The base model pretrained and fine-tuned on 960 hours of Librispeech on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz. For more information check Facebook's Wav2Vec2.
streamlit==1.9.0
streamlit_webrtc==0.37.0
torch==1.10.1
torchvision==0.11.2
torchaudio==0.10.1
SoundFile==0.10.3.post1
librosa==0.9.1
transformers==4.19.1