>Mike found that it was easier to take a modified recording and "dumb it down" than to "amp up" a stock recording; Forza stopped using stock recordings!!
Your little addendum at the end of this quote is a bit misconstrued
My "addenda" are in italics, everything else is paraphrased except where quotation marks are used, as best as I can remember.
If you're referring to the exclamation marks, it was long rumoured to be the case and this was the proof to settle it.
I also consider it further evidence of what I call the content problem, not being able to produce enough content to cover every eventuality, and this was also the implication Mike made when talking about representing tuning changes.
for low level (and especially turbocharged) cars (think like Fiesta ST) the sound you're going to get out of a dyno is just going to be some airy wind with no tone or excitement. In those cases, using a modified car will sound more "correct" and more importantly more "enjoyable" to the majority of people. However in the case of truly unique or already-aggressive cars (like ferrari F12 or like racecars) then the stock sound is already rich and exciting enough that it needs no modifications to convey the excitement.
In my experience, the disparity is due to missing sources and unnaturally rigid recordings, particularly the staticness of the coloration.
Missing sources in most games include the structure-borne sound of the car, aeroacoustic noise (both in terms of gas flow in and out of the car, as well as around it), and even the sound of your own body (with the baffles out of my motorbike, the base of my skull becomes a transducer and ear plugs are useless).
In GT, you can add intake to that list as well.
PGR4 sounded just fine with stock recordings, even with the more mundane turbocharged cars. That's the dedicated interior recordings (missing sources thus represented) doing their job.
Another thing that needs to be considered when discussing car sound reproduction is the volume disparity between a real car and your TV listening level. The 787b at 5' behind the exhaust (ie 3rd person camera position) is pushing 112 decibels. Your TV listening level if you're playing loud is probably around 75-80 decibels. Keeping in mind that dB is a logarithmic scale where for every increase of 6dB you've doubled the energy, the real car is something like 36-37 times more energetic than what you're going to be listening to. As a result a lot of racing game designers will introduce some audio effects to try to sell that excitement you're not getting from the quieter volume level - Turn 10 has experimented with real-time distortion for example - to make it feel more like the real experience of the car. Using a more aggressive version of a car can have the same effect. PD is obviously also experimenting with harmonic distortion as demonstrated by the Nissan GTR-LM video above.
This should already be covered with HDR mixing, which PD have implemented, and is also common in middleware. In fact, it's caused a few issues in the recent past with certain effects drowning out others in low dynamic range listening environments. So a more "designed" mix is needed for the low range case, and everyone then still gets the option of a more open mix instead of being limited by the assets. The large theatre mode in PS3 GT games is really something to behold, just in terms of its dynamic power and the affect that has on the sense of space.
The GTRLM, like the other AES cars, has a currently overdriven (
inharmonic) distortion effect based on a structural model of "rasp". That overdriving is most likely to compensate for missing sources, as above. The model should sound a little nicer when toned down, once the full model is rolled out.
When you're driving around in a racing game you shouldn't be asking "is this a recording of this exact actual car" but rather "does this experience feel like i'm driving this car" - because for most medium- to low-level cars those answers will be quite different.
Indeed, this is why I've fallen out of love with samples and prefer the interactivity of synthetic models, despite the usual drop in "photorealism".
The beauty of such models is that missing sources that are hard to record in a manner conducive to reproduction in an interactive medium can be generated based on the interaction directly. Take the exhaust flow noise PD has been generating since GT5P, which is coloured according to the situation in game. It wouldn't work so well with a recording, certainly not with the resources befitting a mere garnish.