Need information!!

[Quantik_Music]

Okay so, I'm currently absorbed and interested in complex computer based sound design and processing. The thing is though I have no idea where to find resources and sources of information to help me understand it.
It all started through producer and composer BT. He mentions being able to process audio files (his voice for example) using granular re-synthesis and time correct the new files so they sync with the rest of the project and don't sound completely random and all over the place.
He says he uses phase vocoding, wavelets and a whole bunch of other techniques I've never heard of to do this time correction. So...where do I find the information I need to understand what he is talking about and how to use these techniques myself?

[Quantik_Music]

94 views and not one piece of advice? o.O I know this is gearslutz guys and not sound designslutz but the reason I posted the question here is because of the huge number of forum members and because I've learnt so much from this site already. All I need i a lead! Maybe the name of a forum dedicated to this kind of thing or a contact who knows about this area.

[Preston135]

well i can't understand what you're trying to achieve. are you wanting to time the vocal up? put effects on? from what source? what gear are you wanting to do it with?

[Quantik_Music]

Originally Posted by Preston135

well i can't understand what you're trying to achieve. are you wanting to time the vocal up? put effects on? from what source? what gear are you wanting to do it with?

Information is what I'm after. What I'm talking about is really esoteric, unless you understand concepts like phase vocoding, granular re-synthesis and how BT might use other time correction modalities I'm afraid you probably won't be able to help. You can help however if you know of any good forums or places where I can find information about the aforementioned concepts and techniques.

[Yoozer]

Originally Posted by Quantik_Music

94 views and not one piece of advice? o.O

Here's why - and I'll be blunt, but get used to it: it's because your topic title sucks. 90% of the people who start topics here "need information" or "need help", so it's completely superfluous to put that in the topic title.

Here's what'd work better; "BT production techniques - any clues about granular re-synthesis, phase vocoding?" because then a lot more folks would answer.

Originally Posted by Quantik_Music

He mentions being able to process audio files (his voice for example) using granular re-synthesis

Granular synthesis chops up a sample into very small bits - called "grains". These are then played back in a so-called "grain cloud" - a bunch of these bits chosen somewhat randomly, played back at different pitches and at different times. Check out the demo of New Sonic Arts

"Re-synthesis" is sort of odd to use in this context, because you're not really re-synthesizing something; that's usually reserved for additive synthesizers. These accept a sample and split it up in a (limited) set of sinewaves at specified frequency intervals. Additive synthesizers, such as Camel Audio Alchemy, Virsyn Cube, ImageLine Morphine and various others allow you to load up a sample, and then it'll try to recreate it with a bunch of sinewaves. This enables you to stretch the sound - because you're not dealing with a sample that is slowed down but with 512 volume envelopes, stretching becomes a lot easier. It does have artefacts though - the result is going to be somewhat warbly and glassy.

The principle of a wavelet is explained here: Wavelet - Wikipedia, the free encyclopedia - but it's not a thing by itself. The article mentions convolution; convolution is used to mimic EQs, compression, and reverbs (and various other things).

Phase vocoder might simply be a fancy term for what Antares Auto-tune does. It also has a link to a page of IRCAM; putting that through Google shows SuperVP | Ircam Anasynth - and then it suddenly becomes a lot easier because that paper shows the demo of changing the gender of someone speaking into it. By adjusting formants in Ableton Live's timestretch, you can also do this.

It helps if you have link to the interview; a lot of times it may simply mean that you have to cut through the buzzword jungle to get something useful out of it.

As for tools to achieve this; higher end modular/programmable signal processing is your friend. Eventide, Kyma, etc.

[Preston135]

Originally Posted by Yoozer

Here's why - and I'll be blunt, but get used to it: it's because your topic title sucks. 90% of the people who start topics here "need information" or "need help", so it's completely superfluous to put that in the topic title.

Here's what'd work better; "BT production techniques - any clues about granular re-synthesis, phase vocoding?" because then a lot more folks would answer.



Granular synthesis chops up a sample into very small bits - called "grains". These are then played back in a so-called "grain cloud" - a bunch of these bits chosen somewhat randomly, played back at different pitches and at different times. Check out the demo of New Sonic Arts

"Re-synthesis" is sort of odd to use in this context, because you're not really re-synthesizing something; that's usually reserved for additive synthesizers. These accept a sample and split it up in a (limited) set of sinewaves at specified frequency intervals. Additive synthesizers, such as Camel Audio Alchemy, Virsyn Cube, ImageLine Morphine and various others allow you to load up a sample, and then it'll try to recreate it with a bunch of sinewaves. This enables you to stretch the sound - because you're not dealing with a sample that is slowed down but with 512 volume envelopes, stretching becomes a lot easier. It does have artefacts though - the result is going to be somewhat warbly and glassy.

