THE PROBLEM
How it looked before
Voice dictation for smaller languages is poorly served by mainstream tools, and many organisations can't send audio containing client or patient information to third-party cloud services.
WHAT WE BUILT
The solution
- Desktop dictation application running speech-to-text on the local machine
- MLX-Whisper for transcription, local language models for processing
- No audio leaves the device at any point
WORTH KNOWING
This is the answer to the data residency objection. When data can't leave the country or go to a third party, this is a working example of AI deployed entirely on-premise.