some model need lot of time to generate. with this feature, user can hear the speech without wait the generation fully completed. but it may perform worse if generation speed is low. so please add a option for this rather than force on
some model need lot of time to generate. with this feature, user can hear the speech without wait the generation fully completed.
but it may perform worse if generation speed is low. so please add a option for this rather than force on