Knight Ridder/Tribune Business News delivered by Newstex

In Any Tongue

by Martha McKay 8.10.06

Ever dream of speaking French like a Parisian? Or Swahili like a Kenyan?

A Manhattan-based company called Voxonic can make your linguistic longings come true.

Computers these days are relatively adept at translating foreign language text, and can even make your spoken voice appear as text on a screen without much trouble.

Voxonic Inc., however, has put computer technology to work on a new translation task.

Using an intriguing blend of its own software, human voice actors and some audio smarts, the company can replicate any person's voice in any language.

Want to hear yourself recite the Gettysburg Address in, say, Navajo?

No problem, says Arie Deutsch, the company's president of entertainment and son of Fred Deutsch, a New York entrepreneur and real estate investor who founded Voxonic in 1998. Fred Deutsch came up with the idea for the company after watching Harrison Ford in a film dubbed in French. Why, he reasoned, couldn't a computer be used to get Harrison Ford's voice to speak French?

Voxonic, which launched commercially last year, has sights on a range of markets, including politicians who want to produce a speech in another language (Arie Deutsch says at least one U.S. governor traveling in Asia used Voxonic; he won't say who it was, but did confirm that it was not New Jersey's Jon Corzine).

Corporate executives' speeches are another potential market. Berlitz Cross Cultural hired Voxonic to create an electronic press kit that was introduced last month.

And of course, there is the entertainment and music industry.

After a chance meeting last year with Arie Deutsch, Uptown Records founder Andre Harrell, who is credited with discovering Grammy Award-winning recording artist Mary J. Blige, signed on to became a partner with Voxonic.

Harrell and the 26-year-old Arie Deutsch see a potential market in everything from rap star ring tones translated into Chinese or Japanese, to full-length songs translated for foreign markets.

On a recent visit to the company's midtown recording studio, I saw how it all works.

Arie Deutsch instructed me to speak for about 10 minutes, in English, into a mike.

I read a bit from a novel, then talked about having a baby -- the audio engineer's wife went into labor that morning, so it seemed timely. The idea was to capture the pitch and inflection of my voice -- the specific mouth moves that make me sound like me.

A day later, I sent him a few lines of text I wrote up in English, which he outsourced to a professional translation company. I could, amazingly, choose from more than 1,600 languages.

I decided on French, partly because I'd taken about seven years of the language but never mastered it. The thought of finally hearing me parle francais was appealing. And besides, if I chose a language I wasn't somewhat familiar with, I might not be able to judge the quality of the Voxonic end-product.

After Deutsch received the French text, he hired a female voice actor to speak the lines in French. Voxonic then applied its patent-pending software, which uses filters every millisecond to map the sounds of my voice with the actor's recording, basically melding the two into a final version that sounds like me.

I received a CD with an audio file several days later. Arie Deutsch had included several snippets of me speaking English from the recording session, and then came the French.

At the studio, Arie Deutsch aired several demos, including one of a Bill Clinton speech that Voxonic translated into Spanish. The sound was uncannily like Clinton, but just a tad off somehow. Not exactly quite like him, but very close.

And that's how I sounded. It was me speaking like a native Frenchwoman, and it sounded a lot like me, but there were moments when it was not precisely my voice.

Several colleagues I shared this with who understood the Voxonic process agreed that it didn't seem to sound like me, but it was close.

One colleague, who speaks several languages, pointed out that when she speaks French, she doesn't sound exactly like herself. "My voice takes on tones, pitches, intonations, characteristics that I never use in English," she said.

Interestingly, some co-workers who had no idea what I was up to listened to the recording and were impressed I could speak French so well.

In some ways, Voxonic's service reminded me of the first time I realized the power of digital photo manipulation. Is the picture real? Or did a computer modify the image? In the case of Voxonic, is the voice real, or did software re-create the voice?

The questions raise some serious issues about how this particular technology might be used or misused -- Voxonic has a "template" of my voice, and can make recordings of "me" saying just about anything.

Arie Deutsch is a stickler about confidentiality, which is good, and the company has said it aims to keep control over the technology.

As for the commercial viability of Voxonic's service, it's hard to say. The company scored a coup by signing up Harrell, who is well-known in the music industry, and it may be that non-English-speaking fans will clamor to hear their favorite stars singing or rapping in their native language.

At about $1,000 per minute, Voxonic does not come cheap (Arie Deutsch says entire songs fully translated could run as much as $10,000). The company does not appear poised to revolutionize the movie-dubbing industry yet.

But Voxonic, which is backed by Fred Deutsch's fortune, is not desperately seeking venture capital, which can be an advantage. The company can perfect the technology and explore various markets, and see where it takes them.