lspahija / AIUI

AIUI is a platform enabling seamless two-way verbal communication with AI.
MIT License
325 stars 75 forks source link

The parameter responsible for the length of waiting for the start of the response #20

Open chicager opened 1 year ago

chicager commented 1 year ago

Autorun without a start word is a very cool feature! It's like talking to a real person! But sometimes when you start talking and stop for a second to think to continue to formulate a sentence, the bot finishes listening and starts responding, even though the sentence is not finished yet. Is it possible to add a parameter with which you can adjust the delay of waiting for listening before responding? This is especially true when you speak different languages, because it is not always possible to formulate a sentence right away.

arashmh commented 1 year ago

@chicager i second this suggestion. I think it could be even cooler if we could make a lightweight system for detecting when the sentence has ended actually. pos tagging or something like that. Now combine that with a relatively short waiting time and i think it will feel very human like

Braunfeltd commented 1 year ago

it would require some sort of Voice Activity Detection (VAD). I have been playing around with a lot of things, this is on my list to look into when I have some time too. you could always try to work with GPT to come up with a solution. thats how I get around a lot of things but can take a lot of time of back and forth if you are not a programmer before you get a working result that works properly.