IT Italiano English

Laboratorio · where it starts

A radio voice in every call, without knowing anything about audio.

«I wanted my microphone to sound as if it went through a well set up broadcast chain, without having to learn anything about audio. In every call, live, on Linux and on macOS.»

Paolo MainardiCTO · tells Voice Strip
The Voice Strip faceplate with its VU meters
How I approached itThe advanced view: spectrum, loudness and the nine modules

«I built it largely with coding agents, steered and reviewed by me. The rules live in AGENTS.md: no allocations, no locks, no I/O on the audio thread, and every change closes with a measurable test.»

‘Sounds better’ is never an automated acceptance criterion.

AGENTS.md
What came outVoice Strip in the Omarchy panel

«Nine processing stages and a virtual microphone that Zoom, Meet or OBS see like any other. Everything stays on the computer: no account, no cloud.»

1.5 msof processing delay; neural noise removal adds 20 ms, only when you turn it on

Voice Strip.
Built with
Rust, coding agents
Runs on
Linux with PipeWire, macOS
Needs
A microphone
Status
Research project, Apache 2.0