Learning center
Voice infrastructure, explained
Resources on voice infrastructure and how phone calls work, from SIP to AI voice agents.
AI voice agents
What changes when the thing on your end of the call is a model, not a person. Latency budgets get tight and the failure modes are new.
What is an AI voice agent?
Software that holds a spoken phone conversation by chaining speech recognition, a language model and speech synthesis in a real-time loop.
How do AI agents make phone calls?
The full path from model output to a ringing phone: where SIP, numbers and carriers fit under platforms like Vapi, Retell and LiveKit.
What is bring your own carrier (BYOC)?
Pointing an AI voice platform at your own carrier trunks instead of the numbers it resells, so you keep control of rates, routes and caller ID.
What is turn latency?
The gap between a caller finishing a sentence and the agent starting its reply. Past roughly 800ms, people assume the line dropped.
SIP and VoIP
The protocols underneath every internet call. SIP sets the call up and RTP carries the audio. Knowing which does what explains most problems.
What is SIP?
Session Initiation Protocol, the text-based signaling protocol that negotiates calls. It reads a lot like HTTP, with requests and response codes.
What is a SIP trunk?
Connecting a phone system to the public network over the internet instead of physical lines, with capacity measured in concurrent calls.
What is the PSTN?
The public switched telephone network: the global system every phone number belongs to, and the thing your VoIP calls eventually land on.
What is RTP?
Real-time Transport Protocol, the stream that carries the actual voice packets. SIP sets the call up, RTP is the call.