The Sound Update Thread (The Return)

  • Thread starter TayeezSA
  • 1,178 comments
  • 140,465 views
2,534
South Africa
Johannesburg
Tayeedz1
This is pretty much a return of the Sound Update Thread question from the GT6 Kaz Q&A forum by @FrzGT , since that sub-forum is now closed but it is archived so you can go have a gander at that right here

So there have been a few developments as of late, Developments like: cars which are sounding quite fetching to the ears, sound sources that are viewpoint dependent, The Pitstop blog and even the experimental AES sound generation system which apparently splashes the tried and trusted sample methods out of the water. I'll leave it to @Griffith500 to explain the technical tid-bits

I guess one case in point for the AES system is the astounding LM55

Astounding I tells you

Now with the recent statement from Kaz stating that the PS4 Will Make An Incredible Difference , the question now on everyone's minds is that is it safe to say that we can expect an incredible difference in sound too? We were told that we'll get improved sounds in GT6 for the PS3 (Yes, they did get their improvement somewhat.... LM55 'nuff said) and that the big update is set for GT7 on the PS4.

Besides correct the vehicle samples and sounds, what more do you all think we can expect?
various Blow off valve sounds, turbo spools, VVTI engagements and most of all, induction noises?
Could PD possibly recreate something like this perhaps?


Also, how can some things actually be performed let alone make life easier?

Speaking of that, one example I thought of is how to record the various turbo and blow off noises. The problem there though is you want to record those items as a separate source from the engine so you don't get conflicting sounds. Now to make life easier for PD , why don't they just rig up a closed compressed air system similar to that of a turbo system in a car, run different turbos and blow off valves at different RPMs and capture the sounds for that. Let's say you upgrade the turbo or blow off valve on your vehicle in the game, the sound of it will change yes? PD can now record just that. In theory of course. Genius suggestion no?

Another thing that needs mentioning, small things can also make a big difference. We've seen in other games like Driveclub recently, that within the Audio options, you can adjust the sound volumes for various features like the environment, engine etc. Do you think maybe GT could do justice with that?
On the note of Driveclub, plenty was achieved in terms of sound, heck even DSG Farts


So what's the point of this thread then you may ask, share your thoughts on the Sound, maybe something has sparked in your brain and you wanna share it. Who knows, maybe PD could be looking at this.

To end off, I leave you with some tasty motor sounds, to expect of course


RULES

-The vacuum cleaner joke is tired and dead so can we please stop it.

- NO bashing, you may critique but not bash

- If you're going to post a sound from X game then post a real life counterpart as well to determine accuracy
 
Last edited:
First thing's first: somebody on CFW swap the LM55's sound onto the 787B, then go for a lap or three using H-pattern. :D
Oh, and most importantly, record it for the rest of us. :P

The automated gear changes in the game really need looking at.


A quick note: AES is also an encryption method, the very one that Sony uses on all its files (in a few different forms)... Irrespective, the game has this new sound synthesis technology in it - I could call it Geoff, and it wouldn't deny its existence.

Regarding the capabilities of PD's "AES", a demonstration would be better than my drawn-out explanations. ;)


So, to get things going, here's a nice video showing a fully synthetic engine sound.



This was created using "offline-simulation" (to create traditional samples, played back in the usual way), whereas PD have created something that works in real-time (hence the quality gap, for the time being).
 
So @Griffith500 , according to that vid you posted, essentially that should be a really good example of how AES and its flexibility should function? It's not only for exhaust and engine but also for induction noise too provided PD figures out how to tune it accordingly?

I'm thinking with the greater power of the PS4 over the PS3, we should see a smoother AES system and it being able to handle plenty of sounds. I suppose with that smoothing out we can get sounds like this:



So really what's holding PD back is staff and maybe time?
 
