Give us better sounds - PLEASE !!

[sparkytooth]

One of my favorite cars ever in GT5 sounds wrong.

Here it is in GT5 (Stock, buying another exhaust makes it worse ):



Here it is driven by Tiff Needell around Rockingham Motor Speedway:



I really hope that "Magic" GT6 patch fixes this thing, the 787B (The most brutal racecar sound I've heard) and countless others /)_-

 

[Chippy569]

Interesting. I always thought that was a conscious design decision, to make the sounds be raspy and not so static.

Here's the first tutorial vid I could find that had actual examples on what phasing sounds like (you can skip to 2:44 if you want):

 

[RACECAR]

I asked you this in another thread but you never answered. Where do you work?




I'm sorry, but the American V8 cars are the most disappointing of all the cars in gt5.

Here's a good example. GT5 Corvette GT2 ("RM"). Pay close attention to when the view flips to "replay camera" at about 1:41.


The first thing I hear is some really, really bad phasing. Phasing happens when you play the same sound twice, and one is slightly delayed compared to the other, which gives a comb filtering effect that sounds "swirly" or "wooshy" in an unnatural way. I noticed it the worst at 1:47 but it's all over the place in external camera. Phasing is also the cause of the sound that's a bit like a jet whipping by at ~2:39.
Second, ask yourself how aggressive that feels, compared to the sounds of like the tires. Part of it is a mix/balance issue, sure, but right at 1:53 for example, the car slides, the tires squeal a bit, and then the engine comes on full-power, but it's quieter than the tires just were.

Here's a good reference for the real car. Check out the shot starting at :32.


To my ear, I hear that the real car has a lot more "tone" - more bassy mids and no where near the amount of airy high-frequency sounds the GT clip does. The shifts are vastly more dramatic in the real car, too. you can hear the direct sound of the car for a lot longer distance-wise than in the GT5 clip as well.


And then... on-board. The kind of microphones you get with "on-board" clips are less-than-desirable, so let's listen to a handful of clips.

Gopro on the driver's helmet (skip to 5:00):


Unknown driver-mounted camera (the saturation in this clip sounds really good):


Lemans on-board:


Those three clips all sound rather different so it's amazing to think they're all the same car (differences in year notwithstanding). Which one is the most accurate? That's a silly question, they're all the exact car. Which one is the most exciting? To me I'd vote #2.

Pretty sure the Corvette video (the one at Monza) is the GT1 car (not sure why the user calls it the GT2 car when theres a difference).

 

[Johnnypenso]

Those three clips all sound rather different so it's amazing to think they're all the same car (differences in year notwithstanding). Which one is the most accurate? That's a silly question, they're all the exact car. Which one is the most exciting? To me I'd vote #2.

2 is the best to me. A lot of the difference is in the position of the mic and the quality of the recording though..

 

[Eager Snake]

Chippy569, you are certaintly right in some of your points, but as Johnnypenso said, it's a matter of quality in the recordings and, of course, the position of the mic in those videos as well. Your point to the onboard videos is also correct, they are all different, we need to hear that ourselfs to be able to say, ok, that's correct, but we can't. At least I can't jajaja.

They are lacking the... emotion of the cars. You can have the sounds bad, in fact they do jajaja, but AT LEAST give us something exciting to hear.

I'm not saying that the sound in GT is the best out there, because that's not the case, we all know that, but it also depens on your point of view to the audio itself. When I heard the Corvette, the regular car on the track, I was right in front of the car, and the sound was exactly the same.
I remember talking with my father in law, saying the same but again, it is a strange topic to talk about.

As far as the videos you posted, the thing is, the car it's not the same, BUT, it's a good reference overall regarding the different kind of quality you may get depending on your equipment.

Raggadish, thanks, and yes, I agree

sparkytooth, The LFA was "perfect", at least in the acceleration section, and then they updated the sound...

 

[Griffith500]

...
I'm sorry, but the American V8 cars are the most disappointing of all the cars in gt5.

Here's a good example. GT5 Corvette GT2 ("RM"). Pay close attention to when the view flips to "replay camera" at about 1:41.
www.youtube.com/watch?v=wMow_Nlmfsg

The first thing I hear is some really, really bad phasing. Phasing happens when you play the same sound twice, and one is slightly delayed compared to the other, which gives a comb filtering effect that sounds "swirly" or "wooshy" in an unnatural way. I noticed it the worst at 1:47 but it's all over the place in external camera. Phasing is also the cause of the sound that's a bit like a jet whipping by at ~2:39.
Second, ask yourself how aggressive that feels, compared to the sounds of like the tires. Part of it is a mix/balance issue, sure, but right at 1:53 for example, the car slides, the tires squeal a bit, and then the engine comes on full-power, but it's quieter than the tires just were.

