The rise of Artificial Intelligence is taking us to points never imagined. Meta launched this week its new Artificial Intelligence, that allows you to translate 100 languages by voice and writing.
That’s how it is SeamlessM4T.
As requested by the user, the AI can translate speech to text, speech to speech, text to speech and text to text for a hundred languages.
“We are publicly releasing SeamlessM4T under a research license to allow interested parties to develop this work,” said Mark Zuckerberg’s company. it’s a statement.
“We are also launching SeamlessAlign metadata, the largest open multimodal translation dataset to date, with a total of 270,000 hours of voice and text alignments extracted”, added.
SeamlessM4T builds on Meta’s advances in the search for a universal translator. It is the evolution of NLLB (No Language Left Behind), a text-to-text machine translation model, from Universal Speech Translator and Massively Muntilingual Speech.
The largest data to translate to date, says Meta about its Artificial Intelligence
Multilingual and multimodal translation are possible thanks to Artificial Intelligence SeamlessM4T. For this they used, according to Meta, tens of billions of sentences in text format and 4 million voice clips.
The researchers created a dataset called SeamlessAlign, which linked 443,000 hours of speech-to-text and 29,000 hours of speech-to-speech recordings. “It is the largest multimodal translation dataset to date,” Goal points.
According to the company, SeamlessM4T’s unique system reduces errors and delays, increasing the efficiency and quality of the translation process. “This allows people who speak different languages communicate with each other more effectively.” concludes Mark Zuckerberg’s company.
You can try the demo of SeamlessM4T in this link.