tensorflow>=2.15
gradio>=5
facenet_pytorch
numpy
opencv-python-headless
mtcnn
moviepy
librosa
torch
torchvision
transformers
accelerate
soundfile
noisereduce