Review – Virta: Voice Controlled Synthesis from Madrona Labs
Want to use your voice (or other sound source) to control a powerful synthesizer? Then try Virta.
by Warren Burt, May 2016
Want to use your voice (or other sound source) to control a powerful synthesizer? Then try Virta, Madrona Labs new semi-modular:
Madronalabs, $89 USD.
Madrona Labs, the company of software and hardware developer Randy Jones, is the makers of softsynths Aalto and Kaivo, and the MPE (Multidimensional Polyphonic Expression) equipped keyboard/controller Soundplane. Always a creative developer, he has now released Virta, which is billed as a “Voice-Controlled Synthesizer.” It is indeed that, and more. It uses the same patching template as Madrona’s earlier synths, Aalto and Kaivo – a row of control modules across the top, a patching panel in the middle, and a signal path through sound making modules on the bottom.
Virta continues this idea with a new family of modules, which are optimised for sound making with external, usually vocal, controls. Although any kind of sound can be used as the input for Virta (and this is the opportunity for a LOT of sonic exploration), the input module is optimised for voice, or other monophonic sound.
Here’s a screen-shot of the single page GUI of Virta. Notice the Audio module on the top, and the Formants and Delay modules on the bottom.
Virta, by the way, continues Randy Jones’ tradition of using Finnish words for the names of his synthesizers. “Virta” means current, as in both electricity and flowing streams. “Aalto” is Finnish for wave, and “Kaivo” is Finnish for “well.” The tradition of water-soluble names at Madrona Labs continues.
Looking at the modules on the top from left to right, reveals some old friends, and some new additions to the Madrona line-up. First on the left is an old friend, the “Key” module. This has a slot for a Scala file to tune the output to any scale. You can use the included selection of Scala files, or you can add your own, by adding a directory of them to the Scales folder in your Documents area. On my Win8.1 machine, this is found in “Warren/App Data/Roaming/Madrona Labs”. Your location may vary, depending on operating system and machine. Under the Scala file slot are controls for Number of Voices, and Glide. Virta can generate up to four voices, but for some of the patches, the plugin is very CPU hungry and you may find yourself running out of power at four voices. If that happens, just dial down the polyphony and things should be fine. Mac versions require OS 10.7 or higher. Windows versions require Windows 7 or higher. Windows XP and Vista are likely to work but are not officially supported. All versions require a 1.6GHz Intel Core Duo processor or better. On my ASUS Core i5 Win8.1 laptop, I find that I can always run Virta monophonically, and in most cases 2 voiced. Three and four-voiced operation depends on the modules used in the patches.
Also included in the module is a MIDI meter, which says which of the four voices is currently playing, a unison toggle (to make thick sounds with all four voices), a bend control (you can select up to + / – 24 semitones maximum for the glide range, a polyphonic aftertouch output, and a mod cc# control, which sets the number of continuous controller that will be sent out the MOD output at the bottom of Key. You also get two other CC outputs, +1 and +2 MIDI CC numbers above the initial cc# chosen.
The next module is the heart of the synth – this is the Audio module. The well written and clever manual has a delightful diagram which explains the function of the module very well.
As you can see, an incoming sound signal is analysed with a number of processes. From top to bottom, or left to right on the module, you have the unprocessed sound (pre), the sound through a low-cut and high-cut filter and compressor (comp) a gate (gate) signal which occurs when the sound crosses a settable preset threshold, an envelope tracing (env), a signal proportional to the pitch of the sound (pitch), a quantization control (quant) which allows the pitches of the input sound to be quantized to whatever scale is specified in the Key module (this alone would justify the purchase, in my eyes), a signal which gives a value between 0 and 1 depending on how noisy the sound is (noise), and a “peak” signal (peak), which is a measure of the overall brightness of the sound, the center of the dominant spectral content. This gives quite a number of signals derived from the analysis of the input sound, and these can all be used in different ways and combined in numerous ways in the patching panel.
A Low Frequency Oscillator then follows the Audio module. This is a standard LFO, but it has a wide range (100 seconds per cycle up to 220 cycles per second), and has a couple of unique waveforms, such as the blip, which is like an exponentially decaying envelope, and the noise waveform, which is one cycle of a noisy waveform which always repeats (so it’s actually pseudo-random) as well as the usual assortment of sines, saw, triangle, and square waves. The LFO can be synced, and the rate and volume of it can be externally controlled. Having such a wide range means that the LFO can double as another oscillator, or act as an auxiliary envelope, or as a control source.
