Installation

Voice Mirror is in alpha and under active development.

Voice Mirror is Windows-first today: the see-and-drive App Preview, push-to-talk, and native-app driving currently run on Windows. macOS and Linux support is planned but not available yet.

Windows Installer (recommended)

Download the latest installer from the releases page and run it. Installed builds auto-update to new stable releases.

The alpha installer is unsigned, so Windows SmartScreen will warn you on first run. Click More info → Run anyway. Code signing is planned.

Release channels

Channel	What you get	Where
Stable	Tagged releases with release notes; auto-updates	Latest release
Nightly	The dogfood channel — newest features, rougher edges	Nightly build

Nightly is what we run ourselves day to day. If something breaks on nightly, tell us in the Discord — that’s the point of the channel.

Build from Source

The project lives under the GitHub org contextmirror. The commands below assume the repo at github.com/contextmirror/voice-mirror — if the exact repo name differs, check the org page for the current one.

Prerequisites

Node.js (with npm)
Rust toolchain (install via rustup.rs)
System dependencies for Tauri 2 — see the Tauri prerequisites guide

No Python required. The entire voice pipeline (STT, TTS, VAD) runs natively in Rust.

Clone and Run

git clone https://github.com/contextmirror/voice-mirror
cd voice-mirror
npm install
npm run dev

The repo has a root-level src/ (Svelte 5 frontend) and src-tauri/ (Rust backend). npm run dev rebuilds the voice-mirror-mcp Rust binary, starts the Vite dev server, and launches the Tauri app with hot reload.

Build for Production

npm run build

This produces a platform-specific bundle in src-tauri/target/release/bundle/. Unlike the shipped installer, local builds aren’t code-signed and don’t auto-update.

Voice Engine

Voice features are built into the Rust backend — no separate installation is required.

Component	Default	Notes
STT	Whisper (local, whisper.cpp)	Model `base`; `tiny` / `base` / `small` downloadable
TTS	Kokoro (local ONNX)	Runs locally
TTS (fallback)	Edge TTS	Free Microsoft cloud voices; used if the local model isn’t present
VAD	Silero	Voice activity detection (ONNX)

AI Provider Setup

Voice Mirror connects to AI in three ways. You can switch providers without restarting, and API keys are auto-detected from your environment.

CLI Agents

Terminal-based agents that run inside Voice Mirror’s integrated terminal:

Claude Code — Anthropic’s CLI agent
OpenCode
Codex
Gemini CLI
Kimi CLI

Local API

Local model servers, auto-detected with no API key required:

Ollama
LM Studio
Jan

Cloud API Keys

Direct HTTP connections to cloud providers. Keys are picked up automatically from environment variables:

Variable	Provider
`ANTHROPIC_API_KEY`	Anthropic
`OPENAI_API_KEY`	OpenAI
`GOOGLE_API_KEY`	Gemini
`XAI_API_KEY`	Grok (xAI)
`GROQ_API_KEY`	Groq
`MISTRAL_API_KEY`	Mistral
`OPENROUTER_API_KEY`	OpenRouter
`DEEPSEEK_API_KEY`	DeepSeek

Configuration

Voice Mirror stores its configuration in the platform app data directory:

Windows: %APPDATA%\voice-mirror\
macOS: ~/Library/Application Support/voice-mirror/
Linux: ~/.config/voice-mirror/

All settings are accessible through the in-app Settings panel. Configuration is persisted via Tauri’s Rust backend with atomic writes.

Verify Installation

The floating orb appears on your desktop
Say “Hey Claude” — the orb pulses red when recording
Wait for the response — the orb pulses blue when speaking
Press Ctrl+Shift+V to expand the full workspace (editor, terminal, and App Preview)