Spectral MPhatik

[vectorwarrior]

I was doing some tests with a multiband Phatik and it created some really interesting creative effects. It's similar to vocoding or morphing but cleaner and closer to the original than either of them. Unfortunately 6 bands isn't enough and doing more bands requires repeat crossover filter and with sharper filters... which starts to sound pretty gross on both input signals. So I wondered if a spectral Phatik could be a possibility.

I'm curious to know how different it would sound to Morph, as there's obviously some similarities. But my understanding is that Morph is essentially just a spectral vocoder... which means it doesn't actually match signals, but instead acts more as a massively multiband envelope follower... which isn't the same as matching as there is no 'comparative' analysis going on.

Example, with morphing, if the main input contains a loud 1khz signal it allows the second input's 1khz signal to pass through.... but it doesn't MATCH the two signals together. There is therefore less conformity between the two signals than this hypothetical spectral phatik.

I'm thinking I may be wrong and morph IS doing this, but I wanted to ask. It's incredibly difficult to know what's really happening behind the scenes.

[jmg8]

Yes, I have thought about this too. Although I always assumed that MMorph IS spectral MPhatik.
He said that MMorph was similar to a spectral vocoder but not the same. MVocoder goes up to 100 bands and has morph modes, so to me, it would be too similar to MMorph. Also he released MPhatik around the same time as MMorph, maybe this is because they share the same tech? I might be wrong, but id be interested to find out.

[MeldaProduction]

Hmmm, well, seems that fact that I know how the plugins work prevent me from understanding this . I quite don't see too much MPhatik and MMorph have in common really . MMorph is indeed based on some extended spectral vocoding (it's rather more complex though), while the point of MPhatik is to apply a dynamics of one signal to a processed version of it (or even some other signal). I quite don't know what spectral version would do, but it would be a huge CPU hog. And if you'd loose the follower it would essentially become a convolution . So, well, I just don't know ...

[jmg8]

Hmmm, well, seems that fact that I know how the plugins work prevent me from understanding this . I quite don't see too much MPhatik and MMorph have in common really . MMorph is indeed based on some extended spectral vocoding (it's rather more complex though), while the point of MPhatik is to apply a dynamics of one signal to a processed version of it (or even some other signal). I quite don't know what spectral version would do, but it would be a huge CPU hog. And if you'd loose the follower it would essentially become a convolution . So, well, I just don't know ...

But convolution is static. This would be like a constantly evolving IR.

[vectorwarrior]

Hmmm, well, seems that fact that I know how the plugins work prevent me from understanding this . I quite don't see too much MPhatik and MMorph have in common really . MMorph is indeed based on some extended spectral vocoding (it's rather more complex though), while the point of MPhatik is to apply a dynamics of one signal to a processed version of it (or even some other signal). I quite don't know what spectral version would do, but it would be a huge CPU hog. And if you'd loose the follower it would essentially become a convolution . So, well, I just don't know ...

A constantly updating convolution would be very interesting!

But yeah, i know that the ponit of MPhatik is to apply dynamics from one signal to another, but think about doing it in multiband... and then extend that thought to doing it spectrally across the whole signal. Matching each frequency's signal constantly with another is a pretty interesting idea... especially if it's distinct from MMorph.

Anyway, it was just an interesting idea that I was thinking out. I know you're busy and have a master plan for the future, but was curious what you and others thought of the idea.

[MeldaProduction]

Actually it depends on what you are convolving . Try Mixer module in MXXX, 2 sources, set the mode for the second one to Convolution .

[jmg8]

Actually it depends on what you are convolving . Try Mixer module in MXXX, 2 sources, set the mode for the second one to Convolution .

You said it would be a CPU hog. Plus I've tried the mixer mode and it sounds nothing like MConvolution.

[vectorwarrior]

Yeah, I was super excited when I saw convolution in the mixer module in MXXX when it first came out, but am yet to get anything remotely useful out of it.

Anyway, no worries, just wanted to throw this idea out there

[pone]

"Example, with morphing, if the main input contains a loud 1khz signal it allows the second input's 1khz signal to pass through.... but it doesn't MATCH the two signals together." I think what you're looking for it SpectralDynamics. Or do you mean that the frequencies of the original would shift closer to that of the sidechain?
ReaFIR has realtime L/R convolution and as expected it's a cpu hog and is filled with artefacts since it's realtime.

[vectorwarrior]

"Example, with morphing, if the main input contains a loud 1khz signal it allows the second input's 1khz signal to pass through.... but it doesn't MATCH the two signals together." I think what you're looking for it SpectralDynamics. Or do you mean that the frequencies of the original would shift closer to that of the sidechain?
ReaFIR has realtime L/R convolution and as expected it's a cpu hog and is filled with artefacts since it's realtime.

No, SpectralDynamics set to side chain doesn't do any comparison between input and sidechain... it just reacts to the dynamics of the sidechain in comparison to the transfer graph... and places that reaction on the main input. What I'm talking about is comparing the input to the sidechain and making the necessary adjustments to match them across all frequencies. So... close but no cigar

[pone]

I still don't understand then what is getting matched and what is compared. It sounds like a spectral vocoder, which it can do and morph kind of does. But, nevermind....