Waves released Vocal Bender

[yellowmix]

It achieves zero latency by delaying the signal a bit. You can test it yourself. Duplicate audio to another track. Process it with Vocal Bender and render it. Compare the original to the processed version. You'll see, by my count about a 512 sample difference. That's a common FFT block but the accuracy is kind of low so it's probably doing wavelet or granular processing.

512 samples is a little over a millisecond at 48 kHz so it's otherwise not really noticeable, but you may still want to realign it. This also has implications for mixing in a send configuration (I assume its internal wet/dry is compensated but have no way of testing). Little Altarboy, with 2255 sample latency, would incur 4.6ms at 48 kHz which is noticeable which is why it's honest, so the DAW can perform compensation.

It sounds good and the modulation gives it a lot of creative flexibility, and the presets are interesting. Very sure we'll be hearing them in published songs soon. GUI is resizable in increments from 75% to 200%.

[pixel85]

Hmm sounds like mumble rappers will have new toy after autotune: thisplugin with 1/16 LFO on pitch and formants. Honey for my ears!

[Aloysius]

Wasps in other peoples ears.

[briefcasemanx]

Why would one cycle delay be necessary?

Read up on different analysis methods. They exhibit a definite uncertainty between time localization and and frequency resolution. Even fast, variable window analysis methods will need a full cycle to resolve the pitch of a signal. This will have (guessing based on other methods) a variable but very small latency depending on the fundamental frequency of the input. That's about as good as you can wish for unless you tell the plugin what pitches to expect.

I'm not well versed enough to fully understand everything you just said. I'm not sure why you'd need to know the fundamental frequency or do any analysis of the signal. You can't just take the audio stream and pitch everything in it up or down regardless of frequency content within the stream?

I definitely remember using pitch and formant shifting plugins way way back from companies/devrlopers that I'm reasonably certain didn't have any pitch detecting algorithms is n their arsenal. Like way back when pitch detecting wasn't really a thing.

Pitch detection is usually a part of a "clean" pitch shift. Ideally you want to be shifting "grains" or whatever terminology/ method at the frequency of the fundamental or some multiple of it. Maybe you've played with simple granular pitch shifters and noticed that with small grain sizes or weird widowing, you get a pile of AM type artifacts. With the window size set high enough these go away but then lead to a "stuttering" sound like old hardware samplers' time stretching algorithms. The better you can match these, the smoother you can make your output. Many things use an envelope follower to track the input so high frequency transients shrink the window and sustained bits use a longer window.
With sample/ audio playback the sample can be analysed on loading and then played back optimally. On a realtime signal you still want to collect enough data to have your window or buffer at the optimal size. This is one reason why most real-time/ almost real-time shifters or tuners want you to pick the pitch range of your input. It keeps them from looking outside of the most meaningful values of input and wasting time/ processing.
If you do some reading on fft and wavelet transformation as well as some adaptive granular stuff it becomes clear how analysis leads to a smooth processing of a signal.
Anyway at a minimum a "vocal shifter" is going to be looking only in the range of fundamental frequencies that can be achieved with the human voice and setting the buffer to either a sliding scale within that range or maybe the lowest expected range at all times.

JJ

Cool, that makes sense. And yeah I remember bad artifacts like you're describing on super old pitch shifting algos.

Thanks for the explanation.

[Cancel Culture Club]

Wasps in other peoples ears.

I believe you might be confusing your devices...

[teilo]

I have Little Alterboy and MTransformer.

MTransformer never seems to get much attention, but it is an amazing plugin. It is far more flexible, less grainy, and comes with all the modulation typical for Melda plugins. It can do typical pitch and formant shifting, and a whole lot more, with control over many more parameters. Since you can adjust the transformation curve and the spectral envelope, it is capable of a lot more exotic effects, such as literally making you sound like a Transformer, or a demon, etc. I'd love to see a head to head with MTransformer and Vocal Bender.

[Ploki]

MTransformer has latency tho

[teilo]

MTransformer has latency tho

Yes, it is a concern, and the big selling point for Vocal Bender. I use MTransformer in web meetings all the time, and it works fine there. Most of the time I use it in post anyway, so latency is not a concern for my use cases. I'd still like to see a head-to-head.

[Cancel Culture Club]

Try doing this...

[Aloysius]

Wasps in other peoples ears.

I believe you might be confusing your devices...

[pixel85]

Try doing this...

Tell it to Infected Mushroom

[accessdune3]

I wonder how they define zero latency. The DAW alone will introduce latency already... I don’t know of any pitch shifting technique that would allow zero latency... Technically at least one cycle in its frequency range would be necessary... (Heisenberg I hear you callin’...)
But anybody who knows better could explain the DSP principles behind a zero latency pitch shift. Maybe its a marketing principle though...

Guitar rig 5/6 (pitch shift modules), vocal synth 2, manipulator all do pitch shift in zero latency

Love this plugin and the gui is actually nice looking (which is a surprise) and not analogue gimmicky looking

[Weasel-Boy]

QuikQuak’s Pitchwheel has been around for 10 years and does the same job...with a few extra wrinkles.

https://www.quikquak.com/Prod_Pitchwheel.html

[h66]

I expected these results from my useless Waves OVox

[Ploki]

gotta be honest it has a cool limited feature set (Freeze, pitch and formant) and formant sounds fairly clean