For the sake of reference as to what is possible without reinventing the wheel. Note, the video is less than a minute long and it's all sound, no video, so just for fun, turn on your surround system and play the video with your eyes closed and see if you can tell whether the first or second half of the video is game or real:

 
Last edited:
So @Griffith500 , according to that vid you posted, essentially that should be a really good example of how AES and its flexibility should function? It's not only for exhaust and engine but also for induction noise too provided PD figures out how to tune it accordingly?

I'm thinking with the greater power of the PS4 over the PS3, we should see a smoother AES system and it being able to handle plenty of sounds. I suppose with that smoothing out we can get sounds like this:

www.youtube.com/watch?v=hcFmjcW0mC0

So really what's holding PD back is staff and maybe time?

It's that performance thing, yes. I said back in the GT5P days that for sound improvement on PS3, PD needed to swap the memory requirement of samples for more processing power, i.e. to generate the detail instead.

They've got a lot to balance, and late sound "frames" can be a bit more distracting than late video frames, initially - especially if there's "nothing" (e.g. random discarded memory fragments; noise) to be "read" at the time it needs to be read. Coursemaker will be having an effect, because the new type (Ronda) needs a new renderer.

I dont know the status of PD's intake synth (it's definitely planned, the stuff is there alongside the exhaust configurationsalready used, it's just all zeroed out), but the mechanical engine sound is easily handled with additive synthesis plus shaped noise.

You can't rush programming; doing new stuff is basically a really stupid idea in the world of software development. :D
In addition, the nature of the subject of engine sounds is that you learn a lot of things as you work on it, how a lot of the "established wisdom" is a load of rubbish, and how much of the real info is out there in plain sight, if you could only "translate" it to suit your own ends! So you start out doing one thing, reasonably confident based on the "wisdom", and end up trying another seven as the reality of the phenomena reveals itself - and that's before you consider the difficulty of implementing it in code.

This feature seems to have been long enough in progress that it was initially planned for PS3... Hopefully it's in better shape for PS4.
For the sake of reference as to what is possible without reinventing the wheel. Note, the video is less than a minute long and it's all sound, no video, so just for fun, turn on your surround system and play the video with your eyes closed and see if you can tell whether the first or second half of the video is game or real:

www.youtube.com/watch?v=8DpvxmA4ifM
There's a lot more detail (subtle changes) in the real version, and the throttle transitions give the samples away the most. Still a very good effort, for one car, in one carefully reproduced situation - samples still have the drawback of only being able to accurately reproduce precisely what was captured, so you need to capture and store and process quite a lot these days.

But, what happens when you want to modify that car? That's effectively the same as having many more cars to produce content for. What if you also have a lot of cars, that nearly all of which can be modified? That careful attention to detail becomes much harder to maintain.

Why not, instead, use a method that is universal and, once set for the stock case (using as much detail as is available, the rest can be close enough in many cases), can give believeable changes to anything that is a control within the generation system?


The "old way" will persist for some time in many corners, because it's convenient, and thats fine. But the bigger players will follow GT's lead, in some form or other, eventually.

It's not accurate to call it "reinventing the wheel", implying it's superfluous effort. It's more like inventing the pneumatic tyre. ;)
 
There's a lot more detail (subtle changes) in the real version, and the throttle transitions give the samples away the most. Still a very good effort, for one car, in one carefully reproduced situation - samples still have the drawback of only being able to accurately reproduce precisely what was captured, so you need to capture and store and process quite a lot these days.

But, what happens when you want to modify that car? That's effectively the same as having many more cars to produce content for. What if you also have a lot of cars, that nearly all of which can be modified? That careful attention to detail becomes much harder to maintain.

Why not, instead, use a method that is universal and, once set for the stock case (using as much detail as is available, the rest can be close enough in many cases), can give believeable changes to anything that is a control within the generation system?


The "old way" will persist for some time in many corners, because it's convenient, and thats fine. But the bigger players will follow GT's lead, in some form or other, eventually.

