August 15, 2008

Zenph Studios

When Art Tatum sat at a studio piano in 1933 to commit, for the first time, some of his solo performances to wax, the science of sound recording, while already half a century old, was still in its technological infancy. Louis Armstrong had made his first records with the Hot Five and Hot Seven only eight years before, and the scratchiness and poor fidelity of those acoustically recorded sides -- which make them hard to listen to for some -- hadn’t markedly diminished by the time Tatum entered a studio. Brunswick Records released the four songs Tatum recorded that day on 78rpm platters, and more than 20 years later Columbia Records (which by then owned Brunswick) put them on an LP, along with eight solo performances Tatum had recorded live at the Shrine Auditorium in Los Angeles in 1949. That LP, Piano Starts Here, is still in print on CD. But the 1949 recordings weren’t much better than the 1933 tracks. Indeed, as Richard Cook and Brian Morton noted in The Penguin Guide to Jazz Recordings, "(Tatum’s) whole discography is marred by inadequate recordings." Even Tatum’s late recordings, from the 1950s, lack the sonic sparkle that an artist of his stature deserves.

To bring Tatum’s piano playing into sharper focus, to present a clearer idea of how he sounded, to give his work the level of sound it justifies, is the sort of challenge Zenph Studios takes very seriously. Zenph’s "re-performance" of Piano Starts Here follows its highly acclaimed work, released last year, on pianist Glenn Gould’s 1955 recording of J.S. Bach’s The Goldberg Variations.

Zenph Studios is a software company based in North Carolina that, as stated on its website, "specializes in the algorithms and processes for understanding -- and re-creating -- precisely how musicians perform." John Q. Walker and Peter J. Schwaller established Zenph in 2002, two years after they and two other partners sold another successful software company they had developed. Walker’s extensive academic background in mathematics, computer science, and music uniquely qualified him to pursue Zenph’s goals.

It was while studying piano at the University of Illinois, Urbana-Champaign, that the scientific and artistic strains of Walker’s education began to coalesce. One of his piano teachers there was Ruth Slenczynska, who had studied with Sergei Rachmaninoff. "Ruth used to show me examples of playing," Dr. Walker told me in an e-mail, "and say things like ‘Mr. Rachmaninoff showed me this,’ and ‘here’s how Mr. Rachmaninoff would play this passage.’ And I thought how amazing it would have been to sit in the room and hear Rachmaninoff himself play, except that he had passed away years before I was born."

Walker thought of other great musicians he hadn’t gotten to hear live, including "Leopold Godowsky, Gershwin, and Fats Waller. Glenn Gould retired before I could hear him. As a computer scientist, this looked like a ‘virtual reality’ problem that I thought we could address when the time came." Walker was waiting for hardware and computer capability to "cross the reality-acceptance gap -- good enough to convince the ears that they were hearing a real musician playing live in the room."

The piece of piano hardware that convinced him he could bring his idea to fruition was the Yamaha Disklavier Pro, an actual acoustic piano that can interact with MIDI and PC software to re-create a human performance. "When we saw and heard this extraordinary piece of Japanese engineering being used in a piano competition -- where the subtlest of nuances are extremely important -- we knew we could move forward." After that, it was a matter of developing software that could re-create the smallest shades of expression in a musician’s performance, then programming that software to interact successfully with the Disklavier.

The development of Zenph’s software was based on close observation and documentation of how musicians approach their craft. "We spent a few years bringing in live musicians to give live performances, so that we could record them microscopically -- to understand the range of human playing. Musicians know how to do these things -- and their teachers explain it to them and can give examples -- but we found there was no science, no formulae, for what it meant to play staccato or to pedal in a certain manner." With this research as its basis, the company uses the original recording as a baseline to painstakingly re-create what the musician originally played. The result is what Zenph calls a "re-performance."

Walker compares Zenph’s approach to that of Pixar, the successful computer-animation studio. "They have an intense amount of high-resolution computer hardware and software specialized to the animation tasks, but they still have highly skilled animators who guide the final results," he explained. "The computers are at the service of video experts. We similarly have professional musicians, in this case experts at how the piano is played, who work with high-resolution computer software to assure the accuracy of the re-performances."

