zlacker

Can they make that claim if SJ voice exists in the training data before fine tuning? We don’t know what they train on