HARTMANN NEURON VS

Several years ago, German company Hartmann launched a formidable synthesizer keyboard called, simply, the Neuron. From the high-tech toy shop (or is that
Author:
Publish date:

Several years ago, German company Hartmann launched a formidable synthesizer keyboard called, simply, the Neuron. From the high-tech toy shop (or is that science lab?) of Axel Hartmann — the man behind other high-end instruments such as the Waldorf Wave and Q and the Alesis Andromeda — the Neuron represented all there was to love, and to hate, from a pioneering technology. Sonically, the Neuron stood out for its ability to create complex multidimensional “sound clouds” (to use Hartmann lingo) that often defied easy description or natural source and could be transitionally melded into each other in a way vaguely similar to vector synthesis. Based instead on another exciting and advanced form of synthesis known as resynthesis, it allowed for unprecedented design and control of real-world sound models derived from algorithmic analysis of traditional samples. Not to be confused with physical modelers or virtual-analog synthesizers, the Neuron was the first product to conveniently package this previously research-only technology into a consumer-feasible interface.

Still, the beast was a far-out design (putting it mildly) — with numerous hardware shortcomings — stocked with all sorts of odd-looking and strangely arranged buttons, scroll controls and joysticks as well as a bevy of pretty, pretty lights, each having equally puzzling labels and parameter assignments. Intuitive, it was not. Just looking at the instrument was daunting enough, let alone the $5,000 price tag. But thanks to the power of today's computers, the Neuron concept has been adapted for regular people in the form of a significantly more affordable VST or Audio Units instrument that runs on PC or Mac.

Neuron VS features an extremely elegant software interface that closely mimics the various control sections of the hardware version's front-panel layout. Functionally, little is different between the two. Whereas the keyboard was multitimbral in operation, the plug-in is not, but multiple instantiations cure that. Likewise, being a VSTi, Neuron VS has no audio input; although the hardware Neuron did have physical audio inputs, strangely, they were never implemented in its software.

No doubt, a major sales incentive of Neuron VS will be its cute little Nuke controller. A copy-protection dongle in disguise, the hand-size Nuke connects to your computer via USB and provides essential tactile control of many of Neuron VS' interrelated edit parameters. As you'll soon appreciate, once you explore the unique and complex synthesis structure of Neuron VS, the Nuke's soft-assignable stick controller (identical to that found on the hardware version) and four rotary encoders greatly streamline and make easy to navigate other-wise impossibly demanding edit gestures during performance.

I tested Neuron VS on a Pentium 4/3.2GHz, with 2 GB of RAM, running Steinberg Nuendo 2.0 as my host application under Windows XP. The setup is a snap. Included are two installation discs: a CD to install the Neuron VS plug-in and preset patches and a DVD to install more than 2 GB of factory models. Having the Nuke connected is mandatory to run the software, so I plugged in the supplied 10-foot USB cable and was appropriately presented with Windows' New Hardware Detected dialog. I selected Automatic Installation from the hardware wizard, and the installer picked up the custom Nuke driver from the CD. This process is not necessary on the Mac, as the Nuke is ready to use as soon as it's plugged in.

ANOTHER HEADSPACE

In case you're not up on the original Neuron and how it generates sound, here's a quick primer: The Neuron produces its sound in a way unlike your conventional physical-modeling synths, samplers or virtual analogs. Using a fairly heady (pun intended) process — a specialized form of artificial intelligence and pattern recognition that Hartmann calls Neural Networking — the Neuron analyzes sample data to produce algorithmic models. Now is a good time to note that when discussing synthesis, Hartmann likes to use original, often ambiguous names and terminology for familiar processes.

Neuron models are a combination of musically meaningful values, including the formative parameters of a sound (formants), categorized into groups called Scape and Sphere. Scape is best thought of as the excitation or action stage of an instrument or audible occurrence being sounded; this would include plucking a string on a guitar, striking a drum or blowing into a wind instrument. Conversely, Sphere parameters represent the variety of extrinsic sound-shaping factors and influences that may be forced upon that sonic energy after the initial excitation stage: guitar body, drum shell or flute size, for instance. Once a sound model is created, the sample data is discarded and only the model information is used to resynthesize the sound. This results in sounds that can be freely manipulated in ways impossible with sample-based synthesis or even granular synthesis.

