Archive

Posts Tagged ‘formant’

Cyber Attaque Mentale

12/05/2013 Commentaires fermés

Inutile de vous le redire, je me suis penché le moins dernier à cinq ou six reprises sur la conception de formant filters en natif sur Renoise, vous avez pu vous en apercevoir via de multiples articles en anglais et aussi sur le forum trucs et astuces de http://www.refra.fr.

Maintenant, il s’agit de mettre en application. Pourquoi donc ne pas simplement balancer des impacts de basse en dent de scie et y appliquer ce formant filter ?

C’est ce que ce morceau de Drum & Bass très classique dans sa forme vous propose :

Pour étudier le fichier source .xrns du morceau, rien de plus simple, il suffit de cliquer [ici] et de laisser Renoise se lancer.

Merci pour votre écoute et à bientôt !

Maxime

Easy Formants for Renoise 2.8 !!! V2 !!! [English]

08/04/2013 Commentaires fermés

Hey folks I’ve just received the help of Bit_Arts himself to improve the concept !

Even more easy than easy, now.

What’s new ?
– 4 Formants Filter Bands instead of 3, and everything is still packed a just 1 chain
– Virtual Keyboard Based Formants Control
– Formants Feedback control
– Auto Q
– Auto Bot Voice
– fixed RExciter ™ position 😉
– replaced the detuned wobble sample by a more typical chipsound

FAQ

How to install it in my own song ?

  • Right-click somewhere on the Track DSP chain.
  • Select Device Chain => Save As…
  • Select your target directory and give a name to the chain.
  • It must have the .XRNT file extention.
  • Then when you edit your song, right click somewhere on your Track DSP tab.
  • Select Device Chain => Load…
  • Find your .XRNT file, load it.
  • That’s all.

How to modify Formants ?

  • Formants here are « Keyboard Controled ».
  • Use the C-0 to C-2 Keys to modify formants.
  • Go into the Sample Keyzones.
  • Modify your Sample Layer so that it doesn’t use notes from C-0 to C-2.
  • By this way you’ll be able to both play your sample & control the filter in the same track.

How exactly did you placed formants on the low notes ?

  • Formants are composed of 5 cycles of 5 ‘aeiou‘ vowels.
  1. the first cycle is adapted to « Basses » sounds
  2. the second cycle is adapted to « Tenors » sounds
  3. the third cycle is adapted to « Contra-Tenors » sounds
  4. the fourth cycle is adapted to « Altos » sounds
  5. the last cycle is adapted to « Sopranos » sounds
  • Selecting the right cycle can’t be automatic.
  • Just use your own ears for that !

How to sound lika robot ?

  • Completely drag the Humanize slider to the left.

I don’t like the robot sound. How do I delete it ?

  • Don’t delete the DSP
  • Drag a bit the Humanize slider to the right.
  • Or uncheck « Bot Voice »

What is the Feedback parameter ?

  • This is something that adds some personnality to the sound.
  • Warning to high parameters will produce a glassy/metallic effect, stange but interesting.

What is the Auto Q parameter ?

  • The Auto Q parameter slightly increases the Q factor of bandpass filters when formants deal with higher frequencies. You can switch it off if you want to use the Bandwidth slider like a Wet/Dry signal slider.

My filtered sounds is poor compared to yours.

  • It’s logical. Some harmonics have been removed by the chain. Put the *Exciter DSP and the end of the chain, and raise sliders to revive the flat sound.

Download
HERE.

Credits

Still featured in the www.refra.fr french portal.

have fun with Renoise !
kurtz/lapiNIC

Easy Formants for Renoise 2.8 [English]

07/04/2013 Commentaires fermés

THIS TIME IT’ EASY ! ALL THE SYSTEM WORKS IN JUST ONE TRACK NOTHING ELSE.

* * *

Since I’ve deleted by mistake the Bit_Arts initial work that showed it was possible to build a native formant filter in Renoise, I had high hopes to rebuild another similar one from the start. I’ve read some interesting articles for that.

So for example, this demo is once again based on this S.o.S webpage :
http://www.soundonsound.com/sos/mar01/articles/synthsec.asp

Formants can be seen as a set of 3 parallel bandpass filters ; 5 voyels are emulated by defining specific bandpass frequencies in 3 mixed BP filters.

