A model for predicting audio quality and aesthetic scores, part of the AudioBox family for audio generation evaluation.