Serum 2 - Spectral Sampler

[DanH]

@griffinboy it's probably as CPU friendly as it can be. Steve Duda said he 'spent millions' on Serum 2. Here's me holding one note on a short sample. Once you add unison and fx it goes up quick.

This is on a brand new M4 mac mini...

So did you build your own sampler to import into?

[clevername27]

@DanH Is that your plugin?

[griffinboy]

@DanH

pfff yeah like I thought
XD

Very expensive cpu wise. It's unavoidable. I've been looking at ways of doing a mixture of offline and online processing to achieve something similar. FFT / Additive stuff is always expensive.

[DanH]

@clevername27 I wish

[clevername27]

@DanH What does it do that you want to do?

[DanH]

@clevername27 sounds cool

[clevername27]

@DanH Ha, OK. If there is a specific technology you'd like to use in your plugin, I can tell you how to do it. I can't tell you how to do it in HISE—but I can tell you how to do it. Feel free to ask away…

[d.healey]

@DanH said in Serum 2 - Spectral Sampler:

One of the features people seem to be going most wild for is the Spectral Sampling. I wonder if this is something we might see in Hise one day?!

What is spectral sampling?

[DanH]

@d-healey sorry should have written spectral sampler. It's actually a spectral re-synthesis engine that takes whatever sample you throw into it and re-synthesises it, with interesting results. You can then warp it using multiple filter / fm options.

[clevername27]

@DanH @d-healey (My plugin basically does this.)

[DanH]

@clevername27 pray tell....

[d.healey]

@DanH said in Serum 2 - Spectral Sampler:

It's actually a spectral re-synthesis engine that takes whatever sample you throw into it and re-synthesises it

You can do this with Loris (GPL)

Also Paraphrasis (which uses Loris) might be of interest to you.

[clevername27]

@DanH See the videos here. The main thing my software does is convert audio recordings to these models—which can then be resynthesised. To be clear, much of this processing occurs outside of the HISE plugin. (Please excuse the idiot in the photo.)

Home - Under Construction - Virtual Audio Workstation

The virtual audio workstation A Music Production Methodology About the researcher Bill Evans, PhD Evans is an audio scientist, engineer […]

Virtual Audio Workstation (virtualaudioworkstation.com)

[DanH]

@clevername27 This is an awesome pug-in! Congrats! When you say the processing happens outside the Hise plug-in where does it happen exactly?

[clevername27]

@DanH Thank you, kindly. It happens in other software I've written (stand-alone, command-line, boring-looking).

[Allen]

I have an unfinished prototype does "spectral granular" with warp functions. Been considering including that in my synth for a long time though. I'll try if i can find a way to implement that in hise.

[clevername27]

@Allen I'd be interested in learning more…

[DanH]

@Allen very interested in the warp functions!

[Allen]

@clevername27
It's actually not really complicated, just a bunch of voices playing back different positions of a spectrum in different speed ratios.
I got this idea from a m4l device called "Pathways"
https://youtu.be/ljljrQ9QPiY?feature=shared

[Allen]

@DanH
I store the whole spectrum data in a N x 512 grayscale matrix, the pixel with the highest value in each column will be the fundamental frequency. And basically you can manipulate the spectrum according the pixel values in each column like shifting the vertical relative distance between each selected pixels and the fundamental frequency, or smearing the values horizontally.
Most functions which processing columns can be done in the play back stage by directly changing the frequency and amplitude of partials but those functions processing rows (time axis) will definitely need a separate buffer and some interpolation to make it smooth.