Here's a good reference for the real car. Check out the shot starting at :32.
www.youtube.com/watch?v=bor-tA9ir04

To my ear, I hear that the real car has a lot more "tone" - more bassy mids and no where near the amount of airy high-frequency sounds the GT clip does. The shifts are vastly more dramatic in the real car, too. you can hear the direct sound of the car for a lot longer distance-wise than in the GT5 clip as well.


And then... on-board. The kind of microphones you get with "on-board" clips are less-than-desirable, so let's listen to a handful of clips.

Gopro on the driver's helmet (skip to 5:00):
www.youtube.com/watch?v=Kf0hSVL6y9A

Unknown driver-mounted camera (the saturation in this clip sounds really good):
www.youtube.com/watch?v=jR-YG1D6VVg

Lemans on-board:
www.youtube.com/watch?v=evmIoocgthg

Those three clips all sound rather different so it's amazing to think they're all the same car (differences in year notwithstanding). Which one is the most accurate? That's a silly question, they're all the exact car. Which one is the most exciting? To me I'd vote #2.

I think the main source of the phasing on that particular car (it's certainly not universal) is because it has two exhausts set far apart. Unfortunately, they don't have multi-channel samples, so they're playing the same sample back from each position on the car.

Doubly unfortunately, their use of an exhaust sound effect (namely what I call the flow noise from the "tip") has a listening-angle-based filter effect to it. Each exhaust "tip" presents a different angle because it's in a different position on the car, though, so each one sweeps through its filter at different rates as a result. A single frequency sweep sounds very much like a variable phaser already (and you can hear that effect walking around a real car's exhaust), but combining two slightly out of phase (in control terms) with each other would probably sound very phasey with the exact same input.

This effect, and the tyre noise etc., are all realistic components of the sound, but the samples already have some of that in them already, and they're just not quite the right format for it to work as well as it could. The effects generally add extra life and character to the stronger samples, though.
The Corvette unfortunately has one of those dirty samples (part throttle recording), which is perhaps provoking the filter-y thing to generate a bit too much high frequency energy.

In the 2.02 update, they tweaked this filter thing a bit (as well as the mix, which may have a larger effect, depending on what you're listening through):



I think the SLS illustrates the effect the best, and is also coincidentally the closest to the Corvette in terms of engine, (intake) and exhaust configuration.

Presumably these extra layers are designed to work with PD's new sound synthesis, and hopefully that'll allow different input to each tip, because, when you do that, the phasing sounds much more natural.

 

[research]

I'm just gonna throw this in the discussion every few weeks...

/

Just some simple "sliders" for the sound effects during gameplay could REALLY help us out.

Tire noise ....l..
Engine Bay .l.....
Cabin sounds ...l...
Air rush ...l...
EXHAUST NOTE ......l
Transmission ...l...
Crash effects ....l..
Environment ...l...
Crowd Noise ....l..

etc...

 

[silicon1138]

Here's a little clip of the recording of a car for Forza. This is obviously in the garage but i wonder if they record it outside on the track too. The amount of reverb in the garage will exaggerate the overall sound of the car i think, even though it's close mic'd. http://www.youtube.com/watch?feature=player_embedded&v=s7fstY4gZWo

 

[CarreraGT]

One of my favorite cars ever in GT5 sounds wrong.

Here it is in GT5 (Stock, buying another exhaust makes it worse ):



Here it is driven by Tiff Needell around Rockingham Motor Speedway:



I really hope that "Magic" GT6 patch fixes this thing, the 787B (The most brutal racecar sound I've heard) and countless others /)_-

Funnily enough I actually thought the LFA was one of the better sounding cars in GT5 - just need to crank up your speakers (and bass). You're right though, exhaust upgrades certainly make it sound worse, in fact the whole exhaust upgrade thing is pretty hit and miss in GT5.

Some upgrades make a car sound completely different, some cars use the same sound file for upgraded exhaust etc. It's all over the place. I say scrap all the standard model cars, upgrade a bunch to premium and re-record all of engine sounds with stock and aftermarket exhausts.

