posted 23 October 2006 07:19            

I posted this as a 'programming challenge' because I suspect this tool will appeal to other Kyma users, and could be programmed collaboratively? If you're interested then read on!

As far as I understand it, these are the main points to BBCut:

BBCut has a mode which can 'intelligently' detect transients to work at a synchronized bpm.

It can cut any source audio in real-time: live, buffered, rhythmic or static.

There are several algorithms within the BBCut library (link at bottom for a comprehensive review of it's capabilities by Nick Collins), which cut the audio into slices forming new rhythms with optional fills.

Highlights include groove variation, independent control over tempo/pitch, organic amplitude variation, panning variation, and pitch effects. (There is also a DutyCycle control).

There is a lot more to BBCut, but these are the main features as I can see them.

I plan to start working on this challenge today. Off the top of my head, the following Prototypes seem appropriate:

Sample
Memory Writer
Analog Sequencer
Amplitude Follower
Pan
PitchShifter
Noise

If you'd like to collaborate, then please reply to this post, and share your ideas on how to best build this tool! We can use this thread to attach Sounds to, and make comments etc.

Thanks!

Jack

This is the link to Nick Collins review:
http://www.cus.cam.ac.uk/~nc272/papers/pdfs/bbcutlib.pdf

posted 23 October 2006 07:35            

The plugin is called Livecut and was programmed by mdsp, for smartelectronix. It does not feature the full capabilities of BBCut, but will give you a good idea of what's possible...

Jack

posted 23 October 2006 17:29            

wavesets.kym

This seems interesting!
I'd be curious to hear what this library produces. I have not been able to use the plugin, it seems the AudioUnit is broken in the current version.

Attached is a sound I've been scratching my head for a long time. It is a waveset stretcher. I think the idea is kind of the same as the BBCut library, but in a microscopic viewpoint. These sounds can be a source of inspiration for working on your idea. You will need Pete's DSP algorithms in order to make them work.

Building the waveset stretcher has been really worth of many learnings. I find cutting and pasting sound pieces is not a trivial task in realtime DSP systems because it involves careful tracking of indexes and memory management, even when you use such an easy user interface as the one Kyma offers. What I mean is that those algorithms are more like Event driven algorithms, and not Data driven algorithms. Time passes and you need to store events in a sort of database and process them, not like for example a reverb algorithm which process the incoming samples one at a time.

Something important to consider (I think) is the precision of the cuttings. For example, if you use the AmplitudeFollower prototype, you will need to use Hot Variable (the ones that begin with a exclamation point, like !Amplitude). The thing to know with these variables is that they are updated a thousand times per second. This is because they are control signals, not data signals (which run at full sample rate, e.g. 44.1 kHz). Then, your cuttings are limited to the precision of 1000 events per second, which can lead to bigger errors when the algorithm applies many operations on a badly times event.
The solution to that problem is to use Pete's DSP prototypes. They offer many ways to process data at sample rate.

It is not necessary to make the previous statement a general rule: the wavesets stretcher needed that resolution because one work on very smal grains of sound. For the BBCut library, control rate resolution (1 kHz) is certainly enough. If you compare 120 BPM to 1 kHz, there is a ratio of 1/2000, which make very little error, knowing that the BPM is internally computed from a much mor precise clock, that has virtually no jitter.

So here were my 2 cents about cutting and pasting sound pieces ...

Best,
Camille

posted 24 October 2006 19:32            

I also haven't decided how to store individual input 'hits' into memory buffers, so that the Sound can access them to assemble new beats. I'm going to try and figure out a system whereby a Peak Tracker (along with some probability scripts) aids in deciding which hits to store (and into which of say 8 memory buffers to store them), and use Analog Sequencers and Samples to re-sequence those hits into new phrases.

But that's off the top of my head and I'll have to think about it properly later on.

Thanks again for you perspective!

Jack

[This message has been edited by JackRosete (edited 24 October 2006).]

posted 25 October 2006 01:45            

I might be up for helping out on this project. I've known about the BBCut stuff for awhile now but haven't checked it out - but now I've downloaded it and will try brush off my small bit of SuperCollider knowledge and take a look.

I'm guessing that it is is built on C code which is made into external objects for SuperCollider (or is that unit generators in SC-speak?). So I'm not sure how well that will translate to Kyma without delving into creating microsounds, though maybe Pete's objects will do much of what is needed.

Anyway, some of what you are describing sounds like it might tie into an idea I had for analyzing incoming audio for pitch and rhythm and storing that info in tables (global controllers or memory writers) which could then be accessed by Kyma in generating complimentary material in a random or a quasi-intelligent fashion.

I'm not sure how to procede though. I think I'll try to look at what BBCut over then next few days, and maybe I'll get some ideas.

