Recordings of spoken words and phrases can be analysed to try and make eSpeak match a language more closely. Unlike most other (larger and better quality) synthesizers, eSpeak's data is not produced directly from recorded sounds.