Case Study — Desktop App
SAUTITYPE
Transcription on your own machine, private by design
The problem
Cloud transcription services charge per minute and require uploading private audio to someone else's servers — a real problem for interviews, lectures and client work.
What was built
A native Windows app that transcribes entirely on the user's own machine using Whisper: 99 languages with auto-detection, live microphone transcription, file and link import, karaoke-synced playback where clicking a word seeks the audio, low-confidence highlighting, and exports to TXT, SRT, VTT, DOCX, PDF and JSON.
Native Windows transcriber using on-device Whisper: 99 languages with auto-detect, live mic, karaoke-synced playback, click-word seek, low-confidence highlighting, exports to TXT, SRT, VTT, DOCX, PDF and JSON.
Challenges & solutions
Running capable speech models on ordinary laptops without a GPU server, and keeping very large files usable — solved by running everything on-device with a model manager that lets the user choose the speed and accuracy balance.
Outcome
A private, offline-capable transcription app: the audio never has to leave the machine.