One of Zenph’s first projects was a re-performance of a recording from the 1920s by pianist Alfred Cortot. The results were so accurate that they caught even Cortot’s mistakes. But it was with Sony Classical’s 2007 release of the company’s re-creation of Glenn Gould’s debut recording that Zenph became more widely known. Gould was 23 when he recorded The Goldberg Variations in 1955 in New York, and its release the following year secured his reputation as a pianist of unique interpretive and technical skills. He would forever be associated with this Bach work, and recorded it again in 1981, the year before he died. The 1955 recording was monaural; Zenph recorded its re-performance in stereo, binaural stereo, and multichannel: the first two for CD, all three for SACD. This time around, the music emanated from a Disklavier Pro in Glenn Gould Studios, in Toronto, Ontario.

It isn’t hard to hear why Gould’s original performance was so compelling and popular. It was intensely emotional yet technically astounding. Even on vinyl, however, the recording contains a lot of tape hiss and other noise, including what sound like floorboard creaks, intakes of breath, and occasional vocalizations by Gould. The re-performance cleans away all of that to give us Gould in three dimensions. In addition, the piano sounds more tonally correct; on the original, it sounds thin in the upper register. In a close comparison, the performance on the new recording sounds uncannily like Gould’s original. The binaural tracks, designed to be listened to through headphones, let you hear what Gould heard as he played.

With the Tatum recordings, the recording quality of the source material was even worse than with the Gould. I was surprised at how good Tatum sounded in the 1933 performances on my LP, which was probably pressed sometime in the 1970s. The piano rings out clearly, and the recording captures the excitement, the drive, even the subtleties of his technique. There’s no denying, however, that the source recording contains a lot of surface noise. For me, that places it in its time and gives it a period charm, but others find the noise distracting. The 1949 live recordings at the Shrine contain a lot of tape hiss and actually sound flatter than the earlier studio performances. The Zenph version removes all that fog and lets you concentrate on Tatum’s dazzling runs and his astonishing left hand, which plays quick stride lines one moment, complex chord changes the next. The sound is more three-dimensional, and Tatum’s playing is coming from a quality instrument, which was not always the case in real life.

Dr. Walker and his team carefully research the original sources for their re-performances. He described some of the things they found when sifting through the Tatum recordings: "In working with the Tatum album, we discovered that some flaws had crept in over the 50+ years since the original release. So, we worked from the most recent CD release, but then we made an effort to obtain all the original LPs. This is one of those places where eBay became quite helpful -- someone selling off grandpa’s old record collection had a copy of the very first LP release. Finally, we worked closely with Arnold Laubich, who was Art Tatum’s discographer. His book points out a flaw in the track titled ‘The Man I Love.’ Two minutes of Tatum’s performance had been excised -- and Arnold had a copy of the original performance, from which we could restore the lost material." The Zenph release includes those two minutes, and corrects a speed anomaly in the original release -- on 12 of the 13 tracks, Tatum actually played faster.

I asked Walker about the discomfort many might feel about re-performances -- which remind me of something from a Philip K. Dick novel. "In our opinion, all the conversions, remasterings, and edits that have intervened have perturbed the original performance," he wrote in his e-mail. "We go back to what exactly the musician performed, and start all over with new recordings -- from a variety of perspectives. Before Sony’s binaural recordings, there wasn’t a way to hear what Tatum himself was hearing as he played -- and it’s quite a revelation. . . . So, our best recommendation is simply to listen and discover a wealth of revelations."

I wondered about future Zenph projects, and if John Q. Walker foresees his company bringing its magic to vocal performances. "You’ve seen just a few albums from us so far," he wrote to me, "but we want to soon be making hundreds each year. We plan to expand into all instruments and genres, with voice being the most difficult. It is the playback and parameterization of voice that is still so hard -- describing exactly what it is that makes each voice distinctive and being able to faithfully replicate those, with all their colors and phonemes. So, yes, we look forward to hearing some of the greatest voices in history in re-performance: Caruso, Bessie Smith, Billie Holiday, Louis Armstrong, and the Beatles should all be singing again within a decade."

The two Zenph discs released so far are wonderful sonic experiences that let you hear more clearly than ever what Glenn Gould and Art Tatum were playing. For me, Gould’s vocalizations and the sounds in the studio are part of what’s enjoyable about his 1955 recording of The Goldberg Variations. The noise on Tatum’s early recordings reminds us that the ability to record and hear forever what a musician has played is a great gift, even in its most primitive form. Those recordings will always be there. Zenph gives us an exciting alternative.

. . . Joseph Taylor
josepht@soundstageav.com

 


All contents copyright © Schneider Publishing, Inc.; all rights reserved.
Any reproduction, without permission, is prohibited.
SoundStage! is part of Schneider Publishing, Inc. and the SoundStage! Network