In listening to the demos of MIR, one thing that I noticed that really calls attention to the fact that I'm listening to samples is the extremely low noise floor. There's no sense of having a number of people in a room (breathing, etc.). Has anyone dealt with that issue totally within MIR, or is must one take the output of MIR and then mix in some amount of "scoring noise" afterwards? If so, it would be a shame if the noise wasn't subjected to the same convolution.
Sorry if this has already been explored (and if so, please point me toward a link).