Voice/Singing Conversion: ignored. Text-to-Speech: the text or lyrics to generate. Singing Editing: the edited lyrics. Singing Style Conversion: Input Audio's own transcript — content stays the same, only the style changes. Melody Control: the lyrics to sing.

Voice/Singing Conversion: not used. Text-to-Speech: transcript of Input Audio. Singing Editing: transcript of Input Audio. Singing Style Conversion: transcript of Reference Voice. Melody Control: transcript of Reference Voice.