Skills qwen-image-2
🎨

qwen-image-2

Safe

Generate AI Images with Qwen-Image-2

Also available from: inferen-sh

Create professional AI-generated images and edits using Alibaba Qwen-Image-2.0 models. Generate text-to-image art, edit with reference images, and render complex text through Claude with the inference.sh CLI.

Supports: Claude Codex Code(CC)
📊 69 Adequate
1

Download the skill ZIP

2

Upload in Claude

Go to Settings → Capabilities → Skills → Upload skill

3

Toggle on and start using

Test it

Using "qwen-image-2". Generate a futuristic cityscape at sunset with flying cars

Expected outcome:

AI-generated image showing a sci-fi city skyline at golden hour with flying vehicles, rendered in photorealistic style

Using "qwen-image-2". Create a poster with 'Summer Sale!' in bold red text, '50% Off' in blue, beach background

Expected outcome:

Vertical poster image with crisp red title text, blue subtitle text, and tropical beach scene, using Pro model for typography

Using "qwen-image-2". Make the person from image1 wear the outfit from image2

Expected outcome:

Composite image combining the person from the first reference image wearing clothing from the second reference image

Security Audit

Safe
v1 • 3/12/2026

Documentation-only skill with no executable code. Static analysis detected 53 patterns from markdown code blocks and documentation links - all are false positives. The skill contains only markdown documentation with bash examples for using the external inference.sh CLI tool. No security risks, malicious intent, or prompt injection attempts detected.

1
Files scanned
185
Lines analyzed
0
findings
1
Total audits
No security issues found
Audited by: claude

Quality Score

38
Architecture
100
Maintainability
87
Content
22
Community
100
Security
83
Spec Compliance

What You Can Build

Marketing Visual Creation

Generate product images, social media graphics, and marketing posters with text overlays using the Pro model for professional typography

Concept Art and Design

Create concept art, illustrations, and design variations quickly using the fast model for rapid iteration and exploration

Image Editing and Compositing

Edit existing images by combining elements from multiple reference images, such as changing clothing, backgrounds, or poses

Try These Prompts

Basic Image Generation
Generate an image of [describe subject] in [style] style
Text-Heavy Poster Design
Create a poster with title '[main title]' in [color] [size] text, subtitle '[subtitle]' in [color] below. [Background description]. Use qwen-image-2-pro model with prompt_extend: false
Multi-Image Edit
Edit image to [desired change]. Use reference images: [Image 1 URL] shows [subject], [Image 2 URL] shows [element to apply]
Controlled Generation with Negative Prompt
Generate [description] with seed [number]. Negative prompt: [unwanted qualities]. Use [model] variant at [width]x[height] resolution

Best Practices

  • Use the Pro model (qwen-image-2-pro) for text-heavy designs requiring precise typography and professional text rendering
  • Set prompt_extend to false when you need exact control over the generated output and don't want the model to modify your prompt
  • Provide specific style descriptors like 'photorealistic', 'watercolor', or 'digital art' to achieve consistent aesthetic results

Avoid

  • Don't exceed 800 characters for your main prompt or 500 characters for negative prompts - the model will truncate them
  • Avoid requesting output dimensions larger than 2048x2048 total pixels, as the API will reject the request
  • Don't use the fast model (qwen-image-2) for complex text rendering - use the Pro model instead for professional typography

Frequently Asked Questions

Do I need to install anything to use this skill?
Yes, you need to install the inference.sh CLI tool. Run 'npx skills add inference-sh/skills@agent-tools' to get installation instructions. The skill itself just provides documentation for using this external tool.
What is the difference between qwen-image-2 and qwen-image-2-pro?
The standard qwen-image-2 model is faster and good for general image generation. The Pro model offers professional text rendering, fine-grained realism, and enhanced semantic adherence, making it better for posters and text-heavy designs.
Can I edit existing images with this tool?
Yes, you can provide 1-3 reference images to guide the editing process. Use the reference_images parameter in your input to specify which elements from each image should be combined or modified.
What image sizes can I generate?
You can generate images between 512x512 and 2048x2048 total pixels. Common sizes include 1024x1024 for square images, 1536x1024 for landscape, or 1024x1536 for portrait orientation.
How do I get consistent results across multiple generations?
Use the seed parameter with a specific integer value (0-2147483647). Running the same prompt with the same seed will produce identical output, allowing for reproducible results.
What is the negative prompt used for?
The negative prompt specifies content you want to avoid in the generated image, such as 'low resolution, blurry, deformed, oversaturated'. This helps improve output quality by explicitly excluding unwanted characteristics.

Developer Details

File structure

📄 SKILL.md