qwen-image

What This Skill Does

The qwen-image skill is a professional-grade image generation utility for OpenClaw that leverages the power of Alibaba Cloud's Bailian AI models. It acts as an intelligent orchestrator, automatically selecting the most appropriate model—either Z-Image or Qwen-Image—based on your prompt's intent. If the system detects human subjects or photographic requirements, it defaults to the z-image-turbo model, which excels in skin texture, film grain, and realistic portraiture. For creative illustrations, complex typography, or artistic compositions, it routes the request to qwen-image-max. This seamless abstraction allows users to focus on creative expression rather than technical model selection.

Installation

To integrate this skill into your environment, use the OpenClaw package manager by running the following command in your terminal: clawhub install openclaw/skills/skills/irron/qwen-image-skill After installation, you must authenticate the agent by providing your Alibaba Cloud DashScope API key. You can set this globally by adding DASHSCOPE_API_KEY: sk-your-key-here to your TOOLS.md configuration file, or by exporting the environment variable DASHSCOPE_API_KEY in your shell profile. Ensure your API region matches your account configuration to prevent connectivity errors.

Use Cases

This skill is ideal for a variety of professional and creative tasks. Use it for generating social media assets that require high-fidelity text rendering in Chinese, creating consistent high-quality character portraits for digital storytelling, or producing photorealistic stock imagery with specific cinematic aesthetics. Its ability to render complex Chinese characters makes it particularly powerful for posters, signboards, and graphic design mockups. Developers can also utilize the advanced flags to bypass auto-selection for custom workflow requirements.

Example Prompts

1. "Generate a portrait of a young professional woman in a modern office, cinematic lighting, shallow depth of field, 4:5 aspect ratio."
2. "Create a stylized comic-book illustration of a cat scientist working in a laboratory, including a sign on the wall that says '智能实验室' in big bold characters."
3. "Design an atmospheric landscape of a misty mountain range at sunrise, using high-quality landscape photography style."

Tips & Limitations

• Prompt Engineering: While the skill supports prompt expansion, your initial prompt should be descriptive. Keep positive prompts under 800 characters and negative prompts under 500 characters.
• Watermarks: Use the --no-watermark flag to ensure cleaner outputs if the default configuration includes model signatures.
• Aspect Ratios: Always choose the resolution that matches your target platform (e.g., use 9:16 for mobile, 16:9 for presentations).
• Model Override: If you feel the auto-detection is not meeting your artistic needs, use the --model flag to manually lock in the engine of your choice.