Plug-ins have certainly come a long way from being regarded as "one-cell brains" and the poor stepchildren of hardware processors. The development of sufficiently powerful DSP chips and, eventually, even more powerful native computer CPUs has hastened their evolution, permitting the incorporation of increasingly complex mathematics. While earlier Sonnox (formerly Sony Oxford) plug-ins such as their Oxford EQ and Dynamics are still regarded as among the best, until recently they still required outboard DSP engines such as TC's PowerCore or Pro Tools|HD's TDM engine to run them. Today's multi-core CPUs have changed that; those two plug-ins are now available in native versions.
Native CPUs also support the kind of complex maths and code writing that would be extremely difficult, if not impossible, to program for standard audio DSP chips. The latest example is Sonnox's new SuprEsser, an amazing application that combines a simple and intuitive de-esser with an extremely sophisticated frequency-specific compressor. It's available for Macs and PCs, with RTAS, VST, and AU compatibility. Its heritage derives from the compressor section of Oxford Dynamics, with linear phase crossover filters based on code from Oxford EQ.
As a de-esser, SuprEsser can treat vocal sibilance and fricatives as well as whistles and various artifacts associated with wind and string instruments; it can also remove low-end plosives and thuds without affecting neighboring frequencies. Its graphic display features a real-time FFT (Fast Fourier Transform) of the input signal, showing exactly what the plug-in is doing at any given instant — but the incredible thing is that its action is auto-tracking. In other words, the threshold and central FFT peak are constantly adjusting to the audio input; the FFT shows only the frequency band being treated, so it changes with adjustment of the octave and frequency controls. The spectral change is shown in red whenever gain reduction occurs. If one clicks on another button labeled "Access," the FFT window shrinks somewhat and additional controls appear, which provide a complete set of compressor/limiter adjustments.
Oxford SuprEsser is supplied with a comprehensive set of twenty-eight presets in seven categories: Vocals, Winds, Strings, Bass, Standard Compressor Modes, Analysis, and Effects. Although designed as starting points, most are remarkably effective as is. The vocal presets are called "Gentle and Natural," "Harder but Natural," "Harder and Lithping," "Vocal Soften," and "Whistle Control." One can "derasp" a sax, "deharsh" a trumpet or a cello, and reduce breath noises and key hits on wind instruments. Acoustic guitars can be "detwanged," electric guitars can be "debitten," and basses can be "deboomed." One can also set SuprEsser up to act as a brickwall limiter or a traditional compressor with or without side-chain EQ.
Fast Facts Applications
Studio, project studio, audio post, audio for broadcast, live; treatment of aggressive frequencies in vocal and instrumental pick-ups.
28 presents in seven categories — Vocals, Winds, Strings, Bass, Standard Compressor Modes, Analysis, and Effects; automatic level-tracking follows energy level; intuitive graphic display.
Sonnox | www.sonnoxplugins.com I started my tests by purposely making a sibilant recording of a female vocalist. It took less time than it takes to type this sentence to remove the sibilance completely and absolutely transparently. Once I selected the Gentle and Natural preset (using Bias Peak Pro XT 6 as host), I tweaked the center frequency, width and slope, and then adjusted the threshold downward. When I liked what I heard, I watched as the plug-in tracked the rest of the vocal automatically! I was previously accustomed to the limited adjustments possible with the standard analog hardware de-esser in my Manley VoxBox, so using this "automatic digital parametric de-esser" was quite a revelation.
One of SuprEsser's coolest features is its ability to monitor "inside" or "outside" the mix. One can hear the output of the bandpass filter, as processed by the compressor—i.e., "inside," or the output of the band-reject filter—i.e., everything "outside" the bandpass window, and also simultaneously adjust the wet/dry mix, since SuprEsser's architecture permits mixing the original untreated signal with its processed outputs—analogous to parallel compression techniques. These three controls, plus the three mentioned above, allowed me to produce perfect results quickly.
All this power is not without a price. The maths, much of which involve convolution, introduce processor delay. Since SuprEsser has two filters, their "kernel size" is directly proportional to both latency and accuracy at the low end. Accordingly, Sonnox supplies SuprEsser in three versions: the standard Oxford SuprEsser, with a delay of 1,044-3,092 samples, SuprEsser HR—designed for work at the higher sample rates—with delays between 4,116 and 12,308 samples, and SuprEsser LL, the low latency version, with delays between 276 and 3,072 samples. This one is not recommended for processing below 400Hz, but can be used easily while tracking.
I also processed a recording I had made of a local flutist, whose sound, although beautiful, also contained a bit too much air. Despite having recorded her using the most flattering of vintage tube microphones, I was still not satisfied. Enter SuprEsser's "Reduce Embouchure Blow" preset and violà! There was much less breath noise and, amazingly, it didn't touch the high end on the piano. Just watching the gain reduction meter was a revelation. It dipped to -12dB in the "worst" places when the flute was playing but stayed right at 0dB for each piano interlude —and the results sounded almost identical to the original, only minus the breathiness. Product Points Plus
- Works transparently, produces excellent results, easy to use.
- Relatively high latency in high-resolution mode.
It's simply the best de-esser plug-ins presently available — but it's a lot more, too.
For my final test, I ripped the famous 1956 Spike Jones hit, "All I Want for Christmas Is My Two Front Teeth" from CD, which features an erstwhile "kid" singing as if he were missing his two front teeth, making quite loud whistles each time he tries to pronounce an "s," as in "Sister Suzy" or "For Goodness Sake." I've concluded that the master must have been edited, because many of the whistles precede the S sounds too much, and they're also almost at full level. Nonetheless, I decided to see if SuprEsser could eliminate the whistles in "For Goodness Sake," which is near the very end of the tune.
I ended up using two instantiations of SuprEsser's Whistle preset, adjusted slightly differently (since each has a maximum of 27dB gain reduction). The whistle frequency was 4.9kHz, and I used widths of .4 oct. and .3 oct., with a sharp slope of 72dB/8ve. I lowered each threshold to -40dB. This double treatment just about eliminated the whistling, with quite minimal side effects, but left some breathiness. So I added a third instantiation, employing the aforementioned "Reduce Embouchure Blow" preset, and that lessened the breath considerably.
I admit that this exercise was an extreme one — and Spike Jones' hit was all about the whistles — but it definitely shows the power that lies underneath SuprEsser's hood. If it can de-whistle "Two Front Teeth," it can do just about anything!
Oxford SuprEsser is the new standard for de-essers and redefines the frequency-specific compressor category. I highly recommend it!
Dr. Fred Bashur has received credits on hundreds of recordings during the past 30 years. He is also a regular contributor to Pro Audio Review.