aria-segmentation 0.1.0

Speaker segmentation with pyannote models, for aria-kit
Documentation

Aria hears you, writes what you say, and reads text aloud. It's a voice assistant for macOS and Linux, with all speech processing on your machine. It also talks to your AI agents and keeps the conversation going. They get your words, never your audio.

aria dictate          # speak; the text lands in the focused app
aria ask              # speak a question; your agent answers
aria say "Done for today."
aria listen start     # transcribe system audio, with speakers

What it does

  • Dictate into any app on macOS, or to the clipboard on macOS and Linux.
  • Ask your AI agent by voice, and keep the conversation going.
  • Speak text aloud.
  • Wake on a word you choose, with your own model.
  • Listen to system audio on macOS and transcribe it, labeled by speaker.
  • Learn your project's vocabulary, to help spell its names right.

Get started

cargo build --release --bin aria
target/release/aria setup       # choose engines and fetch their models
target/release/aria doctor

A source build runs as target/release/aria in the examples above.

just app builds the macOS menu bar app with the daemon inside, into clients/build/mac/Build/Products/Release/. It needs Xcode and XcodeGen.

Engines

  • Speech recognition: Whisper, Parakeet, Moonshine, Nemotron, Parakeet MLX
  • Speech synthesis: Kokoro, Qwen3-TTS, system voice
  • Wake word: openWakeWord, livekit-wakeword

Toolkit

You can use these Rust crates in your own projects, without the Aria daemon.

  • aria-kit: audio capture, and the traits every engine implements
  • aria-stt: speech-recognition engines
  • aria-tts: speech-synthesis engines
  • aria-wake: wake-word detection
  • aria-speaker: speaker embeddings
  • aria-segmentation: speaker segmentation
  • aria-web: the dictation pipeline in the browser, through WebAssembly
  • aria-primitives: process, file, and socket building blocks

License