OAI Waifua project by kyon |
|||
around mid september oai released speech to speech for their mobile chatgpt app. this was both great and terrible news. i wanted the apis dammit, i wanted to build a waifu that could converse with you in real-time.
oai-waifu
this was just an experiment to front run realtime voice implementation along with vrm in browser expression + gesture handling by extracting streaming audio features (prosodic, mfcc etc)
i’m gonna reverse engineer chatgpt app to find new voice api calls, if you have experience dm me
@roon bro you could save me so much time
shown while vrm + anims load in the bg
still lots of minor issues to fix:
