Vowel Filter

Would love to see the envelope vowel filter implemented as a dsp device.

use the chorus or flanger device, depth zero 100% wet, no feedback.

I was aware of that id still like to see a vowel filter dsp with its own set of parameters.

Sure, being able to somehow edit the formant frequencies and map keyframes on a slider (or xy) range would be a very welcome tool for me. Look at the zynaddsubfx vowel filter for the full nerdy implications. Right now we only have those fixed male croaky mid/high ish vowels. But the used filters do have a certain cool, talkboxish sound when driven and resonating, I think, unlike other formant filters. And I often wondered how they could sound with different formant data.

At least being able to shift the formants harmonically up and down by ratio would be a grand addition. I.e. drive them down to the monster realms, or up into mickey mousedeer’s squealings.

That would be the dream,i would settle for a basic vowel dsp though.

You can do 80s vox sound with the ring mod device or transposer Melda EQ doofer, too:

Attachment 6904 not found.

Patterns:

1: korg saw in renoise sampler

2: ring mod device on top, following key

3: Transposer-MeldaEQ doofer (following key)

Someone could modify transposer doofer to exactly match formant filter coefficients like (A,E,I,U,O and so on), using more bands. It’s quite fast.

Some stuff I miss in Renoise, too (off topic):

ringmod device IMHO should be integrated on waveform modulation level, so you could control it using adsr, on voice basis. Also a sync device…

EQ5 lacks of step filters like melda EQ has.

doofer only 8 knobs…

Melda stuff doesn’t load cross platform (melda adds .bin to executables on OSX)

cheers for that

IDK it sounds a bit weak. The ringmod is just for the grit? And why keytrack the formants, I thought the definition of a “formant” is that they are fixed frequency boosts (changing with the vowel definition though) of the carrier, and not ones that will move with the carrier freq, those would be elements of “timbre”?

I did 2 instruments using eq 5, for example with low the and high shelf building a -20 db reduction floor, and the 3 peaks tracked on formant frequencies. But with simple 2-vowel setups, because of the tedium to do more by hand. If you repeat the eq step by doubling it (with different q or even slight detuning), it will sound stronger, more defined. It can sound quite OK, a bit like the renoise vowel filter, but it is real tedious work to set up all the frequencies with lfo tables hydras etc in a doofer that you would wish to try it only twice, and then you start thinking about using a vst or coding a tool for that or so. Also the eq peaks can get badly piercing and/or glitching unless you hit sweet spots.

A cheap vowel type shit you often hear in dubstep is the “yoi” called stuff, you can create it by setting a sample and hold effect (lofimat “rate”) at 650-1500 hz to taste from monster to mickey mouse, and then put a reso lp or bp filter before it rising from 1/4 to 1:1 or even above of the snh rate frequency. I call it YOUURAAAUWRGH! heheh, gritty. If you are clever you can watch the peaks and generated aliasing reflections, and simulate it with a bunch of resonant filters to get the same guttural sound but cleaner without the bitcrush ratelimit mayhem.

Real champion design can be obtained by using parallel processing and bandpass filters or overlapping setups that are like bandpass or so. Again with formant data, or endlessly (but knowingly) tuned to sound like monsters. So, use a lot of sends and send channels (instrument FX are cool for this), and have lots of filters in parallel, and you can even give each formant its own sugar. You need a strong CPU for this.

Formant data (the frequencies and intensities of the peaks) can be found in boring fashion for first experiments on the web, if you wish to customize a lot more you can use speech processing/analysis software (or just a spectrum analysis and good eyes/nerves) on custom samples to find your data for experimentation.

Vowel or guttural like sounds are the real stuff for me. More intense anthropomorphic like speech like elements nothing can ever get. And any distortions on them you will instantly associate with distortions on human features. You already hear it in muffled ways in the belly of the mom before you ever even see any light of this world. It is the second thing you probably feel it is like you, right after the heartbeat.

1 Like

Well I really like the result using the transpose eq. It sounds 80s. never really found good singing example with a vowel filter. Can you show me? Regarding the transposing, it’s for easy modulation of age and maybe gender, too. And it will ensure that it doesn’t sound statically. A vowel filter kinda lacks of nice resonance to my ears. Do you know those 80s synth choirs on korg machines? It are just samples, so it’s filter frequencies are fully transposed, too! It’s not about sounding like a real human (such vowel filters never will), it’s about sounding nice.

There is transposer eq5, too, but since renoise eq filters lack of steepness and variations, it will not sound too convincing.

The ringmod example sounds a bit weak to my ears, too, but I just wanted to give you the idea. And it’s more as about adding interesting overtones to a sound. And IMHO it sounds weak because of the limitations of the ringmod device. If it was integrated in the sampler layer process just as it is in a real synth, it can sound very convincing.