Google Translate’s camera can now automatically detect languages

Starting today, Google Translate’s camera can automatically detect languages so you can point your camera at a flyer or sign and get results in your native tongue even if you don’t know what language you’re reading. TEST.

Google Translate’s computer vision translation capabilities also got a heavy shot in the arm today with the addition of 60 new languages for instant translations. This brings the total number of languages you can translate by simply pointing your camera at words to 88.

The new editions will benefit travelers visiting countries in Europe, Africa, and Asia, with additions like Hindi, Latvian, Persian, and Shona.

Google Translate’s static image translation — which requires users to upload an image, then use their finger to select the text they want translated — got 13 new languages in fall 2018 like Arabic, Bengali, and Vietnamese.

Instant translations with Google Translate and smartphone cameras got its start in 2015. NMT was made available for Google Translate results on the web in 2016.

News of a new and improved Google Translate follows the addition of on-the-spot reading and translation of more than 100 languages in May for Google Assistant’s computer vision service Lens.

Approximately 25 of the newly added languages, including Arabic, Swahili, and Urdu, are available for offline translations and require no network connection, a Google AI spokesperson told VentureBeat in an email. Roughly one year ago, Google Translate got neural net-driven offline translations in near 60 languages.

Prowess in language translation has become an important demonstration of the capabilities of some of the largest AI companies in the world, as part of language understanding operations that power things like AI assistants, local commerce, or social media interactions.

For its part, Google introduced Instant Translations with Google Home speakers in January, a way to translate conversations on the fly.

In late 2018, Facebook introduced ways to use unsupervised learning to translate languages that don’t have large data sets available for training. Baidu also showcased its simultaneous translation abilities, and Microsoft’s LinkedIn brought translation of 60 languages to its enterprise networking services.