I don’t think Convai outputs emoji’s, but maybe you could in code examine the transcript and if the textual description of an emoji is found you could replace it with an actual emoji.
Maybe try putting into her speaking style section on the Convai website that she speaks in a brief and concise manner, add something to example dialogue section too.
I don’t think there is any out of the box way. You could try a different voice as I think some of them are faster than others. Or you could get the audio output in code before it is played and slow it down.
Sorry if these are not much help!
EDIT : I’ve just realised, this is not a Unity or Unreal or Web SDK project is it? If it’s the Avatar Studio then I don’t think some of the above applies.