Quote:
His speech is very loud, pressing the peak level, and his music is very flat, not very loud.
For this, I believe you would want to use the BIMP -- Built-in Microphone Processor addon.
This allows you to use Input 2 as a separately processed (microphone) input that goes over the existing (music) programming.
You may want to set the BIMP to use the voice processing you prefer, and the rest of Stereo Tool to process the music (more flatly or quietly). Then you can choose how much louder to inject the BIMP ahead of the standard processing.
I hope this gives you some ideas.