hackernewsaudio-processing
Build a singing-practice app after live pitch extraction falls short
At a glance
A developer uses Claude Code to build OnKey and changes the design when live singer-pitch detection fails. This workflow is listed under audio-processing, ai-assisted-development, prototyping, music.
- Agents / tools
- Claude Code, OnKey
Building a tool for these workflows? See weekly rates.
What they did
Claude Code first produced Python that displayed the user's pitch in a terminal. The developer then built a Mac app that listens to a song once, learns its vocal pitch track and scores later sing-along practice.
How it went
The developer released an open-source app and a singing demonstration after the real-time approach fell short.
Worth knowing
Noisy backing tracks and octave errors remain limitations. The release requires Apple Silicon and macOS 14.2 or later.
Save ideas in this browser. Upvotes count across the board.