Concatenative Synthesis

nordmodular · Post by **nordmodular** » Sun Feb 10, 2008 12:23 am

Hey all,

Is there anyone who could perhaps point me in the right direction as to where I could find info about Concatenative synthesis? I making a Concatenative synthesiser in SuperCollider for my third year project at uni and wondered if I could find any hints/tips from someone who may have used/made one.

Thanks in advance

Yoozer · Post by **Yoozer** » Sun Feb 10, 2008 12:27 am

Concatenative synthesis

Is that generating several waveforms and then gluing them together?

nordmodular · Post by **nordmodular** » Sun Feb 10, 2008 12:50 am

Principle

Corpus-based concatenative sound synthesis uses a large database of source sounds, segmented into units, and a unit selection algorithm that finds the sequence of units that match best the sound or phrase to be synthesised, called the target.

So the answer is yes

The main problem i'm having is trying to come up with a system that organises the samples in a neat array which can constantly be updated and allows for an easy change in the samples...

synapsecollapse · Post by **synapsecollapse** » Sun Feb 10, 2008 12:54 am

im currious: how is this different from wave sequencing via the wavestation? (or is it at all?)

what are you studying and where. i am working on mt BSEE in the US and i dont think i will ever have a school project that neat sounding.

nordmodular · Post by **nordmodular** » Sun Feb 10, 2008 1:03 am

I'm studying at uni doing Music Informatics. Haven't used a wavestation but by the sounds of it that would be a very basic form which woould have to be preset by the user. Concatenative synthesis uses an algorithm to 'match' the users desired sound from a bank and changes their pitch, filter etc...

GeneralBigbag · Post by **GeneralBigbag** » Sun Feb 10, 2008 1:08 am

So is this basically taking a granular approach to sampling, i.e., you take your original waveform, load it up in the editor, the editor then trawls through a huge database of sound grains, matching spectra as much as possible, until it has recreated the sound, at which point, you can take the output and play it at any pitch or duration you want because it is made up of these grains?

nordmodular · Post by **nordmodular** » Sun Feb 10, 2008 1:11 am

yeah! it's kind of a stage on from a granular synth. It's just taking me a while to get my head around organising it all

Post by **nathanscribe** » Sun Feb 10, 2008 1:11 am

Ah, but can it sound like a bassoon?

crystalmsc · Post by **crystalmsc** » Sun Feb 10, 2008 9:25 am

you might find some useful information from the Nord Modular book as the synth is also a great resource for educational needs and the SOS workshop. as for the VST, you may want to check the Crusher.

Yoozer · Post by **Yoozer** » Sun Feb 10, 2008 10:06 am

nordmodular wrote:Concatenative synthesis uses an algorithm to 'match' the users desired sound from a bank and changes their pitch, filter etc...

Pitch is not that much of an issue; frequency content is. You could consider making a "fingerprint" of each fragment over a fixed lenght (block) so you get an average frequency plot. Then the trick is to compare those fingerprints and see which looks the most like the other. You could either build "strands" of a series of lookalike plots that follow a fixed evolution or graphs where you connect certain fingerprints that can serve as lookalikes for two different strands.

I don't think Csound does all of this so you might want to consider using C++ with a FFT library and homebrew algorithms to generate fingerprints.