Skip to content

Acknowledgements

Rex is built on the work of these open-source projects, models, and artists.

Libraries

  • WhisperKit by Argmax - on-device Whisper speech recognition for Apple silicon. MIT License.
  • FluidAudio by FluidInference - Core ML inference for Parakeet transcription, Silero voice activity detection, and speaker diarization. Apache License 2.0.
  • webrtc-audio-processing
  • the freedesktop.org/PulseAudio packaging of Google's WebRTC Audio Processing Module, which provides Rex's echo cancellation. BSD 3-Clause License. Bundles Abseil (Apache License 2.0).
  • Swift Argument Parser by Apple - command-line parsing for the rex CLI. Apache License 2.0.
  • TOMLKit by Jeff Lebrun - TOML parsing for config.toml, built on toml++. MIT License.

Models

Models are downloaded on first use and run entirely on your Mac.

  • Whisper by OpenAI - speech recognition models (Core ML conversions by Argmax). MIT License.
  • Parakeet TDT by NVIDIA - fast English and multilingual transcription (Core ML conversions by FluidInference). CC BY 4.0.
  • Silero VAD - voice activity detection. MIT License.
  • pyannote community-1 - speaker diarization (Core ML conversions by FluidInference).

Icon

The Rex dinosaur is based on the Dinosaur rex icon by Lorc from game-icons.net, licensed under CC BY 3.0, with modifications.

Documentation

This site is built with Zensical by the Material for MkDocs team.