The principle of a wavelet is explained here: Wavelet - Wikipedia, the free encyclopedia - but it's not a thing by itself. The article mentions convolution; convolution is used to mimic EQs, compression, and reverbs (and various other things).

Phase vocoder might simply be a fancy term for what Antares Auto-tune does. It also has a link to a page of IRCAM; putting that through Google shows SuperVP | Ircam Anasynth - and then it suddenly becomes a lot easier because that paper shows the demo of changing the gender of someone speaking into it. By adjusting formants in Ableton Live's timestretch, you can also do this.

It helps if you have link to the interview; a lot of times it may simply mean that you have to cut through the buzzword jungle to get something useful out of it.

As for tools to achieve this; higher end modular/programmable signal processing is your friend. Eventide, Kyma, etc.

AKA Time stretch, is this what is Wanted?

i just could'nt understand the post. i guessed that was why 88 people looked and never replied.

[Rimwolf]

The Computer Music Tutorial, Curtis Roads

Computer Music Journal

[Quantik_Music]

Hello guys and thank you for the replies. Yeah I will work on the thread titles..not the best I will admit. I've always heard it referred to as granular re-synthesis considering you aren't actually synthesizing a new sound merely cutting an existing one into grains or particles. I will post the section of the BT interview so you guys can sees the exact terminology he uses. It may be a time stretch that he is talking about but I'm not sure as he calls this 'time correction' which makes me think instead of aligning off-beat audio like a granulized file so that it sounds rhythmical and works with the time signature and tempo of your project. Here is the section in which he talks about it all, maybe someone can decode what he is talking about into something more easily understood? Thanks for the help guys I really do appreciate it.

Interviewer- When you mix organic sounds with the electronic, how do you get them to jell together?
BT- So the essential thing for me about making acoustic instruments jell with electronic instruments is the proprietary time-correction modality that I’ve come up with over the period of 20 years, quite honestly. So it’s a thing that I’ve done many lectures on at different universities. It’s a very complicated technique and complicated in that there are many different types of time stretching that I like to use, from phase vocoding to granular to wave width, to something that I call wave-cycle repetition—on things like analog synthesizers where you take a single bi-polar, uni-polar wave and repeat it.

[acreil]

He's being totally vague and buzzword-y so as to appear smart. There's no actual information to decipher. He could have just said "It's like time stretching and stuff lol, quite honestly."

He allegedly used Csound a lot. You can do the same with Max/MSP or Pure Data or whatever. Lots of DAWs already do time correction and audio quantization.

What he calls "wave-cycle repetition" is probably this: PSOLA - Wikipedia, the free encyclopedia. You could use that to "freeze" the input, or whatever.

[Quantik_Music]

Okay can anyone help me understand how these time correction (time stretch) techniques would be applied? Let's take BT's track every other way for example. There are some really interesting granular textures in the track and I want to know how he would have time corrected them so it wouldn't sound like a random mess.
BT featuring JES - Every Other Way - YouTube Granular bit starts at 0:52
So in essence with a 4/4 track at a tempo of roughly 125-130 bpm or what ever it is how would BT keep those granular stems in time with the track?

[Quantik_Music]

Originally Posted by acreil

He's being totally vague and buzzword-y so as to appear smart. There's no actual information to decipher. He could have just said "It's like time stretching and stuff lol, quite honestly."

He allegedly used Csound a lot. You can do the same with Max/MSP or Pure Data or whatever. Lots of DAWs already do time correction and audio quantization.

What he calls "wave-cycle repetition" is probably this: PSOLA - Wikipedia, the free encyclopedia. You could use that to "freeze" the input, or whatever.

Thanks very much for helping me with that one. BT tends to talk himself up quite a lot which makes it extremely difficult to understand what he is talking about sometimes. He loves to make things as esoteric and complex sounding as possible

[acreil]

Originally Posted by Quantik_Music

BT tends to talk himself up quite a lot which makes it extremely difficult to understand what he is talking about sometimes. He loves to make things as esoteric and complex sounding as possible

I'm glad I'm not the only one who gets that impression. I much prefer a self-deprecating intellectual over someone who actively markets himself as a genius. He comes across as the personification of pretentious... with stupid frosted hair.

Anyway, I imagine if you could analyze a file for transients, and then pick events A, B and C to map to positions X, Y and Z on a timeline, you could use that to control your choice of time stretching algorithms (including reversing, freezing, etc.). There's a variety of ways to do time stretching: granular, phase vocoder, etc. But they're already well known and not really rocket surgery. If you could also detect and control pitch, correct or deliberately shift formants, and automate other time stretch parameters, I expect it could be a really powerful interface. Of course you can do all this stuff already, but it's a pain to do it manually.

