EMZETT.VoIP
In short: Voice over IP — telephony over internet protocols instead of the classic circuit-switched telephone network.
In more detail: Speech is digitised, packed into data packets and transmitted like any other network traffic — mostly over UDP, because in voice transmission an occasionally lost packet (a brief crackle) is less disruptive than the delay caused by a repeated request with TCP. Enables cost savings and additional functions (video calls, chat integration) compared with classic telephony.
In Depth
VoIP usually uses a two-protocol model: a signalling protocol (mostly SIP — Session Initiation Protocol) sets up the call, clarifies the codec capabilities of both sides and ends it again; the actual audio data then flows separately via RTP (Real-time Transport Protocol), mostly over UDP. This separation allows the signalling channel to be run reliably (possibly also over TCP), while the time-critical audio stream itself is optimised for minimal latency instead of a delivery guarantee.
A central quality feature in VoIP is jitter — the variation in latency between consecutive packets. Even if the average latency is low, strong fluctuations lead to audible dropouts or distorted speech, because the recipient has to play packets at even intervals but has to buffer (or in extreme cases discard) packets that arrive unevenly. Modern VoIP systems therefore use adaptive jitter buffers that continuously adapt to the actual network conditions.