Although Neuron VS comes with 340 sound models from the factory — covering an enormous range including ambient pads and synthetic textures; loops and drum kits; electro-acoustic pianos and guitars; choirs; bass and synth bass; orchestral strings; woodwinds and brass; wavetable, spectrum and sound effects; and more conventional classic-synth waveforms — it wasn't until recently that Hartmann decided to release to VS owners a piece of code called MODELmaker that allows you to import your own samples for conversion into Neuron models (more on that later).

The most basic building block in the Neuron's synthesis structure is something called a Resynator. Inspired by the notion of a resynthesis oscillator, the Resynator is used to render an actual sound from a model description. At the voicing level, Neuron VS contains two Resynators, or oscillators; an exotic mixer section; a generous assortment of envelopes; an LFO; and something akin to a multiband LFO, called a Slicer. From there, signals travel through the polishing tools of a filter and multi-effects section, collectively referred to as Silver. The Neuron VS user interface is divided into three areas referred to as Screens: Resys, Silver and Remote. At the very top on all Screens is the Programmer module, the sound-management center where patches are selected, loaded and saved.

BREAKING IT DOWN

Positioned at left and right on the Resys Screen, the Resynator 1 and 2 modules are identical in design and provide a wealth of parameter and control features. Most notable is the large circular porthole residing smack-dab in the center of each Resynator. A visual feedback of the Nuke's stick controller position, the porthole is cornered by four text/value displays, as well as a button for selecting among parameter levels and a button to switch between Scape and Sphere modes. To understand how these all relate, look at a patch.

Loading preset 126, Romantic Layer, reveals that Resynator 1 is using a model of a “haunted piano” sound while Resynator 2 is using a synth-pad model. Focusing on Resynator 1, you can see what parameters control its sound by exploring the Scape and Sphere modes. For every model, each Scape and Sphere component contains as many as 12 parameters arranged as three groups of four. Within each grouping of four are two sets of associated parameters, aligned opposite each other around the Resynator porthole. In the example here, the piano model of Resynator 1 has the following 12 Scape parameters cross-matched around the porthole: metal versus plastic string, metal versus wood hammer, metal versus wood frame, loose versus solid frame, normal versus torn frame and wobbly versus tense strings. The Sphere mode reveals parameters such as small or large case, metal or wood, dampened or open and so forth. Other model types might include parameters such as friction, torsion, tubular, planar, warm, cold and so on. Not all Hartmann parameter names make sense, but you tend to get the drift once you play a few notes and tweak them.

Using the Nuke stick controller, you're able to literally morph these parameters into and out of each other — not simply mix volumes as you would with vector synthesis à la Korg Wavestation. Rather, Neuron recalculates and interpolates values as you sweep through all 360 degrees of motion. Pressing the selector button directly below the joystick on the Nuke selects whether you control Resynator 1, Resynator 2 or filters or effects in the Silver Screen. A special Stick Animation feature allows you to write short movements into a preset for saving. While on the topic, you can also set the four rotary encoders on the Nuke to control any of the 63 assignable parameters found on the Remote Screen. All stick and knob movements are recordable by host.

The transitions that Neuron makes between Scape and Sphere parameters are not always smooth. Sometimes, a stick movement will result in extreme audio annoyances such as squelches and bursts, and other times, the parameter blends simply don't make contextual sense with what the source models are. Hit or miss, the batting average wasn't bad, but I learned to monitor with my speakers turned way down until I was certain a patch configuration wouldn't fry my ears — or my speakers.

Straight down the center of the Resys Screen is the Blender module, a complex mixer of sorts used to apply dynamic crossfading between two Resynators; it even lets you manipulate parts of one Resynator using another. In the latter sense, you can choose from Blender modes that allow, for example, the dynamic movement of a drum loop to act upon the body size or string tension of an instrument model. The mixing possibilities are pretty wild, and through much experimentation with the Blender module, I still feel it's a hit-or-miss proposition once you boil it down to sonic results.

Finally, sprawling across the bottom of the Resys Screen is the Shapers module. You and I know this better as an envelope generator. In fact, there are three ADSR envelope generators per Resynator, one that is hard-wired for modulation of model parameters, one that is hard-wired to modulate amplifier or Resynator volume, and another free envelope capable of modulating the Resynator pitch or the Blender amount. A special Shaper Mode button allows you to select between normal (one-shot) and repeat triggering.

THE SILVER SCREEN

Next in the line of signal flow is the Slicer module, a simple low-frequency oscillator that is selectable between two basic modes: Vertical and 3D. Vertical mode is nothing more than a fancy name for amplitude modulation, creating oscillation to the output signals of Resynators 1 and 2. In 3D mode, Slicer introduces frequency or pitch modulation into the equation, creating the sense of an LFO meandering its way not only up and down along a volume plane but also simultaneously back and forth along a frequency plane. Modulation shapes for each mode consist of 12 waveform choices controllable by depth (Vertical) or spread (3D) and frequency.

