Tauri + Rust
Native desktop runtime, background behavior, local system integration, and Mac packaging.
Creator
Built by Khalid Irfan.
I built Mynah because voice typing should feel like a Mac utility, not another cloud subscription. Press Fn, speak, release, and your words should land where your cursor already is.
I use local AI every day, and I care deeply about keeping private thinking on my own machine. Mynah keeps the core dictation loop local: microphone, model, transcript, and diagnostics stay on the Mac.
The trigger was simple: I was tired of paying monthly for tools that meter words or move voice data through servers. Mynah is my answer to that friction: focused, local, fast, and built for everyday use.
If you want private Mac dictation without a word quota or cloud dependency, Mynah was built for you too. Try it free for 60 days; if you love it, buy before the trial ends and keep it for 60% of the $29 lifetime price.
Technology
Local stack. Named openly.
Svelte + Bun
Fast app UI and local-first settings surfaces.
whisper.cpp
On-device speech transcription using local GGML model files. No cloud API required.
CPAL
Cross-platform audio library used for low-latency microphone capture during recording.
CoreGraphics event tap
Raw macOS API used to capture the hardware Fn key globally without requiring keyboard shortcuts.
Enigo
Cross-platform input simulation used to paste transcribed text at the active cursor position.
Upstream
Built on open work.
Whispering
Mynah is adapted from Whispering, an open-source dictation app by Braden Wong and contributors at Epicenter. The core recording and transcription pipeline builds on their foundation.
What Mynah adds
Mac-first packaging, global Fn key capture via CoreGraphics FFI, a simplified single-user onboarding path, and local diagnostics without transcript logging.
License
MIT. The Whispering upstream project is MIT-licensed. Mynah's additions are also MIT-licensed at the source level.