Formant synthesis
-
- KVRer
- 13 posts since 8 Mar, 2020
Basically this method means that every single wavetable you ever owned is now a unique formant-vocal wavetable. Just throw it in a granulat synth and fiddle with the grain pitch and grain envelopes.
-
- KVRer
- 13 posts since 8 Mar, 2020
Formant sounds are most often created by filtering on an everyday basis. Tuts for formants out of EQs and phasers are everyday examples,
Oscillator hardsync in some trad. analog synths was a notable exception- formant control (a "magnifying lens on the groups of upper harmonics") out of just an oscillator (a re-synced oscillator with 2 freq inputs and one complex waveform)
Now, we have fast powerful granular synthesis available. With just a sinewave for grain waveform you get somethibg that sounds autonatically like a bandpass or lowpass on a saw - but you can hear the changes in your grain envelope as changes to the filters's personality.
You switch the sine to "Serum_wavetable.wav" and now that bandpass/lowpass has turned into a unique filterbank. Slide the grain pitch around without touching wavetable, and instant yuargles. Layer two voices where the grain pitched go in opposite directions, and the yuargles are now sick growly psytrance leads.
Its actually a lot like oscillator hard sync but with an amplitude envelope applied (the grain envelope).
Grains can overlap or have gaps, which makes for very subtle tuning of the sound.
Formant filters of a "designed vocal sound" are examples of a variation on the filter-as-formant idea. Indeed, the term formants is just an arbitrary definiton from the field of studying real-life filterbanks (vocal cavities) applied to the abstract definition of an instrument. The famous Stradivarian violin "tone", if you want, is made by the body of the violin acting as a filterbank and subtly making formants, so this if you want, Grump, is where the Y axis has been conquered before. If we're not trying to make a Strad but ee want a strong formant at 990Hz and a weaker one at 1776Hz while we noodle around a melody in the sub-bass we want a building block which responds automatically, with the entire spectrum to play with, which cleably translates the frequencies and waveforms we select into formant bands with adjustable shapes. For the simpleet case we select a sine and 880 Hz, we get a single formany at 880. We select a waveform with harmonics 1 & 2 in it - formants at 880 and also around 1760. We select a "crazy waveform" - or we modulate a wavetable index - and we get a wonky shape up and down the spectrum like.
You ever use big wavetables as Impulse Responses to process a synth and get formants? It works erally well, but you have no direct mapping of controls-to-formants. Its all indirect by changing the wavetable.
Granular works a lot like that. But the difference is, in granular your controls work on the granular parameters - which happen to directly form the formants. So you get direct control. With granular, you never have to change the wavetable, but still make "animal noises" by jittering the speed of the wavetable inside the grains. You can also think of it like AM synthesis. No doubt a sophisticated hardsync+AM oscillator can do something very similar, in fact. But with granular you have more control - the grain envelope can be controlled to change the shape of the formant and size, the grain pitch to slide it around the piano roll, from about the second harmonic of your fundamental up to the 4-5kHz and higher. If you use it a sinewave or wavetables built from sine waves this method is also intrinsically bandlimited.
Vosim is a similar implementation of granular with grain pitch controlling the formant frequency.
Oscillator hardsync in some trad. analog synths was a notable exception- formant control (a "magnifying lens on the groups of upper harmonics") out of just an oscillator (a re-synced oscillator with 2 freq inputs and one complex waveform)
Now, we have fast powerful granular synthesis available. With just a sinewave for grain waveform you get somethibg that sounds autonatically like a bandpass or lowpass on a saw - but you can hear the changes in your grain envelope as changes to the filters's personality.
You switch the sine to "Serum_wavetable.wav" and now that bandpass/lowpass has turned into a unique filterbank. Slide the grain pitch around without touching wavetable, and instant yuargles. Layer two voices where the grain pitched go in opposite directions, and the yuargles are now sick growly psytrance leads.
Its actually a lot like oscillator hard sync but with an amplitude envelope applied (the grain envelope).
Grains can overlap or have gaps, which makes for very subtle tuning of the sound.
