As mentioned in the readme, the MIDI emulation is terribly slow. On the other hand, when I use the PC speaker emulation for output and decrease the sample frequency to 8kHz (has to be done in the AHI settings!!!) I got pretty much everything: music, sfx, speech and a decent frame rate.
Then I spotted another tricky point: When you change the ScummVM global settings AFTER you've added some games, the game specific settings tend to switch to override mode which might mess up you config (especially audio). So better double-check everything after you changed anything.
Btw. my hardware is: A1200 AGA, B1260@50MHz, 128MB FastRAM, Paula-audio