Review by: Chris Sanchez

SpectraLayers Pro (h.f. ‘SLP’) is a new audio processing program developed by Robin Lobel.  It is distributed by Sony Creative Software Inc (SCS).  SLP installed quickly and easily on my 2.3 GHZ Macbook pro; it loads fast and works efficiently. As far as what it does: SLP is, essentially, a novel way of displaying and allowing manipulation of audio frequency distribution over time, combined with some pretty smart ‘pattern-recognition’ which allows the user to isolate specific audio-mix-components based on the software’s understanding of naturally-occurring harmonic relationships.  In terms of the visual interface: imagine a cross between a spectrum analyzer and Adobe Photoshop and you are in the right direction.  

As far as what it does: SLP is, essentially, a novel way of displaying and allowing manipulation of audio frequency distribution over time, combined with some pretty smart ‘pattern-recognition’ which allows the user to isolate specific audio-mix-components based on the software’s understanding of naturally-occurring harmonic relationships. 

In a basic sense, SLP gives the user total control of isolating specific frequencies, at specific times, within any audio file: an infinite number of infinitely sharp filters with micro-second control over their operation.   This is all pretty straightforward.  But here’s where it gets interesting:  SLP is designed to recognize and pattern harmonic relationships. So rather than simply allowing the user to visually recognize and then manipulate specific frequency-bands, it actually does a pretty good job of recognizing and isolating specificmix components.  For instance: If you open up a pop/rock stereo mix in SLP, and you can identify which of the visual-information-clusters relates to, say, the lead-vocal track that you are hearing, and you can also identify whether you are looking at the fundamental tone (or which harmonic), SLP will be able to ‘track’ that part pretty well over whatever time-region you drag the mouse.  You can then remove it (via phase cancellation) from the mix, or render it out as a separate layer. 

A bit of background: my current audio work consists mainly of recording and producing rock and pop groups, as well as television underscoring and sound design for commercials and industrials.  I do all of this using Pro Tools HD3 and Pro Tools 9 plus the usual flotilla of plug-ins, microphones, instruments, and outboard hardware processors.  I tested SLP for both sound-design and musical application, and I will describe both of those tests here. 

Since the SONY demo that I watched (more on that later..) was based around a predictable sort of audio-post task, I decided to start my review with a similar challenge.  I’d recently recorded a lengthy bit of bird-sanctuary location-sound for use in a music production.  The bird sanctuary happens to be rather close to an airport, though, so it was difficult to get a recording of significant length without some airplane noise.  This challenge for SLP was therefore: de-plane my birds.  I imported the 4:00 stereo file into SLP and, within moments, identified the airplane sound visually based on where it entered the mix timeline-wise.  Now, I could only clearly see some of the upper harmonics of the airplane noise due to background sound, mic-pre noise, etc; the low-frequency end of the sound-file was pretty much a mess.  But, SLP has got you covered: it’s simply a matter of figuring out (through trial and error) which # harmonic you are clicking on, and then SLP will automatically track back down to the fundamental (and include all the harmonics in-between.  Pretty easy.  Oh, and the doppler shift that happens as the place passes?  Not an issue.  As I said earlier, SLP does not simply track an exact frequency range that the user specifies: it tries to ‘figure out’ how the user-specified mix component is changing and track along.  And, in the case of something as subtle as doppler shift, it can do this very well.  

Result: airplane noise pretty much gone in about sixty seconds. 

