[RamblingCTO]

I don't think so. Overfitting = the model was too closely aligned to the training data and can't generalize towards *unseen* data. I think it saw "silence" before, so it's not overfitting but just garbage in, garbage out.