Any guesses on why the Viziv GT sounds so darn good?

[ibnumgt1]

Because Subarus have engines that causes eargasm.

That's why.

 

[rialler]

@Griffith500 , you think this new method fully implemented will be better that one with proper samples and sounds? I mean, it seems PD is implementing a difficult method to sound "as good" as other racing games.

By now, the best sounding car, sounds like someone playing rfactor inside a church at full volume. Too much reverb (at least with good headphones).

I will be happy if results are near Dirt 3 and raceroom, which I think they play samples.

Why not use better recordings? its not necessary to have all the cars "re-recorded". Only with few, the game changes completely because with a good sounding car, you want to drive all the time.

 

[Griffith500]

@Griffith500 , you think this new method fully implemented will be better that one with proper samples and sounds? I mean, it seems PD is implementing a difficult method to sound "as good" as other racing games.

By now, the best sounding car, sounds like someone playing rfactor inside a church at full volume. Too much reverb (at least with good headphones).

I will be happy if results are near Dirt 3 and raceroom, which I think they play samples.

Why not use better recordings? its not necessary to have all the cars "re-recorded". Only with few, the game changes completely because with a good sounding car, you want to drive all the time.

Eventually, the quality gap between this style of synthesis and "traditional" sample-based synthesis will be closed. Not on PS3.

It is "difficult" to achieve (although no more so than the original research into digital sampling, 40 years ago!) but the payoff is well worth it. Its major benefit is in responding to even the smallest modification in a realistic way; with samples, you'd have to have separate recordings to account for that, and then there's the issue of how to select which sample when. There has been very little success with real time modification of a smaller subset of samples to account for these changes; the best way to do that is really to design a system like PD's, and then you don't need the samples anymore.

There is nothing wrong with the recordings PD have, it's just that the samples were made to PS1 specs. And sampling is flawed in many ways owing to their fixed nature, when engines are continuous in nature: over the rpm range, with throttle position, boost etc. Some of this continuousness might seem discrete, like when the exhaust pipe switches mode in its vibrations, but that doesn't occur in exactly the same place every time. Samples will cause that repetetive behaviour, whereas PD's method already shows lots of natural variation according to engine speed, ascending and descending and throttle input, on and off etc. Another issue is that in "blending" samples to patch that continuum with fixed content, you lose detail and nuance, and the resulting sound is not one the car actually made; the physical model doesn't blend, except where it is a convenient shortcut (like the new off-throttle tone on the Subaru); the difference is that all blending is time-domain, output-based for samples, whereas you can blend anything with the model, such as vibratory modes / formants, transfer functions, pulse shape etc. which don't always sound so obviously "blended", depending on how it's achieved.


In short, if you want to reproduce behaviour with samples, you have to record the behaviour, isolate the specific effects (somehow...) and then apply a convoluted, non-physical control basis that will be different per-car (trial and error tuning). With a physical model, the effects will be inherent (as long as your model incorporates its causes), and the behaviour will naturally extend to any car according to the physical parameters supplied. In practice, the physical model will still have non-physical aspects, due to approximations and shortcuts, but that will diminish as time goes on.

 

[GTboyz]

Is there any chance that GT7 will get major sound improvement and characteristics ? like rpm idle ?

 

[sparkytooth]

Is there any chance that GT7 will get major sound improvement and characteristics ? like rpm idle ?

If you're referring to idle noise, that's already there. It was greatly improved for 1.14 too.

 

[GTboyz]

If you're referring to idle noise, that's already there. It was greatly improved for 1.14 too.

i didnt know that Nice....This is why i need to buy the game.

My only hope then is that they improve the downshifting. It sounds way too fast

 

[SavageEvil]

All I know is the potential of a sound system that you can plug in figures and have it simulate engines is crazy. Next they have to actually record actual engine sounds for reference to ensure that they are on point with the engine sounds. This sounds like one crazy algorithm, I wonder who designed that program. Once they fine tune this thing, oh boy I have a feeling that we'll be in for a treat. I wonder what the program code looks like and I can only image the amount of parameter vectors they use when creating points for sound variation and vibration as well as pitch and modulation. I'm going to buy a 7.2 system just to accompany my eventual purchase of my PS4 in expectation of GT7.

