Home / Articles / Technology

Codecs and HD voice explained: why some VoIP calls sound so much better

G.711, G.722, G.729, Opus. The codec your system negotiates decides whether a call sounds like a 1990s landline or like the person is in the room. Here's how they compare.

The short version

  • A codec turns your voice into data and back. It decides call quality and how much bandwidth each call uses.
  • "Narrowband" codecs (G.711, G.729) carry the same audio range as an old landline. "Wideband" or HD voice (G.722, Opus) carries roughly twice the range and sounds far clearer.
  • HD voice only works end to end if every leg of the call supports it. A call out to the regular phone network usually drops back to narrowband.
  • Opus is the most adaptable modern codec and copes best with patchy networks.

What a codec does

When you speak into a VoIP phone, the audio is sampled, compressed by a codec, split into small packets (typically every 20 milliseconds) and sent across the network. At the other end the codec decompresses the packets back into sound. The two ends agree on a codec at the start of each call, a process handled by SIP and SDP. If they share nothing better, they fall back to the lowest common denominator, usually G.711.

The codecs you'll actually meet

CodecAudio bandCodec bitrateApprox. bandwidth per call*Where you'll see it
G.711 (µ-law / A-law)Narrowband, 300–3,400 Hz64 kbps~87 kbps each wayThe universal default; SIP trunks to the phone network
G.729Narrowband8 kbps~31 kbps each wayLow-bandwidth links; slightly lower quality
G.722Wideband, 50–7,000 Hz64 kbps~87 kbps each way"HD voice" on most desk phones
OpusNarrow to full band6–510 kbps (voice typically 16–40)~40–70 kbps each wayWebRTC, softphones, Teams-style apps, modern PBXs
AMR-WB / EVSWideband / super-widebandVariableN/A (mobile network)VoLTE and Wi-Fi calling on mobiles

*Including IP, UDP and RTP headers plus Ethernet framing at 20 ms packets. Exact figures vary with packet size, VPNs and link type. Try our bandwidth calculator.

Why HD voice sounds so different

Traditional telephone audio is cut off at about 3.4 kHz. That's enough to understand speech, but it strips out the higher frequencies that distinguish "s" from "f" and give a voice its character. That's why people spell their names on the phone ("S for Sydney"). Wideband codecs like G.722 extend that to around 7 kHz, and the difference is immediately audible: less fatigue on long calls, fewer misheard numbers, and better results from speech recognition and AI transcription.

Why your HD calls sometimes aren't HD

HD voice needs every device and network in the path to support a wideband codec. Calls between two extensions on the same system, or between two apps on the same platform, are usually HD. A call that leaves your system via a SIP trunk to a landline or a different mobile network will normally be converted to G.711 narrowband somewhere along the way.

Opus: the adaptable one

Opus is an open, royalty-free codec standardised by the IETF and built into every modern web browser for WebRTC. It changes bitrate on the fly, covers everything from narrowband speech to full-band music, and includes forward error correction that can rebuild lost packets. On a congested home Wi-Fi network or a mobile hotspot, Opus will typically sound noticeably better than G.711 or G.722, which have no built-in loss recovery.

Practical recommendations

  • Desk phones on a decent network: prefer G.722, then G.711. Put G.729 last unless bandwidth is genuinely tight.
  • Softphones and remote workers: prefer Opus where the platform supports it.
  • SIP trunks: G.711 A-law is the norm in Australia. Ask your provider whether they support G.722 or Opus on the trunk, and whether HD carries through to their network.
  • Avoid transcoding (converting between codecs mid-call) where you can. Every conversion adds delay and loses a little quality.
  • Bad call quality is usually caused by the network, not the codec. If calls are choppy, read our call quality guide before changing codecs.