It's not accurate to call it "reinventing the wheel", implying it's superfluous effort. It's more like inventing the pneumatic tyre. ;)
To my untrained ears, not listening to it as a sound technician, it sounds beautiful. In uncompressed high fidelity, with Dolby Surround, in game, surrounded by all the sights and sounds that accompany the exhuast/engine/intake notes, it's absolutely porn for the ears and every single untrained ear that has heard it live says they can't distinguish it from the real thing.

IMO the old way persists because it works and it's convincing. The new way to me is the easy way, develop a single program to cover everything, punch in a bunch of numbers and you have the exhaust/engine etc. sounds for unlimited numbers of cars. It remains to be seen how accurate it becomes, and to my again untrained ears, it needs a huge amount of work just to get to the level of sample based sound generation. IMO of course.
 
For the sake of reference as to what is possible without reinventing the wheel. Note, the video is less than a minute long and it's all sound, no video, so just for fun, turn on your surround system and play the video with your eyes closed and see if you can tell whether the first or second half of the video is game or real:



At first I thought the First half was in game. I usually can tell by the way gear changes sound and also just the fidelity of the sound. Very identical in that video. Well done I must say.

To my untrained ears, not listening to it as a sound technician, it sounds beautiful. In uncompressed high fidelity, with Dolby Surround, in game, surrounded by all the sights and sounds that accompany the exhuast/engine/intake notes, it's absolutely porn for the ears and every single untrained ear that has heard it live says they can't distinguish it from the real thing.

IMO the old way persists because it works and it's convincing. The new way to me is the easy way, develop a single program to cover everything, punch in a bunch of numbers and you have the exhaust/engine etc. sounds for unlimited numbers of cars. It remains to be seen how accurate it becomes, and to my again untrained ears, it needs a huge amount of work just to get to the level of sample based sound generation. IMO of course.

Man's got a point. The Untrained ears will hear the sample as beautiful and what works with minimal critique on the emission and other boffin bits. I won't lie though, the video Mr J posted is on point for a sample. The thing is with AES , while I think it is good and covers a large spectrum with promising possibilities, it seems now were only getting the crumbs of what's supposed to be the greatest cookie to ever eat.

@Griffith500 I'm trying to wrap my brain around everything and train my ears to a certain degree. I think once I hear it all together then it should make sense.

I like the fact that PD is venturing into new territories and at the same time taking a big risk. It could be the new standard in sound for all we know.

Another topic I wanted to bring up is gear change sounds, I think that's the term. In the current GT, it sounds like the cars have tiptronic systems and like the cars don't build up to a gear change. Particularly for the manual cars, it's like there's no pause in between the sound for gear changes. In GT5P and GT3 from what I remember it was there. Could it be a physics thing maybe?
 
Gear changes are 100% a physics / control issue. The sound engine just does as it's told; the moment you start adding fakery into the sound part to cover these physicalities, you start chasing your own tail and creating lots of special cases, when what you really need for 1000+ cars is (yes) a general model. Scipted physical gear changes fall into the category of "fakery", also, and technically exhibit the same "special cases" problem. That's maybe not a problem if you only have a handful of cars to worry about, and no modifications possible.

Using sound fakery also separates the audio feedback from the actual physical situation, adding a layer of garbled communication between the virtual car and the player. Some games fix that separation by forcing the physics to copy what the sound engine demands, but that's hardly appropriate in this case, where it should be the physics ruling all else.

So the gear changes might sound more involving etc., but to do so in this superficial way, you either have to make the cars harder / less responsive to drive (ambiguity of feedback) or you make the underlying simulation less physical. What's more important is a matter of opinion, and genre definition, but with the model you can at least get the better sounding gear changes with no detriment to the physicality or the feedback for a very large number of cars.

There are hints that PD are planning a drivetrain overhaul, as there are lists of transmission types, some of which aren't differentiated in the game as-is. The hope is that it implies a future physics change, which is greatly needed for the drivetrain modeling anyway. Once that happens, the physicality of gearchanges should be much better represented and differentiated per car.

