Gamespot Playstation 3 Audio Article

I read this article earlier. Its very interesting. I'm also extremely happy that the Sound Director for MGS4 is putting alot of emphasis (and even gives examples) of how important sound is. I also like his comparison to past generation consoles in relation to PS3 and the possibilites of the PS3.
 
Each SPE on the Cell processor consists of 21 million transistors, two thirds of which is dedicated to SRAM, leaving seven million transistors for logic. SRAM acts as a small data cache and the logic portion performs all the calculations. In comparison, Creative's new X-Fi sound processor has 51 million transistors, and its previous chip, the Audigy, had 4.6 million transistors.
Very nice article... didn't know the above info before.

Edit: just realised.. where is RSX handling the sound, if it is?
 
Tahir2 said:
Very nice article... didn't know the above info before.

Edit: just realised.. where is RSX handling the sound, if it is?

RSX will likely take the final Audio and mix it into HDMI/TV-Out.

Anyone else find it curious they first talk about generating sound on-the-fly (bottle falling down) and then stating they are recording sounds through all the world !?!?
What for, if the sounds are adopted/generated on-the-fly.
 
You'd still like a (sound) sample as your base material, wouldn't you? ;)

It makes little sense to generate a crashing bottle on the fly using synthesis...

Edit - typo.
 
Last edited by a moderator:
Npl said:
Anyone else find it curious they first talk about generating sound on-the-fly (bottle falling down) and then stating they are recording sounds through all the world !?!?
What for, if the sounds are adopted/generated on-the-fly.
You don't generate sounds on the fly. Audio synthesis is pretty weak still and you won't get realistic sounds in most cases. They're talking about starting with a base sample and processing it, such as applying filters.
 
Shifty Geezer said:
They're talking about starting with a base sample and processing it, such as applying filters.

Yup. And adding reverb when needed, doing real time 5.1 positioning, etc.
 
Shifty Geezer said:
You don't generate sounds on the fly. Audio synthesis is pretty weak still and you won't get realistic sounds in most cases. They're talking about starting with a base sample and processing it, such as applying filters.

They dont generate those samples on the fly, but they claim to apply the environment on the fly, I hope thats better worded now ;)
ie. they dont need to record gunshots in a bathroom and gunshoots in the plains, instead they record the sample once (in a clean studio environment likely) and then filter/distort/bend/abuse them in realtime. But they still need "clean" samples for that, why fly around and record samples which are already affected by the environment ?

Actually thats exact the kind of additional work I thought sophisticated effect processing would eliminate.
 
Last edited by a moderator:
Npl said:
But they still need "clean" samples for that, why fly around and record samples which are already affected by the environment ?

They do need 'dry' samples for your example, but maybe they're talking about the sounds in a rain forrest?

Who knows? :)
 
They hinted at strong aural focus in their TGS description also:

With next-generation hardware, Hideo Kojima intends to undertake the task of creating the inner qualities of these elements. When objects break, the sound generated from the breakage differs according to the weight and hardness of the object, and the content of the objects becomes exposed.

http://www.konami.jp/gs/kojima_pro/english/mgs4_02.html

Good to hear more about it. I'm not an audio tech-head so I've no idea how advanced, or not, this is, but it sounds good.
 
Npl said:
They dont generate those samples on the fly, but they claim to apply the environment on the fly, I hope thats better worded now ;)
ie. they dont need to record gunshots in a bathroom and gunshoots in the plains, instead they record the sample once (in a clean studio environment likely) and then filter/distort/bend/abuse them in realtime.
Just for info, that's nothing particularly new either. Reverb and filters appear in dsome games already. The Amiga would use it's dolby high-pass filter on occassions such as driving through tunnels or for underwater. Some games on PS2 apply reverb on the fly for different environments, and Creative Labs EAX has been providing that for years (when people cared to use it!)
But they still need "clean" samples for that, why fly around and record samples which are already affected by the environment ?
Environmental audio. How else are you going to get the sounds of the Amazon Rainforest in full surround if you don't get the sounds from there? There's no point recording the dry studio sound of a WajaBeebee bird and then applying an 'Amazon Rainforest' filter to 30 played back samples simultaneously if you can grab a copy of the bird in that environment in one sample. I don't imagine when they say they are travelling all over the world to get samples, they're flying out to a Wild West Saloon to sample a bottle on a carpet, and then flying off to Dubai to record the sound of a cork popping, and then flying down to Rio to sample 'man falling in a swimming pool'! The only reason to travel round the world is to get audio you can't get elsewhere, environmental audio, which doesn't have to be clean and dry, but in fact you want it for having all those environment qualities already.
 
