I can only echo Chippy569's comments regarding pCARS. I don't particularly like the gear wobble effect, though - it seems too deep and too slow generally, but I think that's a limitation they've inherited in their engine, and is probably too deeply ingrained to remedy in one step.
I'm really pleased SMS are at the stage that their progress is tangible, perhaps the flak they're getting on their forums will subside now!
I just wish I could try it out myself (dead comp), I think I hear some interesting reverb...
I was very surprised at how reactive the highs were. I was going in with a view to putting a bit of ugly in to the sound when suddenly.... "hey, what are you getting up to over there?". I was not working with anything overly precisioned, but that was kind of the point in a way.
I just wonder if the PD team spent a whole lot of money on recordings that ended up being just too bare bones, and lifeless. I then go on to wonder if the issues with treble arose from trying to manufacture a sound stage that was not there in the recording. Treble can become very tempting when trying to add space. With the "twice cooked" samples, I was wondering about how they could make the most of a possibly "soulless" recording, while awaiting the complete re-do.
Sounds reasonable to me. This sort of voodoo is beyond me, frankly, I've always been scared of eq.s! I have heard that the highs add "air", whatever that means!
(I know, I should play).
I'd rather the space was added in the mixing, not cooked into the samples, but at least it's only subtle here; although that does mean the samples are transparent still, and betray their age.
Nice observations. 👍
To be honest, I am not up on all the ins and outs of how these recordings are put together, but certain concepts make sense to me. I do wonder about sampling at lowest and highest revs, and interpolating the difference. Rather than merely pitch shifting a base sound, the audio would be morphing from one known quantity to another. That may be primitive thinking though, I don't know. I was also wondering about how linear an accelerating car's sounds are, and if there are subtle interplays that are not being captured. I suppose that could be where fusing in granular components could come in.
Interpolation is really difficult. There is no physical basis for taking, say, one sample at 2000 rpm, another at 8000 rpm, pitch-bending them both to 4000 rpm, and blending them. That is guaranteed to be a sound that the real car never made.
The nature of pitch shifting is that it is indiscriminate, so that all frequencies are shifted, when in reality as the engine speed rises, only the harmonics shift (and their relative amplitudes shift slightly, too) - there is a lot of resonant colouration that stays roughly the same, or only switches mode at certain points etc., that is shifted, too, when it shouldn't be.
It's these shifted "formants", we may call them, that interfere when you blend samples together (and the harmonics are just interpolated, when perhaps in reality they should go through some complex pattern). It's very difficult to avoid this, and it's even harder to avoid the popping you get if you try to minimise the blending regions (and it can also increase the perception of the chipmunk effect from the pitch bending).
Using lots of samples mitigates this to some extent by minimising the pitch bending required and the timbral differences between separate samples, but that's not ideal either given how finely you have to do it, and it can become overwhelming from a control standpoint.
The Portishead example was just a fun way of talking about processes of adding character. I find processing chains fascinating. The ways in which doing a, b, and c in different orders can produce such dramatically different results. The sounds being not merely just the sum of their parts.
Absolutely, more inventive ways of using signal / processing chains can work for realistic sounds, as well as more artistic / abstract sounds. There are inklings of this sort of thing coming through in games now, which is exciting.
An option to demo the sound each of the exhausts makes before you purchase one would be nice, save having to buy each one in turn to find the best sound. Dream world would be a sample editor built into a custom exhaust designer, so you could morph the sound to your taste from any given engine type.
I think the exhaust (and intake) sound should be largely independent of the tuning applied, in the case of recorded samples. I.e., you should just be able to pick from a number of sounds according to the cylinder count etc., because they'll all more or less be achieveable on a given car without drastically changing the performance.
Ideally, you would indeed have a full model; I've been playing with my own version of such a system, based on a fairly low-fi synth, and it's surprisingly fun! It's really quite satisfying to just look at an engine configuration and be able to get the right kind of sound from the model by inputting the corresponding settings, or changing a few things to make it sound different, or making up unique configurations etc. I could see a community setting itself up around such a feature, as you would a livery editor, or the photomode.
But the sound is just sickening... especially going over 200 mph on those Group C cars..
. Gt 500 standerd Nsx's sound awsome but premium Nsx autobachs horrible. GT has good sound, PD needs to pay attention to those details in game.
