A Push-to-talk voice relay tool for Windows and Linux. Speak into your mic, get clean synthesized audio back through a virtual microphone — in real time. Currently in very early state of in WIP.
EchoRelay transcribes what you say, fixes the grammar, runs it through a TTS engine, and routes the result to a virtual mic so other apps pick it up as if it were you talking normally without actually exposing your real voice.
- Push-to-talk capture — hold a hotkey, speak, release. That's the whole flow.
- Speech-to-text transcription — your voice gets converted to text
- Grammar correction — the transcription gets cleaned up before it hits the TTS
- TTS synthesis — corrected text is spoken back through your chosen engine
- Virtual microphone output — the audio routes to a virtual mic, usable in any app
Three engines are planned, each with a different tradeoff:
| Engine | Type | Notes |
|---|---|---|
| gTTS | Google Translate TTS | Lightweight, broad language support |
| Edge Neural TTS | Microsoft Edge neural voices | High quality, natural-sounding |
| Piper Voices TTS | Local neural TTS | Fully offline, low latency |
- Linux — PipeWire / PulseAudio virtual sink routing
- Windows — VB-Cable or similar virtual audio cable
Current state considered as work in progress. Core pipeline is being built out — transcription → correction → synthesis → routing. TTS engine switching and config will come once the base is stable.