The input sound should be « pulse-modulated » / distorded (shift model) or (why not) Lofimatized (Bitcrushed) for better results.

The 3 Frequencies are defined in a 3×5 matrix :

  •      « ee » sounds like « leap » F1=270Hz F2=2300Hz F3=3000Hz
  •      « oo » sounds like « loop » F1=300 F2=870Hz F3=2250Hz
  •      « i » sounds like « lip » F1=400 F2=2000Hz F3=2550Hz
  •      « e » sounds like « let » F1=530 F2=1850Hz F3=2500Hz
  •      « u » sounds like « lug » F1=640 F2=1200Hz F3=2400Hz
  •      « a » sounds like « lap » F1=660 F2=1700Hz F3=2400Hz

If you define for any LFO an amplitude of 50% and an Offset of 42.30% then you can convert the previous values like that :

  • « ee » sounds like « leap » F1=0% F2=70.755% F3=81.509%
  • « oo » sounds like « loop » F1=2.222% F2=34.34% F3=70%
  • « i » sounds like « lip »  F1=9.811% F2=65.283% F3=70%
  • « e » sounds like « let » F1=18.113% F2=62.264% F3=74.151%
  • « u » sounds like « lug » F1=23.962% F2=46.038% F3=72.453%
  • « a » sounds like « lap » F2=25.094% F2=59.057% F3=72.45%

The problem was to « store » somewhere this 3×5 frequencies matrix. The solution came from Ragnar Aambø (TheBellows) that created a nice piece of random tune called « Invisible Melodies« . In this tune some notes are stored in custom LFOs with « points » and those custom LFOs points are triggered (one shot mode) with the modulated « Reset » button.

So following this example, Renoise can store the 15 specific frequencies (corr. to each vowel) in 3 custom LFOs and the frequency selection has to be made with a Hydra device pointing to the « Reset » of the LFOs.

Concerning the 3 filters, they can be found in the new Renoise « MultiTap » device that has « filtered » delay features. If you simply block delays, you can just use the filterings there without using routings or send tracks anymore (as show in previous techniques).

It makes the installation & setup of the Formants Filter Mechanism EASY to INSTALL and to TWEAK, you’ll just have to copypaste the chain contained in FMTS CTRL in any track you want, the track does all the job by itself.

Note that if you link the LFOs containing formant values, directly to the MultiTap Filters, then you’ll encounter a small problem of Robotized output, because transitions between formants are not sweet enough. That’s where kraken/gore comes 🙂 and brought us the « Inertial Slider » Secret *Formula Device Trick. If you check closely the chain you’ll see that I’ve inserted the Inertial Slider between the Filters and the Custom LFOs, so that transitions are more sweet and human. Since it sounds like the Intertia parameter found in the *Filter DSP device, I’ve renamed it Inertia directly but it’s a pure copypaste of kraken/gore’s work.

Concerning the values’ storage, in the LFOs, I’ve used the new « External Editor » to adjust precisely the point values : a PITA(!) but that hopefully worked.

DOWNLOAD IT HERE.

* * *

USAGE

At first, SAVE the chain contained in the FMTS CTRL track. Give it the name you want. The file extension must be .XRNT.

Then IF you need a Formant Filter in your own module, THEN LOAD this previously saved chain in the desired track. It can be a standard track or a SEND track, it works too.

The Formant control is EASY and SIMPLE.

MODULATE the 3 parameters there :
(1) Formants – self explicit.
(2) Bandwidth – increase/decrease formant filters
(3) Humanize – human/robotic vocal transitions

Modifying Formants With the dedicated slider:

0 % ======================================>> 100 %
« ee » =====> « oo » =====> « i » =====> « e » =====> « u » =====> « a »

BTW : you can AUTOMATE the Formants slider as you wish, with lfos, with automation envelopes, with a key-tracker, anything works.

