Paul's stretch

[k brown]

I wonder what Martin's house looks like, with all that code 'lying around' ?

[Spogg]

My scheme takes a snapshot every 512 samples and calculates the spectrum (using a windowed 2048 point FFT). Based on the extracted frequencies and amplitudes, it will drive sine oscillators for the next 512 samples. After that, the next snapshot is used to update the osc frequencies and amplitudes, and so on. There is some sooth interpolation regarding the amplitudes in order to avoid jumps every 512 samples.

So not exactly granular synthesis. How would you synthesise an arbitrary (not single ppitched) sound from grains?

I agree it’s not exactly granular synthesis by some definitions. But the principle of breaking down audio into small chunks and processing those could be described as granular (by me anyway ).

Use of the term synthesis, which I take to mean assembling a whole from parts, can be misleading in this context. Really what we have here might be better described as a granular effect. An input is processed and modified but the output is dependent on the input and how it’s processed, rather than an output being wholly generated. It’s just semantics!

When I said that Paulstretch had a more granular sound, my ears tell me that there is more than 1 sound playing simultaneously, like overlapping grains. Another clue is that he provides a window setting, and the shapes available are classic for granular techniques to crossfade grains to avoid discontinuities and thus clicking. Of course it could be that the chosen window is applied to the wav sections used for FFT, to avoid high frequencies which would affect the FFT outcome.

End of waffle!

[martinvicanek]

BTW this is the PaulStretch algo: https://46dd8e9b-a-d30f3c5d-s-sites.goo ... 673301.png

[Spogg]

Hey thanks Martin.

I hadn’t seen that before and it’s what I was trying to deduce.

I no longer need to puzzle about this!

[martinvicanek]

With Paul's scheme lying there in front of me, I could not resist the temptation to put it into practice. So here is the result, maybe somebody will have fun with it.

[Spogg]

Once again… AMAZING!

Out of the box this one sounds like Paulstretch which I guess is not surprising given it’s using the same algorithm.

Messing around with the offset values creates some interesting results, as does the FFT size. Dare I say it actually gives a wider variety of results?

Bravo!!

[BobF]

Well Martin,

This is better than what I asked for, going back several years now (that is live time stretching, viewtopic.php?f=3&t=5915&p=28267&hilit=time+strech#p28267). Could this be applied to that principle you came up with way back then? I think it's been about 5 years now, you have really learned a great deal more. This is really great , thank you so much!

Having great fun with it, BobF.....