The Sound Update Thread (The Return)
- Thread starter TayeezSA
- 1,178 comments
- 146,375 views
- 8,705
And there's no external reverb to speak of yet. That should be breath-taking with their ambisonic mixing.
The reason for that is the synthetic sounds are a blank slate and can be made to sound like they're anywhere, whereas the recorded sounds already sound like they're in a specific place (which we actually mainly interpret subconsciously). Adding one "spatial colouration" over another can work, but it's very hard to get right; and it's very hard to remove that initial colouration as well (in fact, they're basically the same task).
Yes and no.
You get the same ability to represent practically any engine configuration, state of tune and so on, but it's debatable as to whether the underlying technique / process is at all similar. There will be similarities, because there are only really a few ways to efficiently generate a broadband, expressive sound, but how that generation step is informed and controlled is critical.
I seem to remember, from earlier clips of the in-development product, that this was an "empirical" approach: starting with the target sound and working backwards. Regardless, AES is "forwards". That is, you start with the physical parameters, and it generates a sound which is close; then you work backwards to tweak bits and bobs (mainly to cover the deficiencies of any performance-oriented approximations in the synthesis).
This is actually a good point. However, the same effect can be added onto synthetic sounds more easily than with the usual "close- mic'd" recordings (where quiet, stock sounds aren't an issue).
The reason for that is the synthetic sounds are a blank slate and can be made to sound like they're anywhere, whereas the recorded sounds already sound like they're in a specific place (which we actually mainly interpret subconsciously). Adding one "spatial colouration" over another can work, but it's very hard to get right; and it's very hard to remove that initial colouration as well (in fact, they're basically the same task).
Yes and no.
You get the same ability to represent practically any engine configuration, state of tune and so on, but it's debatable as to whether the underlying technique / process is at all similar. There will be similarities, because there are only really a few ways to efficiently generate a broadband, expressive sound, but how that generation step is informed and controlled is critical.
I seem to remember, from earlier clips of the in-development product, that this was an "empirical" approach: starting with the target sound and working backwards. Regardless, AES is "forwards". That is, you start with the physical parameters, and it generates a sound which is close; then you work backwards to tweak bits and bobs (mainly to cover the deficiencies of any performance-oriented approximations in the synthesis).
- 2,534
- Johannesburg
- Tayeedz1
Interesting post I saw from @Zer0
Like what they said from the Pitstop blog, they did try the dynamometer but it was quite noisey. I guess that video is some evidence to show that PD really have been testing different methods and approaches over the years.
@Strittan Music to my earholes.
From the news page, interesting watch. It's long so get a sandwich and some juice
Like what they said from the Pitstop blog, they did try the dynamometer but it was quite noisey. I guess that video is some evidence to show that PD really have been testing different methods and approaches over the years.
@Strittan Music to my earholes.
From the news page, interesting watch. It's long so get a sandwich and some juice
Last edited:
- 4,992
- Nowhere special
- SweetLilPoppet
- SweetLilPoppet
- 3,440
- Keasbey, NJ
- Smurfybug
- Smurfybug
Hehehe nopeYou clearly haven't played the best gt game of them all then
- 8,705
Incidentally, @Chippy569 do you recognise whatever it is that is attached to the wires run over the rear bumpers towards the exhaust in these two pics? It's been bugging me for ages. (And holy crap, I only just noticed that wheel at the top of the LFA pic...!)
I mean, I have an idea of what it could be, based on the direction that I think PD are taking, but I'd appreciate your expert eye on this one. It actually looks like the one in the LFA shot is spliced into an XLR lead; so far, so good.
I mean, I have an idea of what it could be, based on the direction that I think PD are taking, but I'd appreciate your expert eye on this one. It actually looks like the one in the LFA shot is spliced into an XLR lead; so far, so good.
- 2,534
- Johannesburg
- Tayeedz1
Incidentally, @Chippy569 do you recognise whatever it is that is attached to the wires run over the rear bumpers towards the exhaust in these two pics? It's been bugging me for ages. (And holy crap, I only just noticed that wheel at the top of the LFA pic...!)
