When i self host avatars/tts/llm/stt - should i host these in us-east or close to the user?
m
many-fall-81099
12/18/2025, 6:50 PM
Close to user helped. Self hosted my agent to us regions and theres a noticeable latency difference in agent initialization times for users far away (India). This is for selfhosted livekit agent tho
🙏 1
r
refined-appointment-81829
12/18/2025, 7:49 PM
If you have to pick one or the other the most important thing is to have your agent as close to LLM/TTS/STT/Avatar as possible. Generally speaking there is a lot more round trips with those service when compared to the interactions with the user.
🙏 1
s
salmon-book-23491
12/18/2025, 7:56 PM
What bothers me most is the connection from tts to the avatar via livekit - how does the livekit routing work here? Eg if the user is in San Francisco, stt/llm/tts are in san francisco and I have my avatar hosted in us-west - Is the routing via us-east or does it stay in west?
r
refined-appointment-81829
12/18/2025, 8:00 PM
data has to get to where ever data has to get to. So if you have things that need to talk to each other in different places the data has to go there.
refined-appointment-81829
12/18/2025, 8:04 PM
But the routing is via webrtc so that is going to be pretty low latency in general