How it works

A voice packet is mostly overhead. aitcore removes as much of it as possible while keeping the call clear.

Anatomy of one 20 ms G.711 packet

50 of these leave every second, in each direction, for every call. That is 87.2 kbps per call.

After aitcore: one 40 ms G.729 packet

25 packets per second, 50 bytes each. That is 10 kbps per call.

Codec transcoding

Calls arriving as G.711 are transcoded to G.729 or low-rate Opus at the edge of the WAN and restored on the far side, so phones and PBXs keep their existing settings.

Packetization tuning

Grouping 40 or 60 ms of audio into each packet cuts header overhead by half or more. aitcore adjusts this per link to keep added delay inside your latency budget.

RTP header compression

On point-to-point WAN links, cRTP and ROHC replace the 40-byte IP/UDP/RTP header with a 2 to 4 byte context reference.

Silence suppression

Voice activity detection stops sending packets during pauses and fills in comfort noise. On typical conversations this saves a further 30 to 40 percent.

Call multiplexing

Between two aitcore nodes, many calls share one tunnel and one outer header, which keeps savings high on Ethernet links where cRTP is not available.

QoS and monitoring

Voice is marked EF and prioritised ahead of data. Each call reports MOS, jitter and loss so you can see quality as well as savings.

What to expect

Savings depend on what your calls look like today. Networks running G.711 over leased WAN links typically see 80 to 88 percent less voice bandwidth. Networks already on G.729 save less, mainly from packetization and header compression. The savings calculator shows the figure for your own setup.