I mean, I have an idea of what it could be, based on the direction that I think PD are taking, but I'd appreciate your expert eye on this one. It actually looks like the one in the LFA shot is spliced into an XLR lead; so far, so good.
I think those are lapel mics or like those stereo noise cancelling microphones that people put on Go Pros.
- 8,705
- 2,534
- Johannesburg
- Tayeedz1
Indeed, but I wonder if there exist such small mics suitable for recording exhaust sounds. Shame we can't see what's in the car.
Oh yes indeed, those mics can record stuff like this. The Rode Mics I know are very good for this because they can take quite a bit of punishment.
One thing that irks me with AES is that can it produces sounds like Turbo Spool, External Wastegates, Blow off valves etc
The sounds you hear from these vids basically. Yes, they're Subarus... That's one car's sound which I know very well in reality
Or are those elements that you still have to sample? That's the thing, doesn't AES have a limit to what it can make live?
- 8,705
They are indeed good mics, but that is quite far away from the exhaust, unlike what PD have done (recording so close like that affects the colouration of the final sound). PD don't have the restriction of only using what they can attach to the car, either, so I'm curious as to why they chose to do that when there are a couple of properly positioned, "full-size" stand-mounted mics as well.
Well, if PD decide that AES includes those things, then yes, it can produce those sounds. Otherwise I'm sure they could conjure up some other three-letter acronym for the incidental effects.One thing that irks me with AES is that can it produces sounds like Turbo Spool, External Wastegates, Blow off valves etc
The sounds you hear from these vids basically. Yes, they're Subarus... That's one car's sound which I know very well in reality
www.youtube.com/watch?v=VW5Z6_cdNw0
www.youtube.com/watch?v=Smo-56fE6Jw
Or are those elements that you still have to sample? That's the thing, doesn't AES have a limit to what it can make live?
For best results, you should inject the sounds, whether sampled (after pre-processing) or generated, into the AES "signal chain" for better cohesion. Things like wastegates and blow off valves are a modeling concern, and affect engine performance, so they're really physics issues first - how that's handled will inform how the sound part is handled. Turbine and compressor noise, along with overrun / ignition cut pops and bangs, are all just incidental effects that can be thrown in any old way and it'll be convincing, because they'll be coloured by the AES system from the point they're injected.
There is a computational limit within which all the sounds must fit, but there's no real limit to what kinds of things can be synthesised, and by what method. Of course, it makes more sense to use certain techniques for certain sounds, for efficiency, but that's not a limitation as such. There is a bit of compromise in the sense that you may as well use an existing (already running) generation method for as many additional effects as possible, but you can still use several different methods if needs be, and if you can afford to.
- 301
- Washington
- Chippy569
- Chippy569
That would be a cardoid-pattern lapel or "lavalier" microphone. Taping one to the bumper near the exhaust is a popular (and quite old) car recording technique, particularly for when cars are not on a dyno (though in this particular orientation, were it on a track it would have a lot of road noise.) I'm not sure of the exact make/model but the blue windscreen should be a giveaway, I just don't recognize it.
For really loud cars I have a pair of DPA 4062s (which are the same mics as the inside of the Formula 1 camera pods, which they were designed for) and I'll mount them in a similar way. They're also great when taped inside the intake runners. The capsule itself is really small, smaller than the fingernail of my pinky, so they're quite easy to get into weird places.
Regarding mic placement with respect to proximity of the exhaust, you actually want to get the mic outside of the "wind" or air coming directly out of the pipe, partly because it's quite hot and partly because you'll only capture wind noise on your microphone. Finding mounting points for microphones, particularly on moving cars and not on dynos, is really difficult because you have to consider road noise, car-moving wind noise, exhaust wind, etc. etc. etc.
- 301
- Washington
- Chippy569
- Chippy569
Synthesizers would be great for a very tonal sound like the spool of the turbo, however because of the way games work, things like BOV/gates are better served by a simple recording. Sure you could try to modulate on a white noise generator and whatnot, but really it's so much faster/easier/cheaper to just play a one-shot of a recording that it's not worth the effort.
