I think that a lot of posters here are missing the point - this is just the tip of the iceberg, and even simple improvements in prompting (infer the meaning of the lyrics, then generate footage related to that) would create a better result.

There is no going back from this, only forwards, for better or worse.