Comment by tristanc

Use any model trained on the AudioSet dataset. There is one called EfficientAT i think that I use regularly and is pretty reliable