Baiyin Vocal and Accompaniment Separation

Problem addressed

When vocals and instrumental parts are mixed, extracting the vocal stem or generating a clean accompaniment is not trivial. baiyin-track-separation-skill wraps the Baiyin Open Platform track separation API into an executable workflow: after receiving an audio file, it handles upload, task creation, status polling, and result delivery, reducing manual API calls and ad-hoc task checks.

How it works and limits

• Core capability: supports separationType: 1 for vocal/accompaniment separation, returning vocalTrackUrl, accompanyTrackUrl, and zipUrl for bulk download.
• Key steps: if BAIYIN_API_KEY is missing, it prompts the user to create one on Baiyin Open Platform; if the input is a local file, attachment, or other non-submittable source, the skill uploads it first to obtain data.url, then calls /api/open/v1/track-separation/split.
• Status checks: uses /api/open/v1/tasks/:taskId to observe queued, processing, succeeded, and failed, so incomplete tasks are not reported as final results.
• Notes: supports MP3, WAV, and FLAC; files are recommended under 100MB. Result links are kept for 7 days by default, so download them promptly. If balance is insufficient, the API returns INSUFFICIENT_BALANCE rather than a successful link.