Problem

The existing virtual-mic options on macOS were either paid, abandoned, or shipped as kernel extensions that fought every OS update. I wanted a low-latency virtual input device for routing AI-generated audio into meetings — and a clean way to install and update it.

Approach

A small Swift daemon that registers a CoreAudio aggregate device, plus a Homebrew cask formula so installation is one command. No kext, no signed installer dialog, no third-party preference pane. The daemon runs under launchd; updates land via brew upgrade like any other tool.

Stack

Swift + CoreAudio for the audio path, launchd for lifecycle, GitHub Actions for cask updates on every tag. About 1,200 lines, tested on Apple Silicon from Ventura through Sequoia.

What shipped

Cask available since mid-2024. Used by a handful of podcasters and a research team running text-to-speech evals. Zero kernel panics reported, which is the bar I held myself to.

What’s next

A small companion menu-bar app for level monitoring. Optional — the daemon is the product, the UI is convenience.