v0.5.0 - Parakeet Transcription Engine

What's New

Parakeet Transcription Engine

• Added FluidAudio Parakeet as alternative to WhisperKit
• Parakeet v2: ~110x realtime, 1.69% WER, English
• Parakeet v3: ~210x realtime, 1.8% WER, 25 languages
• Faster and more accurate than Whisper on benchmarks

Features

• Voice-to-Text: Cmd+Opt+Z (offline) or Cmd+Opt+X (Gemini cloud)
• Text-to-Speech: Cmd+Opt+S with Gemini Live streaming
• Screen Recording: Cmd+Opt+C with visual context transcription
• History: Cmd+Opt+A to view past transcriptions
• Paste Last: Cmd+Opt+V to re-paste last transcription

Settings

• Unified model selector showing all engines in one list
• Download models directly from Settings
• Engine preference persists across restarts

Requirements

• macOS 14.0+
• Gemini API key (for TTS and cloud transcription)
• ffmpeg (for screen recording)