As both a listener and a player, I can confirm that the presence and
size of an audience makes a huge difference. It presumably depends
largely on the propotion of volume ( and surface area..) the audience
occupies in relation to the total. For a large cathedral, probably the
differences would be minor. An excessively reverberant and bright empty
church can become very pleasant indeed with a full house, and a pleasant
empty hall can become uncomfortably dry with an audience added. 

I do wonder about the use of impulse responses for this very reason
(excepting the obviously unothodox ones, such as drain pipes, etc!). The
decay time will be lowered, the hf absorption will increase, and
probably the echo density will decrease, while lf dispersion, if
significant, may even increase. The density changes are the ones I would
expect to be the most difficult to apply to an impulse response; I
suppose the others could be achieved by filtering and enveloping.
Perhaps that would be good enough in most cases.

The recording technique must also be a factor. Professional engineers
are adept at reducing excessive reverberation from a recording venue
(directional microphones, spot placement, even acoustic baffles, if it's
not a public concert), whereas for a true record of the impulse response
you would want at least omni-directional microphones to be used, if not
a full ambisonic Soundfield system.

Richard Dobson

Fred Floberg wrote:
> 
> John mentioned the particularly good sound of the cathedral at Smolensk
> when listening to a recorded performance of a choir there. If a person
> is trying to create an impulse response file in order to faithfully
> reproduce the qualities of a certain space, based on hearing a
> performance there, then shouldn't the presence of the audience be taken
> into account?
> 
> It would seem that the presence of several hundred people would modify the
> acoustic properties of the space quite a bit. That being the case, an
> impulse response file created in an empty hall might give unexpected
> (and possibly disappointing) results compared to what had been heard
> in a recording of a performance in that same hall.
> 
> Can this be accounted for after the fact in an impulse response recording?
>