Finally, create long shadows and warm lighting on all surfaces'

LoRA Models​

  • Use LoRA models for consistent style application
  • Adjust strength_model based on desired intensity (0.5-1.5 typically works best)
  • Lower values (0.5-0.8) for subtle enhancements
  • Higher values (1.2-2.0) for dramatic transformations

Performance Tips​

  1. Resolution: Start with lower resolutions for testing (512x512), then increase for final outputs
  2. Batch Processing: Use webhooks for multiple edits to avoid blocking
  3. Caching: Store successful prompts and settings for reuse

Error Handling​

try {
  const result = await client.images.generate({
    prompt: 'Transform into watercolor painting',
    model_type: 'qwen-image-edit',
    width: 1024,
    height: 1024,
  });
  
  if (result) {
    console.log('Success:', result.image_url);
  } else {
    console.log('No content returned (204)');
  }
} catch (error) {
  if (error.code === 'VALIDATION_ERROR') {
    console.error('Invalid parameters:', error.message);
  } else if (error.code === 'INSUFFICIENT_CREDITS') {
    console.error('Not enough credits for this operation');
  } else {
    console.error('Unexpected error:', error);
  }
}

Integration with Other Features​

Combining with Other Models​

You can chain Qwen Image Edit with other models for complex workflows:

// Step 1: Edit with Qwen
const editResult = await client.images.generate({
  prompt: 'Transform into fantasy landscape',
  model_type: 'qwen-image-edit',
  width: 1024,
  height: 1024,
});

// Step 2: Upscale the result
if (editResult) {
  const upscaleResult = await client.images.upscale({
    original_run_id: editResult.run_id,
    scale_factor: 2
  });
}

Working with Run IDs​

The returned run_id can be used to:

  • Track the generation status
  • Reference the image in other operations
  • Retrieve the image later
  • Use as input for other transformations

Limitations​

  • Maximum resolution: 2048x2048 pixels
  • LoRA strength range: 0-2
  • Processing time varies based on complexity and resolution
  • Credits are consumed based on image dimensions

See Also​