Meta Launches AI Model for Voice and Text Translation in 100 Languages

'SimulSpeak M4T' That Translates Spoken and Written Language
Meta Plans to Open Source This AI Model Too

Meta Platforms, the parent company of Facebook, has launched an AI model capable of recording and translating in up to 100 languages.

On the 22nd (local time), Meta announced the release of its language translation AI model, 'SeamlessM4T.' This model, which stands for Massively Multilingual and Multimodal Machine Translation, translates in approximately 100 languages.

[Image source=Reuters Yonhap News]

When voice or text is input, it can be translated into either voice or text. For text-to-text translation, 100 languages are supported, while for voice-to-voice or voice-to-text translation, input in 100 languages can be translated into 35 languages.

The supported translation languages have not yet been disclosed. It appears that English, Spanish, German, and others are included.

Mark Zuckerberg, CEO of Meta, said he envisions this AI model facilitating communication among users worldwide in the three-dimensional virtual world metaverse, into which Meta is heavily investing.

Meta plans to make this AI model available to the public for non-commercial use, similar to its proprietary large language model (LLM) LLaMA, which underpins generative AI. CEO Zuckerberg stated, "Providing the AI model as open source rather than charging fees for its use will yield greater benefits," adding, "An open AI ecosystem is better for Meta."

This content was produced with the assistance of AI translation services.