I build Kekoso, a macOS app for on-device dictation and transcription, on my own: the app, the licensing backend and the website.
The app runs three families of speech models locally through one protocol, so the engine can be chosen per job: Parakeet TDT v3 and SenseVoice through FluidAudio, and Whisper through WhisperKit on the Neural Engine. Nothing is uploaded and dictation works with the network off. Written in Swift and SwiftUI, layered behind protocols, with pure decision logic kept apart from the hardware so it can be tested: 145 source files and 1444 tests.
Beyond dictation it transcribes audio and video files, YouTube links and recorded calls, exports SRT and VTT, and applies a user dictionary in post-processing. I also handle the release side: Developer ID signing, notarisation, Sparkle updates and a DMG.
The backend is a static Astro site plus Cloudflare Pages Functions over D1, issuing and validating licence keys with signed responses.