As for the digital recreation of the sound, I don't know much about that. What I do know is that what we have now is rubbish. Gearshifts are perfectly smooth, autos sound the same as manuals sound the same as DCT's, no backfiring, all cars sound tinny etc. etc. you know what I'm getting at

 

[Chippy569]

I think the main source of the phasing on that particular car (it's certainly not universal) is because it has two exhausts set far apart. Unfortunately, they don't have multi-channel samples, so they're playing the same sample back from each position on the car.

That's not a particularly good reason. It would be far better to condense a stereo file down to mono and either put its emitter in the center of the two exhausts, or scrap the point-source behavior.

Doubly unfortunately, their use of an exhaust sound effect (namely what I call the flow noise from the "tip") has a listening-angle-based filter effect to it. Each exhaust "tip" presents a different angle because it's in a different position on the car, though, so each one sweeps through its filter at different rates as a result. A single frequency sweep sounds very much like a variable phaser already (and you can hear that effect walking around a real car's exhaust), but combining two slightly out of phase (in control terms) with each other would probably sound very phasey with the exact same input.

surely this is done with a low-pass and not a parametric though, right? That in and of itself would not create a phasing effect.

This effect, and the tyre noise etc., are all realistic components of the sound, but the samples already have some of that in them already, and they're just not quite the right format for it to work as well as it could.

I can only imagine what the sampling rate is on these files... 24k?

Also, I've put together a video explaining what I think the crux of the "gt5 sound" issue really is. this is my personal opinion, and i have no evidence of it. The piece I used is part of the FMOD Studio demo which is free and available at fmod.org if you want to play with it yourselves. You can hear the audio framerate drop pretty hard in this vid - my old mac was struggling to video record and audio render at the same time.


Given my theory, the best way to address it is to have more loops, which means allocating more RAM for audio. I'm really looking forward to what PD can do when they move up to the PS4 if they continue with this loop-based technique.

CarreraGT above me gets at the other issue: sounds in shifting or other dynamic moments, which is what makes real cars sound interesting, but that's a whole different chapter to start down.

 

[Johnnypenso]

Given my theory, the best way to address it is to have more loops, which means allocating more RAM for audio. I'm really looking forward to what PD can do when they move up to the PS4 if they continue with this loop-based technique.

CarreraGT above me gets at the other issue: sounds in shifting or other dynamic moments, which is what makes real cars sound interesting, but that's a whole different chapter to start down.

Thanks so much for the video. It explains least your theory of sounds in GT5 better than any number of written posts can. I know it took some work and just wanted to say thanks for that. Well done:tup:👍

Everyone that has an interest in sounds in the GT series should watch this 7 minute video on Youtube!!!

 

[Chippy569]

like i said, if you want to play around with it yourself, it's for free with the FMOD 1.2 installer. You won't get the same framerate drops that my computer did (ie the "stuttering" when the rpm moved too quickly) You can grab that here: http://www.fmod.org/download/

Because I know someone will ask, FMOD is a program that works between a sound designer and a programmer. It's called "middleware" and it allows the sound designer to dictate how sound behaves based on game parameters. It also handles bank management, mixing, asset management... lots of things really. There is another common middleware called Wwise that is more popular these days, with most of your "indie" games and a few of the smaller 3rd party studios using either of these middleware solutions. Other studios make their own custom tools for audio creation. I know Codemasters has their own, as does ISI and Simbin. The Frostbite engine also has its own audio management for battlefield/NFS.



Critically, I don't think anyone in this thread has said anything that PD doesn't already know and hasn't already thought of. GT6 sound will be the best it can given the time and budget. For me personally, it's just such a big disappointment against an otherwise compelling game, especially in a franchise i grew up with.

 

[Jedi2016]

Thanks for that. Seems like the crux of the problem isn't that they're necessarily doing anything wrong, just that they don't have near enough samples of what the engines actually sound like at different speeds.

 

[AnAL0G]

Given my theory, the best way to address it is to have more loops, which means allocating more RAM for audio. I'm really looking forward to what PD can do when they move up to the PS4 if they continue with this loop-based technique.

CarreraGT above me gets at the other issue: sounds in shifting or other dynamic moments, which is what makes real cars sound interesting, but that's a whole different chapter to start down.

Nice video!

Technically though, those wouldn't be loops, they would just be individual samples. They're only loops if you, well... loop them.

I do think that you are right though in that what PD does is take a small number of samples, maybe even just one sample in some cases, and instead of smoothly fanning out several real recordings of engine sounds, they instead just pitch shift the one or few original samples.