Credits/References

  •  Bit_Arts – showed that doing it in renoise was possible – many thanx for removing your links, it truely pushed me to move my ass and understand the whole mechanism by myself
  •  Sound On Sound – the BIBLE of hi-tech music recording industry
  •  The BellowsInvisible Melodies
  •  kraken/goreInertial Sliders (Secret *Formula Device Powered) Trick

Final Notes :

Thanx for reading
have fun with Renoise !

kurtz/lapiNIC

Catégories :Renoise Étiquettes : , , , , , , , , ,

A 5 Bands NATIVE Formant Filter for Renoise [English]

07/04/2013 Commentaires fermés

This post follows the previous one about the right way to build a 3 bands native formant filter in Renoise. I previously based my work on a Sound On Sound webpage, but I’ve just found a more complete description of formants where I’ve got not only 3 bands but 5 bands !

Check this very interesting webpage :
http://www.csounds.com/manual/html/MiscFormants.html

Increasing the number of bands and increasing the « range » of formants could be a nifty solution for those who can not obtain proper formants with higher tones/sounds (alto / soprano sounds for example).

Formants can be seen here as a set of 5 parallel bandpass filters ; the 5 vowels are emulated by defining specific bandpass frequencies in 5 mixed separate (i.e. send) tracks. The input sound should still be « pulse-modulated » / distorded (shift model) or (why not) Lofimatized (Bitcrushed) for better results.

The 5 Frequencies are defined in a 3×25 matrix :

1. Basses

Bass ‘a’         f1   f2   f3   f4  f5
frequency (Hz)   600 1040 2250 2450 2750
Bass ‘e’         f1   f2   f3   f4   f5
frequency (Hz)   400 1620 2400 2800 3100
Bass ‘i’         f1   f2   f3   f4   f5
frequency (Hz)   250 1750 2600 3050 3340
Bass ‘o’         f1   f2   f3   f4   f5
frequency (Hz)   400  750 2400 2600 2900
Bass ‘u’         f1   f2   f3   f4   f5
frequency (Hz)   350  600 2400 2675 2950

2. Tenors

Tenor ‘a’        f1   f2   f3   f4   f5
frequency (Hz)   650 1080 2650 2900 3250
Tenor ‘e’        f1   f2   f3   f4   f5
frequency (Hz)   400 1700 2600 3200 3580
Tenor ‘i’        f1   f2   f3   f4   f5
frequency (Hz)   290 1870 2800 3250 3540
Tenor ‘o’        f1   f2   f3   f4   f5
frequency (Hz)   400  800 2600 2800 3000
Tenor ‘u’        f1   f2   f3   f4   f5
frequency (Hz)   350  600 2700 2900 3300

3.Contra-Tenors

CounterTenor ‘a’ f1   f2   f3   f4   f5
frequency (Hz)   660 1120 2750 3000 3350
CounterTenor ‘e’ f1   f2   f3   f4   f5
frequency (Hz)   440 1800 2700 3000 3300
CounterTenor ‘i’ f1   f2   f3   f4   f5
frequency (Hz)   270 1850 2900 3350 3590
CounterTenor ‘o’ f1   f2   f3   f4   f5
frequency (Hz)   430  820 2700 3000 3300
CounterTenor ‘u’ f1   f2   f3   f4   f5
frequency (Hz)   370  630 2750 3000 3400

4. Altos

Alto ‘a’         f1   f2   f3   f4   f5
frequency (Hz)   800 1150 2800 3500 4950
Alto ‘e’         f1   f2   f3   f4   f5
frequency (Hz)   400 1600 2700 3300 4950
Alto ‘i’         f1   f2   f3   f4   f5
frequency (Hz)   350 1700 2700 3700 4950
Alto ‘o’         f1   f2   f3   f4   f5
frequency (Hz)   450  800 2830 3500 4950
Alto ‘u’         f1   f2   f3   f4   f5
frequency (Hz)   325  700 2530 3500 4950

5. Sopranos

Soprano ‘a’      f1   f2   f3   f4   f5
frequency (Hz)   800 1150 2900 3900 4950
Soprano ‘e’      f1   f2   f3   f4   f5
frequency (Hz)   350 2000 2800 3600 4950
Soprano ‘i’      f1   f2   f3   f4   f5
frequency (Hz)   270 2140 2950 3900 4950
Soprano ‘o’      f1   f2   f3   f4   f5
frequency (Hz)   450  800 2830 3800 4950
Soprano ‘u’      f1   f2   f3   f4   f5
frequency (Hz)   325  700 2700 3800 4950

It IS a PURE PITA to build it but anyway I’ve nothing to do this week end.

