Formed in 2011, Zynaptiq are a German company specializing in high-end signal analysis and processing. They attracted some high praise earlier this year with their first release, the real-time, polyphonic pitch processor PITCHMAP, which was seen to combine the best aspects of the established daddies of pitch-correction, Melodyne DNA and Autotune.
The company's second product, UNVEIL, seeks to make an equally ground-breaking contribution to audio processing, and to further stretch the accepted limits of what can be achieved with recorded sound. This software, its makers claim, can reduce—or boost—the level of reverb present within any recording. Zynaptiq describe it as a "real-time de-mixing" tool which intelligently separates a signal into "foreground" and "background" elements, before giving the user control over both. In metaphorical terms it could be thought of as an audio sponge, soaking up reverb and other unfocussed "filler" within a signal.
UNVEIL is currently only available for the Macintosh AudioUnits format. Zynaptiq are promising to release VST and AAX versions in future updates, but for now users will need an AU-compatible host (although UNVEIL also functions as a standalone app), as well as at least a dual-core Mac. Quad-core machines are recommended: this plug-in has a serious appetite. 64-bit operation is supported, and the software can handle any sample rate, dependent on the host CPU.
Before learning my way around the software, I had hypothesized that it must use a combination of dynamics processing, phase cancellation or de-noising algorithms. In fact, Zynaptiq have proudly avoided these processes, instead employing their own artificial intelligence system, Mixed-Signal Audio Processing. Earlier this year, Zynaptiq acquired various dormant technologies and intellectual property from Prosoniq, including the artificial neural network (ANN) algorithms behind the company's APEM de-reverberator. It is these which enable the software to "listen" discerningly to incoming audio and distinguish between its different elements, just as a human would.
Eager to discover if UNVEIL could deliver what it promised, I threw an exhaustive range of material at it: drums and synths (single sound sources); spoken dialogue; and polyphonic stereo and mono recordings including a solo piano, symphony orchestra, live band rehearsal (hastily recorded with room mics), and a variety of commercial pop and rock mixes.
Opening the plug-in and the provided quickstart guide, I was struck by their simplicity. The GUI provides only a handful of controls, dominated by five circular faders and a customizable display which illustrates the original and output signals; the guide has just two pages of practical instruction. This is one of UNVEIL's strengths: despite the complexity of the processing going on, its front-end is very intuitive. My first test of the software was a theatre piece involving two actors who had been recorded in different locations. One had been recorded in a much larger room than the other, and as a result their dialogue was sonically mismatched. Just a couple of minutes after opening UNVEIL for the first time, I had the recorded ambience under control and the audio sounding uniform throughout the scene.
Reverb and other filler within a signal is manipulated using a single control, Focus. This has a central "0" position whereby no processing is applied; moving the fader up attenuates the reverb level, while moving it down has the opposite effect, boosting the background elements and creating a thick, reverberant texture: instant "wall of sound." The other controls serve to tweak the processing as appropriate to the incoming signal. Refract and Adaptation are broadly comparable to the envelope controls of a dynamics processor, affecting the software's time-based response and helping to achieve a natural-sounding result. Localize allows the user to target specific parts of the incoming signal as defined by time and frequency, but can result in artefacts at extreme settings and with complex input signals. The Focus Bias controls affect the extent to which different frequency bands are processed: for example, low frequencies, which often don't carry much reverb, can be bypassed.
Users can monitor the processor's behaviour with the I/O Diff button, which flips the output to expose the signal elements on which UNVEIL is working. There is also a Transient Threshold control, with which users can specify and monitor whether or not transients are being processed.
I found that attempting to understand the scientific theory behind the main controls was not as important—or as effective—as simply getting a feel for them through experimentation. The documentation helpfully suggests starting points for the controls from which the user can tweak until the desired effect is achieved. Again: the software has been designed to be very simple to operate, and it's easy to forget just how hard it's working behind the scenes—until your CPU grinds to a halt. Real-time operation comes at a price, and when using UNVEIL on a couple of software instruments (which were themselves fairly processor-intensive) my CPU activity immediately spiked, playback glitched and my only recourse was the Track Freeze button—the first time I've ever used it on my 8-core Mac Pro.
The most dramatic evidence of UNVEIL's capabilities came when applying it to a stereo recording of a full symphony orchestra. This type of signal is extremely complex, with constant shifts in timbre, texture and dynamics; and natural reverb, in varying proportions, is of course an integral element. UNVEIL took all this in its stride, effectively pulling 80-plus musicians out of the space in which they had been recorded, and doing so in a way that sounded almost completely natural and transparent. To reiterate: it's not just reverb tails that are dealt with (i.e. reverb that can be heard during gaps in the musical texture), but crucially, the background ambience that exists while the orchestra are playing. A second acid test for the processor was feeding it a slice of vintage dub reggae, heavy with spring reverb and other background mud which, again, UNVEIL was able to strip away, leaving a much cleaner and more focussed signal.
The functions that UNVEIL performs have perhaps a narrower appeal than real-time, polyphonic pitch manipulation, but the software is no less impressive for that. It represents an incredible technical achievement—which is reflected in its pricing—and will prove enormously valuable to many audio professionals across a range of applications. I was incredibly impressed at both the quality of results that UNVEIL delivers, and the ease with which they can be achieved.