Internal sound are amazing BUT the external sound are a bit of a let down... it was so amazing with the Judd beemer teaser and its awesome dopplers... on this track, the replay sounds are
quite boring, I don't know if it because of the cameras or the Aquila's engine, but it's disapointing... and where the hell are the dopplers? they don't sound quite as good as the beemer's replay video.
www.youtube.com/watch?v=4Stv-iHOm0E
That is a very promising step forward, they're actually updating the audio control input at a sensible rate, so there is none of the usual (e.g. in FM4, the Shifts, other Simbin games etc.) "stepping" as the inputs change rapidly (free revving, close fly-bys etc.). Just a shame that there's only one external source, because the looped samples are excellent.
That's the next thing to tackle, convincing sound propagation from the various sources on the car - as a proof of concept, I applied a very approximate representation of the frequency-dependent directivity of a speaker (the exhaust is approximately a circular plane-source, much like a speaker diaphragm), and that alone made quite a difference to the expressivity of the exhaust sound as the car approached and drove by - it also gives that nice phasing effect as you "walk around" the car.
GT5 has the directivity, but it's not frequency-dependent, so the sound just gets louder or quieter as the angle changes (although their excellent exhaust flow-noise thinger does phase slightly; it's partly that, combined with the tyre noise, and the high-resolution audio control that makes the fly-bys so good in GT5). Then there's other cool stuff like the way the car's body itself radiates sound, and there are ways to collect all of these effects into one technique.
One thing that can never be solved with ordinary sample-based synthesis is the stretching of frequencies that should be static (often called "formants") - these are things like the resonance of the intake manifold, or the harmonic rattling of an exhaust (its natural frequencies don't change, it just switches mode as the energy input changes) - these things should be held roughly constant (or modeled separately somehow, e.g. filters) whilst only those harmonics that are coming from the engine should be shifted as the revs change. In
this video, notice that the overall "envelope" of the sound stays roughly the same (showing the formants) as the pitch increases, which can be seen in the "spikes" (harmonics from the vocal cords) moving left-to-right - the last sound is simple pitch shifting, notice how it moves all frequencies indiscriminantly and as such it sounds unnatural.
There has been an attempt at this already with musical instruments, based on "spectral morphing". This is where you extract the highly-structured harmonic sound (e.g. of an oscillating reed, or embouchure) and separate it from all the noise and the (pitch-invariant) formants present in a recording, and then adaptively recombine them using some form of
additive synthesis. Additive is usually very expensive, though, so you have to be a bit clever - knowing the target sounds can help shortcut a few things at the expense of generality of application.
There's also
this, which uses "grains" (small snippets) of the sound of a car accelerating (and / or decelerating), and loops them according to the detected engine speed. It has the disadvantage at the moment of not sounding great at steady state (audible looping.)