Note

This « 5 bands » based version is different from the  « 3 bands based version » (a bit lighter and probably more CPU friendly) proposed in the « Basic Formants for Renoise » module, released previously this week. Warning : since it contains some chained meta devices, and some complex routings, this module could quickly become CPU intensive, if you’ve got some problems with that, please grab the « Basic Formants for Renoise » version.

Technique

Same thing as the one described here before. Storing this 3×25 frequencies matrix uses exactly the same Custom-LFO based trick, that I’ve discovered while playing the Ragnar Aambø (TheBellows) tune called « Invisible Melodies » (remember that in this tune notes are stored in custom LFOs with points and those custom LFOs points are triggered with the modulated « Reset » button.

DOWNLOAD

  • Download the .XRNS Standard Module HERE
  • Download the .XRNS Advanced Version Module HERE (contain Smoothly Amp Controlled Filters)

Usage

It’s the same usage as mentionned in the previous « Basic Formants for Renoise » ; it’s based on 3 specific hydra devices, that I’ve renamed.

At first SEND YOUR SOUND to ControlF.

Then MODULATE the 3 parameters there :
(1) Formants – self explicit.
(2) Bandwidth – increase/decrease formant filters
(3) Transitions – human/robotic vocal transitions

Modifying Formants :

0 % ======================================>> 100 %
« ee » =====> « oo » =====> « i » =====> « e » =====> « u » =====> « a »

PROBLEMS / TO DO LIST

The sound is BETTER (of course there are 5 fucking bands).
BUT
Rebuilding this into new modules on demand is just another PITA, too complex ! It’s CPU intensive, I should really find a better way to add the usefull info in just one chain and simplify the way to install / remove it.

Credits / Thanx

  • Bit_Arts – showed that doing it in renoise was possible and then pushed me to move my ass and understand the theory and find the tricks by myself
  • www.csounds.com – cool information there
  • The Bellows – Invisible Melodies

Thanx for reading
have fun with Renoise !

kurtz/lapiNIC

Catégories :Renoise Étiquettes : , , , , , , , , ,

A Basic Native Formant Filter for Renoise [English]

06/04/2013 Commentaires fermés

Bit_Arts initial work gave me high hopes to build a native formant filter in Renoise by myself, and I’m actually on the path  to rebuild another similar one from the start and I’m reading some articles for that.

I’ve read this very interesting webpage :
http://www.soundonsound.com/sos/mar01/articles/synthsec.asp

Theory

It says that formants consist in a set en 3 parallel bandpass filters ; 5 voyels are emulated by defining specific bandpass frequencies in 3 mixed separate (i.e. send) tracks. The input sound should be « pulse-modulated » / distorded (shift model) or (why not) Lofimatized (Bitcrushed), for better sounding results. Vowels morph because parameters are slowly modulated so I guess that we would need some « Inertia » in the formula.

What we’ve got in our hands with Renoise

  1. the DSP *Filter (it has an internal BP mode that could help me to rebuild the setup)
  2. the *multiTap device can be another option : behind the 4 delays mechanism, it has 4 internal Filters with a BP option but it has apparently no inertia so it should probably produce not human vocal results
  3. the *multiband device, that’s able to route the signal in 3 separate channels in which I could put the desired BP *Filters
  4. or the typical *Send routing device, that can be chained with Keep source 2 time, and Mute source for the last routing.
  5. the *Hydra meta-device, that can just control 3 parameters with 1 input, could be usefull to change vovels more easily
  6. and much more… 😉

Back to S.o.S site, the 3 Frequencies are defined in this 3×5 matrix :

  • « ee » sounds like « leap » F1=270 F2=2300 F3=3000
  • « oo » sounds like ‘loop’ F1=300 F2=870 F3=2250
  • « i » sounds like like « lip » F1=400 F2=2000 F3=2550
  • « e » sounds like « let » F1=530 F2=1850 F3=2500
  • « u » sounds like « lug » F1=640 F2=1200 F3=2400
  • « a » sounds like « lap » F1=660 F2=1700 F3=2400

The problem is to « store » somewhere this 3×5 frequencies matrix.

