"multilingual-speech"

Called ‘SeamlessM4T,’ the single model can perform speech-to-text, speech-to-speech, text-to-speech, and text-to-text translations.