It'll have the added benefit that PD can plug in the physical values of the cars (still) and everything should just work, with only fine tuning required, rather than hand-authoring what are basically special effects for every car (or having cars sound too similar as that unique treatment diminishes).

To my untrained ears, not listening to it as a sound technician, it sounds beautiful. In uncompressed high fidelity, with Dolby Surround, in game, surrounded by all the sights and sounds that accompany the exhuast/engine/intake notes, it's absolutely porn for the ears and every single untrained ear that has heard it live says they can't distinguish it from the real thing.

IMO the old way persists because it works and it's convincing. The new way to me is the easy way, develop a single program to cover everything, punch in a bunch of numbers and you have the exhaust/engine etc. sounds for unlimited numbers of cars. It remains to be seen how accurate it becomes, and to my again untrained ears, it needs a huge amount of work just to get to the level of sample based sound generation. IMO of course.
I never said it didn't sound good. ;)

The main advantage with using a model instead of hand-authored content is its sheer flexibility. You cannot use samples to represent differences in cam timing, unless you have those samples to hand. You can do it with the model, and 1:1 simulation sounds as good as the real thing, if you set it up that way (Sonory).

In time, real-time implementations will have the same level of perceived quality despite the truncations and approximations, but those compromises will stay because they use less computational power than the 1:1 simulation: just as it is with graphics, and the existing sample-based synthesis.

You seem to think this is some kind of aesthete vs. technocrat issue; it isn't. It's a content issue.
I'm a bit confused about you labeling real-time synthesis of physicaly accurate engine sounds as the "easy way", though. Surely just copy-pasting pre-made sounds (a.k.a. sampling) is the "easy way"? Unless you mean in terms of content generation (once you have the tools prepared, and the relevant data isolated). ;)
 
One thing they really need to work on is idle sounds and in between gear change sounds.

The Mazda video above still has issues with that (oh and it also sounds like it's echoing in an empty room). When it's downchanging there's no sound at all in the intermission.

Similarly for all cars at the start countdown they should be burbling away or whatever idle sound they make and there is none unless you rev the engine.
 
Kaz did say GT7 will bring major update in sound to me so will be interesting to see what they achieve. Hopefully they can pull off best in genre sound, like the way they are heading. They were working on it for a few years already back in 2013 so by now hopefully they achieved the breakthrough they were aiming for. I think the Vision GT unveiling videos give a glimpse off what new sound engine will sound like.
 
One thing they really need to work on is idle sounds and in between gear change sounds.

The Mazda video above still has issues with that (oh and it also sounds like it's echoing in an empty room). When it's downchanging there's no sound at all in the intermission.

Similarly for all cars at the start countdown they should be burbling away or whatever idle sound they make and there is none unless you rev the engine.
Idle sounds are there since forever, and in the current versions of GT6 are louder. You can hear them at their loudest during the start countdown at Goodwood when the camera gets really close to your machine.
 
So @Griffith500 , Can someone like @Johnnypenso expect fidelity like this? Closely related to samples



What I'm saying is, maybe this could be used for an AES example because it's taking separate sample sources like that and integrating/generating them into a unified sound no? Plus you hear the induction sound which PD so desires.
 
So @Griffith500 , Can someone like @Johnnypenso expect fidelity like this? Closely related to samples



What I'm saying is, maybe this could be used for an AES example because it's taking separate sample sources like that and integrating/generating them into a unified sound no? Plus you hear the induction sound which PD so desires.

Recordings.

There is the slightest hint of intake sound in the exhaust recording, especially noticeable at the VTEC transition. The exhaust can only really be heard off-throttle in the intake recordings. It is very well isolated nonetheless. The third source is the mechanical noise of the engine audible mainly in the intake recordings, off throttle. (Only three sources is simplifying slightly.)

PD's samples use the mechanical noise (usually off- / part-throttle engine recordings) as one source, and the exhaust recordings as a second source - in many cases, a strong hint of intake timbre can be heard in the exhaust and / or the engine samples.

