Haken Continuum as a Theremin Effect

https://soundcloud.com/rk53-1/theremin-through-haken-continuum-biqmouth-test

I'm starting this thread to store my experiments using the Haken Continuum as an effect processor for the theremin.

Experimented tonight with creating vocal formants - ala talking machine - though I did not try and create a natural sounding voice. Also added reverb in the continuum.

Ran the line out of the Etherwave Plus into the AES input of the Haken continuum. From there you can channel the audio input into the extremely powerful EaganMatrix sound engine and then with the Continuum Editor create a Matrix patch of your choice to do a zillion different things with the theremin input. The EaganMatrix has a set of biquad modal filters that can be set different ways. One is called BIQMOUTH. This is a chain of five filters that lets you do a variety of formant processing on the input (which here is a theremin - Ed Eagan just released a video where he played the Continuum through MIDI from an EVox and used formants that he fed from internal continuum sound sources). 

To do this you will need an A/D converter to convert the theremin analog line out into AES digital format that the continuum uses for an input. Many may not have a way to do this. I use a Behringer Ultramatch Pro which does A/D and D/A with applying different clocking as well - which is very affordable for what it does. There is not much in the EaganMatrix manual for this but it works when set to 96Khz, 16 or 24 bits, no dither.  Most decent preamps by the way will also have AES out with adjustable bit rates (my ART vocal preamp works as well). But you should use a digital XLR cable (standard audio XLRs may not work well as the continuum manual points out).

Didn't have much time to do anything but an experiment here that runs through 10 or so different formant vowels - a whole slew are programmable through the matrix with one of the BiqMouth parameters.

Also as Ed Eagan states in the manual - do not experiment with the EaganMatrix with volume way up as some setting cause major sonic explosions - yes they do! - always experiment with volume way down.

There is world of sonic possibilities to process the theremin in many interesting ways through the EaganMatrix. I'll post more here as I get the chance. If anyone wants to know the EaganMatrix settings I use, I'll post a graphic of the programming.

Thanks Rich!  I don't know what it is about vocal synthesis that's so spooky, but it surely is.  Ghost in the machine or something.

Dewster,

I need to try playing around a bit more to see if I can get a nice vocal effect like you can with the Talking Machine. I think the Thermeini might actually work better as my observation is that the more complex spectral content you have for the input to the formant generator the better quality voice you get out - more harmonics to work with I think.

But the real beauty of this is that with come of the Continuum's spectral shaping "modules" you might be able to create all kinds of different instrument sounds from the theremin input.

Rich, how complex can you make the formant generator?  A chain of 5 biquads (state variable?) will do recognizable vowel sounds, but perhaps isn't sufficient to do detailed vocal textures?  (Kinda talking out of my ass here as I haven't done any vocal synthesis myself yet.)

Dewster:

Well getting a true voice modeling of a throat is likely beyond this. But lets look at the Talking Machine's features versus the parameter controls of the EaganMatrix BiqMouth module (of which you can use two if you want to to get even more complexity). With the Talking Machine you twiddle dials until you get something you like.

Talking Machine:

• Vocal Vowel Filter w/ 7 vowel modes (similar modes to continuum - only continuum models male and female with morphing capabilities)
• Wah and Baseballs modes
• Adjustable Sensitivity control allows both up and down filter sweeps
• Attack and Decay knobs control the speed of the envelope follower while it sweeps the filters
• External effects loop capability for placing the effects on the filters without changing the envelope follower's dynamics

• With the Continuum/Eagan Matrix you approach the problem in a parameterized fashion. I have not experimented with all of the parameters and exactly how they affect the sound yet. But you have to approach this in a much more rigorous fashion than just twiddling dials. And this is just one function that can be combined  with other things in the EaganMatrix. This may well spell the eBay ending of my Talking machine. But "programming" this through the EaganMatrix is also quite interesting. It's a very unique (and as Ed points out historically old) method of configuring a synthesizer - but with a very new modern and extremely flexible twist.
• BiqMouth Settings:
• Audio Input: Audio input into the vocal formant filter bank, a series of five biquad formant filters (this basically lets you run an input through the format generator)
• Mouth Shape: A series of mouth shapes, selectable intervals with smooth morphing between adjacent shapes: (Father) ä, (Made) ā, (Food) o͞o, (Go) ō, (Pad) a, (Merry) e, (Mud) ə. These are defined in a larger matrix allowing for male and female models of each that you can morph between.
• Frequency Spread: Multiplier on formant frequencies
• Frequency Offset: Offset on formant frequencies
• Bandwidth Spread: Formant bandwidth multiplier, 1.0 means each formant has default bandwidth, > 1.0 means formants farther from Centre (see below) have increased bandwidth, < 1.0 means formants farther from Centre have decreased bandwidth.
• C Centre: BiqMouth Centre control, 0 means first formant is centre, 0.625 means last (5th) formant is centre.
• Amplitude Spread: BiqMouth amplitude spread, 0 means no amplitude scaling of formants, > 0 means reduced amplitude of formants

