Timestretching/Pitchshifting

[MikeB]

Will there be a timestretching/pitchshifting function in HISE at some point?

[d.healey]

@MikeB We all hope so. It's actually possible now but you have to implement a custom C++ node and bring in some library like Rubber and or elastiqe and this is out of reach for most of us.

[MikeB]

@d-healey Isn't Rubberband free for the time being under GNU General Public License - as long as no commercial applications are created?

[d.healey]

@MikeB GPL doesn't prevent commercial releases, only proprietary releases whether you charge a price or not.

[Soundavid]

@MikeB There is a Simple Pitch Shift in the Faust Libraries, is called Transpose I think.

[Christoph Hart]

Alright, I think the timestretching is ready for testing so anyone that wants to check it out should pull the feature/time_stretching branch.

GitHub - christophhart/HISE at feature_timestretch

The open source framework for sample based instruments - GitHub - christophhart/HISE at feature_timestretch

GitHub (github.com)

There's a simple core.stretch_player node that can pitch / stretch a single audio file and the sampler module has a timestretching parameter that can be set to different modes.

It's highly experimental and you might experience glitches with different samplerates / buffer sizes, but feel free to play around and let me know how it works / sounds for you.

[d.healey]

@Christoph-Hart Yippeee! Will give this a go very soon. Did you manage to optimize the FFT stuff?

[Christoph Hart]

@d-healey yes I swapped the fft with the juce wrapper which uses IPP / vdsp / FFTW. However the performance on my Windows system was already pretty good (about 2% per voice).

I‘m curious how the sound quality holds up against the alternatives - in my tests it‘s pretty good.

[d.healey]

I'm getting some errors, I'll post some below. They are all in the same file, pixman-arm-neon-asm.h.

In file included from ../../../../../HISE/hi_sampler/sampler/components/SamplerSettings.cpp:20,
                 from ../../../../../HISE/hi_sampler/hi_sampler.cpp:56,
                 from ../../JuceLibraryCode/include_hi_sampler.cpp:9:
../../../../../HISE/hi_rlottie/src/vector/pixman/pixman-arm-neon-asm.h:214:27: error: stray ‘#’ in program
  214 |     mov     TMP1, VX, asr #16
      |                           ^
../../../../../HISE/hi_rlottie/src/vector/pixman/pixman-arm-neon-asm.h:218:42: error: stray ‘#’ in program
  218 |     add     TMP1, mem_operand, TMP1, asl #1
      |                                          ^
../../../../../HISE/hi_rlottie/src/vector/pixman/pixman-arm-neon-asm.h:219:27: error: stray ‘#’ in program

Huh actually it's really weird because that file hasn't changed since you added it...

Oh I get it, it's something to do with the include in SamplerSettings, but I don't know what...

[d.healey]

Meh I commented out the line and it compiled just fine.

I'm using FFTW, the time stretching sounds really good. With a guitar loop I'm getting a little bit of what I'd call a bassy humming sound when it's stretched but with a percussive loop there are no issues.

Here's some of the loops I'm testing with - https://e.pcloud.link/publink/show?code=XZji1BZ3qTkeAWhYWkoVhYP2yNmVpaFRtIy

The bug where HISE's internal BPM is off by 10 is still there. When the project is compiled though it's fine.

Is it possible to go down to 25%? I know it won't be as "pleasant" but can be useful for more sound design stuff.

[Christoph Hart]

@d-healey haha, I'm using a new tool within VisualStudio (Resharper C++) and I think I've pressed a wrong shortcut and it inserted this completely idiotic include statement. I also got the compilation errors but I thought it was a glitch so I removed the file locally...

I can't reproduce the BPM issue - is this Linux only?

And yes the timestretching algorithm can cope with extreme ratios (and it even starts to randomize the phase outside of the 0.5...2.0 range to get a PaulStretch-like sound. I've just added this limitation arbitrarily so I don't have to deal with buffers that get too big...

[d.healey]

@Christoph-Hart Yeah it could be a Linux thing, I haven't tested on other systems.

[audiolyf]

@Christoph-Hart Very cool! Where can we poke around to lower the limit? And is there a way to reach the stretch ratio of the Sampler module via scripting ?

[Christoph Hart]

And is there a way to reach the stretch ratio of the Sampler module via scripting ?

Sure:

Sampler.setTimestretchRatio(newRatio);

[Christoph Hart]

I've updated the docs to include a chapter about how it works:

HISE | Docs

(docs.hise.audio)

[d.healey]

@Lindon Need to update your VI thread :)

[ulrik]

@Christoph-Hart Much appreciated Christoph!

[audiolyf]

@Christoph-Hart Nice! Is getTimestretchOptions() not implemented yet?

[d.healey]

@audiolyf I can see it in the repo - https://github.com/christophhart/HISE/commit/3c8711f0713f445d197fa6b7d8f9e4819de114f6#diff-92b4316aaf5f04995603daa0550fbbb9156f53dbb5061d9608eacb18fb356d73

[Lumi283]

Thank you Christoph Hart for this feature. My project was stuck, but now the problem is solved! Don't you think @d-healey that it would be interesting to offer a tutorial on your YouTube channel to explain how to properly implement time stretching? It could help more than one person.
Thanks,
Remi Dufeu