huggingface / speech-to-speech

Speech To Speech: an effort for an open-sourced and modular GPT4-o
Apache License 2.0
3.47k stars 361 forks source link

I cannot understand the purpose of should_listen Event #94

Closed xjlv1113 closed 3 weeks ago

xjlv1113 commented 1 month ago

can you tell me

rs545837 commented 1 month ago

Hi xjlv1113,

Main purpose is to prevent listening itself, when it's giving the audio output.

How is it being used in different scenarios:

andimarafioti commented 1 month ago

Exactly! Without the 'should_listen', the model would listen to itself and begin talking to itself (unless you're wearing headphones). I guess there's a way to improve that but both I and @eustlb looked into it individually and weren't happy with the solutions we found.