Formant filters of a "designed vocal sound" are examples of a variation on the filter-as-formant idea. Indeed, the term formants is just an arbitrary definiton from the field of studying real-life filterbanks (vocal cavities) applied to the abstract definition of an instrument. The famous Stradivarian violin "tone", if you want, is made by the body of the violin acting as a filterbank and subtly making formants, so this if you want, Grump, is where the Y axis has been conquered before. If we're not trying to make a Strad but ee want a strong formant at 990Hz and a weaker one at 1776Hz while we noodle around a melody in the sub-bass we want a building block which responds automatically, with the entire spectrum to play with, which cleably translates the frequencies and waveforms we select into formant bands with adjustable shapes. For the simpleet case we select a sine and 880 Hz, we get a single formany at 880. We select a waveform with harmonics 1 & 2 in it - formants at 880 and also around 1760. We select a "crazy waveform" - or we modulate a wavetable index - and we get a wonky shape up and down the spectrum like.
You ever use big wavetables as Impulse Responses to process a synth and get formants? It works erally well, but you have no direct mapping of controls-to-formants. Its all indirect by changing the wavetable.
Granular works a lot like that. But the difference is, in granular your controls work on the granular parameters - which happen to directly form the formants. So you get direct control. With granular, you never have to change the wavetable, but still make "animal noises" by jittering the speed of the wavetable inside the grains. You can also think of it like AM synthesis. No doubt a sophisticated hardsync+AM oscillator can do something very similar, in fact. But with granular you have more control - the grain envelope can be controlled to change the shape of the formant and size, the grain pitch to slide it around the piano roll, from about the second harmonic of your fundamental up to the 4-5kHz and higher. If you use it a sinewave or wavetables built from sine waves this method is also intrinsically bandlimited.
Vosim is a similar implementation of granular with grain pitch controlling the formant frequency.
-
- KVRian
- 1189 posts since 11 Jun, 2019
That is my latest practical Approch to the Topic. FM with Multisamples + X within the MSF. Tests are ongoing. It´s for sure Trial and Error, but Yamaha´s Results are promising and with FM (RM, Sync, ...) I can give the static Samples some spectral Dynamics back.
Halion is my only "Synth" that has a Grain Engine, but I fear it lacks some of the mentioned Controllers. Do you have any Suggestions concerning a VST that can perform the mentioned Approach?
Very inspiring by the Way. I have to state that I´m often too much concerned with all the Materials that I have produced already. And with all these spectral and Mastering Effects for sure
VOSIM
- but not my Builing Site!-
- KVRian
- 1189 posts since 11 Jun, 2019
Little Demo File: Dull/muted VOX Pad + 2% Sync with an other VOX Pad (+32 ST) > new high Formants. Very promising, but I´ll have to develop more complex Models.
cheers
cheers
You do not have the required permissions to view the files attached to this post.
-
- KVRian
- 629 posts since 15 Jun, 2017
Going from the definition a Formant, a formant is just a peak in a spectrum.
https://en.wikipedia.org/wiki/Formant
That said, there are many ways to create and manipulate these peaks. So, generally, just about any synthesis technique can be used to achieve formant synthesis.
Some of them:
- using filters (to attenuate what you don't wont, leaving what you do want)
- using a resonator (to amplify what you want to amplify, might be a resonant filter)
- generating peaks using additive synthesis (not just per partial, "true" additive synthesis, but also just mixing oscillators at specific pitch ratios).
- generating peaks using wave shaping functions like FM/PM/AM (where the ratio between carrier and modulator is generally some harmonic interval which generally effects specific regions in the spectrum most)
- generating peaks using oscillator sync (where the sharp cut/jump in the waveforms amplitude/shape will generally generate many high amplitude partials at that point, resembling resonance)
- of course you can also use the spectra of different samples e.g. via a wavetable.
https://en.wikipedia.org/wiki/Formant
Generally sounds are characterized by the relative amplitudes in the spectrum. Some of those relative amplitudes of (clusters of) frequencies can be characterized as formants.In speech science and phonetics, a formant is the spectral shaping that results from an acoustic resonance of the human vocal tract. However, in acoustics, the definition of a formant sometimes differs as it can be defined as a peak, or local maximum, in the spectrum. For harmonic sounds, with this definition, the formant frequency is therefore that of the harmonic partial that is augmented by a resonance. The difference between these two definitions resides in whether "formants" characterise the production mechanisms of a sound or the produced sound itself. In practice, the frequency of a spectral peak differs from the associated resonance frequency, except when, by luck, harmonics are aligned with the resonance frequency.
