When Was Autotune Invented? In 1996, Built to Go Unheard

The question went around a lunch table at the 1995 National Association of Music Merchants trade show: “What needs to be invented?” Andy Hildebrand had asked it of a few friends and their wives, and one of the women half-jokingly answered, “Why don’t you make a box that will let me sing in tune?”

Nobody took it up. “I looked around the table and everyone was just kind of looking down at their lunch plates,” Hildebrand told Priceonomics in 2015, “so I thought, ‘Geez, that must be a lousy idea’, and we changed the topic.”

Hildebrand, a flutist with a doctorate in electrical engineering, had spent more than a decade in the oil industry. He spent the next six months on other projects, until the suggestion came back to him. “It just kind of clicked in my head,” he said.

When was autotune invented?

Auto-Tune was invented in early 1996, when engineer Andy Hildebrand wrote the program over a few months, according to Priceonomics, and it went on sale in 1997. His company, Antares, released it as a plug-in for Pro Tools studio systems, built to correct off-key singing without anyone hearing the fix. Its robotic side became famous in 1998, on Cher’s “Believe.”

The Recording Academy dates the release to the spring of 1997, and Sound On Sound reviewed it in its August 1997 issue, when it was “only available for Pro Tools TDM systems,” as reviewer Paul White recalled in 1998.

Date	What happened
1995	At a NAMM show lunch, a guest asks for a box to help her sing in tune
Early 1996	Hildebrand writes Auto-Tune on a specially equipped Macintosh
Spring 1997	Antares releases it as a Pro Tools plug-in
September 19, 1997	A stand-alone version ships
October 27, 1997	The provisional patent application is filed
October 1998	Cher’s “Believe” is released
October 26, 1999	US patent 5,973,252 is granted
October 2018	The patent expires

September 19, 1997 is the date of a later version

Wikipedia gives Auto-Tune’s release date as September 19, 1997, citing, among other sources, a news page from Antares’s website archived in 2000. That page’s entry for 9/19/97 reads: “Auto-Tune is now stand-alone. It is shipped with AudioStream, the Antares stand-alone host program and can run on any Macintosh.”

Sound On Sound had already reviewed the Pro Tools plug-in in its August issue.

Hildebrand learned the math mapping oil fields

After his PhD at the University of Illinois in 1976, Hildebrand joined Exxon, where, Priceonomics says, he was “tasked with using seismic data to pinpoint drill locations.” The job, as he described it, was to send sound into the ground, “listen to reverberations that come up,” and “try to figure out what the shape of the subsurface is.”

In 1982 he co-founded Landmark Graphics, which built 3D mapping workstations for oil exploration, a field whose survey data also outlined the crater under Mexico’s cenotes. He retired in 1989 to study composition at Rice University. As an oil engineer he had used autocorrelation, which measures how closely a signal matches a delayed copy of itself.

The oil-drilling origin story is about the inventor, not the software

Auto-Tune is often described as oil-exploration technology put to work on music. Its patent never mentions oil or geophysics, and it states its aim plainly: “The purpose of the invention is to correct intonation errors of vocals or other soloists in real time in studio and performance conditions.”

What crossed over was the inventor’s math. Antares’s company history says he realized “the same techniques he’d used to map oil fields could track and correct the pitch of a human voice.”

Autocorrelation finds the repeat in a sung note

A sung note is a sound wave that repeats, and its pitch is set by how often it repeats. Hildebrand’s patent, US 5,973,252, puts it in one line: “Determining the pitch of a sound is equivalent to determining the period of repetition of the waveform.”

Slide a copy of a repeating wave along the original, and when the shift equals exactly one period, the two line up again. Hildebrand says the method is “never fooled by the changing waveform.”

The obstacle was arithmetic. “In practice, the auto-correlation function has not been used to search for periodicity or track existing periods, because it requires a high level of computation to generate,” the patent notes. Hildebrand says he realized that “most of the arithmetic was redundant,” and that his simplification “changed a million multiply adds into just four.”

With the pitch measured, the software resamples the audio, adding or dropping whole wave cycles, to move the voice onto the closest note in a chosen scale.

Hildebrand tells the zero-setting story two ways

A speed control sets how fast the correction happens. As Hildebrand put it, “I built in a dial where you could adjust the speed from 1 (fastest) to 10 (slowest). Just for kicks, I put a ‘zero’ setting, which changed the pitch the exact moment it received the signal.”

Talking to the writer David Friedman in 2014, he credited the setting to his director of marketing, who asked, “What’s the harm in letting it go to instantaneous?”

“I disagreed. I thought nobody in their right mind would ever use it that way,” Hildebrand told Friedman.

The patent leans toward the second telling. It allows a setting of zero “giving instantaneous pitch changes,” and calls instant changes “objectionable when the human voice is being processed.” At the fastest settings, the gradual slides a singer makes between notes disappear, and each note is, in Simon Reynolds’s phrase in Pitchfork, “pegged to an exact pitch.”

The vocoder credit on “Believe” was a cover story

Cher’s “Believe,” released in October 1998, spent seven weeks at number one in Britain. Its producers, Mark Taylor and Brian Rawling, used Auto-Tune’s zero setting to make Cher’s voice sound robotic.

When Sound On Sound covered the recording in February 1999, Taylor said he tried a 1970s Korg VC10 vocoder, found “the results just weren’t clear enough,” and got the effect from a DigiTech Talker pedal instead. The magazine called the effect “basically down to vocoding and filtering.”

A note the magazine added later says the producers were “apparently so keen to maintain their ‘trade secret’ process that they were willing to attribute the effect to the (then) recently-released Digitech Talker vocoder pedal.” Reynolds calls it a “cover story”: the vocoder story came from the people who made the record.

A musical vocoder needs a second sound, such as a synthesizer, as its carrier. Auto-Tune uses the singer’s own voice and moves its pitch to the nearest note. The same Sound On Sound note says the sound became known as the “Cher effect,” and Priceonomics reports that Antares marketed Auto-Tune under that name.

Records used it discreetly for about a year first

Sound On Sound’s note calls “Believe” “the first commercial recording to feature the audible side-effects of Antares Auto-Tune software used as a deliberate creative effect.” Auto-Tune had been on the market for about a year by then, and “its previous appearances had been discreet, as its makers, Antares Audio Technologies, intended,” Reynolds wrote.

“Studios weren’t going out and advertising, ‘Hey we got Auto-Tune!’” Hildebrand said. Wikipedia names Aphex Twin’s “Funny Little Man,” from the 1997 Come to Daddy EP, as one of the earliest tracks to use it. As with the first video game, the answer depends on what counts.

The robot voice came back in the mid-2000s

Wikipedia credits T-Pain with reintroducing Auto-Tune as a vocal effect in pop on his 2005 album Rappa Ternt Sanga. By June 2009 it was common enough for Jay-Z to release “D.O.A. (Death of Auto-Tune),” whose lyrics address its overuse. One inspiration, he said, was hearing the effect in a Wendy’s commercial.

The box a lunch guest asked for in 1995 was meant to correct pitch “without artifacts in a seamless and continuous fashion,” in the patent’s words. The sound that made it famous is the artifact, from a zero setting its inventor, in one telling, said nobody in their right mind would use. Strictly FYI.