Seedream 4.5 图像生成

What This Skill Does

The Seedream 4.5 Image Generation skill enables the OpenClaw AI agent to interface with ByteDance’s powerful Seedream 4.5 model. This skill supports both text-to-image synthesis and sophisticated image-to-image editing, allowing for high-fidelity visual content creation. Whether you are generating art from scratch with detailed prompts or performing complex edits—such as changing backgrounds, manipulating subjects, or modifying artistic styles—this tool provides a seamless workflow. It includes support for various aspect ratios, multi-image reference handling, and resolution presets like auto_2K and auto_4K.

Installation

To integrate this skill into your OpenClaw environment, execute the following command in your terminal: clawhub install openclaw/skills/skills/hexiaochun/seedream-image Ensure your configuration has access to the required API endpoints provided by the source repository.

Use Cases

• Creative Design: Generating concept art, illustrations, or marketing assets from descriptive natural language prompts.
• Image Editing: Transforming existing photos by swapping backgrounds, modifying lighting conditions, or applying stylistic changes using cross-image references.
• Rapid Prototyping: Creating high-resolution visual mockups for UI/UX or product design workflows without leaving your development environment.
• Personalized Media: Creating customized social media content based on user-provided source images.

Example Prompts

1. "Seedream, generate a high-quality, ultra-realistic photo of a futuristic coffee shop in Tokyo during a rainy night with neon signage reflected on the wet pavement, set the size to landscape 16:9."
2. "Use this image [URL] as reference: please replace the background with a serene mountain landscape at sunrise, maintaining the lighting consistency on the subject from the original photo."
3. "Create four variations of a sleek, minimalist desk setup with a mechanical keyboard and a high-end monitor, focus on a clean aesthetic with warm desk lamp lighting."

Tips & Limitations

• Reference Syntax: Use the 'Figure X' notation (e.g., 'Figure 1') to precisely reference images uploaded in the image_urls list for editing tasks.
• Resolution Control: Utilize the image_size parameter to match your output requirements, choosing auto_4K for high-resolution output or standard auto_2K for faster generation.
• Safety: The model features an integrated safety checker; verify that your prompts adhere to content policies to ensure successful processing.
• Queue Management: Always use the task_id returned by the submit_task call to poll for completion using get_task, as image generation can take time depending on server load.