That said, there are many ways to create and manipulate these peaks. So, generally, just about any synthesis technique can be used to achieve formant synthesis.
Some of them:
- using filters (to attenuate what you don't wont, leaving what you do want)
- using a resonator (to amplify what you want to amplify, might be a resonant filter)
- generating peaks using additive synthesis (not just per partial, "true" additive synthesis, but also just mixing oscillators at specific pitch ratios).
- generating peaks using wave shaping functions like FM/PM/AM (where the ratio between carrier and modulator is generally some harmonic interval which generally effects specific regions in the spectrum most)
- generating peaks using oscillator sync (where the sharp cut/jump in the waveforms amplitude/shape will generally generate many high amplitude partials at that point, resembling resonance)
- of course you can also use the spectra of different samples e.g. via a wavetable.
-
- KVRian
- 1189 posts since 11 Jun, 2019
In other Words: all Ways lead to Rome and all Aninmals are equal. Although some are maybe a little bit longer and some Synths are somewhat more voxy than others :/
What I dislike a little bit: my Kind has not been mentioned here. I think about that D-word and I have a Dream that one Day the People will understand that Sampling + FX Processing is a Synthesis Method, too!?
-
- KVRian
- 629 posts since 15 Jun, 2017
Haha, well, some roads are dead ends, or may take you on a huge detour, or a rough ride. But often you will find yourself having all kinds of fun and great experiences on the way.GRUMP wrote: ↑Wed Jun 24, 2020 4:04 pmIn other Words: all Ways lead to Rome and all Aninmals are equal. Although some are maybe a little bit longer and some Synths are somewhat more voxy than others :/
What I dislike a little bit: my Kind has not been mentioned here. I think about that D-word and I have a Dream that one Day the People will understand that Sampling + FX Processing is a Synthesis Method, too!?
A definition of Synthesis is: "the composition or combination of parts or elements so as to form a whole." So, in short combining various means to achieve a goal.
And about "my Kind has not been mentioned here". You can safely add Sampling + FX Processing to the list. Though both Sampling (recording) and FX Processing (transforming) are pretty generic terms in this context.
-
- KVRian
- 1189 posts since 11 Jun, 2019
Synthesis: simply true - and four+ Oscillators good, two Oscillators bad. I mean - it is difficult or maybe even impossible to control 1000+ Partials over an Y Dimension - but the more (let´s say) Layers we have, the easier we get an Impression that reminds us of this very natural Timbre.Kwurqx wrote: ↑Wed Jun 24, 2020 5:41 pm Haha, well, some roads are dead ends, or may take you on a huge detour, or a rough ride. But often you will find yourself having all kinds of fun and great experiences on the way.
A definition of Synthesis is: "the composition or combination of parts or elements so as to form a whole." So, in short combining various means to achieve a goal.
And about "my Kind has not been mentioned here". You can safely add Sampling + FX Processing to the list. Though both Sampling (recording) and FX Processing (transforming) are pretty generic terms in this context.
I want to encourage the People to look into the Future and demand more from our Developers. And not to look backwards and save them Months of Research. A Diva can neither sing nor breathe and no Starlet will ever make her.
But anyway - Sampling is generic by Nature. Yes. But we should never forget, that the Measure (concerning Quality, Sales Quantity, Likes, ...) are still ROMplers = lots of Sampling and Processing (however)
-
- KVRer
- 13 posts since 8 Mar, 2020
It is true that the literal term 'formants' doesn't describe one type of sound at all, it's just a jargon term for a resonant band in any sound's timbre.
Nontheless when I hear people ask about formant sounds in synth music it often seems they are looking for those ratchet ooh-ee-ah-wei-yo mid bass kind of timbres to use as a musical element with a whacky hey-is-that-synth-talking tone. Specifically, people are looking for easy ways to do this controllably... just like OP said,. Controllably, because as is well known, those talking-synth sounds are notoriously fickle and elusive. You might manage to tune in one good yoi-yoi bass, but the artists you want to sound like some how come up with a distinct vocally timbre for every song, and it takes you ages to tune in iust one. Sure you can achieve the effect with phasers, of maybe with FM... but just how exactly? Its absolutely esoteric.
