The Massively Multilingual Speech (MMS) is a project from Meta building a single multilingual speech recognition model.