Support for text and voice conversion in 1100 languages. Meta launched the open source AI project MMS.

CTOnews.com, May 23, Meta launched the Massively Multilingual Speech (MMS) project, which provides advanced language-to-text and text-to-voice services for more than 1100 languages, and supports nearly 4000 unmarked spoken languages.

Meta integrates self-supervised learning work wav2vec 2.0 and a new data set for the MMS project. It is found that MMS is better than the existing model and supports 10 times the number of languages.

The key to Meta's creation of this achievement is the introduction of a large audio data set containing more than 1100 readings of the New Testament (New Testament). Meta said that the voice model resulting from the integration of the audio data set performed equally well in male and female voices.

CTOnews.com is enclosing a detailed original report of Meta, which can be read by interested users.