The major difference is that PD are using these samples to reproduce the sound by volume changes alone. So the exhaust sound always has that intake timbre, it's a constant component, whereas it should be variable according to throttle position - it's not easy to separate that out (the most effective way is to use a model of the engine...)

The engine sound has the same problem, where there is intake bleed on those samples, and it does not respond in the proper way to throttle position.


Sampling

The "usual" way to handle on- and off-throttle is to use two banks of recordings, containing all the sources down-mixed as one in either case.

In the off-throttle case, there's little to no intake, the exhaust timbre is the off throttle timbre and the mechanical sound is at its most audible.
On-throttle, the intake is loudest, the exhaust has its on-throttle timbre and the mechanical noise is background noise.

This has the disadvantage that the spatial reproduction is poor (the balance of sources shifts slightly, and experience spectral changes also - you can't do that with down-mixed samples), and you usually need separate sample sets for interior and exterior, totalling four.


A different method might be to use those four separate sample sets in a different way: intake, exhaust on-throttle, exhaust off-throttle and engine. These can be placed in their physical locations and mixed accordingly, as well as separately controlled according to what the throttle is doing - it's possible to account for boost in this way as well, to some extent. But you now have four sources to mix, instead of two in the previous case (whether internal or external).

Whilst, for the down-mixed case, it is very difficult to filter the samples to create interior sounds from exterior sounds, with the sources separated in this way, it is much easier to apply the different spectral changes required for each source separately, much as it is for the external case (the internal view can be thought of as a special case of the spatial mixing problem, but certain "unimportant" sources in the external case become much more important for the internal viewpoint).

However, "filtering" the sources to apply the internal spatialisation results in sub-par results compared with dedicated interior recordings, because there is no relaible way to separate the external spatialisation from the samples first.


Enter simulation

The exhaust sound PD generates using AES has no spatialisation, it must be applied in real-time according to what's happening in the game. That means it is much more receptive to the external and internal spatialisation tricks. Currently, PD's interior spatialisation is primitive, but that needn't remain for PS4. The exterior spatialisation is genre-leading. They also only need to mix three sources (intake, exhaust, engine), because the variation is inherent to the generation method.


As for the fidelity, that can be separated in a number of ways. One relevant way is to separate the source and mixing fidelity. E.g. you can create a totally artificial sound, not meant to resemble anything, and "mix" it in a "high fidelity" way to make it seem like that sound really exists in the space presented visually. This has an under-appreciated immersive and interactive power, in my experience.

Another aspect is the source fidelity itself, but that can be seen to be implicitly tied to the spatial fidelity - because that's how we hear sounds in the real world. One particular facet of the source fidelity is its responsiveness to input, and that is always lacking with samples: they always sound the same. The model PD are already using has emergent properties, where each throttle transition can be different according to the history of said input. I call this "expressivity", in line with how synthetic musical instruments are described.


Summary

Put succinctly, the spatial mixing of sources in the game combined with the dynamic expressivity of the individual sources will alone provide a new kind of fidelity not experienced in a game. Whether that produces "photo-real" snapshots is kind of irrelevant in a sense, much as the original rotoscoped computer graphics animations weren't "photo-real" aside from their movement.

That said, all that is missing in PD's AES model of exhausts, relative to that S2000 recording, is the use of a non-random distortion method for the rasp, and a kind-of "smoothness" whose real-world source I ascribe to a combination of higher-order (high frequency) structural vibrations and non-linear acoustic effects. All of those things require a bit more processing power, but simply adding intake sound will complete the picture.
 
Yup, judging from what was posted above, sound isn't as easy as many claim it to be while I do have to say that it is long overdue though. One thing I really agree with @Griffith500 is that the exhaust needs some smoothness to it, one vehicle that needs it especially is the Lotus 97T and the Subaru sound on the VGT. In real life, they seem way softer than their AES counterparts though the sample I'd say is spot on.

Interesting to hear what Forza has done with the interior noise on the 2016 MX5



One thing that bothers me though, is the current interior sound in GT6. The way I imagine interior sound for let's say a modern car is that in the interior it should sound muffled. Like me trying to shout putting a cloth over my mouth. The road and wind noise is now audible but not in the way I expect it to be.

The reverb that's there already would rather suit a stripped out racecar. I feel it's too much now.
Driveclub I rate has the best interior sounds right now and audio controls for such are coming soon even.
Inspiration maybe?



I guess the challenge for interior maybe is to blend all the sources and keep them muffled at the same time.

I suppose since AES should become a general thing for the Sound Department at PD, instead of plugging in numbers to make the car sound, wouldn't they just plug whatever samples they recorded on the Dynapac and the AES replicates the samples, betters it and makes it live? I'm thinking that would be the more streamlined route

In other news, I quite like this April Fools vid :D

 
Yup, judging from what was posted above, sound isn't as easy as many claim it to be while I do have to say that it is long overdue though. One thing I really agree with @Griffith500 is that the exhaust needs some smoothness to it, one vehicle that needs it especially is the Lotus 97T and the Subaru sound on the VGT. In real life, they seem way softer than their AES counterparts though the sample I'd say is spot on.

Interesting to hear what Forza has done with the interior noise on the 2016 MX5
www.youtube.com/watch?v=cLMvpQgPZGk

One thing that bothers me though, is the current interior sound in GT6. The way I imagine interior sound for let's say a modern car is that in the interior it should sound muffled. Like me trying to shout putting a cloth over my mouth. The road and wind noise is now audible but not in the way I expect it to be.

The reverb that's there already would rather suit a stripped out racecar. I feel it's too much now.
Driveclub I rate has the best interior sounds right now and audio controls for such are coming soon even.
Inspiration maybe?
www.youtube.com/watch?v=Qm4ei8gaR9I

I guess the challenge for interior maybe is to blend all the sources and keep them muffled at the same time.

I suppose since AES should become a general thing for the Sound Department at PD, instead of plugging in numbers to make the car sound, wouldn't they just plug whatever samples they recorded on the Dynapac and the AES replicates the samples, betters it and makes it live? I'm thinking that would be the more streamlined route

In other news, I quite like this April Fools vid
www.youtube.com/watch?v=IznYvy2yJ0I
Forza Horizon 2 seems to still use the separate intake and exhaust sources approach, with a specific mix (no extra samples) for interior as an approximation of the actual interior sound. Notice it's mostly the raw intake samples that can be heard in the interior (compare to the views looking backwards onto the car from the front).
One thing that's odd is that they modulate the volume of the intake source on the rev-limiter, which is physically inaccurate (the throttles don't close).

Driveclub instead seems to use dedicated interior recordings, just like PGR4 did all those years ago.

At this point, it's clear you can't combine both approaches - i.e. this is still a necessary compromise on the current consoles.



The lack of smoothness in the 97T in particular is due to excessive use of what I (well, PD) called "phase noise" before - it has the highest level of this noise out of all of the AES cars, and it dominates its sound more than normal.
They don't use samples, though - that's the firing order and manifold model you can hear. ;)



PD's interior modeling is a basic "filter" at this point. However, in theory, you can reproduce any space by using what's called a convolution reverb (convolution is a mathematical process, and can be used as a general method of applying a filter, i.e. see "FIR filter").

The problem with convolution is that it's expensive to do in real time, even on parallel hardware (the PS3's SPUs, the PS4's GPU). That practically limits the length and quality of the impulse response that you can convolve with your source. That PD can't do it for interiors (impulse response decays in much less than a second) implies there's no room for such "realistic" filtering at all on PS3. It's not in FH2, either.

Using "filter primitives" (e.g. poles and zeros) to approximate an impulse response is a tricky matter, and the subject of a lot of research in many fields, from data transmission to image manipulation, never mind audio reverb. Automation of "fitting" such simplified filters to complex phenomena is of particular interest. I can't imagine PD would waste time creating content production problems in the form of temperamental specific tunings per car and potential instabilities in the filter structure itself, that may need ironing out and new presets re-making for all cars so far.

