Suppose you call a colleague from an app on your laptop. The microphone picks up your voice, and the app converts the sound into a stream of numbers. It sends that stream as a series of small packets, each addressed to the other end. Your colleague's device puts the packets back in order and plays them as sound. The same thing happens in the other direction at the same time.
Two published standards show how the work is divided. The Real-time Transport Protocol, or RTP, carries the audio. It gives each packet a sequence number and a timestamp so the receiver can rebuild the order. Its specification also says what RTP leaves to the network underneath. RTP doesn't guarantee that a packet arrives, or that it arrives on time. So the network a call crosses affects how it sounds, along with the devices at each end and the way the audio is encoded.
SIP handles the call around the audio. It locates the other party and negotiates the session, then closes the session when someone hangs up. The setup messages and the audio can travel by different routes.
We follow a call the whole way in how VoIP works, and from the handset's side in how a VoIP phone works.