There is a logical reaosn why vocaloid-bass-lead-synth timbres are elusive and coy. Because the formant patterns of human vowels are not simple, getting a synth to warble its formants in a convincingly stoopid humanoid-like way was always half alchemical science and half sheer dumb luck. You can't just wiggle a filter or a phaser around, because generally the formants in human vocals do a dance that I call "you go up, I go down". To get from one vowel sound to the next in a typical human utternace like "yoi-yoi-yoi-wee-oo-ee-uhh" some formants move up in pitch at the same tine others are shifting down. It makes for a bery headache-inducing experience, even if you have a chart of observed formant frequencies, it might not be so hard but the results still need so much work to sound properly nasty. Hopefully someone on here gets whay I'm saying. The main reason synth people get into formants is to make "those sounds" but making "those sounds" has traditionally been a mysterious art, for reaosns.
But it doesn't have to be this way any more.
In fact it can all be as simple as 1,2,3 now using advanced granular synth. The best ripping-lead-yoi-wobble patches are just two or more of the granular-wavetable streans I described above. If you take a single static wavetable and read it from a granular oscillator in the formant-generating configuration I described previously, each variation in the grain pitch (formant pitch) produces a different complex shape (a formant pattern?) yer all the sounds emitted will still sound similar yet different. THEREFORE "formant-splitting" becomes possible if you have two independent (parallel) granular streams reading the same wavetable for their source, but with formant pitches that do a "you go high, I go low" dance. This recipe always yields a continual stream of yoi-yoi bass sounds, that change subtly whenever you swap one wavetable for another or modify the exact modulation of the formants and range of formant pitches. But while yielding a constant stream of different yoi-basses, they are all yoi-basses - a winner every time, no "trial and error" necessary like it is with phasers and FM. Any old wavetable/waveform works. Just put two independent (parallel) granular streams together, and make one's formants go up while the other go down. All of the formants should be turned up pretty high to begin with so you get that nice squealing ee-ooh-woy-oy-oy-oy greasiness.
Now if you say you care aboutt other kinds of formant manipulation, fine, I really haven't said anything very important, as this only applies to nasty vocaloid synth timbres. But if you *happen* to be a freak looking for nasty vocaloids, look into granular synths with wavetables, and make one stream go up while the other goes down and you'll get amazing wobbles every time. Granular ftw.
Nontheless when I hear people ask about formant sounds in synth music it often seems they are looking for those ratchet ooh-ee-ah-wei-yo mid bass kind of timbres to use as a musical element with a whacky hey-is-that-synth-talking tone. Specifically, people are looking for easy ways to do this controllably... just like OP said,. Controllably, because as is well known, those talking-synth sounds are notoriously fickle and elusive. You might manage to tune in one good yoi-yoi bass, but the artists you want to sound like some how come up with a distinct vocally timbre for every song, and it takes you ages to tune in iust one. Sure you can achieve the effect with phasers, of maybe with FM... but just how exactly? Its absolutely esoteric.
There is a logical reaosn why vocaloid-bass-lead-synth timbres are elusive and coy. Because the formant patterns of human vowels are not simple, getting a synth to warble its formants in a convincingly stoopid humanoid-like way was always half alchemical science and half sheer dumb luck. You can't just wiggle a filter or a phaser around, because generally the formants in human vocals do a dance that I call "you go up, I go down". To get from one vowel sound to the next in a typical human utternace like "yoi-yoi-yoi-wee-oo-ee-uhh" some formants move up in pitch at the same tine others are shifting down. It makes for a bery headache-inducing experience, even if you have a chart of observed formant frequencies, it might not be so hard but the results still need so much work to sound properly nasty. Hopefully someone on here gets whay I'm saying. The main reason synth people get into formants is to make "those sounds" but making "those sounds" has traditionally been a mysterious art, for reaosns.
But it doesn't have to be this way any more.
