# Polyglot-Audio

> Polyglot-Audio is an open-source desktop application for bidirectional, low-latency, direct speech-to-speech translation between physical audio devices and applications with selectable audio inputs and outputs.

Canonical website: https://polyglot.vox.ad/
Source repository: https://github.com/dutz-ro/Polyglot-Audio
Downloads: https://github.com/dutz-ro/Polyglot-Audio/releases
Version described by this site: 0.1.1
Target operating systems: macOS, Windows, Linux
License: GNU GPL version 3 only, with separate commercial licensing available by written agreement

Android companion: Polyglot-RX
Android source: https://github.com/dutz-ro/Polyglot-Audio-Android
Android downloads: https://github.com/dutz-ro/Polyglot-Audio-Android/releases

## What The Application Does

- TX route: physical microphone -> direct speech translation provider -> virtual microphone -> voice application.
- RX route: voice application output -> virtual audio input -> direct speech translation provider -> physical headphones.
- TX and RX have separate provider sessions, audio streams, controls, buffers, and failure handling.
- The application supports compatible voice chat, conferencing, browser calling, streaming, and recording software with selectable audio devices.

## Important Technical Facts

- Polyglot-Audio requires providers with streaming audio input and streaming translated audio output.
- It does not silently substitute an STT -> text translation -> TTS pipeline.
- The audio engine is written in Rust and the desktop shell uses Tauri 2, React, and TypeScript.
- Audio callbacks do not perform networking, logging, locks, resampling, or expensive allocations.
- API credentials are stored in the operating system keychain and are not returned to the UI.
- Raw audio, transcripts, credentials, authorization headers, device names, and device identifiers are excluded from logs and diagnostics.
- A local direct audio-to-audio mock provider is available for route and failure testing without network access.

## Polyglot-RX For Android

- Polyglot-RX is an RX-only Android companion, not the bidirectional desktop application.
- It captures speech with the phone microphone and plays translated speech through connected wired or Bluetooth headphones.
- It uses direct streaming audio input and translated audio output, without an STT -> text translation -> TTS fallback.
- It does not provide TX, a virtual microphone, or desktop virtual-device routing.
- Signed ARM64 prerelease APKs are published at https://github.com/dutz-ro/Polyglot-Audio-Android/releases.

## Platform Routing

- macOS: two distinct, separately installed BlackHole routes.
- Windows: two distinct, separately installed VB-CABLE route families.
- Linux: separate TX and RX PipeWire or PulseAudio-compatible virtual routes.
- Polyglot-Audio does not bundle or install third-party virtual audio drivers.

## Accuracy And Limitations

- Version 0.1.1 is an early project release.
- A build that compiles on one operating system is not evidence of runtime support on another.
- Cloud translation depends on provider authorization, billing, model availability, terms, retention policy, and network access.
- Physical headphones are recommended to reduce feedback.
- Download buttons lead to the official GitHub Releases page; no binary URL should be inferred when no release artifact is published.

## Primary Sources

- Overview: https://github.com/dutz-ro/Polyglot-Audio/blob/main/README.md
- Architecture: https://github.com/dutz-ro/Polyglot-Audio/blob/main/docs/ARCHITECTURE.md
- Provider details: https://github.com/dutz-ro/Polyglot-Audio/blob/main/docs/providers.md
- Security model: https://github.com/dutz-ro/Polyglot-Audio/blob/main/docs/security.md
- Setup tutorials: https://github.com/dutz-ro/Polyglot-Audio#installation-tutorials
- Manual acceptance plan: https://github.com/dutz-ro/Polyglot-Audio/blob/main/docs/manual-test-plan.md

## Search Concepts

real-time voice translator; live audio translation; direct speech-to-speech translation; bidirectional voice translation; two-way speech translator; virtual microphone translator; translated microphone; multilingual voice chat; translate call audio; Romanian French voice translation; BlackHole audio translator; PipeWire voice translator; VB-CABLE speech translation; open-source desktop voice translator