openvino>=2024.4.0
huggingface_hub
torch>=2.1
gradio>=4.19
librosa==0.9.2
opencv-contrib-python
opencv-python
IPython
tqdm
numba
numpy

openai-whisper
sounddevice 
scipy

transformers>=4.35
torchvision>=0.18.1
onnx>=1.16.1
optimum-intel @ git+https://github.com/huggingface/optimum-intel.git
openvino
openvino-tokenizers
openvino-genai
datasets
soundfile>=0.12
python-ffmpeg<=1.0.16
nncf>=2.13.0
jiwer

gtts