Use Zeph as a microphone
Turn the device into a system audio input any app can pick — Zoom, Meet, or anything with a mic dropdown.
The companion adds a virtual microphone to macOS. Point your voice at it, then pick it as the input in Zoom, Meet, or any app with a microphone dropdown — and that app hears your device.
This carries live audio, not text. There is no transcription and no speech provider involved: the app on the other end gets your voice, the same as any other mic.
The path your voice takes
Device to app, as a real input
Set it up
Install the audio driver
Open Settings → Audio in the companion. If the virtual microphone is not installed yet, an install card sits at the top — installing it prompts for your macOS password once. After that the companion can offer virtual audio devices; the first one is named Zeph Audio, and you can rename it or add more.
Send your voice to it
Open a device's Glances, and in the Microphone section choose how your voice reaches the mic:
- Push to talk — hold the main button to talk, release to stop. You bind the hold to the virtual microphone in Buttons.
- Continuous — always listening. Speech is detected automatically, so silence doesn't stream.
- Both — always listening, and holding the main button still opens a clean turn.
For Continuous or Both, set Send my voice to the virtual microphone. A live level meter below the picker confirms audio is actually flowing.
Pick it in the other app
In Zoom, Meet, or anything with a microphone setting, choose the virtual microphone (Zeph Audio) as the input. That app now hears your device.
macOS only
The virtual microphone is a CoreAudio driver, so this feature is macOS-only today. The rest of the companion is portable, but the audio input is not.
How the app hears your device
The companion installs a CoreAudio plug-in that publishes a virtual audio device — a real macOS input, indistinguishable to other apps from a USB mic. The device's captured audio is cleaned up on your Mac and written into that device's ring buffer, which any app reading the input drains. Each virtual device is per-device and renameable, so you can run more than one and route them independently. Nothing here transcribes: the virtual microphone needs no speech provider — that's only for turning speech into text at your cursor.