You don't generate sounds on the fly. Audio synthesis is pretty weak still and you won't get realistic sounds in most cases. They're talking about starting with a base sample and processing it, such as applying filters.

Modern sound synthesis techniques are very powerful and things like physical modelling can create incredibly accurate sounds.

If you're using an SPE for sound you are going to be doing virtual synthesis. That's a very powerful group of synthesis techniques which includes physical modelling.
You are also not limited to one particular method of synthesis. You could if you want, write a specific engine for creating breaking glass sounds and have a different sound every time glass breaks.

For existing sounds physical modelling might be too heavy but something like re-synthesis could be used, this is where sounds are created from a description of a sound rather than a sample. An MP3 player effectively does a limited version of this when it's playing music.

I don't know how much game companies have looked into this but it looks like a ripe area for middleware.
 
Shifty Geezer said:
Just for info, that's nothing particularly new either. Reverb and filters appear in dsome games already. The Amiga would use it's dolby high-pass filter on occassions such as driving through tunnels or for underwater. Some games on PS2 apply reverb on the fly for different environments, and Creative Labs EAX has been providing that for years (when people cared to use it!)
Sure, but you expect nextgen Games to use this concept in more sophisticated ways than the `Miggys lowpass-filter.
Shifty Geezer said:
Environmental audio. ...<snip>...
Youre right, my bad. But using some looping samples aint something really spectacular either, the context of the article was about using dynamic sounds, ie. something else than playing back prerecorded 5.1 sounds.
 
ADEX said:
Modern sound synthesis techniques are very powerful and things like physical modelling can create incredibly accurate sounds.
Physical modelling is both not particularly well developed at the moment (last I checked, anyway) plus way way way to intensive for what's mind-numbgingly undemanding if you just play back a few samples to the same effect. For different sounds all they'd really need to be convincing is a few basic filters on the audio. Pitch shift and low/high pass filters + reverbs would be enough to turn a brick breaking into a glass breaking I expect. Just mixing up some pitch and filters would add a lot of much needed variety. eg. Your average footsteps is one of three or four step samples for each surface. Next-gen hopefully it'll be more a case of 2 samples blended and processed so you don't hear the repeats of the audio.
 
I think Konami should focus on their weapons sound which doesn't sound realistic at all especially that M4. When I saw that MGS4 trailer, the M4 really got my attention because it just sounded stupid.
 
NANOTEC said:
I think Konami should focus on their weapons sound which doesn't sound realistic at all especially that M4. When I saw that MGS4 trailer, the M4 really got my attention because it just sounded stupid.

It may be a case of recycling MGS3 assets. It sounded very similar to the M16 sound loop of Snake Eater. A metallic "hitting" sound.
 
drpepper said:
It may be a case of recycling MGS3 assets. It sounded very similar to the M16 sound loop of Snake Eater. A metallic "hitting" sound.

Could be but it just sounds dumb. I've never heard of a real M4/M16 sounding like a squeaky wheel.
 
Last edited by a moderator:
NANOTEC said:
Could be but it just sounds dumb. I've never heard of a real M4/M16 sounding like a squeaky wheel.

I think that's just the Kojima touch, everything seems to have a different feel to it from reaslity. The sounds for the guns in MGS2 were even worse. :smile:

Anyways, this thread prompted me to play the trailer all over again, man it looks awesome, E3 can't come soon enough.
 
Curious how they compared transistor count to Creative cards but didn't factor in clock speed.

But I like the boombox on Snake's shoulder.
 
drpepper said:
...
Anyways, this thread prompted me to play the trailer all over again, man it looks awesome, E3 can't come soon enough.

Getting off-topic, but I prefer last E3 trailer, THAT was something...
(Everyone was showing CGI videos but Kojima made a fun trailer using PS2 graphics, "think different, think Kojima" ;) )
 
Back
Top