When you pitch shift a signal like that, you are actually making it less complex. An actual car revving at 8000rpm is putting out a lot of complex sound/noise. You will have both low and high frequencies, and a ton of waves interacting, phasing, etc. But when you pitch shift a signal, you are narrowing the complexity of the sound's spectrum more and more the further you go, and that's why it sounds more and more synthetic as the GT engines turn more rpm's. All of its naturally occurring partials/harmonics are being distorted, so it starts to sound manipulated and synthetic.


This is why when you are working with a really good sampler, something like Kontakt 5, and you look at the instrument controls, you will see that for each single note on the keyboard they will have loaded multiple samples of the same note, and those different samples are recorded at actual real life attacks and velocities. The program then selects the proper sample to play back based on how hard you hit/press the keys. If you look, there is a yellow block selected in the middle of the screen. It is a single sample for a single key on the keyboard, but the range of samples for that single key extends above and beyond that particular recording. It's set to play back only when the key is struck at a certain attack/velocity.

A simple analogy is to think of the PS3 headset controls when they do the voice changing. Your voice may change to the natural range of say a female's voice, but instead it sounds like a chipmunk because of the narrow spectrum of harmonics and partials. The tones are at the same pitch, but they have very different timbres.

 

[Imari]

This is possibly the best post in the whole thread.

 

[Johnnypenso]

This is possibly the best post in the whole thread.

Agreed, nothing replaces visuals overlayed with explanatory audio tracks:tup:

 

[Raitziger]

Video makes sense. Btw is there a car game on PS3 with better sounds than in GT5?

 

[Imari]

This is probably going to be a daft question with no real sensible answer, but what sort of size differences are we talking about between a set made of two or three samples and a set made of a dozen? Is it substantial? I assume it is when games go to lengths to use less samples.

 

[Cosbuster]

Btw is there a car game on PS3 with better sounds than in GT5?

Almost all of them? I really can't recall many high budget games that managed to do worse.

 

[Chippy569]

This is probably going to be a daft question with no real sensible answer, but what sort of size differences are we talking about between a set made of two or three samples and a set made of a dozen? Is it substantial? I assume it is when games go to lengths to use less samples.

This is really dependent on the type of compression that GT uses, and I have no idea about that... but if I had to guess it would be similar to OGG format and would come out to 1kb per second per sample. So 3 2-second loops is 6kb. I believe GT uses an exhaust loop set and an engine loop set, so per car 12kb for the "car" stuff, and then add the transmission, tires, air noise, and multiply by unique cars on track.

For comparison, "uncompressed" (ie 48k .wav linear pcm) is roughly 100 kb per second per sample, so the answer could be anywhere between there as well depending on what compression type GT uses.

 

[Raitziger]

Almost all of them? I really can't recall many high budget games that managed to do worse.

The reason why I ask is that PD is pointing that issue is with memory of PS3 and not their coding.

 

[Cosbuster]

The reason why I ask is that PD is pointing that issue is with memory of PS3 and not their coding.

The memory is indeed very limited but in the end it's just a decision of making compromises. I do believe that with the way they've decided to do things there's not much room left for proper sound but that's not to say they couldn't have done things differently like other developers manage.

 

[Imari]

This is really dependent on the type of compression that GT uses, and I have no idea about that... but if I had to guess it would be similar to OGG format and would come out to 1kb per second per sample. So 3 2-second loops is 6kb. I believe GT uses an exhaust loop set and an engine loop set, so per car 12kb for the "car" stuff, and then add the transmission, tires, air noise, and multiply by unique cars on track.

For comparison, "uncompressed" (ie 48k .wav linear pcm) is roughly 100 kb per second per sample, so the answer could be anywhere between there as well depending on what compression type GT uses.

Thanks.

I was going to say it didn't seem like much, but I think I can see how even very small samples could easily add up to a lot with all those other sounds as well. I seem to recall seeing a presentation by Nick Wiswell where I was really surprised by the amount of different sounds that needed to be available at the same time.

I guess it bodes well for next gen though, if sounds don't get squished out by more makey shiney.

 

[Viperman240]

Video makes sense. Btw is there a car game on PS3 with better sounds than in GT5?

And now a Ps2/Xbox/Pc game

 

[CarreraGT]

The reason why I ask is that PD is pointing that issue is with memory of PS3 and not their coding.

I don't know much about the tech side but F1 2012 has awesome sound (admittedly there's only a handful of different cars and they all sound pretty similar). Otherwise Supercar Challenge was really good too - cars sounded really raw, gear changes varied, plenty of transmission whine, race cars backfired etc. The game itself was lacking but I found myself playing it just for the sound.

