Sotto vs Hearsy: Local Mac Dictation Apps Compared

Sotto is a macOS dictation app that runs speech-to-text locally using WhisperKit and Parakeet. Hold a hotkey, speak, and release to insert text system-wide. It costs $49 one-time covering up to 3 Macs with lifetime updates.

Sotto offers a wide selection of transcription engines: WhisperKit variants (Tiny through Distil Large V3) running on Apple's Neural Engine, Parakeet v2 (English) and Parakeet v3 (Multilingual) locally, plus optional cloud transcription via OpenAI or Groq APIs.

Distinctive features include a custom vocabulary dictionary for specialized terms, audio file import and transcription, re-transcription of recordings with different models, per-language hotkey assignment, and automatic text cleanup rules (local, pattern-based). AI-powered functions for email drafting, code prompt optimization, and translation use cloud APIs (OpenAI or Groq).

The 3-Mac licensing at $49 is a standout — most competitors license to a single machine.

Hearsy is a menu-bar dictation app that runs entirely on your Mac. Press a global hotkey from any app, speak, and transcribed text is pasted at your cursor. No internet connection is used during transcription. Audio is processed in local RAM by one of two AI engines: Parakeet TDT for English (under 50ms latency) or Whisper Large V3 for 99 languages. Optional AI cleanup runs locally via Qwen 2.5 by default, with no API call required.

Sotto's local transcription via WhisperKit and Parakeet processes audio entirely on your Mac — nothing is transmitted. For core dictation, Sotto and Hearsy are equivalent on privacy.

The difference is in AI enhancement. Sotto's pattern-based cleanup rules (filler word removal, grammar fixes) run locally and offline. AI-powered functions — email drafting, code prompt optimization, translation — use cloud APIs (OpenAI or Groq) and require sending text to those providers.

Hearsy's local LLM (Qwen 2.5 via MLX) handles contextual AI cleanup on-device with no API call. For users who need the full pipeline — transcription plus AI formatting — to stay local, Hearsy provides that without cloud dependencies.

Both apps support Parakeet, so English latency is essentially the same — under 50ms on Apple Silicon. For most English-speaking users, raw transcription speed will feel identical between the two apps.

The Whisper implementations differ: Sotto uses WhisperKit (Neural Engine via Core ML), Hearsy uses whisper.cpp (GPU/CPU). Performance is comparable on Apple Silicon, but Sotto's Neural Engine path may be more power-efficient for sustained dictation on laptops.

Sotto's wider Whisper model selection (Tiny through Distil Large V3) lets you trade accuracy for speed on older or lower-RAM hardware.