Since MusicGen cannot synthesize vocals and this model was fine-tuned with songs of NewJeans with vocal track unseparated, glitchy vocal artifacts might be included in output tracks.