Formant filter

[baconpaul]

Chatting with @JackyLigon on slack this morning and he suggested a new filter type, the Formant filter.

I said

And Formant filter - you mean just ‘3 poles medium-tight set of bandpass in one filter spread to human sounds’ type thing? Or something more than that

And he said “yes”

And then I suggested we would code up this table

From https://www.phon.ucl.ac.uk/courses/plin0065/week5.php and he said “yup”

Then we concluded

The DSP for that is super easy to construct in surge. Need to think about how it would fit into the filter banks. Presumably we would want controls for where in Formant space you were though so more than just cutoff and resonance. Something like ‘vowel sound’ and ‘base frequency’ and ‘resonance’ seems to be what you would want

In theory my skinning engine in 1.7 lets us add a parameter. So we might be able to do it. It is a nice small unambiguous change for 1.7 that proves our point too

And ended with

As we know, routing a noise component, as well as the oscillators, through a format filter, can produce some quite extraordinary timbres of a vocal tract nature.

[mkruselj]

One thing that would make this more usable is being able to choose between which vowels you really want to morph.

Just going through a list of vowel sounds as in that table is repetitive and boring. But if we could pick, say, 5 different vowels, and have a parameter to morph through them, then we're in for a lot more fun!

Of course, then it's not adding just a single parameter, alas...

[baconpaul]

most of the research seems to show the vowels laid out in an x y plane with 'dead spots'. probably some idea there.

[mkruselj]

Here's a brilliantly executed formant filter:

https://youtu.be/rP7ILrztu80?t=90

Important stuff starts at 1:30 onward.

[mkruselj]

@baconpaul if this is added as an FX first, then it likely doesn't require param list change, maybe? :)

[baconpaul]

true. but then it's not as fun since you can't do polyphonic framptons

[mkruselj]

We could borrow MorphEQ from Shortcircuit here. It's bog-standard biquads, but maybe we could update it with something a bit nicer/more modern (not necessarily non-linear, but perhaps ZDF) with @jatinchowdhury18's help. 🙂