Soul Painter

Problem

Converting a regular photo into a loose black-and-white sketch does not require photorealistic fidelity. The goal is to keep the subject’s expression, posture, and a few recognizable cues, then produce a humorous image for avatars, stickers, social posts, or creative visuals. This skill is intended for that kind of “soul painter” style output.

How It Works

The skill accepts a base_image, either a valid image URL or Base64 payload, and combines it with soul_level (childlike, lively, refined) and aspect_ratio values such as 1:1 or 16:9 to build a constrained prompt. It first uses a multimodal model to extract the subject’s action, expression, and 1–5 identifying visual features, prioritizing spirit over shape. It then submits an asynchronous reference-to-image task through ShowAPI and polls the returned task_id until an image URL is available. A simple HTML preview can also be rendered. If no showapi_appkey is configured, the skill enters mock mode and returns a one-shot preview result instead of a real generated image.

Limits

The input image must be accessible and suitable for playful reinterpretation. Real tasks are strictly asynchronous, and polling state should be surfaced to the user in real time. Mock output is preview-only and should not be treated as final generated content. The style is exaggerated and sketch-like, so it is not ideal for strict brand assets, high-detail realism, or sensitive formal material.