In this project, we will fine-tune the Facebook/wav2vec2-Xls-r-300m model on the Urdu subset of the common_voice dataset licensed under CC-0. This model architecture has enabled data scientists to develop state-of-the-art speech-to-text applications in over 80 languages.