Some models, like Bark, 
can also be conditioned to generate non-verbal communications such as laughing, sighing and crying, or even add music.