voice

Like say, but with Kokoro TTS and Moonshine STT. A command-line speech tool for macOS, powered by MLX on Apple Silicon.

Install

Pre-built binary (recommended)

Install with cargo-binstall to get a pre-built binary — no compilation required:

# Install cargo-binstall if you don't have it
cargo install cargo-binstall

# Install voice
cargo binstall voice

Build from source

Requires Git LFS for embedded voice/model data:

# Install git-lfs if you don't have it
brew install git-lfs
git lfs install

# Clone and build
git clone https://github.com/rgbkrk/voice.git
cd voice
cargo install --path crates/voice-cli

Note: cargo install voice compiles from source on crates.io, but the Metal shader library path can break — see the main README for details. Use cargo binstall voice or build from a local clone instead.

Usage

# Just talk (backward compatible — no subcommand needed)
voice Hello world

# Text-to-speech with the say subcommand
voice say -v am_michael "How are you today?"
voice say -f script.txt -o output.wav
echo "Hello" | voice say
voice say --markdown -f post.mdx

# Speech-to-text from microphone
voice listen
voice listen --continuous

# Transcribe an audio file
voice transcribe recording.wav

# JSON-RPC 2.0 server on stdin/stdout
voice serve -v am_michael

Options

Top-level

Usage: voice [OPTIONS] [COMMAND] [TEXT]...

Commands:
  say         Speak text aloud (default when no subcommand given)
  listen      Record from microphone and transcribe (speech-to-text)
  transcribe  Transcribe a WAV audio file
  serve       Run as a JSON-RPC 2.0 server on stdin/stdout

Arguments:
  [TEXT]...  Text to speak (shorthand for `voice say <text>`)

Options:
  -q, --quiet  Suppress progress output
  -h, --help   Print help

voice say

Usage: voice say [OPTIONS] [TEXT]...

Options:
  -f, --input-file <FILE>        Read text from a file (use - for stdin)
      --phonemes <IPA>           Raw phoneme string (IPA)
  -v, --voice <VOICE>            Voice name [default: af_heart]
  -o, --output <PATH>            Write WAV to file instead of playing
  -s, --speed <SPEED>            Speech speed factor [default: 1.0]
      --markdown                 Strip markdown/MDX formatting before speaking
      --sub <WORD=REPLACEMENT>   Word substitution (repeatable)
      --sub-file <PATH>          Load substitutions from a file

voice listen

Usage: voice listen [OPTIONS]

Options:
      --continuous  Record and transcribe segments continuously

voice transcribe

Usage: voice transcribe <FILE>

Voices

American: af_heart, af_bella, af_nicole, af_sarah, af_sky, am_adam, am_michael

British: bf_emma, bf_isabella, bm_george, bm_lewis

See the full list in the main README.

License

MIT