I feel like music will be much easier then text. Highly complicated texts will be easier to spot inconsistencies, where with songs it's pretty common to have a high margin of errors/linguistic interpertation making it a much better candidate for current approach of generative ai.