About this skill

Problem

Calling image-generation APIs can stall on ambiguous parameters, local reference images, and task polling. This skill wraps the Baiyin open-platform image task flow with stricter model, parameter, and upload rules.

How It Works

Fixed scope: creates and queries image generation tasks only; it does not handle video or digital-human tasks.
Fixed models: nano-banana-2 and kling_v3_omni_image; if the user does not choose one, it guides selection between these two only.
Required inputs: prompt, resolution (1 / 2 / 4), and aspect_ratio, with model-specific allowed ratios.
Reference images: multi_image accepts only an array of public image URLs, up to 3 items; local paths are uploaded through the Baiyin file-upload endpoint and replaced with data.url.
Task results: returns taskId and status after creation, then polls for imageUrl or the full images array when available.

Boundaries

It requires BAIYIN_API_KEY and uses the fixed Baiyin open-platform BASE_URL. It will not create a task until required parameters are confirmed, and it will not query extra model lists. Errors such as 400, 401, 404, or failed should be handled by correcting parameters, API credentials, task ownership, or backend errors.

Screenshots

Use Cases

Confirm model, resolution, and aspect ratio before creating a poster image task.
Upload a local product image to a public URL before submitting it as multi_image.
Poll the created task and return the final imageUrl once it succeeds.
Validate up to 3 reference URLs and reject base64 or local paths.

Best For

Backend engineers calling Baiyin image APIs: create and poll image tasks consistently.
Design-collaboration engineers making posters or product images: collect model, resolution, and ratio.
Agent developers integrating multimodal workflows: upload local reference images to public URLs first.
Integration-test engineers handling task polling: query task status and extract imageUrl.