Anyone besides me want to get down to the guts of PD's simulations for engines and sounds after hearing about this new apparently PD designed sound replication system. PD needs a second team, perhaps cloning since it seems that the guys who created the system are the same guys building the game.

The Viziv sounds good but there is some odd distortion so I guess that the system isn't fully refined as yet or it's just really intensive on the PS3's SPU's. I'm just tickled that this means no more large rack of sample sounds for anyone's information GT6 has over 30k in sound samples alone...yea that's absurd. Also this also allows for real time addition of variables like air temperature causing the engine to change sound slightly(assuming PD includes that). I'm just amazed at the things these guys are working on despite having one of the largest driving games on their production table.
As Griffith said seem like intake is missing, this could be an oversight or it might be downplayed, who knows but it is needed in order to get sounds right. I wonder if we ask PD about it will they divulge anything? Hey Griffith where'd you come upon this information, I would like to read up on it myself just to add to information in my head since I like learning things for the hell of it.

 

[Griffith500]

All I know is the potential of a sound system that you can plug in figures and have it simulate engines is crazy. Next they have to actually record actual engine sounds for reference to ensure that they are on point with the engine sounds. This sounds like one crazy algorithm, I wonder who designed that program.

You can buy at least one book that details a simulation approach from the physical principles, although the one I have glosses over the numerical computation part (which is an engineering core subject anyway, these days, and is the chief difference from an engineering solution and a game solution). The trick is shortcutting that simulation to get it to work in real time.

I've played with cheap dyno software that is probably based on such information (and a whole lot of original research, and data for correlation / accuracy), and its output can be coerced into producing decent sounds, albeit band-limited - that would be easy to fix with access to the source code. It's not suitable for a game application, though.

Once they fine tune this thing, oh boy I have a feeling that we'll be in for a treat. I wonder what the program code looks like and I can only image the amount of parameter vectors they use when creating points for sound variation and vibration as well as pitch and modulation. I'm going to buy a 7.2 system just to accompany my eventual purchase of my PS4 in expectation of GT7.

Hopefully they'll demonstrate the full-fat version on PS4 before you make that purchase, just in case!

Anyone besides me want to get down to the guts of PD's simulations for engines and sounds after hearing about this new apparently PD designed sound replication system. PD needs a second team, perhaps cloning since it seems that the guys who created the system are the same guys building the game.

There have been two "sound simulation" credits since GT5P (which raised my suspicions back then, and I've been crowing about the potential ever since). It's likely it is they who are working on the system, and will probably be DSP / physical modeling specialists.

The Viziv sounds good but there is some odd distortion so I guess that the system isn't fully refined as yet or it's just really intensive on the PS3's SPU's.

I think that's the main reason, it's deliberately low precision to save calculation costs. It's also compensating for the lack of intake.

I'm just tickled that this means no more large rack of sample sounds for anyone's information GT6 has over 30k in sound samples alone...yea that's absurd. Also this also allows for real time addition of variables like air temperature causing the engine to change sound slightly (assuming PD includes that). I'm just amazed at the things these guys are working on despite having one of the largest driving games on their production table.

It's actually the most sensible approach to take, given the car count is only set to climb, and we're used to customisation, as effective as it is or isn't at the moment. In order to expand that customisation further, this is really the only viable option no matter how many cars there are.
PD already have an exhaust temperature parameter; however it's the same for all cars I've seen, and I don't know why they've included it.
It has a few effects in real life, all of them subtle in a racing context.

As Griffith said seem like intake is missing, this could be an oversight or it might be downplayed, who knows but it is needed in order to get sounds right. I wonder if we ask PD about it will they divulge anything? Hey Griffith where'd you come upon this information, I would like to read up on it myself just to add to information in my head since I like learning things for the hell of it.

I expect PD will explain what they did, as a continuation of the blog articles at some point. The intake is a deliberate omission for the time being, I'd say, because the game has never had intake, and adding 16 extra sources with their computational cost of synthesis is probably not easy with the other changes still due (rendering for the course maker, AI for B-Spec).