Lots of "audio quantize" and tempo changing stuff that's included in DAWs now does almost the same thing. But it's generally meant to transparently work on rhythmic material, where as here you'd be using monophonic tonal sounds.

[zerocrossing]

Originally Posted by Quantik_Music

It all started through producer and composer BT. He mentions being able to process audio files (his voice for example) using granular re-synthesis and time correct the new files so they sync with the rest of the project and don't sound completely random and all over the place.
He says he uses phase vocoding, wavelets and a whole bunch of other techniques I've never heard of to do this time correction. So...where do I find the information I need to understand what he is talking about and how to use these techniques myself?

I think what I'd do if I were you is get a copy (maybe subscription) of Computer Music magazine and start reading. Computer Music has a lot of tutorials and comes with a nice free suite of tools that'll help you achieve what you're talking about. Then maybe download a demo of Native Instrument's Reaktor and run it inside Ableton Live. Live is particularly good at quantizing audio in time and pitch. Reaktor is like audio "Legos." So while I'd not recommend starting to build your own stuff in Reaktor quite yet, it comes with a really good library of instruments and effects such as granulators that'll keep you busy for a long time. Also there is a great user library of stuff others have made that is free to use for license owners.

Later, when you start grasping the basics, post specific questions here. Don't be afraid to play and make mistakes. Exploration is key to figuring out the tools and stumbling upon new techniques that you may not have thought of.

[zerocrossing]

Originally Posted by Quantik_Music

Okay can anyone help me understand how these time correction (time stretch) techniques would be applied? Let's take BT's track every other way for example. There are some really interesting granular textures in the track and I want to know how he would have time corrected them so it wouldn't sound like a random mess.
BT featuring JES - Every Other Way - YouTube Granular bit starts at 0:52
So in essence with a 4/4 track at a tempo of roughly 125-130 bpm or what ever it is how would BT keep those granular stems in time with the track?

First, frost hair.

Second, get a Delorean and a blonde singer

Third, Granite. New Sonic Arts

[bluegreengold]

complex audio manipulation howto:

1) Learn what it is your trying to do.
2) Learn a scripting language such as Perl or Python
3) Learn to use a library of audio/DSP algorithms -- ie RTcmix, supercollider, chuck, Csound etc...
4) Write a script to analyze the data from your audio files
5) Write a script to manipulate the data

[Quantik_Music]

Thanks for all the replies all! I think I understand now that this isn't stuff you can just 'implement' right away. Quite clearly understand DSP and synthesis a lot more would be more beneficial to me than just trying to use a plug in. I can also see now that after more research into the more 'esoteric' fields of computer music it will allow me to know what it is that I'm trying to accomplish instead of only knowing half of it. Big thanks to everyone =D

[Reptil]

adding to the list some more specific stuff:
Audiomulch
Max DSP
Kyma
Eventide
Kurzweil VAST

[audioconsult]

Originally Posted by Quantik_Music

Okay so, I'm currently absorbed and interested in complex computer based sound design and processing. The thing is though I have no idea where to find resources and sources of information to help me understand it.
It all started through producer and composer BT. He mentions being able to process audio files (his voice for example) using granular re-synthesis and time correct the new files so they sync with the rest of the project and don't sound completely random and all over the place.
He says he uses phase vocoding, wavelets and a whole bunch of other techniques I've never heard of to do this time correction. So...where do I find the information I need to understand what he is talking about and how to use these techniques myself?

i guess he is just babeling around to cloud a bit that he is just using ableton lives warp markers to make a sound fit...

[Quantik_Music]

Haha the last post wins so much! Loving the answers so far guys.
Even though there is a possibility BT is just using the warp markers in Ableton would there be a reason for him to use phase vocoding/wavelets to time stretch? Is it because it would result in less artifacts that are normally left when conventional techniques are used?
Every time I have used granular effects on like a vocal for instance I can't use it in conjunction with the rest of the track because it goes completely off beat.
I'm just wondering how BT manages to lock his granular affected tracks in rhythmically. The the mapping to X-Y post up above was a really interesting idea. Does anyone think that's how BT does it so well?

[gremlin moon]

Originally Posted by Quantik_Music

Every time I have used granular effects on like a vocal for instance I can't use it in conjunction with the rest of the track because it goes completely off beat.