I tried to emulate a Talking Machine like effect on a tonal/modal musical example tonight running the Etherwave Plus through the Continuum's vocal formant module that I set up.

https://soundcloud.com/rk53-1/theremin-thru-haken-continuum-vocal-formant-generator-test-2

I find at least in the small amount of time I've been playing with this that I get more quick clipping of freqs and seems like less dynamic range (less time from piano to forte and quicker saturation). But I have not exhausted playing with all the parameters yet. I'm trying to set the female vocal mouth shapes (to set a constant value like 3.5 set .5 in one direct X parameter and 3 in the other and they will be added) as the EaganMatrix appears to emulate the higher vocal sound much better than the Talking Machine - though I can't quite get that dynamic tenor sound you can on the Talking Machine. I have high hopes that using the continuum as a vocal processor will let me much better emulate counter-tenor, alto and soprano than the Talking Machine can. And the Talking Machine can only get an acceptable "ah" sound as I tried to emulate here. The continuum, I believe will be able to generate a better range of vowel sounds for musical use. 

Ya, the Talking Machine seems to be a one trick pony - but it nails that one pony dead. 

I'd rather play with something more malleable and varied, such as the synth in the Continuum.

I long ago gave up any hope of winning popularity contests in the theremin community, so I will go ahead and ask……rkram, are you aware how exasperatingly off pitch your theremin playing is on the above soundcloud sample? 

One of the unfortunate realities of the theremin is that if it is played sufficiently badly, the performance becomes funny and whatever the player was trying to demonstrate is lost in a tsunami of giggles. I couldn’t get the image out of my head, of a badly constipated tenor sitting on the pot!

I have often heard thereminists admit they do not play very well, and ask their listeners to hear beyond poor technique and focus on some other musical point they are trying to make. This is like a chef asking diners to ignore the fact that he accidentally spilt the entire salt pot into the soup, and focus instead on the delicate flavor of his bisque! 

As for the TALKING MACHINE, speaking as the first person to introduce the thing to the theremin community, the only sound that is of any use is the open “AH”. All the others are comical, particularly the diphthongs. 

P.S. The lute sounds really good.

Hmm. The human mind being what it is, this leaves me with the singular thought...

Wow. Yea! Peter liked the lute! :-)

At least I got your attention. Frankly it was late and my mind was only interested in the issue of what it takes to run an analog theremin signal through the continuum to get a vocal sound - something I have not seen anyone do yet. The real question is: Does this have any practical musical application? 

But I need to go listen to it. And if I can't tell it's crap then I need to get a hearing aid. So my task this weekend is to try and use the continuum's vocal formant module to produce something musical with a theremin - because my goal before I die is to create something that will elicit Peter's singular comment: "Rich - I really liked that!"

P.S. Wow - I just listened to it and yes it is crap but I think I'm not hearing that lute too well. That hearing loss of mine is starting to really scare me. A good reason to want to write for the instrument. Perhaps my sample library doesn't sound quite so bad now. 

By the way, for those interested in Continuum/EaganMatrix "programming", here's the EaganMatrix program that I used to run the theremin through the vocal formant generator. In the matrix itself, inputs are on the left top to bottom and outputs are across the top. Placing a constant or formula in the intersection will patch an input to an output (Ed Eagan got this idea from old matrix patched synthesizers). "1" normally signifies a unity value. The AES inputs are attached to the BiqMouth 1 module's inputs (little triangle up top) which are assigned to bank A as noted under the BiqMouth label. AES input up top is turned up to the input volume I want. The L&R outputs of Bank A are assigned to main outs (that are also processed by the internal reverb)  SL and SR. Then reverb time dial at the bottom is dialed up to what I desire. Finally the Direct parameter inputs to the BiqMouth module are set to various values I'm experimenting with based on the chart above. Since I'm just setting fixed parameters to the BigMouth and not using the fingerboard at all, I don't need to assign any formula equations at the right. You'll notice next to the BiqMouth is a module called BiqBank. Both of these are selectable to a variety of modules that do different things. I set the one I am using to the BiqMouth setting.