I have slowly assembled knowledge from trying to do the exact same thing PD has achieved. For an insight into how you might approach the synthesis, even if not from physical simulation (e.g. old school additive synthesis), it might be best to start with engine configurations; it's where I started, at any rate. E.g. why does a boxer sound like a boxer? Or, where does the V8 burble come from?
From there, you just look into how the engine produces sound in the first place (gas dynamics), and what things (other than configuration) affect the final sound, and how.

Most of it is in the domain of automotive engineering, specifically engine design and research. Ricardo (as in Sir Harry Ricardo) have a few technical papers that might give an insight as to the depth of understanding out there, and it should be clear that it's already far in excess of what is required for a game. Their WAVE program is representative of the kind of software in use at most car / engine manufacturers, so the art is well developed already, including real-time applications.

 

[phosphor112]

The idea is that you provide data about the engine configuration and its state of tune, and the synthesiser does most of the rest to produce the base sound. From there, you specify the exhaust and intake configuration, etc. and tweak the silencing attributes using real world characteristics such as lengths and diameters of tubes etc.

It is best to refer to recordings in order to fine tune the bits that were approximated, because you can't rely on a 1:1 physical model due to the computational cost. So recordings are still necessary for reference, for stock sounds especially.

Where this system comes into its own is in its inherent ability to reproduce changes in sound according to physical changes made to the engine - because it is physically based in the first place. That means there are theoretically limitless variations to the sounds, within the bounds of the physical changes that can be made in the game. So, cam changes, exhaust and intake changes etc. mainly.

The trouble is, PD can't let us change everything. They'll have to provide an interface for whatever customisation they choose to expose, or else sick to the presets approach.

Any word on how accurate it is? You do say they need recording still to make sure it sounds correct, but mostly it should generate it itself?

Anyway, I can't stop using the Viziv. It sounds so much better than all the other cars. The cockpit view sounds like a recording from an onboard camera. Not only that, but even though this is a fully "synth" setup, it sounds less synthetic than the recordings (in my opinion).

Also, do you work for PD? Just wondering.

 

[RedBaron]

Nothing from PD is synthetised yet. Still samples, this time few too distorted and few really right chosen are doing their job pretty well on Viziv sounds. I don't thing distortion is the way, the sample quality and quantity and good transitions between them is the way. Samples for opened and closed intake, just the sound after shifting that Forza uses and GT is just attacking volume knob for a short while is making whole world of a difference. Everyone could hear it.

 

[Griffith500]

Any word on how accurate it is? You do say they need recording still to make sure it sounds correct, but mostly it should generate it itself?

Anyway, I can't stop using the Viziv. It sounds so much better than all the other cars. The cockpit view sounds like a recording from an onboard camera. Not only that, but even though this is a fully "synth" setup, it sounds less synthetic than the recordings (in my opinion).

Also, do you work for PD? Just wondering.

The accuracy won't be apparent until it's the complete version. The accuracy of synthetic sounds is not really in question, see sonory. They generate samples for use in the normal way, but they're fully synthesised from scratch. The difference is doing that in real time, which is where the quality suffers, but that might not impact the accuracy too much.

The Subaru works well with the sound update because the update was intended for the sound method it uses, not the samples. When you generate sounds, there's no spatial signature to it yet. So you have to add it; if you try to add that same spatial signature to a sound that already has one, it just sounds like a bad reverb. The spatial signature helps with sound location, and a sense of space that is mostly subconscious. Recordings inherently have a signature according to mic placement, and the environment.

I don't work for PD; if I did, there's no way I'd be this chatty about it all!

Nothing from PD is synthetised yet. Still samples, this time few too distorted and few really right chosen are doing their job pretty well on Viziv sounds. I don't thing distortion is the way, the sample quality and quantity and good transitions between them is the way. Samples for opened and closed intake, just the sound after shifting that Forza uses and GT is just attacking volume knob for a short while is making whole world of a difference. Everyone could hear it.

The Subaru does not use samples. Save a best lap replay and view it in the datalogger - those are the samples attributed to it. No amount of distortion can create the change of timbre from those samples to what you hear when driving it, never mind the expanded expressiveness of the AES method generally.

