Miaoyin AI Music Assistant

Problem Solved

AI Music connects natural-language song creation and post-production to the Miaoyin AI music service through an HTTP REST API. It is not a local inference tool or a web-automation wrapper; it is aimed at engineers who need to submit prompts, query tasks, and process audio assets inside an agent workflow. Typical cases include generating pop songs from style descriptions, continuing an existing clipId, re-styling a cover, creating video/MV output, exporting WAV, and separating vocal/instrumental stems.

How It Works

The skill uses https://ai.growingth.com/api as the fixed Base URL and authenticates with an api-token header backed by GROWINGTH_API_TOKEN. The main flow is asynchronous:

• Song generation: submit prompt, tags, and title, with optional model, make_instrumental, and negative_tags.
• Continuation: extend an existing track using audio_id / continue_clip_id, then merge the continuation.
• Cover and video: reuse the original clipId relationship for re-styling and MV generation.
• WAV and stems: submit conversion or separation tasks first, then query the result; stem modes include two, twelve, and dry.
• Status polling: retrieve audio_url, video_url, lyric, and image_url once the task completes.

It emphasizes standard HTTP calls only, error reporting without fabricated responses, unchanged lyrics and titles, and limited polling frequency to avoid agent-side request storms.

Boundaries

Best suited for teams that already have a Miaoyin AI token and want to embed music generation into a workflow. It is not suitable for local generation, offline inference, or bypassing API constraints. Preserve clipId as the context for continuation, cover, video, WAV, and stem operations, and expect asynchronous tasks to require waiting.