- 301
- Washington
- Chippy569
- Chippy569
not particularly relevant, but probably interesting: vehicle sound designer from T10 will be on a twitch stream for maybe the next hour or so? http://www.twitch.tv/forzamotorsport - might be of interest to people who find this thread of interest.
[edit] stream is now over but it can be watched here: http://www.twitch.tv/forzamotorsport/b/659464470 - adam was on for the first hour and a half.
[edit] stream is now over but it can be watched here: http://www.twitch.tv/forzamotorsport/b/659464470 - adam was on for the first hour and a half.
Last edited:
- 8,705
I took basic notes over a week ago and forgot to share:
Overall approach:
High-level "game-designer" standpoint - design the experience first, and the shape implementation around that.
Results obtained using a "signal flow" approach - understand the mechanics of what makes vehicle move (real or imagined), and how that affects the sound produced.
Do you target realism or "sci-fi"; what about a blend?
Mike prefers to make something "awesome" by offline (traditional media production) methods, then figure out how to implement it in real time. This is primarily for impressing producers, and being sure they get the green light to implement the features they want.
Apparently, there are five basic sensations for most vehicles:
- "What makes it sound like I'm hauling ass"
- "How does sound make me feel like I'm slowing down"
- "I'm changing direction"
- "I'm losing control"
- "Idle physics"
- Engine - the sound of power generation (given priority)
- Surface sounds / friction behaviour of vehicle
- Contact with surface (and / or air)
- "Violence layer" - giving the "go fast" experience some teeth. (He said "visceral")
- Throttle - "energy input" (?!)
- Braking
- Speed
- RPM
- Direction - steering
Main types of "implementation" (the code used to reproduce the sound)
- Loops - standard steady-state recordings processed into loops and played back blended with other loops.
- Granular - one long recording of an "RPM ramp" (up or down), playback consists of tiny "grains" (very short portions of the recording) sampled around the RPM point. Blend acceleration and deceleration ramps, or any other ramps as required.
- Synthetic - "the future"; recordings used for reference only. (No further details given.)
A special consideration is the Internal Combustion Engine, which has the following main sources of sound:
- Induction
- Valvetrain, engine bay (ancillaries)
- Exhaust
Mike found that it was easier to take a modified recording and "dumb it down" than to "amp up" a stock recording; Forza stopped using stock recordings!!
Recording is better done on a "dyno", esp. "dynapack" (rollerless). Good idea to record on the track (as well). Must record close-up and from afar: just like a drum-kit. (This is related to the colouration / spatialisation / acoustic transfer I'm always going on about).
Mike sometimes recorded cars at partial throttle for "better sound". (Many of the irritatingly obviously looping sounds in GT are due to part-throttle recordings.)
Another advantage to the dyno is that some designs can "motor" the car, i.e. drive the engine via the rollers at a constant speed, for any throttle opening.
Other sound sources:
- Tyres
- Transmission
- Turbos / Superchargers
- Suspension / Chassis noise
- Propellers
- Flaps / Ailerons
- Rudder
- Wheel linkage (e.g. bogey)
- Suspension sfx
- Braking
Tyres get special consideration:
- Traction! It is a very important channel of player feedback.
- Scrub / Howl -> approaching peak slip
- Screech -> beyond peak slip
- Intensity of screech increases the more a tyre slides
- Load affects pitch: lower load = higher pitch.
Some tricks to try:
- Sustained vs. one-shot; e.g. engine loops plus thruster sound effects as "sweeteners"
- Use unpitched audio for "intensity"; e.g. filtering white noise
- Cycles per second not always pitch; e.g. one shots for train wheel sounds.
- Try Shepard tones for simple systems (constantly rising / falling, psychoacoustic trick). Or "endless gearbox", below.
Fakery! Audio-only physics fakes.
- Using spring mechanics to introduce instability (e.g. gear change "wobbles")
- RPM variance at redline to prevent "flatness" (I think this means rev-limiter modeling)
- Idle and Redline RPM offsets: introduces the concept of "audio rpm" as distinct from the physics rpm - most middleware uses 0 rpm as idle!! If recorded samples don't stretch to redline, offset what you have upwards so that you don't get stretching in the top end! (!!)
