Clever Custom Software Used to Reconstruct First Recorded Sounds


Patrick Feaster easily recalls the moment he first heard the French lullaby "Au clair de la lune" crackle out of the speakers of his home computer. Bleary eyed, sunrise fast approaching, Feaster could clearly pick out the grainy, ethereal voice of a young girl making herself heard for the first time since she was recorded a century and a half earlier, in 1860.

Feaster is a hunter of ancient sounds stored in recordings that have been impenetrable to modern methods of playback -- or that were never even intended to be heard. But with a clever method the recent Indiana University PhD developed in his spare time, Feaster breathes life into timeworn stores of aural information.

"I stayed up all night correcting this sound file," Feaster says of the lullaby. "Gradually, over the course of the night, this voice begins to take recognizable form, until somewhere around 3 or 4 o'clock in the morning, I can hear the whole thing, and it's immediately recognizable -- I know how 'Au clair de la lune' goes. And I sat there thinking, 'I'm hearing someone singing a song before the outbreak of the American Civil War, and I'm the only person alive who has heard this.'"

The rest of the world got its chance soon after, when the played-back sound was presented to eager ears at a recording conference at Stanford in 2008. (Feaster refers to the presentations as "unveilings.") The show resonated with the audience's historical bent, because at 150 years old, this particular rendition of 'Au clair de la lune' -- recorded on a phonautograph, a device created by Édouard-Léon Scott two decades before Edison's 1877 invention of the phonograph -- became the oldest known recorded sound.

In the years since then, Feaster has been lifting the veils off of still older "recordings," if Scott's phonautograms even deserve the name. The device Scott patented in 1857 uses a stylus to trace a line onto a soot-covered cylinder, producing a visual representation of the sound unplayable by any device, contemporary or modern. But Feaster and his sound-chasing co-hobbyists, who style themselves the First Sounds Collaborative, adapted software to reconstruct the path of the stylus by analyzing images of the sooty trace. As if pulling an earthquake out of the readout from a seismograph, Feaster educes whatever sound is represented by the path of the stylus, playing the trace like any sound wave.

A trove of Scott's phonautograms was discovered in France a few years ago, but at the time, without the technology to play them, they appeared to be of mere historical value. The "virtual stylus," the name for the computer software that could finally play the phonautograms, lit a fire under sound seekers like Feaster to unearth and unlock more of the auditory caches.

"It's not everyday you get the people in charge of an archive popping open a bottle of champagne," Feaster says. "There was a reason to hunt these things down now, because we knew we could turn them into sound." (Further investigation also revealed that the 'Au clair de la lune' phonautogram had been played back at twice its intended speed, and the young girl was actually Scott himself.)

But while the search for old recordings has continued -- First Sounds was recognized with two Grammy nominations in 2008 for a naughty collection titled, "Actionable offenses: Indecent phonograph recordings from the 1890s" -- Feaster has set his eyes on even more ambitious projects.

This September, Feaster published his riposte to a challenge from a German colleague, who wondered whether Feaster could play back an 1889 recording by Emile Berliner, inventor of the gramophone. The challenge came with a twist: The physical recording itself had been lost for years -- all that remained was a print of the record in Berliner's scrapbook at the Library of Congress. Could Feaster play back the print?

"There was a lot of skepticism among my First Sounds colleagues," Feaster says. "The amplitude fluctuation on the scan of the gramophone recording is so small that we were wondering whether there would be enough information there to play it. And on top of that, we didn't even know where to start, because like an LP, this gramophone recording was a spiral."

To free the recording's voice, Feaster commissioned a high-resolution scan from the Library of Congress and then brought out the big guns: An experimental method he developed on his own by co-opting software that converts bands of varying width into playable sound, reverse engineering the age-old practice of drawing and scratching on a film soundtrack to create sound for animation. Optical film soundtracks vary a sound's specified amplitude by letting in more or less light, so Feaster edits the grooves and traces he wants to play back to make them look like optical film soundtracks, which the software easily converts into sound.

Though never applied to a spiral -- let alone a print of one -- the method had worked its magic before. Last year, Feaster unveiled a "sound photograph" taken by the noted Brown University physicist and inventor, Eli Whitney Blake, Jr. Blake had managed to capture sound wave-forms on film by bouncing a beam of light off a mirror hooked up to a telephone's mouthpiece and recording the light deflections on a photographic plate. The setup, Goldbergian though it sounds, was an improvement on an alternative model, which used the ear of a human cadaver in place of the mouthpiece.

He published the wave-forms in an 1878 issue of the prestigious scientific journal Nature, with an accompanying transcription: six vowel sounds, followed by the phrase, "Brown University, how do you do?"

Though Edison had invented the phonograph a year earlier, no recordings from then can be heard today, which means the Blake "recording" -- which, like Scott's phonautograms, was never intended to be heard -- is today the oldest playable English speech.

To tackle the tougher spiral recording, Feaster essentially unwound the image of the disc with photo editing software, and then analyzed each of the 57 concentric rotations separately, stitching them together at the end.

Unlike most playbacks, "We had no idea what the recording was going to be -- none," Feaster says. "Was it going to be singing? Speech? One of those obscene stories like from our CD? This was one of the most exciting playbacks."

To his and his colleagues' delight, the spiral recording, called "Schalldruck," was one of the clearest pieces they've unveiled:

But what it is remains a bit of a mystery, even to Feaster, who refers to its content as "sort of a spoken letter, but more like a sampler of experimental test pieces." Berliner can be heard talking, singing, laughing, and counting to ten in a number of languages.

"It's someone who's looking for something to put on a record, doesn't really care what it is, but needs something," Feaster says. "It's fascinating to see what he chose."

This, for Feaster, is what makes his hobby worth the effort. Though there is historical value in the possession and playback of our oldest recorded sounds, Feaster is interested in how early performers interacted with a device unlike any they had seen -- one that would play back the performance at a later date.

Feaster points out that a number of subtle problems had been solved, and their solutions taken for granted, by the time movies, radio and television reached the popularity they have today. Music recordings used to begin with an announcement of the recording's content, for example, but producers wondered if the voice should announce what the band was about to play (though they would have played it by the time the listener heard the recording) or what that band has already played (even though they hadn't played it yet). And should you include applause at the end of a performance, or let the listener decide?

The record is more than a piece of hardware -- it is a complex mix of technical features and social conventions that allow listeners to make sense of what they're hearing.

"I like to think of these [recordings] as the very earliest examples of a new way to use media. Until the phonograph, there wasn't any equipment out there that was designed to record a person speaking, acting, doing something, and then reproduce that somewhere else," Feaster said. "Now we take that for granted with television, movies, radio -- but they all came after the phonograph."