What DAW and sampler/granulizer are you using? Are you syncing the grain parameters to an LFO/envelope/step sequencer synced to the the BPM of the DAW (i.e. not running "free")? The effects in the track I heard can be done using the per-step repeat function in Geist and in S t u t t e r Edit (obviously), Alchemy, Ableton. [I haven't given Granite a go yet as their website was down when I tried to get the demo.] I don't think there is any need to go too deep into a programming language at this point since their are tools that do it for you (albeit that cost $200).

[I read a BT interview where he was talking about nano-quantizing the sound of scuba tank bubbles and how he was going to write a book about it. Personally I can't STAND the sound of unquantitized bubbles. There were some birds outside of my window the other day chirping out of time in some psuedo-polyrhythmic jibber-jabber and I seriously had to put in ear plugs.]

[acreil]

Originally Posted by Quantik_Music

Even though there is a possibility BT is just using the warp markers in Ableton would there be a reason for him to use phase vocoding/wavelets to time stretch? Is it because it would result in less artifacts that are normally left when conventional techniques are used?
Every time I have used granular effects on like a vocal for instance I can't use it in conjunction with the rest of the track because it goes completely off beat.
I'm just wondering how BT manages to lock his granular affected tracks in rhythmically. The the mapping to X-Y post up above was a really interesting idea. Does anyone think that's how BT does it so well?

Time stretching algorithms always have artifacts, it's just nice to have a selection of them because they have different artifacts when used on different types of material. Something chosen as a general purpose tool won't necessarily be well suited to the way you want to use it.

Also granular synthesis is totally general, but a lot of plugin or DAW implementations aren't terribly flexible. If you're going for a very specific effect, it's sometimes better to just make it yourself.

Originally Posted by gremlin moon

There were some birds outside of my window the other day chirping out of time in some psuedo-polyrhythmic jibber-jabber and I seriously had to put in ear plugs.

Christ, I bet they weren't even in key with the traffic noise. Save us, BT!

[bluegreengold]

Originally Posted by gremlin moon

I don't think there is any need to go too deep into a programming language at this point since their are tools that do it for you (albeit that cost $200).

I disagree here. If you have the curiosity, desire, and the will to learn how to do this kinda stuff, it's a good investment to spend some time studying the different kinds of algorithms and learning some programming. Pretty much every trick/tool/technique that is available as some VST plugin can be prototyped fairly quickly using Max/MSP, supercollider, cmix, or even reaktor, if you understand what it is your want to implement. The advantage of using as low level software tools as you can get to grips with, such as going straight to something like rtcmix, is that you have available all the coding libraries / functions available for whatever scripting environment you use.

It's takes some learning, but it is hardly an impossible task to get to the point where it's faster and easier to implement specific ideas using only free software libraries and a little scripting. You'll also gain skills that you can apply elsewhere.

[gremlin moon]

Originally Posted by bluegreengold

I disagree here. If you have the curiosity, desire, and the will to learn how to do this kinda stuff,

I totally agree; I just noticed though the op didn't seem to even do a basic Google search so may instead really be happier with a VST. For instance, just Googling "phase vocoding" the Wikipedia page is at the top of rankings and will give you the history, references and let you know that it is the basis for Autotune (kind of a hard detail to miss). A quick search for "wavelets" brings up Wavelets.org a whole community dedicated to resources about learning and applying wavelets.

[Quantik_Music]

Originally Posted by gremlin moon

What DAW and sampler/granulizer are you using? Are you syncing the grain parameters to an LFO/envelope/step sequencer synced to the the BPM of the DAW (i.e. not running "free")? The effects in the track I heard can be done using the per-step repeat function in Geist and in S t u t t e r Edit (obviously), Alchemy, Ableton. [I haven't given Granite a go yet as their website was down when I tried to get the demo.] I don't think there is any need to go too deep into a programming language at this point since their are tools that do it for you (albeit that cost $200).

[I read a BT interview where he was talking about nano-quantizing the sound of scuba tank bubbles and how he was going to write a book about it. Personally I can't STAND the sound of unquantitized bubbles. There were some birds outside of my window the other day chirping out of time in some psuedo-polyrhythmic jibber-jabber and I seriously had to put in ear plugs.]

Yep, the 'running free' Issue is the exact problem I'm getting, changes made to a file using granular synthesis cause it to go out of time due to the changing parameters. What would be the component that is causing everything to go out of time? Is it grain size that's causing it? I have a feeling if I can sync what ever parameter is causing the timing problems to the track tempo than everything will work out nicely.

Just adding on here. Everyone who has been kind enough to help me this is exactly what I'm talking about ^ Syncing the out of time granulation to the tempo of the project (I actually understand what I'm trying to accomplish now lol) Anyone else want to chime in with some nifty ways to do this?