Moving on to music:  since the demo video did not offer any explicit instruction as far as handling music or even voice, I did not know what to expect from SLP in terms of tracking and isolating specific musical components.  I started with a few different stereo music files: Bob Seeger System’s “Ramblin Gamblin Man,” a very garage-y lo-fi rocker from 1968 which I had imported from a very dusty LP; a circa 1965 Francois Hardy track imported from CD, and Larry Norman’s ‘Lonely by Myself’ which is a well-produced 70’s pop track similar to the solo Lennon ballads in terms of the production.  I started with the Seeger.  In about 15 seconds I was able to isolate his lead vocal track and solo it.  Chills.  No joke. It was kinda scary.  This is way, way beyond anything that you can do with any sort of filter plug-in that I am aware of. In order to do this sort of thing with tools that I had been familiar with before today, you would basically need to run nineteen (SLP can track a max of 19 harmonics) 24db/octave filters, all with identical frequency-automation curves thatyou would have to input manually.  So it is certainly possible to do this sort of thing on Pro Tools, but…  who would bother?  Anyway, back to the chills. I’ve listened to that Seeger track a hundred times, but now I was hearing every little nuance of his lead vocal with perfect detail.  Amazing.  I was reminded of the first time that I heard convolution-filtering demo’d for me when I was an undergraduate (this would have been the mid-nineties).   Our professor played us ‘Golden Slumbers’ by the Beatles, and then immediately played us just the drum part from the recording.  He achieved this through painstaking trial and error with some sort of program that I can’t quite recall.  The effect was remarkable, though, and SLP seems to be the next logical step on this direction.

Now, this software is not a magic-wand: there were chirpy artifacts in the vocal iso, there were times when the organ came in on the tail-end of a word, etc.  But was the quality of the iso good enough that you could use it, for instance, on some sort of club remix?  Absolutely.  BTW I am aware that no one is rushing to make club-remixes of 45 year-old Bob Seeger System songs; but you catch my drift.

I tried the same experiment again with the Hardy and Norman tracks; it worked well with the Hardy, not so much so with the Norman.  The Norman track is very clean and bright, and I could not get the piano to separate fully from the voice.  Similarly, the bass guitar and the piano are so totally in-tune that the software could not distinguish between parts of them. 

As far as criticism: the software seems to have been developed primarily for use by PC-familiar users; some common MAC keyboard shortcuts don’t seem to work, and the overall interface feels a bit foreign to me (I have not used a PC since the 1980s). The documentation that comes with the program does describe the feature-set clearly enough, but strangely there is absolutely no mention WHATSOEVER of what this software might actually be used for.  Not one single indication of an actual real-world situation in which SLP would be the best tool, or even an appropriate tool! I played around with it for almost an hour (after reading the manual) and all I could achieve was weird techno filtering effects.  SCS then turned me on to an excellent one-hour online tutorial in which the demonstrator successfully isolates a poorly placed sound-effect, an annoying telecine transfer machine artifact, and broadband noise from a circa 1950 video clip.  You can watch the tutorial here:  http://www.sonycreativesoftware.com/spectralayersintrowebinar.  

If you are interested at all in this software, the tutorial will give you a good idea of what it can do for you, at least as far as audio-post applications. 

My one real gripe: the software displays time on the horizontal axis and frequency on the vertical, as you might expect.  However, the frequency axis is displayed in a LINEAR as opposed to Log manner (I.E., the span between 19Khz - 20Khz is displayed within the same space a 0 Hz  - 1K hz).  Based on the way that humans actually hear and use sound this is fairly nonsensical.   Since the dawn of audio-measurement, frequency distribution is generally shown in a Log10 manner (IE., the amount of space devoted to 100 hz to 1khz is the same as the space devoted to 1khz through 10khz).  Maybe there is some way to change the scaling, but the online manual did not reveal it.  I found that I really only needed detailed resolution in the 20hz - 2Khz area; this is where the fundamental frequency of most sonic information lies, and given that the program can automatically track and manipulate however many harmonics the user specifies, it seems like log scaling of frequency might be more helpful to users.

 For all of the advances in audio technology that we’ve experienced in the DAW era - autotune, polyphonic guitar tuners, Beat Detective - software is still unable to clearly distinguish one mix component from another with the kind of clarity that humans do intuitively.  Even a small child can clearly identify and ‘hear’ what the drum part, rather than the harmonica part, is doing in a Led Zeppelin song; but try to ask yr DAW to iso Bonham’s drums from ‘When The Levee Breaks’:  it ain’t gonna happen.  Yet.  Convolution filtering was a major step in this direction, though, and I feel like the technology behind Spectra Layers Pro is the next logical step towards truly intelligent audio-computing. 

Chris Sanchez, chief engineer and principal of Gold Coast Recorders in Bridgeport Connecticut.  Chris blogs about audio technology and audio history at PreservationSound.com