Challenge
Speech transcription should be available in any application without context switching. The goal was to build a tool that runs in the background, quickly captures speech and inserts the result exactly where the user is writing.
Solution
The app runs as a Windows tray tool, uses a global hotkey for recording, Whisper for transcription, Vulkan acceleration for performance, and WebSocket streaming or chunked transcription modes depending on the scenario. It inserts the result at the current cursor position.
Outcome
- Fast speech transcription without leaving the current application.
- Flexible architecture for live and batch transcription.
- Practical example of connecting desktop UI, audio, STT and input automation.
