Skip to main content

Meta is building a space-age ‘universal language translator’

When you think of tools infused with artificial intelligence (AI) these days, it’s natural for ChatGPT and Bing Chat to spring to mind. But Facebook owner Meta wants to change that with SeamlessM4T, an AI-powered “universal language translator” that could instantly convert any language in the world into whatever output you want.

Meta describes SeamlessM4T as “the first all-in-one multilingual multimodal AI translation and transcription model.” That’s quite a mouthful, but in simple terms, it means it can convert languages in a range of different ways, such as taking speech audio and switching it into text in a different tongue.

A silhouetted person holds a smartphone displaying the Facebook logo. They are standing in front of a sign showing the Meta logo.
SOPA Images / Getty Images

According to Meta, the tool’s speech recognition and translation features can work in a few different ways:

Recommended Videos

• Speech recognition for nearly 100 languages
• Speech-to-text translation for nearly 100 input and output languages
• Speech-to-speech translation for nearly 100 input languages and 36 output languages (including English)
• Text-to-text translation for nearly 100 languages
• Text-to-speech translation for nearly 100 input languages and 35 output languages (including English)

Please enable Javascript to view this content

Meta says this will “allow people to communicate effortlessly through speech and text across different languages.”

Coming soon to Facebook?

Meta's AI translation tool SeamlessM4T converts Spanish-language input into Vietnamese, both in text and audio forms.
Meta

SeamlessM4T is being released under a research license, and Meta states it’s doing this to “allow researchers and developers to build on this work.” As well as that, the metadata of the dataset that was used to train the translation model, called SeamlessAlign, is also being publicly released. This consists of “270,000 hours of mined speech and text alignments,” Meta claims.

However, Meta did not make it clear where these 270,000 hours of “mined speech” have been sourced. Concerns have been raised over the privacy implications of Meta’s work on AI chatbots, while other AI tools have already been caught stealing protected work. There will no doubt be fears that Meta could have done something similar when it trained SeamlessM4T.

There are already other translation tools like Google Translate that can convert text to text and speech to text, but Meta says its own efforts are superior. SeamlessM4T, Meta argues, “reduces errors and delays, increasing the efficiency and quality of the translation process.”

Meta has not said whether the new tool will be integrated into its apps like Facebook and Instagram, but the company did reveal that it aimed to “explore how this foundational model can enable new communication capabilities” in the future. We’ll have to see what that entails.

Alex Blake
Alex Blake has been working with Digital Trends since 2019, where he spends most of his time writing about Mac computers…
​​OpenAI spills tea on Musk as Meta seeks block on for-profit dreams
A digital image of Elon Musk in front of a stylized background with the Twitter logo repeating.

OpenAI has been on a “Shipmas” product launch spree, launching its highly-awaited Sora video generator and onboarding millions of Apple ecosystem members with the Siri-ChatGPT integration. The company has also expanded its subscription portfolio as it races toward a for-profit status, which is reportedly a hot topic of debate internally.

Not everyone is happy with the AI behemoth abandoning its nonprofit roots, including one of its founding fathers and now rival, Elon Musk. The xAI chief filed a lawsuit against OpenAI earlier this year and has also been consistently taking potshots at the company.

Read more
Google’s new Gemini 2.0 AI model is about to be everywhere
Gemini 2.0 logo

Less than a year after debuting Gemini 1.5, Google's DeepMind division was back Wednesday to reveal the AI's next-generation model, Gemini 2.0. The new model offers native image and audio output, and "will enable us to build new AI agents that bring us closer to our vision of a universal assistant," the company wrote in its announcement blog post.

As of Wednesday, Gemini 2.0 is available at all subscription tiers, including free. As Google's new flagship AI model, you can expect to see it begin powering AI features across the company's ecosystem in the coming months. As with OpenAI's o1 model, the initial release of Gemini 2.0 is not the company's full-fledged version, but rather a smaller, less capable "experimental preview" iteration that will be upgraded in Google Gemini in the coming months.

Read more
The humble bumblebee just messed things up for Meta
A close-up of a bee.

The humble bumblebee has played a part in obstructing an ambitious construction project by Meta, according to a Financial Times (FT) report.

The Mark Zuckerberg-led tech giant has apparently had to abandon a plan to build a nuclear-powered AI data center partly because a rare bee species has been found on the land where the facility would have been built.

Read more