Your voice. Your sheet music.
Read, listen, and sing anywhere.

Documentation of an agentic AI music workflow and skill developed for and with Claude Code as a private code experiment. It is only a “proof of concept”. A choir score goes in, and an “AI-checked” ZIP comes out with 10 practice recordings for every voice.

See the example

4: stages per voice
2: sounds: choir and piano
2: versions: normal and slow
42: recordings in the example

01Workflow

Twelve steps, two stops with questions for you. Click a step below, its details appear here.

02Practice

Every voice gets its own mix. In each mix one voice sits in the middle, the others are spread across the stereo image like the choir stands. You pick a voice, a stage, a sound and a tempo. The headphones show the distribution, and you hear exactly that mix. When the files are created you choose the sound: MuseScore General or GeneralUser GS. It sounds best on headphones. The recordings are mechanically exact, without interpretation.

All My Ducklings

0:00

Soprano

Alto

Tenor

Bass

Tutti, all voices

Voice

Stage

Sound

Tempo

03Example

The result: practice recordings for every voice, guides and logs.

Download the example

The ZIP holds four folders. The voice ZIPs in 2_Practice are ready to hand out, their contents are listed last. Click a folder.

04Rights & AI

Rights

Before scores are processed, the skill asks about the rights. These details are then recorded in the README file. Whether the scores may be processed and passed on depends on the copyright and licence law they are subject to.

AI labelling

Everything the skill creates is labelled as AI content, both with the EU label for AI-generated content and in the text and metadata of the documents.

05Licences

PolyForm Noncommercial License 1.0.0, free for non-commercial use. Third-party components.

Your voice. Your sheet music.Read, listen, and sing anywhere.