In fact it can all be as simple as 1,2,3 now using advanced granular synth. The best ripping-lead-yoi-wobble patches are just two or more of the granular-wavetable streans I described above. If you take a single static wavetable and read it from a granular oscillator in the formant-generating configuration I described previously, each variation in the grain pitch (formant pitch) produces a different complex shape (a formant pattern?) yer all the sounds emitted will still sound similar yet different. THEREFORE "formant-splitting" becomes possible if you have two independent (parallel) granular streams reading the same wavetable for their source, but with formant pitches that do a "you go high, I go low" dance. This recipe always yields a continual stream of yoi-yoi bass sounds, that change subtly whenever you swap one wavetable for another or modify the exact modulation of the formants and range of formant pitches. But while yielding a constant stream of different yoi-basses, they are all yoi-basses - a winner every time, no "trial and error" necessary like it is with phasers and FM. Any old wavetable/waveform works. Just put two independent (parallel) granular streams together, and make one's formants go up while the other go down. All of the formants should be turned up pretty high to begin with so you get that nice squealing ee-ooh-woy-oy-oy-oy greasiness.
Now if you say you care aboutt other kinds of formant manipulation, fine, I really haven't said anything very important, as this only applies to nasty vocaloid synth timbres. But if you *happen* to be a freak looking for nasty vocaloids, look into granular synths with wavetables, and make one stream go up while the other goes down and you'll get amazing wobbles every time. Granular ftw.
-
- KVRian
- 1189 posts since 11 Jun, 2019
Which Product matches your Description?cybilopsin wrote: ↑Mon Jul 13, 2020 4:02 am It is true that the literal term 'formants' doesn't describe one type of sound at all, it's just a jargon term for a resonant band in any sound's timbre.
Nontheless when I hear people ask about formant sounds in synth music it often seems they are looking for those ratchet ooh-ee-ah-wei-yo mid bass kind of timbres to use as a musical element with a whacky hey-is-that-synth-talking tone. Specifically, people are looking for easy ways to do this controllably... just like OP said,. Controllably, because as is well known, those talking-synth sounds are notoriously fickle and elusive. You might manage to tune in one good yoi-yoi bass, but the artists you want to sound like some how come up with a distinct vocally timbre for every song, and it takes you ages to tune in iust one. Sure you can achieve the effect with phasers, of maybe with FM... but just how exactly? Its absolutely esoteric.
There is a logical reaosn why vocaloid-bass-lead-synth timbres are elusive and coy. Because the formant patterns of human vowels are not simple, getting a synth to warble its formants in a convincingly stoopid humanoid-like way was always half alchemical science and half sheer dumb luck. You can't just wiggle a filter or a phaser around, because generally the formants in human vocals do a dance that I call "you go up, I go down". To get from one vowel sound to the next in a typical human utternace like "yoi-yoi-yoi-wee-oo-ee-uhh" some formants move up in pitch at the same tine others are shifting down. It makes for a bery headache-inducing experience, even if you have a chart of observed formant frequencies, it might not be so hard but the results still need so much work to sound properly nasty. Hopefully someone on here gets whay I'm saying. The main reason synth people get into formants is to make "those sounds" but making "those sounds" has traditionally been a mysterious art, for reaosns.
But it doesn't have to be this way any more.
In fact it can all be as simple as 1,2,3 now using advanced granular synth. The best ripping-lead-yoi-wobble patches are just two or more of the granular-wavetable streans I described above. If you take a single static wavetable and read it from a granular oscillator in the formant-generating configuration I described previously, each variation in the grain pitch (formant pitch) produces a different complex shape (a formant pattern?) yer all the sounds emitted will still sound similar yet different. THEREFORE "formant-splitting" becomes possible if you have two independent (parallel) granular streams reading the same wavetable for their source, but with formant pitches that do a "you go high, I go low" dance. This recipe always yields a continual stream of yoi-yoi bass sounds, that change subtly whenever you swap one wavetable for another or modify the exact modulation of the formants and range of formant pitches. But while yielding a constant stream of different yoi-basses, they are all yoi-basses - a winner every time, no "trial and error" necessary like it is with phasers and FM. Any old wavetable/waveform works. Just put two independent (parallel) granular streams together, and make one's formants go up while the other go down. All of the formants should be turned up pretty high to begin with so you get that nice squealing ee-ooh-woy-oy-oy-oy greasiness.
Now if you say you care aboutt other kinds of formant manipulation, fine, I really haven't said anything very important, as this only applies to nasty vocaloid synth timbres. But if you *happen* to be a freak looking for nasty vocaloids, look into granular synths with wavetables, and make one stream go up while the other goes down and you'll get amazing wobbles every time. Granular ftw.
