Overview
Speakit is a local-first macOS dictation app I designed and built to make voice-to-text fast, private, and simple. Users hold a global shortcut, speak naturally, and Speakit transcribes directly into the active application.
Problem
Cloud-dependent dictation introduces tradeoffs across privacy, latency, cost, and workflow interruption.
Product approach
Speakit runs Whisper Small locally in a lightweight Tauri desktop shell. The product deliberately centers one workflow: press, speak, release, and keep working. I owned the concept, UX, architecture, AI-assisted development, local-model integration, desktop permissions, testing, and iteration.
Key lessons
For local AI products, the model can recede into the background. The experience is defined by latency, privacy, reliability, and how naturally the capability fits an existing workflow.
Stack
Tauri · TypeScript · HTML/CSS · Whisper Small · macOS system APIs · global hotkeys · clipboard and text injection