So, here are some points I wonder about:
1. Will VRML 2.0 know about spatialized audio as in convolution? Crystal
River now has a commercial card that uses HRTFs (you can get 'em
personalized, even) to deliver spatialized audio through stereo
headphones. The HRTF (Head-Related Transfer Function) info is contained
within the soundfile itself -- not as a header; it's a process of delay and
equalization that is read as audio information, and you don't need fancy
hardware to decode it: just headphones. (And stereo, and preferably
16-bit, and at least 22050 would be nice....) This is a little ways in
to the future for most folks, but not all that far...
So here's a question: if you've got this radiusOfEffect what happens
when the file itself has HRTF info in it? Does this mean that convolved
audio and VRML will be incompatible? Maybe... if the soundfile's trying
to give you the impression it's coming from up left and rOEff is insisting
on down right....
2.As far as the distance effect (I think it's inverse-square you're talking
about, but I'm no physics teacher, either) -- I like it, I like it...but
something else happens to sounds when they come from far away, which is
that depending on the surroundings certain frequencies get reinforced and
others get attenuated. For example, city noise reads as a loud low
rumble, because the higher frequencies (aka shorter wavelengths) get
trapped between the buildings while the bigger wavelengths can bend
around them, and thus propagate farther. Across a lake, you get a
different set of frequencies that are enhanced; that's why you can hear
people talking from a mile or more away.
3. Also depending on surroundings you get echo/reverb effects. These are
a lot of fun, and I'd love to see VRML develop this capability to some
degree at least. A lot of soundfiles already have reverb/echoes built
in; again, one design question is how will such "canned" effects interact
with VRML's native sound capabilities?
4. Can one register one's own movement audibly? "Footsteps" or "flying"
sounds? Or interact with objects that have noises that belong to them? If
you throw the virtual glass into the virtual fireplace, does it shatter
with a satisying sound?
This means knowing where the edges of objects are and how
fast you're moving when you hit them, which I think you've already got.
This may start down the hairy path of physical modelling....or perhaps
there's an easier way: you've mentioned the "triggering" effect, which
could possibly get controlled by an "amplitude" message to produce a more
symbolic version of a sonically responsive environment...it's not real,
but it kind of sounds right anyhow...
5. It's tempting to use analogies with lights and cameras for sound. And
there are a lot of useful ones, too....but a note of caution: using
ray-tracing algorithms to make echos and to "know" how sound works might
get frustrating, due to the aforementioned fact that sound bends around
corners much more efficiently than light does -- and we rely on it to do
that. So if you're designing a background sound and suddenly you move
behind a sound-opaque object -- do you lose your background entirely, or
do you just get a kind of frequency-dependent attenuation, like you would
in reality?
It gets back to the perceptual level: the kind of information we
want from sound is different than the kind we want from vision. It's for
knowing about what we can't see, for understanding dynamic relationships
with what we can see (how hard does that glass hit the fireplace?) and for
constant contact with the environment around us. Of course we don't want
to be limited by how we've always done things...but it's useful to
remember this is how we "read" sound based on our everyday experience.
I very much like the "privacy" and radiusOfPerception suggestions... And
Gavin, about the looping -- If VRML loops sound files, I assume it will do
so externally to whatever loops the file (aiff, say) might have built into
it...? A loop is a lovely thing, however...
Apologies for the long post from a new voice -- but speaking as a sound
designer, these are some of the issues I've been wondering about for
virtual worlds in general and VRML in particular.
BTW, I'm very much looking forward to Mark P.'s talks here next week.
Cheers, MB
*********************************************************************
Maribeth Back
mbb@gsd.harvard.edu
Cambridge, MA
**********************************************************************