I'm a big Squarepusher fan incidentally, but a little skeptical at how these Squarepusher-in-an-algorithm things work (who knows - maybe that's how he does it, but it doesn't sound like it to me...). But anyway, it's a worthwhile project for wherever it might take us.

posted 25 October 2006 05:05            

Cool!

'I'm a big Squarepusher fan incidentally, but a little skeptical at how these Squarepusher-in-an-algorithm things work (who knows - maybe that's how he does it, but it doesn't sound like it to me...).'

I agree with you, I don't think Squarepusher uses them. My own interest in these algorithms is to build them (for fun) and then adapt them into more experimental, idiosyncratic forms, particularly for live, interactive performance. I'm especially interested in long, drawn out, organic, overlapping micro-looping!

In any case, it should be a fun challenge and I look forward to sharing ideas with you on this. I will post my progress soon!

Jack

posted 25 October 2006 16:09            

Well I have now come up with a basic Sound implementing some of the BBCut features (see attachment). This can serve as a starting point for further development. I will have to rethink my looping structures because the approach I've taken so far (repeat looping) has already hit the wall. The BBCut library (as far as I can tell), samples individual slices and then RE-SEQUENCES them into new structures/loops, as opposed to standard REPEAT looping (what I have done). That's how the pitch-shifting effect is so smooth, because it is actually the sample frequency of each iteration that changes, rather than a pitch-shifter acting on the whole loop.

Yet to be implemented (among others):

Intelligent BPM detector
Beat re-sequencer
Pitching-effect
Better Looping algorithm (including proper !Duty controller)
Max/Min number of Looping iterations controller
Acceleration/Deceleration of loops, conforming to beat subdivisions

posted 26 October 2006 21:08            

I am looking forward trying your Sound.
I am in London right now, so I have no access to Kyma, but I plan to do this next week.

posted 27 October 2006 10:05            

I have made excellent progress on a new Looping algorithm (which I will shortly post into this thread). The new algorithm features:

Better Looping algorithm (including proper !Duty controller)
Acceleration/Deceleration of loops, conforming to beat subdivisions

...and I am currently adding the final touches to the pitching effect!

Yet to be implemented (among others):

posted 27 October 2006 11:49            

BBCutforKyma.kym

Still a way to go, but getting closer

posted 27 October 2006 12:31            

It sounds good. I'll try looking at the code to see what makes it tick, although I haven't yet had much chance to look at BBCut itself still.

I replaced the '10 secs.aif' sampleFileDuration s in CaptureDuration of the memory writer with just "10 s," since I don't have your 10 s audio file. But I don't imagine that it makes a difference, does it?

posted 27 October 2006 16:34            

BPMdetector.kym

It's a pretty poor algorithm however! Any suggestions?

I was thinking about adding a parametric EQ (before the Amplitude follower) to aid in the detection process. I also wondered about writing some code to reject BPM values above say 160/below 80, and instead use the last generated value within that range. Maybe that code could then be adapted to make the algorithm more 'intelligent' towards beat subdivisions.

But I suspect these possibilities would only improve and not solve this problem.

posted 28 October 2006 11:00            

BPMdetector.kym

I attach my latest offering, which is an improvement on the first idea, but I don't feel qualified to take it any further (for the time being).

So I will proceed with the following implementations:

posted 28 October 2006 18:30            

AutoBreakz.kym

These are its main features:

Automatic (live) cutting of beats
Plenty of user-adjustable parameters
Loop stretching mode (via LoopSpeedSmoothing VCS toggle)
Authentic 'Pitch Sweep' effect (via PitchingEffect AND PitchingEffectSmoothing VCS toggles)
Capable of producing some wacky sounds!

I wasn't able to implement a successful tempo tracker for this version, so you'll have to adjust the VCS BPM fader if you want to keep the algorithm synched to your own samples!

Don't forget to check out the VCS presets

Hope you have fun...

Jack

p.s. Thanks to those who contributed ideas!

posted 29 October 2006 13:47            

You're too fast for me. I'm programming a variation of the idea, but I've taken a bit of a different approach (and yet not solved any of the problems that remain to be solved...). I'll post my results in a few days, most likely.

posted 29 October 2006 18:13            

In the mean time, I look forward to trying out your Sound!

posted 30 October 2006 19:57            

I've uploaded my sound to the Tweaky now:
http://www.symbolicsound.com/zzz/pub/Share/Sounds/BeatChopSequencer.kym

It's more of a step sequencer that you can program via the VCS than a randomly generated thing, but it gets fairly complex between the two different sequencers.

Since we're generating the loops differently, I'm not sure how easy it would be to combine our approaches, but it could be interesting to try.