# Belarusian speech recognition

This is a large-scale Belarusian automatic speech recognition model based on the FastConformer architecture, combining Transformer and CTC decoder loss, trained on 1,500 hours of Belarusian speech data.