There are other methods available, that I won't go into, and there is one in particular that has certain "synergies" with the rest of their sound engine; namely the intake and exhaust models, and the "speed of sound model" (recently removed). It also comes in handy for the external reverb: whilst I had limited success with it as an overall scheme, it works well in conjunction with a traditional "algorithmic" reverb. I'm planning a little project in that vein for the original Doom...

It's possible PD are using a low-complexity example of this "synergistic approach" for their interiors, or they're layering in a different, cheap "algo verb" instead (in addition to the one they use for the track ambience and tunnels etc.) Again, this limited complexity can be expanded to something more detailed on PS4.



The automatic generation of synthesis parameters from recordings is certainly possible, but not without a degree of manual data entry first. For an idea of how that might work, read up a bit on "source-filter models" for voice synthesis. You have a similar simplified-filter fitting problem, as mentioned for reverbs, although convolution is always a possibility if you have the computational grunt. In fact, there is a cheap convo-like method I tried which can be used, but only with certain "source" generation methods.

The main sticking point, however, is the underlying model of the "source", which must be assumed first in order to determine what the filter must be to yield the final result. For speech, this source model varies little (e.g. glottal excitation) and can be predicted from a few quick checks and measurements due to the simple pulse shape and stable periodicity.

Engine sounds are not the same; their pulse shapes and apparent periodicities and harmonic content vary with rpm and according to cylinder conditions (affected by throttle position, fueling etc.). You can account for this with a physical model of the cylinders and their immediate plumbing, though, which you would have to feed with data obtained from the real engine. I once used such an approach to modify samples in real time to sound like a different firing order, manifold configuration etc., but there are parts that need separating out to not sound weird, and that's hard to do with samples.

Once a model is set up for a given engine, a quick "starting point" for the filter step can be obtained from data of the rest of the exhaust or intake tract, and it can be automatically refined towards the real recordings using a similar "linear prediction" method (in parts, so manual intervention will be required between each step, perhaps by using more sophisticated, hand-tuned models fit against a simpler model using automation), as used with the voice synths, when the game is made.

The model is really important if you want to change the parts on the car; if you just use naïve "best fit" approaches, you have no concept of the physicality of the system in order to be able to make meaningful changes to it. In effect, PD would be saving themselves work (required to provide a wide array of naïvely-produced "modified engine" presets) by doing the extra work in properly modeling the stock sounds. They could also create realistic sounds without recordings at all, by using the model, for special cases.



But that's only one approach to the synthesis step, which determines the content production step in particular ways. Whilst automation can be used, you cannot separate the human ear from the process, because of the fact that simplification is required (even with samples), and those simplifications need to sound good (in the way that RGB colour still looks "good", despite being an absolute fake).

So really it comes down to what kind of content and how much of it you want; I think we all want a game to tackle the "sound of engine modifications" problem in a convincing way sooner rather than later. A game with only one car could just have lots of hand-authored sample sets for each tuning state, but, for GT, a model-based approach makes the most sense given the huge array of cars.
 
If no changes are made at reveal. :sly:

2w315af.jpg


By me
 
I'm going to stumble into this thread and say I would prefer sounds that are balanced with the gameplay and don't detract from what I'm seeing on the screen. While the sounds in Forza have more depth and 'pop', they push the envelope just a smidge too far into arcade-y fluff. While I think PD can definitely improve their sounds, the overall balance between visual and audio working together is better.
 
I was watching that Mazda LM55 video.Once started, i couldn't stop listening.It sounds very good, although as
a rotary engine relevance, it sounds like a formula one as well.:boggled:

What about the Subaru VIZIV GT?
That's the only real sounding Subaru in GT6.
Now behold 9 minutes of the boxter engine soundgasm.:bowdown:
 
I too think that a sound overhaul for GT6 is long overdue. It's however a pretty different thing to make a mod for a game, than it is to make a whole game.

