by Almost exactly a year ago, Google showed us one of the most mind-blowing tech demos in recent memory: AR glasses that could instantly translate what someone was saying to the user into another language, all in real time. These descendants of Google Glass could give us, as Google claimed, “captions for the world.”
Unfortunately, as is often the case with Google demos, there was a problem: These AR translation glasses were just “an early prototype” that Google had been testing. The Vaporware alarm bells began to ring, and Google has been worryingly quiet on the concept ever since.
It’s a shame, as Google’s AR translation glasses are one of the few hardware ideas in recent years that promise to solve a big problem, rather than a minor inconvenience. As Google said last year, the technology could break down language barriers and help people who are deaf or hard of hearing follow conversations.
But not all is lost. We’re in the run-up to Google IO 2023, which kicks off May 10 and gives Google a chance to tell us how its AR translation glasses are progressing. It is an opportunity that Google must seize, if we do not want to lose faith that the whole concept becomes a reality.
We don’t need Google to announce finished consumer product or shipping dates in IO 2023. But if the keynote goes without a mention of their live translation glasses, we’ll have to assume they’ve been given a first-class ticket to the google graveyard (opens in a new tab). Fortunately, there are good reasons to think that this is not the case.
Lost in translation
Google’s plan to give us the sci-fi dream of live translation technology actually dates back to 2017, when it introduced the original Google Pixel Buds alongside the Pixel 2. Back then, Google said real-time translation it would work “like having a personal translator”. by your side”, but reality did not fulfill that promise.
Even six years later, the process of performing a live translation with the Google Pixel Buds Pro is a tricky experience, as you can see in the demo below. The promise of Google’s AR translation glasses is that they could simplify this, showing you the other person’s translated response as text in front of your eyes, instead of interrupting the flow by throwing it into your ear.
We don’t yet know if the tech stars have aligned to make it happen, but increasingly hot competition (and Google’s own research) seems to be pushing it in the right direction.
Since Google demoed its AR translation glasses, we’ve seen a number of companies demonstrate similar concepts. Arguably the best one yet has been TCL’s RayNeo X2 AR glasses, which we had the pleasure of testing at CES 2023.
TechRadar Managing Editor Matt Bolton called the RayNeo X2 “the most convincing proof of concept for AR headsets I’ve seen yet,” and the key demo is, you guessed it, live translation and subtitling.
While there was a slight two-second delay between a person speaking and their question being translated into text at the bottom of the glasses, Matt Bolton was able to have a full conversation with someone who spoke entirely in Chinese. Ideally, both people should be wearing AR glasses for a full conversation to take place, but it’s a good start.
TCL isn’t the only company producing a working prototype of a concept eyewear similar to the one Google showed off last year. Oppo’s Air Glass 2 managed to outperform rivals like the Nreal Air and Viture One by offering a wireless design, meaning you don’t need a cable to pair the specs with your Oppo smartwatch or phone.
Sadly, Air Glass 2 is unlikely to launch in Western markets, and it’s actually more likely that one of the other tech giants could steal Google’s live translation thunder. Cue Meta’s screeching left turn from the metaverse into his conniving side hustle, announced in February 2022 (opens in a new tab)to make a ‘Universal Voice Translator’.
As the name suggests, this project aims to use machine learning and AI to give everyone “the ability to communicate with anyone in any language,” as stated by Mark Zuckerberg. Crucially for Google’s AR translation glasses, Meta also promised that “with improved efficiency and simpler architecture, direct speech-to-speech could unlock near-human-quality, real-time translation for future devices like glasses.” AR”.
With Meta apparently planning to release those AR glasses, currently dubbed Project Nazare, sometime in 2024, Google is definitely on to start moving with its AR translation specs, and ideally that means announcing some more concrete news at Google IO 2023.
Speech therapy
Will Google really be in a position to announce development for its AR Translation glasses at its big developer conference next week? So far, there have been no rumors or leaks to suggest that it will, but in recent months it has become clear that Google Translate, and languages in general, remain one of its top priorities.
In March, Google’s artificial intelligence research scientists excitedly revealed more information about its Universal Speech Model (USM). (opens in a new tab)which is at the heart of his plan to build an AI language model that supports 1,000 different languages.
This USM, which Google describes as a “next-generation family of voice models,” is already used on YouTube to generate automatically translated subtitles on videos in 16 languages (below), allowing YouTubers to grow their global audience.
Google says this AI has been trained on 12 million hours of speech and 28 billion sentences of text, spanning more than 300 languages. That’s some sizable training data that could hopefully make a significant contribution to AR translation glasses, like the ones Google demonstrated a year ago.
Unfortunately, there has been little evidence that Google’s recent advancements have made their way into Google Translate. While the company announced last year that it had added 24 languages to Translate, bringing it to a total of 133 supported languages, its progress has seemingly stalled, with rivals like DeepL (opens in a new tab) widely considered to be more accurate in certain languages.
Still, Google has made strides elsewhere, revealing in February that Translate will soon be much better at understanding context. For example, he will be able to understand if you are talking about ordering a bass (the fish) for dinner or ordering a bass (the instrument) for your band.
Google added that Translate, for a handful of supported languages, will start using “the correct turns of phrase, local idioms, or appropriate words based on your intent,” allowing translated sentences to match the way a person speaks. native speaker.
All of this again sounds like an ideal foundation for AR translation glasses that work a bit more like the universal Star Trek translator, and less like the clunky, choppy experiences we’ve had in the past.
seeing is believing
Google already has a lot to talk about at Google IO 2023, from the new Google Pixel Fold to the Google Bard, and the small matter of Android 14. But it also feels like a now or never moment for them to grab their AR translation glasses. an important step to become reality.
This year, Google has already come under fire from the ChatGPT boom, and next month it seems likely that we’ll see the long-awaited arrival of the Apple AR/VR headset at WWDC 2023. If you want to be seen as a hardware innovator, instead from a lazy starter, you need to start turning some of your most innovative ideas into real-world objects.
In some ways, Google’s AR translation glasses feel like the ideal contrast to Apple’s mixed-reality headset, which analysts have predicted will throw everything at the wall to see which use case holds up.
Conversely, a simpler wearable that does one thing well, or at least better than recently launched rivals, could be the boost Google needs during a year in which it’s seemingly been attacked from all sides. We look forward to seeing your response at Google IO 2023, and we’ll be keeping an eye out for your hosts’ shows.