 

[SavageEvil]

@Griffith500 Thanks for the links, especially the acoustics Wave site. This is impressive stuff and I know PD tends to get flak for weird things or what we call glaring oversights but it seems that these guys are looking at a much larger picture than we are. There is quite a bit of simulation going on in GT5 and 6, it's almost as if they are trying to simulate every aspect of the game. The weather simulation was impressive when I first took to the track early morning and it was slightly foggy, boggled my mind seeing that in a video game, then came the different surface temperatures which affected grip and changed dynamically as time progressed, then we had them simulating tire smoke(seemed a bit superfluous from the outside) and here we have them simulating engines and exhaust sounds. I am deeply impressed by the obvious geniuses at PD, who are going on about their own business of who knows what and churn out a video game at random points in their work schedule. This is what it seems to me. GT5 and 6 have a lot of things that make you scratch your head but then you get glimpse of implementation of some that is like a wake up dose of cold water. So they weren't sitting around twiddling their fingers, just busy.

It's a sad thing the PS3 can't handle this AES but it definitely has promise, I'm guessing that's probably why Kaz was so liberal with what they can achieve on the PS4 with all that RAM. I wonder if PD will use GPGPU processing for GT7 on PS4 to really spread the power around considering that CPU can bottleneck the system well before the GPU or Bandwidth. That's a whole other topic, but I am optimistic about this AES that PD has and I am eagerly awaiting the obvious incoming PD video of GT running on PS4.
The Boxer engine piece was excellent and really points out why it sounds different on V Dubs and on Subarus. So much readily available information on how engine sound even identical engine types can change the engine sound and exhaust sound. This AES has me absolutely ecstatic about having authentic sounds, even upgrades sounding authentic. I am grinning ear to ear typing this while thinking about the impact of this system. PD, these guys are serious about simulation. I wonder what else they are toiling away at over there that we haven't had the chance to experience?

 

[Griffith500]

Yeah, I think Kaz said a while back their new approach is to "simulate everything". It's just taking a while for that to become a reality.

The flexibility of AES with modifications is going to be the real trick, but that won't be seen until PS4, I think.

PD has made great use of the SPUs, which are superficially similar to stream processors in their small cache, shared store and reduced instruction set, so they'll definitely get in on the GPGPU action with that experience.

And the bigger picture is exactly what I try to bear in mind with PD, frustrating as waiting and their decisions made with "less important things" can be.

 

[SavageEvil]

I just had a thought and what if PD are attempting to build microcosm? Simulate everything that is a bit grand in the scope of things but if they tie this all together and pull it off...imagine the possibilities for virtual reality. Imagine if PD were on PC doing this...mind boggles.

Anyway I enjoy the sounds of the Redbull Jr, they were literally eargasms and the Viziv sounds like it has the very same engine size but since it's a boxer not sure about the firing sequence but that weird noise at the end is a touch annoying but I appreciate this even more knowing more about exactly what is going on at PD. PD should come out and say something to at least quiet the rumblings from the more vocal pundits but then that would be like making an excuse.

I have at times been frustrated with PD at what seemed like a bullheaded direction they were headed in but what a difference a little bit of information makes. I am completely fine with them just molding this game into something before GT7 goes gold, I accept the shortcomings in GT6 and enjoy the things it gets right...now if they would just fix the aero feels like cars without underbody added seem to have the slipperiest Cd in the world and cars with built in underbody someone don't get penalized enough PP points as cars that have to add them.

GPGPU means that GT7 frame rate issues because of CPU stalling, oh man 1080p goodness at a locked 60fps. Mmm sounds delicious.

 

[IneedSchool]

From what I've heard, this is how the Subaru sounds so good:

-There's a debate on here whether the sounds were either created from a keyboard or that dude from the Police Academy movies.

-The sound of the transmission whine is more quiet so that's a plus.

-The "echo" of the sound in cockpit view has been inspired from Pope Benedict XVI revving his Enzo in the Vatican church.

-Remember the large amounts of resumes PD have been recieving since Kaz said they're hiring sound engineers? I think it's working eh.

 

[Connoissuer]