-
- KVRist
- 341 posts since 30 Nov, 2008
I would be interested too which vst you recommendcybilopsin wrote: ↑Mon Jul 13, 2020 4:02 am It is true that the literal term 'formants' doesn't describe one type of sound at all, it's just a jargon term for a resonant band in any sound's timbre.
Nontheless when I hear people ask about formant sounds in synth music it often seems they are looking for those ratchet ooh-ee-ah-wei-yo mid bass kind of timbres to use as a musical element with a whacky hey-is-that-synth-talking tone. Specifically, people are looking for easy ways to do this controllably... just like OP said,. Controllably, because as is well known, those talking-synth sounds are notoriously fickle and elusive. You might manage to tune in one good yoi-yoi bass, but the artists you want to sound like some how come up with a distinct vocally timbre for every song, and it takes you ages to tune in iust one. Sure you can achieve the effect with phasers, of maybe with FM... but just how exactly? Its absolutely esoteric.
There is a logical reaosn why vocaloid-bass-lead-synth timbres are elusive and coy. Because the formant patterns of human vowels are not simple, getting a synth to warble its formants in a convincingly stoopid humanoid-like way was always half alchemical science and half sheer dumb luck. You can't just wiggle a filter or a phaser around, because generally the formants in human vocals do a dance that I call "you go up, I go down". To get from one vowel sound to the next in a typical human utternace like "yoi-yoi-yoi-wee-oo-ee-uhh" some formants move up in pitch at the same tine others are shifting down. It makes for a bery headache-inducing experience, even if you have a chart of observed formant frequencies, it might not be so hard but the results still need so much work to sound properly nasty. Hopefully someone on here gets whay I'm saying. The main reason synth people get into formants is to make "those sounds" but making "those sounds" has traditionally been a mysterious art, for reaosns.
But it doesn't have to be this way any more.
In fact it can all be as simple as 1,2,3 now using advanced granular synth. The best ripping-lead-yoi-wobble patches are just two or more of the granular-wavetable streans I described above. If you take a single static wavetable and read it from a granular oscillator in the formant-generating configuration I described previously, each variation in the grain pitch (formant pitch) produces a different complex shape (a formant pattern?) yer all the sounds emitted will still sound similar yet different. THEREFORE "formant-splitting" becomes possible if you have two independent (parallel) granular streams reading the same wavetable for their source, but with formant pitches that do a "you go high, I go low" dance. This recipe always yields a continual stream of yoi-yoi bass sounds, that change subtly whenever you swap one wavetable for another or modify the exact modulation of the formants and range of formant pitches. But while yielding a constant stream of different yoi-basses, they are all yoi-basses - a winner every time, no "trial and error" necessary like it is with phasers and FM. Any old wavetable/waveform works. Just put two independent (parallel) granular streams together, and make one's formants go up while the other go down. All of the formants should be turned up pretty high to begin with so you get that nice squealing ee-ooh-woy-oy-oy-oy greasiness.
Now if you say you care aboutt other kinds of formant manipulation, fine, I really haven't said anything very important, as this only applies to nasty vocaloid synth timbres. But if you *happen* to be a freak looking for nasty vocaloids, look into granular synths with wavetables, and make one stream go up while the other goes down and you'll get amazing wobbles every time. Granular ftw.
-
- KVRian
- 1189 posts since 11 Jun, 2019
I recently discovered the VOX Qualities of Chorus/Flanger Plugins btw. Either D16 Fazortan or Syntorus offers three Bands with variable Frequencys. Surely interesting for EDM Stuff ...
-
- KVRist
- 341 posts since 30 Nov, 2008
Iam really interested in that grain/wavetable stuff for vocal sounds. Got interesting results using Waldorf Nave.
Wonder if there is a better plugin synth for that.
Wonder if there is a better plugin synth for that.
-
- KVRian
- 1189 posts since 11 Jun, 2019
Halion. Hast both Grain and an other Approach on Wavetable (more Resampling Seq). But no molecular Grain Control here :/greententacle wrote: ↑Tue Jul 28, 2020 11:24 am Iam really interested in that grain/wavetable stuff for vocal sounds. Got interesting results using Waldorf Nave.
Wonder if there is a better plugin synth for that.
-
- KVRist
- 341 posts since 30 Nov, 2008
But no molecular Grain Control here :/
Thats the point.
Thats the point.
