With voiceover translation within Google Meet, Google may have finally come as close as we can to solving real-world translation issues.
We’ve dealt with translating information from one language to another in any number of ways: automatically translating Web pages, rewriting PDFs in another language, or using mobile apps like Google Translate to provide ongoing text translations—either visually or via generated text for another person to read.
At Google I/O, Google’s demonstrating something better. Speech translation with Gemini is a new feature that you’ll be able to trigger within Google Meet. Speak, and you’ll hear your words. But the person you’re speaking to will hear an AI-generated voiceover, with nuance, translating your words into their native language.
“I think we are very, very close to having a natural and free-flowing conversation across languages,” Sundar Pichai, the chief executive of Google and Alphabet, said in a briefing for reporters.
What Google showed off was a brief, canned demo, of course, but it was impressive regardless. For me, the issue will be how the translation software deals with regional languages and slang. I was eager to try out a pair of translation headphones in Taiwan last year, only to discover that they were totally unable to pick up what I assumed to be the regional dialect. Spanish, too, has regional variations just among South and Central America, let alone the differences between those languages and the Catalan spoken in Spain.
There’s a difference, however, between taking a translation device on the street or in a restaurant and sitting down with someone on a video call. In the example below, negotiating an Airbnb with someone in another country feels very much like a problem Google solved for its employees. But speech translation with Gemini will also benefit from the semiformal setting, as well as the quiet indoor environment.
Pichai said that the translation service will be available today for “subscribers,” presumably those who subscribe to Google Workspace. For now, it’s just English to Spanish and vice versa; “many more languages are on the way and coming to enterprises later this year,” Pichai said.