Let’s say I have a hot/open mic without any voice input.
Am I being charged for processing the background noise?
Hey!
You are charged for input and output tokens.
If something triggers a generation from the realtime API, chances are that input tokens were used as well.
A general rule of thumb:
If the AI isn’t saying anything - you’re not charged anything.
So you can just have silence and be fine.
The best thing would be to use server_vad, that way the server “chooses” when you’re talking and when you’re not.
It’s pretty good at figuring out silence.
Cheers! 
2 Likes