Enhance realism for virtual characters by including mouth, nose and throat sounds.

[Dampfinchen]

Hello,

first, excellent job on the latest xTTSv2. So, sophisticated text to speech models like this one already include that breathing sound before some sentences.

I think it would be very cool if the model could include other sounds such as sneezing, coughing, crying, throat clearing, sniffling, whistling, smacking and many more as well. This would take the realism to the next level and let virtual characters powered by language and TTS models really come to life. The TTS model could accept for example "* coughing *" between sentences or in the sentence itself as input and yeah, then it would incorporate a cough realistically. You could also have different intensities of these sounds corresponding to the input. That level of realism is missing currently, even from Elevenlabs.

What do you think?