Open-source, fully native macOS voice-to-text app using local Parakeet TDT v3 models with zero network calls. Press-and-hold a hotkey to record, release to paste the transcript anywhere.
Personal hot take: paying a monthly subscription for cloud voice-to-text (Whisperflow et al.) feels like madness when Whisper v3 Turbo on Groq processes audio for fractions of a cent. Sotto is one example app that lets you bring your own Groq API key — one-time $50 (it was cheaper when I bought it), then ~1–2¢ per month of usage. Same idea, dramatically cheaper than a subscription long-run.
Hex goes one step further: fully open-source, fully native Swift, and uses local Parakeet TDT v3 (default) or WhisperKit — no network requests at all. Transcription quality is a hair below Whisper v3 Turbo on Groq, but it's good enough that I use it daily, and it's free.
The hot-mic feature is the small detail that sells it: instead of warming up the microphone every time you trigger the hotkey, the system keeps the mic hot, so the latency from press → recording is shaved down to milliseconds. Worth it when you're dictating a lot.
Tempted to open a PR adding Groq Whisper as an optional backend, but I need to look at the project's positioning first — it might be against the spirit of the app, which seems to lean hard into local-only.