A more traditional modulation section, appropriately named Mod, is also present on the Silver Screen. The effects of this module can be routed to control volume, pitch and model parameters of both Resynators; the amount of the Blender; and the cutoff frequency of the filter currently selected in the Silver module (discussed momentarily). Once again, there are 12 modulation shapes to choose from, though the modulation rate appears to be in arbitrary units of 0 to 1,000, which equal 0.0 to 20.0 Hz in fine 0.02Hz increments. There is also a delay parameter that allows you to set the modulation effect to fade in and out softly for every note pressed. In a rather cool twist, modulation routing is determined by adjusting the destination's LFO Depth parameter rather than in an active modulation matrix at the LFO source. This allows for more intuitive dialing in of modulation amounts at the particular module that you want to emphasize or effect. Disappointing is that neither Slicer nor Mod LFOs sync to host or MIDI Clock.

The bulk of the Silver Screen is the Silver module itself, which is nothing more spectacular than a rather generic-sounding multimode filter and multi-effects processor. Filter types include 6, 12 and 24dB lowpass and 6dB highpass and bandpass. Each filter type is resonant and modulatable via the Nuke, LFO, velocity and the dedicated Silver Shaper envelope at the bottom of the screen.

On either side of the filter unit is a full assortment of insert effects. To the left is the frequency-based effects unit (Freq FX), which includes a dutiful parametric EQ/compressor, distortion, a ring modulator, a decimator and a frequency modulator called SP Warp. To the filter's right are time-based effects (Time FX) including stereo spread, stereo delay, flanger, phaser and chorus. None of the effects knock me out either in quality or features, but they certainly do add to the Neuron VS experience, especially when you kick the Nuke into action and begin mixing and mashing parameters in 3D.

MODEL SEARCH

As noted previously, Hartmann has kindly released the MODELmaker software as a free download for registered Neuron VS users, making it possible to analyze your own sample material and generate new Neuron models from it. Written for Java, MODELmaker is an external utility to Neuron VS, requiring Sun's JRE Java Realtime Environment to be installed on your system. The graphical user interface appeared simple and innocent-looking at first, but I quickly realized that a trip or two to the PDF manual that came with the download was in order. My first minor frustration with MODELmaker is that it doesn't allow you to sample directly from a sound source, requiring instead that you already have sample files recorded, trimmed and saved on disk. Second, it can only import mono 16-bit, 44.1kHz sample files saved in AIFF format; this is not a huge inconvenience for anyone with a DAW, of course, but it would be thoughtful to include stereo-to-mono, sample and file-format conversion in a future version.

Once I got myself around that little niggle, I found another when I tried loading in a simple three-second-long vocal sample. If anyone else is familiar with the dime-store theory on resynthesis, it should come as a bit of a surprise that you have to map out pitched root samples to create a usable Neuron model that stretches the length of the keyboard — in a similar fashion as you would use to create a multisample for a sampler. Indeed, I was hoping the theory would allow me to load in a single sample and that MODELmaker would do the job of pitching it for me. Disappointingly, what I ended up with was a terrible case of the chipmunks an octave up and merely a glitch-short shreep at the top end. Begrudgingly accepting this, I began importing some pitched vocal samples and proceeded to map them out, setting low and high note values and root keys. I didn't have samples of multiple velocities, but MODELmaker does have provisions for a max of two velocity layers per sample region. Also, I should note that the multisample approach is kind of convenient should you wish to map out percussion sounds or various loops to be modeled kit-style.

With all samples loaded and mapped, I selected the highest Complexity (quality) setting, pressed Process, and away it went. A few seconds later, I was told that a model had been created. Back in Neuron VS, I assigned my newly created model to Resynator 1 and began to experiment. Met with truly mediocre results, I was disappointed that no matter how much I tweaked the various modules, the modeled results of my pristine vocal sample carried only a vague resemblance, never turning out as clean and intelligible as resynthesis theory would have you believe they should. This is not to say, of course, that they weren't usable in their own, newly mutated way, for they most certainly were. Similarly, I had interesting success modeling samples of drum loops at lower Complexity settings.

