Latency of pitch shifting algorithms cage fight

[Chocol8]

Maybe I’m dumb, but aren’t any of the shifters NOT doing that doing some fancy guessing and not truly detecting a pitch?

Yes, although people will debate the wording like guessing or cheating, any low latency pitch shifter is NOT accurately detecting pitch. It shows in the results.

You can embrace the artifacts and make it part of your sound, eg Jack White, you can tune one of the multiple guitars you most likely already own and not use a pitch shifter, or use the need for another tuning as an excuse to buy more guitars.

You cannot solve the problem with more CPU or better algorithms.

 

[Sascha Franck]

Maybe I’m dumb, but aren’t any of the shifters NOT doing that doing some fancy guessing and not truly detecting a pitch?

From my understanding, the only ones at least attempting a certain level of "correctness" are those in fact analysing a full wave cycle. Just that they likely still don't sound great *and* introduce partially massive latency.
Anything else is some kinda plain mathematical operations, sometimes apparently "masked" (no idea how), which often leads to less than desirable results. Hence exactly what most of the existing tools in most modelers suffer from. I wouldn't call it guesswork, though, it's likely math operations that don't work well for the purpose. At least that's how I understand what I've read.

 

[Orvillain]

You cannot solve the problem with more CPU or better algorithms.

There's work going on in the machine learning space to see if neural networks can be leveraged to give better results. But I really don't know a whole ton about that yet.

 

[Chocol8]

There's work going on in the machine learning space to see if neural networks can be leveraged to give better results. But I really don't know a whole ton about that yet.

Sure, you can make better guesses, as banned pointed out strings have somewhat predictable harmonics and there are other “clues” that could be used to improve predictions. That’s going to still add latency, although maybe less, and require significant processing. In that case, more processing power COULD simultaneously reduce latency and improve the predictions but that’s is currently going to be some pricey hardware dedicated to a task that is pretty minor compared to the rest of what a modeler is expected to do.

How much more would anyone pay for a Stadium Plus/Axefx 4 Plus that includes a few Nvidea GPU’s dedicated to a much improved but not perfect pitch shifter that still adds 3-5 ms of latency? It would be cheaper to buy several other guitars.

 

[FuzzyAce]

Yes, although people will debate the wording like guessing or cheating, any low latency pitch shifter is NOT accurately detecting pitch. It shows in the results.

You can embrace the artifacts and make it part of your sound, eg Jack White, you can tune one of the multiple guitars you most likely already own and not use a pitch shifter, or use the need for another tuning as an excuse to buy more guitars.

You cannot solve the problem with more CPU or better algorithms.

To me, pitch shifting + quirks = fun and interesting effect.

 

[Lysander]

Do you need the full wave to detect pitch? Assuming a pure Sine wave

Not a dumb question at all, mind you, but you can stop right there. There's no such thing as pure sine waves in audio

On the other hand, and this is from very cursory knowledge on the matter, assuming sine waves is exactly what you do for FFT-based pitch shifting, which decomposes signals into fundamentals. The problem is that FFT introduces latency of its own.

 

[molul]

So, is the conclusion "accurate pitch shifting algorithms won't ever benefit from more CPU"?

 

[LP59]

Do you need the full wave to detect pitch? Assuming a pure Sine wave (which is not what a guitar low E string is), wouldn't half a wave suffice (0 to peak to 0)? Or even a quarter wave? (0 to peak...would need to detect that maxima and falling edge)

Now assuming a regular E string which will also have a bunch of harmonics - could the harmonic makeup be used to also detect probability that it is an E string without waiting for the fundamental to complete its cycle?

Sorry if dumb questions.

You can get a pretty good correlation from less than a full wave. The fact remains however, that the longer the period, the more samples you require for the detection. Or, as Orvillian put it, the latency is dictated by the frequency at the low end of the range your realtime pitch shifter supports.

 

[laxu]

So, is the conclusion "accurate pitch shifting algorithms won't ever benefit from more CPU"?

Maybe not for detection but for the actual pitch shifting.

 

[Lysander]

So, is the conclusion "accurate pitch shifting algorithms won't ever benefit from more CPU"?

"Accurate", absolutely. What you won't get is latency improvements by horsepower alone. That is not a problem you can just throw CPU at.

 

[Dimiguitar]

Can I expect anything better from plugin pitch shifters than what we have from fractal fm3, Kemper, etc?

Found this -- https://www.polychromedsp.com/hypertune-metal/

 

[AlbertA]

Can I expect anything better from plugin pitch shifters than what we have from fractal fm3, Kemper, etc?

It's going to be a latency tradeoff - Apple's AUNewPitch shift is pretty good - but you wouldn't use it for real-time, it's like 80+ms of latency.

 

[DrewJD82]

 

[Ed DeGenaro]

To get that off the Helix thread:



This is just wrong. I just transposed an E1 up 2 semitones with the HX Poly Capo and it kicks in at 13ms.

You mean you actually bothered to grab a bass?

 

[Frodebro]

This thread is

It’s like Armistice Day never happened and we’re still on the western front…

 

[AlbertA]

You mean you actually bothered to grab a bass?

 

[GLiDE-BPM]

View attachment 55647

ROFLMFAO!!!!!!!!!!!!!!!!!!!!!!!

 

[TheTrueZoltan!]

This thread is an infinite source of new insights and wisdom. I have already learned so much and feel enriched by it in my life. Thank you!

 

[Sascha Franck]

You mean you actually bothered to grab a bass?

I always have two with me. Just in case the bass player has a heart attack or so.

 

[Ed DeGenaro]

I always have two with me. Just in case the bass player has a heart attack or so.

Wishful thinking?