Don't think I've played any others apart from that

 

[CarreraGT]

Man NFS sounds pretty good too, Aventador sound is bang on!

 

[Johnnypenso]

Video makes sense. Btw is there a car game on PS3 with better sounds than in GT5?

You mean , "Is there a car game on PS3 with worse sounds than GT5"?...lol.

When you think about it, GT is the only series on PS that is marketed as a "simulator". It has the biggest development budget of any car game. It's sounds should blow away the rest of the games on PS and yet we are constantly asking ourselves to compare arcade and smash em' up games to GT sound and GT is now failing most or all of those comparisons. That's bass ackwards to me...

 

[RACECAR]

@0:22 You sure did choose an interesting video to respond with.

Out of curiosity, is there any kind of trick (within the constraints) that can alter the effect abit?

 

[Griffith500]

That's not a particularly good reason. It would be far better to condense a stereo file down to mono and either put its emitter in the center of the two exhausts, or scrap the point-source behavior.

No, they're mono samples played from two sources simultaneously - they "need" to do that because the sources are directional. It's pretty obvious if you get the chance to listen to the mix directly. A better approach is to map the directionality much more finely in space, but that'd have to wait for better hardware. Using a different approach to the sound "sources", i.e. the "samples" (or equivalent) can also reduce the issue markedly. For instance, sending only the sound from each bank to each source, as on the real car, helps.

surely this is done with a low-pass and not a parametric though, right? That in and of itself would not create a phasing effect.

A low-pass can be parametric, yes - and a swept filter of any kind does produce a phasing effect; try it. In fact, the classic "phaser" sound is produced by sweeping the delay in a chorus effect, i.e. it's a "swept" (variable) all-pass filter (a pure phase-shift), usually with feedback - basically a swept comb filter once you add the direct path in.

But that's not the only thing (it sounds like) they're doing. I've made a close approximation using shaped noise, modulated by the underlying sound. Theirs sounds better, but it's still broadly using the "pressure" recorded in the sound file to inform "flow" and "turbulence" approximants in the noise shaper (in acoustics, flow and pressure can be considered to be analogous to electrical current and voltage, so the two properties are similarly linked). The issue is that the samples they're using aren't pure "pressures", in terms of the exhaust as a sound source; it works better if they are.

I can only imagine what the sampling rate is on these files... 24k?

No idea, they're not high fidelity, for sure. But that's not what I meant; I meant that they contain flow noise etc. The flow noise excites more flow noise from their flow noise generator, which is why it sounds excessive at times. Presumably, their new sounds won't do this. With more processing power applied, it can be prevented outright, assuming they're doing what I think they're doing.

Also, I've put together a video explaining what I think the crux of the "gt5 sound" issue really is. this is my personal opinion, and i have no evidence of it. The piece I used is part of the FMOD Studio demo which is free and available at fmod.org if you want to play with it yourselves. You can hear the audio framerate drop pretty hard in this vid - my old mac was struggling to video record and audio render at the same time.
www.youtube.com/watch?v=4HkyrjBbW8I

Given my theory, the best way to address it is to have more loops, which means allocating more RAM for audio. I'm really looking forward to what PD can do when they move up to the PS4 if they continue with this loop-based technique.

Nice video! I've tried to get people to try these things themselves, but I suppose you have to just do it for them in the end!
As has been mentioned already, the general consensus, among those that can tell, is that there are 3 samples per car: idle, low and high rpm. You seem to have heard the same thing.
An alternative synthesis scheme could use less memory at the expense of higher computation cost and potentially bandwidth - flex those SPEs.

Nice phasing as those samples blended in and out, though! That's partly why I've gone off samplers in general - the interference when blending has a characteristic sound I've come to loathe.
Any pitch shifting is wrong (unless it's Doppler), but that's not going to change in the wider industry for some time. Until then, the less pitch shifting the better, so more samples are indeed better.

CarreraGT above me gets at the other issue: sounds in shifting or other dynamic moments, which is what makes real cars sound interesting, but that's a whole different chapter to start down.

Yes, quite, and iRacing has shown that it's best handled in the physics engine. Can't wait to see / hear more of that approach from games.

 

[Chippy569]

Ha, the phasing I had in there is more to do with my laptop's mic picking up itself. I, like an idiot, forgot to use my headphones. You can even hear the delay when I hit play or pause.