Vidu Image Generation

Problem Solved

The Vidu Image Generate skill addresses the repetitive work behind image generation calls: choosing a text-to-image or reference-based model, setting resolution and aspect-ratio parameters, selecting the correct API domain, polling an async task, and extracting the final accessible image URL. It turns these steps into a reusable workflow, so the caller only needs to describe the desired image or provide a reference image for generation or editing.

How It Works

• Task classification: plain text prompts use text-to-image generation; reference images trigger reference-based generation or image editing logic.
• Model selection: q3-fast is preferred by default for speed, quality, and special aspect ratios; q2-pro suits high-quality output, q2-fast suits fast generation, viduq2 supports reference generation and editing, and viduq1 is limited to reference-based generation.
• API invocation: it checks VIDU_API_KEY, selects api.vidu.cn or api.vidu.com based on user language, submits the generation task, and polls for status.
• Result return: after completion, it reads creations[0].url, stores images under {baseDir}/uploads/ by default, and returns the link as a Markdown file reference.

Boundaries and Notes

• A valid Vidu API key is required, and generated image URLs are generally valid for 24 hours, so downloads or saves should be handled promptly.
• Reference image limits differ: viduq2 accepts 0-7 images, while viduq1 requires 1-7 images.
• For image editing, set aspect_ratio to auto; oversized inputs, invalid keys, or failed tasks should be retried after reviewing the error message.