- Shift time offsets
- "Endless gearbox" - constantly changing gears, either in a pre-made loop, or via scripted controls. (For arcadey games, to prevent fatigue during full-throttle running).
Mixing concerns
- On vs. off throttle volume (also timbre, e.g. via blending of different samples)
- Camera perspectives: 2D vs 3D driven. Source placement on car and in environment; better to use a well defined 2D mix at times (e.g. interior)?
- Occlusion of external sources for cockpit view.
- Falloffs and Radii - different for replay and racing? Nearby opponents when racing, zoom shots in replays etc.
- Dynamic mixing: how many vehicles do you need to hear? What's more important to hear? Voice count (concurrent sounds being played back) an issue? Mix crowding, losing definition?
Mike's view of the near future:
Hybrid of granular and loops: the hard problems of granular systems need tackling, but Mike is sure that combining both traditional samples with ramps can alleviate the drawbacks of both (I agree with this in principle, as long as it is actually possible, and have experimented myself - it is a hard problem).
Realistic instability in synthetic systems is hard to achieve, but doable. (I agree with this as well - I've experimented with this as well, and there are several ways to tackle it - should be interesting).
- 717
Just before the tunnel entry, those new AES misfire effects?
I've never heard the effect on other cars before.
I've never heard the effect on other cars before.
- 2,534
- Johannesburg
- Tayeedz1
Sounds like backfire to me. AES seems to be evolving if that's the case. Anything interesting going on with the game's insides?
@Griffith500 where you at dawg?!
- 717
Yeah, you're probably right that it's backfire. It sounds different because the exhausts are at the front.
- 8,705
It sounds like excessive "noise" in the base waveform upsetting the rasp generator, to me. Maybe that's all the "misfires" are for now, but it shouldn't happen at full throttle.
It's likely that the misfiring is not (fully) active yet; there are plenty of other AES strings that don't have any practical consequence in the game yet. It might also only affect idle at the moment (improper cylinder filling is technically a misfire, as far as torque generation is concerned; two stroke, rotary and "hot cam" lope is also misfiring).
Note that PD don't call "back-firing" mis-firing; they call it after-firing. Which is more correctly what it actually is (back-firing is back out the intake, mis-firing happens within the cylinder). There are a lot of similarly precise definitions in their "code".
It's likely that the misfiring is not (fully) active yet; there are plenty of other AES strings that don't have any practical consequence in the game yet. It might also only affect idle at the moment (improper cylinder filling is technically a misfire, as far as torque generation is concerned; two stroke, rotary and "hot cam" lope is also misfiring).
Note that PD don't call "back-firing" mis-firing; they call it after-firing. Which is more correctly what it actually is (back-firing is back out the intake, mis-firing happens within the cylinder). There are a lot of similarly precise definitions in their "code".
- 676
- Toronto, CAN
- Guynemer27
Seriously? You have such a low expectation.... I found the sound awful, the vision Peugeot is a tad better but still so far behind the current competition... Buzzzerrrrrrrrrrr.... Couldn't watch the new car's videos more than few seconds before to click "stop" frantically....
- 28,470
- Windsor, Ontario, Canada
- Johnnypenso
- 4,770
- Hope Mills, NC
- GTP_Dylan_GTP
- 2,534
- Johannesburg
- Tayeedz1
Sooooo to the people hating on the GTR and I'm gonna damage control here for GT6's sake because it seems just.
How would you expect a RACECAR, with probably less than a meter of exhaust pipe from the engine, with STRAIGHT PIPES and no mufflers or resonators of the sort to sound? Particularly when it's right near you?
I for one know that it doesn't sound like a super clean Magnaflow install or HKS Silent Hi Power. It will be loud, it will be brash, it will be unapologetic.
Mind you, this is a methos that's still in its preliminary stages.
How would you expect a RACECAR, with probably less than a meter of exhaust pipe from the engine, with STRAIGHT PIPES and no mufflers or resonators of the sort to sound? Particularly when it's right near you?
I for one know that it doesn't sound like a super clean Magnaflow install or HKS Silent Hi Power. It will be loud, it will be brash, it will be unapologetic.
Mind you, this is a methos that's still in its preliminary stages.