As a point of interest, MODELmaker creates two new folders within its own, called Models and Setup. Models is the self-explanatory destination for newly created models whereas Setup stores files containing only raw keymap and velocity-layer information for works in progress or later editing. Because MODELmaker frustratingly imports samples at folder level only, not file, I found it necessary to also create a folder called Samples and various subfolders in which to store my AIFF files for each model. Also, MODELmaker works on a specific file-name convention for importing files, so I had to go and rename each file. As if not enough footwork already, I was disappointed to learn that MODELmaker would not save my new model directly into the Neuron VS user-model banks. Instead, you have to manually move your compiled models into the folder yourself. Needless to say, MODELmaker is currently the Achilles' heel of Neuron VS!

AURAL PLEASURE

Both deep and unorthodox in features, parameters and intellect, Neuron VS is neither something you learn overnight nor appreciate fully straight away. In fact, I was initially quite unimpressed by what I heard in many of the 300-plus factory presets.

Realistically samplerlike, Neuron is not. It does not even come close to competing with physical-modeling synths in this regard — at least, again, not from the presets or factory models that I auditioned. And this was further evidenced in my exploration of MODELmaker. Feeling and sounding closer to Emagic's Logic 7 freebie plug-in Sculpture than anything else, Neuron VS can produce sounds of familiarity but without conveying any sense of realism. This all leads me to believe that Neuron's neural-networking-artificial-intelligence-pattern-recognition whatchamacallit technology is still rather in its infancy and needs further development to become truly formidable.

On the flip side, this same psycho-technology opens some truly exciting doors for a sound designer at heart. The presets that caught my ear were of the far more exotic variety, displaying unique synthetic characteristics that I have never heard before: techno loops passing through violin strings that you can bend and loosen tension on, acoustic drums played through a human vocal tract with airy resonance, chimes passing through bulbous metal kettles that you can dampen or fill with water — the list of craziness goes on. By fusing natural (and unnatural) formants together and freely messing with their attributes, all sorts of amazing-sounding, quasirealistic mayhem results. The orgylike interplay between each part of Neuron's voicing structure is responsible for the weirdness. Maneuvering the Nuke stick, you get all sorts of surprises ranging from disastrous (unless you're into that sort of thing) to freaked-out “musical” noise to the coolest otherworldly tones you've ever heard. The point is, Neuron VS is highly conducive to experimentation and incredibly fun to use. Excelling at mystical-sounding pads; dynamic dreamscapes; and shiny, glistening Hollywood-like sound-effects material, Neuron VS certainly can impress and does so repeatedly; it just took me a little longer to discover its true talent.

The unfortunate reality of it all is, even though Hartmann has set modest minimum system requirements on the retail box, you'll likely barely get off the ground if you aren't on a current-generation system. A single instance of Neuron VS typically chomped away at anywhere between 30 to 60 percent of my P4/3.2GHz, and that's over and above the demands of the host application.

CHECK, PLEASE

Without question, 900 clams is a lot to dish out for a plug-in that consumes a CPU at the blink of an eye. And if you install Neuron VS on a dedicated PC or Mac, consider the price of admission at least doubled or tripled. Likely a minor contributing factor to the significant price point on this package, the Nuke is a welcome inclusion. The Nuke brings the Neuron VS experience that much closer to its bigger sibling and truly polishes the package in a classy way. I can't imagine having to control Neuron VS without it. And its shortcomings aside, MODELmaker is your key to opening Neuron VS to the real world and, frankly, pushes its worth-the-money factor over the edge.

What I love about Neuron VS is how much fun making a mistake can be. By mucking around with Resynator Scapes, Spheres and filter settings, more times than not, you will accidentally strike gold. Best-suited as a sound-design tool and extraordinary live-performance tool on the decidedly avant-garde tip, Neuron VS can mangle a drum loop, warp a vocal track or add that special coating of sugar to an arrangement.

HARTMANN (DIST. BY RUSS JONES)

NEURON VS > $899

Pros: Same powerful resynthesis engine as $5,000 Neuron keyboard. Amazingly fresh, multidimensional sound. Ability to create your own models. Great-looking interface. Hardware controller included. Fun to use.

Cons: No external audio or surround support like with Neuron keyboard. Extremely CPU-demanding. Expensive.

Contact: www.hartmann-music.com or www.rjmg.com

SYSTEM REQUIREMENTS

MAC: G4/800; 256 MB RAM; Mac OS 10.2 or later; VST 2.0 — or Audio Units — compatible host application

PC: Pentium III/850; 256 MB RAM; Windows XP; VST 2.0 — compatible host application