Smoother Dynamic Crossfades

[d.healey]

@VirtualVirgin said in Smoother Dynamic Crossfades:

I can't seem to find any more information on the patent, but if I come across something I will post it.

I have asked them directly about this and they have not responded. As far as I can tell there is no patent. But if there was a patent, which they say was filed in 2004, then it expired this year (assuming it's filed in the US and they didn't get any extension).

Harmonic alignment/phase alignment is not the same as morphing. There are a number of methods for phase alignment, I have a tool that will phase align sustain samples, you can also achieve pretty much perfect phase alignment by flattening the pitches of samples, but it does remove some of the humanity from them so you need to add it back with modulators.

https://youtu.be/I9YvUf_Hzh0?si=YEDcjwmtjIQTkmUb

[Robert Puza]

@Christoph-Hart That's amazing. Hise still surprises me.

[HISEnberg]

Thanks for sharing the video @d-healey !

I think phase alignment is more of an immediate concern for me, personally. Spectral morphing (which would be a bit different then harmonic alignement) sounds like a cool topic, but I imagine i would just go down the wavetable route to accomplish a similar result.

You said you have a "tool" that does this, is this a HISE script which will track the 0-crossings of the sample and delay the transitions of one note to the next (similar to what is described in the Ivy video)?

I am trying to wrap my head around how this would be accomplished. Would you use three different Samplers to enact that transition, or are you able to accomplish it using Round Robins?

I am picturing having to write a MIDI processor that would:

• Track the sample’s playback position.
• Detect zero-crossings in real-time.
• Trigger a phase-aligned transition when a new MIDI note is received.

Am I on the right track or is this over-thought and you have a better method?

[d.healey]

@HISEnberg It's a standalone application I commissioned a few years ago. It takes in multiple samples of different dynamics and outputs the same samples phase aligned to each other. It's not perfect, sometimes the attacks sound a bit weird, but most of the time it does a good job.

@HISEnberg said in Smoother Dynamic Crossfades:

Would you use three different Samplers to enact that transition, or are you able to accomplish it using Round Robins?

I use the built in group crossfade feature.

If you want to build a similar thing in HISE I think Loris would be your friend.

[HISEnberg]

@d-healey yes thank you.

I was just re reading the post on @iamlamprey NEATBRAIN system to see if I could gain some insights on this topic. I haven't really dug into Loris but this seems like a proper opportunity to do so.

[iamlamprey]

@HISEnberg NEATBrain doesn't use any crossfading shenanigans, it just loops a single cycle at its fundamental frequency to simulate a wavetable synth

[clevername27]

@d-healey Isn't morphing just interpolating between the amplitudes of FFT bins?

[d.healey]

@clevername27 I don't know

[Robert Puza]

@Christoph-Hart When I HEARD your snippet - I knew I wanted that filter for sustain. Harm. Filter is suuuuper

[modularsamples]

Excuse my naivety on the subject, but In scriptnode alone we have the ability record, analyse and import wavs, generate sine waves and have extremely large sliderpacks (if I recall, this was bumped up to something like 4000+ in a recent update). All the important stuff needed for spectral resynthesis, no?

[HISEnberg]

@clevername27 I have generally understood it this way, though I think there are a lot of methodologies. From what I can tell there isn't a standard for spectral morphing.

[HISEnberg]

@modularsamples Hypothetically, yes, I think this would be extremely complex and CPU intensive to implement though.

[clevername27]

@HISEnberg @modularsamples @Christoph-Hart @Robert-Puza @d-healey I don't know if this is helpful – I'm using additive synthesis for the sustain portion of my sounds, without phase issues. Here's a screen-capture movie with sound.

[modularsamples]

@clevername27 Nice, that ring is really pleasing. This is kind of attention to detail that I aspire to.

Just gonna throw this into the mix:

Design and Run Real-time Spectral Processing on the Web with Faust
https://inria.hal.science/hal-04507625/document

And to rewind a bit, what about an equal power crossfade? Shouldn't that be a bit more forgiving when using material that' s not been phase aligned? I used to do this all the time in Logics EXS24 and the equal power fade was almost always the way to go.

[d.healey]

@modularsamples said in Smoother Dynamic Crossfades:

what about an equal power crossfade?

EP crossfade it the standard approach that's been used for many years for dynamic layers. It doesn't solve the chorusing issue though which is why phase-alignment (or audio morphing) is necessary. I should say that this is mostly an issue for solo instruments, with ensembles it's less of a problem.

[modularsamples]

@d-healey I notice it quite a bit with synthesised samples too, hence my interest, even more annoying is the tendency for crossfaded sounds to have volume dip between layers. Smoothing helps a bit (smoothed parameter in low pass mode in particular) but it's always there to some extent.

It's a shame physical modelling (of real instruments) never really fulfilled it's promise, but then many of us might not be here, or maybe we'd be talking about HIPME.