I remeber a nice piece of random tune called Invisible Melodies. In this tune notes are stored in custom LFOs with points and those custom LFOs points are triggered with the modulated « Reset » button (One Shot Mode).

  • So following this example, Renoise can store the 5 specific frequencies (corr. to each vowel) in 3 custom LFOs and the frequency selection has to be made with a Hydra device pointing to the « Reset » of the LFOs. If you link the LFOs to Filters and use the « External Editor » to adjust precisely the point values, you can reproduce exactly the previous 3×5 matrix. The only difference is that the input value has to be shown in a % and not in Hz :

With an amplitude of 50% and an Offset of 42.30% the corresponding LFO Values (% instead of Hz) :

  • « ee » sounds like « leap » F1=0% F2=70.755% F3=81.509%
  • « oo » sounds like « loop » F1=2.222% F2=34.34% F3=70%
  • « i » sounds like « lip »  F1=9.811% F2=65.283% F3=70%
  • « e » sounds like « let » F1=18.113% F2=62.264% F3=74.151%
  • « u » sounds like « lug » F1=23.962% F2=46.038% F3=72.453%
  • « a » sounds like « lap » F2=25.094% F2=59.057% F3=72.45%

Download

This is a FIRST test that uses typical SEND BASED routings. Download HERE the .XRNS demo example.

Usage

It’s based on 3 specific hydra devices, don’t worry I’ve renamed them so that they’re easy to understand. I tried to make it as simple as possible.

First SEND YOUR SOUND to ControlF.

Then MODULATE the 3 parameters there :
(1) Formants – self explicit.
(2) Bandwidth – increase/decrease formant filters
(3) Transitions – human/robotic vocal transitions

Modifying Formants :
0 % ======================================>> 100 %
« ee » =====> « oo » =====> « i » =====> « e » =====> « u » =====> « a »

Credits

* Bit_Arts – showed that doing it in Renoise was possible and pushed me to move my ass to understand the mechanism by myself
* The BellowsInvisible Melodies
* MXBResynth 4 is perfect to build the right pulse wave samples with no time

PROBLEMS & TO DO LIST :

That’s fucking painfull to insert this logic in my new modules ! I need to find something EASIER. I need to find a way to put most of the usefull data into one chain, for example.

Still searching for the best trick…

Thanx for reading
have fun with Renoise !

kurtz/lapiNIC

Frogoïd – un chanteur virtuel français pour Renoise

08/01/2013 Commentaires fermés

Le concept de Frogoïd est quasi 99% similaire à celui développé pour Renoid, le chanteur virtuel de Renoise, à ceci prêt que si Renoid est limité aux chansons japonaises, Frogoïd pourra articuler la majorité des textes en français. Renoid et Frogoïd s’appuie sur l’utilisation d’un plugin gratuit appelé KeroVee, dont les principaux atouts sont de pouvoir effectuer une détection/correction du pitch vocal à la volée, comme Autotune (mais pas au même prix), et même de piloter le pitch d’un son à partir d’une entrée MIDI, toujours en temps réel.

Les phonèmes ont été précisément recrées pour permettre un maximum d’articulations verbales en langue française. La banque de phonèmes utilisée est bien plus large que celle du programme initial (254 en tout) et de ce fait, Frogoïd doit dispatcher ses échantillons de façon très différente, en utilisant les fonctions avancées de la tabulation Sample Keyzones sur Renoise. Les samples sont issus d’un programme de TTS (IVONA) – Mathieu pour être précis – mais tous retravaillés, bouclés (forward) et tous précisément accordés sur 440Hz pour faciliter la manipulation du pitch et des formants sous le plug-in KeroVee.

Ensuite, le script LUA Words Pad for Renoid a été partiellement ré-écrit afin de gérer une bibliothèque de samples élargie et fortement stratifiée.

Pour en savoir plus, utiliser Frogoïd, et télécharger les packages utiles, rdv surle portail renoise france dans la section dédiée :

http://www.refra.fr/portail/node/100 !