Pet Personified Portrait Generation

Problem

In pet portrait generation, the hard part is not producing an image, but keeping the pet’s own traits: coat color and pattern, body proportions, facial features, and expression often drift when the style becomes dominant. With only a natural-language prompt, the caller also has to decide whether the source photo is usable, which features to extract, and how to write positive and negative prompts. This skill targets that workflow: it takes an uploaded pet photo, extracts structured features, builds image-to-image prompts, and renders an anthropomorphic portrait while preserving core pet characteristics.

How it works

The pipeline has four stages:
- Feature parsing and quality assessment: reads the pet image, identifies species, coat color and pattern, facial features, occlusion, and clarity, then outputs structured feature keywords.
- Prompt construction: combines the extracted features with the selected style_type to generate positive and negative prompts for image-to-image rendering.
- Asynchronous render submission: uses the original image as reference, applies the requested aspect ratio, starts an asynchronous render job, and returns only a task_id without blocking.
- Status polling: polls every 5 seconds with a 10-minute timeout; on success, returns the image URL, feature keywords, and prompts.

Typical input is a clear front-facing or half-body pet photo, style type, and aspect ratio. Typical output is an HD anthropomorphic portrait plus reusable feature keywords and prompts.

Boundaries and notes

pet_image and style_type are required. For better quality, use a clear, unobstructed front-facing or half-body photo. Rendering is strictly asynchronous, so the caller must poll until completion and surface each returned core payload to the user. Without an appKey, the skill only triggers a mock call for preview and does not produce a real generation; real calls require a configured ShowAPI AppKey.