Meta ML model offers speech-to-speech translation

"The SEAMLESSM4T model, trained on 4.5 million hours of multilingual spoken audio, provides near-instant speech-to-speech translation across 36 languages, streamlining multilingual communication."

"One of the SEAMLESS team's savviest strategies involved 'mining' the internet for training pairs that align across languages, helping the model recognize matching meaning in content effectively."

Read at Theregister