TergeuPro 09 / 25
Speech-to-text & AI statement intake
Records speech, transcribes it with Whisper and turns it into a structured statement document.
Problem
Taking statements by hand is slow; recordings in Kazakh and Russian need reliable transcription.
What it does
- Records audio in the browser or accepts uploads
- Asynchronous transcription with task polling
- Detects Kazakh speech; language auto-detection
- AI extracts statement fields from the transcript; history of statements
How it works
- Upload → task id returned immediately, work continues in the background
- Whisper-based service returns text, language and a Kazakh flag
- Frontend polls status (2-minute processing timeout) and fills the statement form
Stack & models
WhisperYii2 / PHPReact + TypeScript