5 of its AI music model. The cleaner and higher quality the recording, the less data is required. Slop yourself. It lets users train the vocal model on their own voice. Where previous updates focused mostly on improving fidelity and creating more natural vocals, v5.