the base model must be multi speakers data?how many speakers?

[Pydataman]

I use two speakers， but the result is bad！