Children's Idiom Audio Learning Cards

The Problem

Producing children’s idiom learning materials often involves adapting classical Chinese idioms that may contain adult themes, violence, or complex context. Manually rewriting the story, sourcing images, generating narration, and assembling a card is repetitive and hard to keep consistent across age groups and image ratios. This skill takes an idiom, age group, and optional image ratio as input, then generates a child-ready card with a safe story, image URL, TTS audio URL, and core concept notes.

How It Works

The workflow combines multimodal generation with async task orchestration:
- Intent analysis and safety review: extract core educational concepts, filter violent or unsuitable elements, and record safety transformations.
- Age-appropriate rewriting: rewrite the idiom into a Markdown story for the target age_group, and create TTS pacing guidance.
- Image prompt generation: build text-to-image prompts using story keywords, style constraints, and aspect ratio.
- Async generation: real runs submit a task and return a task_id; then poll for image and audio resources.
- Result assembly: output the card title, story, image URL, audio URL, core concepts, and safety transformation record, with optional HTML preview.

Boundaries and Notes

• idiom and age_group are required; tasks are asynchronous and not suitable for synchronous blocking workflows.
• Without showapi_appkey, the system returns simulated results for preview only, not real generated assets.
• If Chinese text rendering is poor, use text_rendering_fallback_suggestion to overlay text locally.
• During polling, surface interim results to the user instead of waiting only for the final JSON.