it will be so cool when models like whisper , also attaches meta data to each word, like tone , pitch, start and end time and recognizes different voices. So that we can feed it back into simple text to voice generator and generate new audio to dub videos. So many anime's , Korean fantasy and sci fi drama, that I would love to listen to instead of reading subtitles. It would also help with creating a star trek like communicate that lets anyone talk to others in in the same tone they intended.