Is the training data annotated with instructions like "(spoken)"?

[allo-]

Does the training data contain instructions like "(spoken)" or "(more quiet voice)" or is it pointless to try something like that in the lyrics?

[allo-]

In general, some more information about the training data annotation would be useful. I've seen in one of the new presets the tag "catchy". I wonder if the training data was annotated with it, or if it is like the stable diffusion "masterpiece" tag something that somehow changes the result (as any change in prompts does), but probably not for the reasons people think and possibly not in the direction they think.