It's kind of like how you can find extremely well modeled character mods for Skyrim. They are very impressive, but the reason the game's own models are worse isn't really because the devs don't have the skills to do it. I consider skyrim to actually be a slight case of quantity over quality, at least when it comes to character models and to some extent environments, though, but I don't mean for this to off topic so I won't say more about that.

-edit-
You're right though. I was probably a bit of a dick in that previous post. I apologize.
 
Last edited:
@Griffith500 never failing to really go in with the explanations :P @Whistle Snap I believe uncovered what the AES parameters and options consist of. Intake parameters were something you(Griffith) brought up may be an issue, but does it seem that PD are close to a break through since it has appeared in the logs?

Another thing I noticed, particularly with the VW Golf Supersport VGTI is that when you start the car in a higher gear and you treat the throttle like an on/off switch, the exhaust actually pops and is audible but very subtle. Use of other parameters maybe?

Just on a side note, I suppose the 'smoothness' of sound for road cars is already present but since we've pretty much only heard AES on racecars, we're hearing only the really edgy and brutal sounds. One car I'd like to hear smoothly is the Supra RZ (Mk4)



The 2JZ-GTE has a really soft and smooth tone, even in person.
 
The intake "pathway", in terms of the mixing step etc. is just not present in the game in its current form (well, I haven't checked since before 1.15).

All we have are exhaust and engine channels in the mixer (plus others not related to engine sounds directly, including blowoff etc.); the parameters that demonstrate that are listed separately from the AES stuff. However, the AES bits and bobs clearly reference intake, and the "timing" aspect is part of the first step in the synthesis chain, so it's definitely intended for inclusion at some point.


The pops are, I think, a deliberate / allowed degradation in quality for performance reasons. The randomness implies numerical error, or something similar, and it's coming from the base generated sound, which affects the next stages in the synthesis chain. It does demonstrate that (for those that need such demonstration; it was already a given, for me) those next stages are responsive enough that, should PD get around to a proper "overrun" effect, it'll work very well with the system already in place.


Regarding the smoothness, the newest Golf VGT is a good car to use. But don't actually drive it anywhere. Hold the handbrake on and gently rev the engine, with minimal throttle input. Note how subtle and gradual the changes in timbre are across the rev range. Then gradually increase the throttle inputs and notice the changes: there's the effect that the rasp generator is having. It's best to use the "bumper" view for this.

If you try it whilst driving, the speeds where such small throttle inputs are workable result in the classic GT "concept car" drivetrain oscillation problem (present since GT2 at least; watch the speedo). That oscillation colours the sound too much, and the real sound of the synthesis "generator" step cannot properly be heard. You can probably do this with the other cars.
 
Every car sounds like it's being revved in an empty room currently. So much odd echoing.
That's the cabin reverb effect. It is a bit odd. It works better for the AES cars, but it still needs fleshing out a bit.

The interior mixing is quite tricky to do using "exterior" sounds (recorded or generated equally so), generally. You effectively need a separate "reverb" for each source, because they're located in different places on the car.

We can expect improvements to that aspect on PS4.
 
I love GT but this quote really worries me - the quality of GT6 sound is frankly rubbish. If they only make the sound great it will be by far the best racing game out there!!!
"Yamauchi further stated that the audio quality for Gran Turismo 6 is already good but they have made it even better with Gran Turismo 7"
 
I love GT but this quote really worries me - the quality of GT6 sound is frankly rubbish. If they only make the sound great it will be by far the best racing game out there!!!
"Yamauchi further stated that the audio quality for Gran Turismo 6 is already good but they have made it even better with Gran Turismo 7"
Out of context quote is out of context. Where did you get it?

What does he mean by quality? Quality generally does not refer to the creative / artistic aspects of sound production, and so is not really a subjective issue.

Kaz is right, the audio quality in GT6 is second to none. But the frozen artistic direction within that high quality framework is what has caused all the problems.
 
Back