Two oscillators follow. These can be set to two modes: “classic” and “vosim.” Classic is what it says, a square to sine to saw oscillator (the waveforms can be morphed), noise can be added, and the oscillator’s pitch can be controlled logarithmically (one per octave) or linearly (good for FM effects). Vosim mode produces a voice-like waveform with controls for the number of formants and which formant is most prominent (peak). This produces changes in the “vowel sound” that the oscillator is producing. This mode, too, can be controlled logarithmically or linearly. An ADSR envelope follows with optional velocity control and a trigger input, and CV controls for all the envelope settings.
Moving across the patch bay, things get very interesting. The first module is called Formants, and has a number of modes. It has four vocoder algorithms, and it analyses the sound at the “program” input, and applies multiband filtering to the signal entering the “carrier” input. As the manual says, this “applies the timbral essence of one signal to another, over time.” Once the sound is analysed, you can shift the analysis of the harmonic content up or down, and stretch or compress the spectrum of the sound and adjust the Q of the multiband filters (up to ringing), so that the spectrum of your input “program” is considerably changed from its original. There is also a “quant” check box. If checked this quantizes the output spectrum of the module to match the scale set in the Key module. This, combined with the “quant” control in the Audio module, means that the spectrum of any input sound can be mapped to the pitches of any imaginable scale. This control has me very excited. But wait (as they say on late night TV), there’s more. There are five different modes for the vocoder (carrier thru, 8, 16, 32, and 256 bands), and there is a “pan” control which can apply seven different modes to the mapping of the harmonics to a stereo spectrum. These are mono, L>R, R>L, spiral, clumps, random and even-odd. Each of these produces a very distinctive spread based on the spatial positioning of the harmonics. The module has two outputs – program (which is just a pass-thru for the program signal) and wet. If you select the “carrier thru” mode, the carrier alone will be routed to the wet output. Otherwise, you get the vocoded sound. This Formants module can produce an incredibly wide range of sounds, and when controlled with the CV signals produced by the Key and Audio modules, can make a huge range of sound types, from subtle vocoding to wild sounds that bear only a peripheral relationship to the input sounds. I’ve been playing with this module for about a week, and I’ve barely scratched the surface of the many types of sounds it can produce. Another cartoon from the manual explains the Formants module quite elegantly.
Next in the chain is a Gate module, which can function as a straight voltage controlled amplifier (VCA) or as a VCA connected to a Low Pass filter. It uses a Vactrol-emulation algorithm, to give it the same characteristics of the old vactrol (a kind of light control) analog amplifiers/filters found in many classic analog synthesizers, such as the Buchla and the Serge. If the output of the Envelope is applied to the control input of this module, it acts like a normal VCA or VCA-filter combo. If other signals are applied to it, the results can be quite striking.
Then comes a Delay module. This features up to 5 seconds of delay, and lo-cut and hi-cut filters, a pitch shifter, separate delays for left and right channels, a diffuse control (which at higher settings can “smudge” the sound a bit, resulting in the sound of an old analog or tape delay), a feedback control, and a LR ratio control, which shifts the balance of the delay times between the two channels. At the mid-point (1.0) the delay is the same in both channels. Shifted to the left, the left delay is shortened while the right is lengthened. Shifted to the right, the opposite happens. This is useful for creating ping-pong stereo delay effects. And all of these controls, except the LR ratio, can be controlled by CV signals from any other part of the synth.
Finally, there’s an output module, which has an Aux In with its own volume and pan controls, as well as an input from the Delay with its associated volume and pan controls as well. There’s also a small oscilloscope, and after the output, there is a limiter which automatically kicks in if your signal goes over 0 dB.
Those are the modules listed, but like any analog or virtual-analog synthesizer, the real power comes from the way the modules are patched together. The Audio module produces a whole raft of very usable signals from any sound input (although it is optimized for voice or other monophonic sounds), and these signals can be combined in many orthodox and non-orthodox ways with the many control inputs of the modules, such as the Oscillators, the Formant module, and the Delay. Although by no means infinite, there are plenty of ways to use this assortment of modules to explore lots of sonic territory. And the addition of quantization possibilities on the Audio and Formant modules means that the realm of fine pitch exploration is added to the world of external sound control, resulting in a powerful small synthesizer that is a proud addition to the Madrona Labs stable.
Apparently, people have been asking Randy Jones to release the modules he’s developed in his stand-alone synths as a patching toolkit. So that one could have the granulator, and physically modelled resonators of Kaivo combinable with the complex oscillator and filters of Aalto, and the formant vocoder and delay unit of Virta all combinable in one package. I agree with this – Randy has spent so much time and effort in bringing these unique modules to perfection; that being able to combine them would produce a very unique and powerful synthesis monster. That’s a monster I would happily play with.
For now, though, at $89, Virta is a real bargain. And given the power for sound exploration and timbral modification that it has, I recommend it highly.