Capture a conversation
Use a short recording from two speakers. The prototype transcribes speech and keeps each speaker's turns separate.
KBTU — Kazakh–British Translation Utility — explores a better way to practise English in real conversations. Record two speakers, review moments of hesitation or language switching, and turn feedback into the next speaking exercise.
Under development · No public app or hosted audio-processing server yet
“I think… hmm… machine learning could help.”
◷ Possible hesitation“Yes, but бізге more examples керек.”
↔ Possible language switch“Let me explain that another way.”
✓ Suggested practice phraseThe local research prototype combines speech transcription and speaker diarization with a review workflow for learners and teachers.
Use a short recording from two speakers. The prototype transcribes speech and keeps each speaker's turns separate.
See possible long pauses, fillers, and English–Kazakh switching. Listen again and accept, edit or reject model suggestions.
Turn reviewed moments into editable speaking prompts and export feedback for each student and the class.
This is an experimental learning tool. Automatic transcripts and detected events may be inaccurate, especially in mixed-language speech.
The working prototype runs on a restricted research server. Public processing, accounts, and a hosted demo are not available yet. This page is for project information and early-access interest only. No audio is uploaded here.
Source code: the application repository is private. There is no public source-code or installer link.
Leave your email to prepare an early-access request. You'll be able to review and send the request from your email application; we don't collect addresses on this static website.
No backend or mailing list is connected yet. Clicking opens an email draft; it does not automatically register you.