Looking for a model that can separate regular voice/speech from A cappella (solo vocal/unaccompanied singing)

[vahid9]

Hi,
so far I was able to remove music and chorus from a performance. But I also want to remove a cappella as well and leave with pure speech/declamation voice without even any solo singing.

It would be nice if the model could remove a cappella if they are mixed with foregraound regular voice simultaneously. Is is possible at all? If not, is it possible at least to detect and to separate the two if they appear at different time intervals (not over each other)?

Could any one suggest a solution?

Thanks.