VoIP Codec Comparison

1. Overview at a glance

The choice between them is rarely about “best codec” in isolation. It is about constraints — bandwidth, processing budget, interop requirements, licensing, and quality target. Each codec has scenarios where it is the right answer.

2. G.711 — the baseline

G.711 is the codec of the legacy telephone network. It samples audio at 8 kHz, encodes each sample with 8 bits using a logarithmic compression curve (PCMU in North America, PCMA elsewhere), and produces 64 kbps. No further compression.

Pros

• Universal support — every SIP endpoint, every PSTN gateway, every legacy phone
• Negligible CPU cost — just a lookup table for compression and decompression
• Royalty-free, no licensing concerns
• Identical to PSTN audio — no transcoding or quality loss when bridging

Cons

• 64 kbps payload + RTP/UDP/IP overhead totals ~80-100 kbps per direction. Two directions = 160-200 kbps per call.
• 8 kHz sample rate limits frequency response to 4 kHz, the same as PSTN. Voices sound “telephoney.”
• Sensitive to packet loss — no FEC, no PLC built in. Lost packets produce audible gaps.

When to use it

Default for any deployment where bandwidth is not a concern and interop is the priority. PSTN bridges almost always use G.711 to avoid transcoding. SIP trunks usually default to it.

3. G.722 — wideband over the same bandwidth

G.722 doubles the audio bandwidth (sample rate 16 kHz, frequency response up to 7 kHz) while staying at 64 kbps total. It uses ADPCM (adaptive differential pulse code modulation) to compress wideband audio into the same payload size as G.711.

Pros

• Substantially better speech intelligibility than G.711 — voices sound natural rather than telephoney
• Same bandwidth as G.711 (64 kbps + overhead)
• Royalty-free
• Supported by most modern IP phones (Yealink, Polycom, Cisco, Snom, Grandstream all support it)

Cons

• Cannot bridge to PSTN without transcoding — G.722 audio must be downsampled to G.711 for PSTN
• Slightly higher CPU cost than G.711, but still light
• Older phones and some softphones do not support it

When to use it

Internal calls between IP endpoints where both sides support it and PSTN bridging is not involved. Modern PBXes often default to G.722 internally and switch to G.711 at PSTN boundaries.

4. G.729 — low bandwidth at the cost of quality

G.729 compresses 8 kHz audio to 8 kbps using CS-ACELP (conjugate structure algebraic code-excited linear prediction). The compression is lossy and significantly reduces audio quality compared to G.711, but cuts payload bandwidth to one-eighth.

Pros

• Low payload bandwidth — 8 kbps + overhead is ~32 kbps including all protocol overhead. Eight times more calls per Mbps than G.711.
• Widely supported in legacy environments, especially older Cisco and Avaya systems
• Annex B (silence suppression) further reduces bandwidth on quiet sections

Cons

• Quality is noticeably worse than G.711, especially with packet loss or low signal-to-noise ratios
• Patent licensing required — many SIP stacks require a separately licensed G.729 codec module
• Bad for in-band DTMF (compression distorts DTMF tones)
• Higher CPU cost than G.711, especially on older hardware that lacks acceleration
• Annex B variants (annexb=yes/no) are a common interop pitfall

When to use it

Bandwidth-constrained deployments — remote sites with limited WAN, high-density trunks where each saved kbps matters. Legacy interop with systems that prefer or require G.729. Increasingly being replaced by Opus for bandwidth efficiency in modern deployments.

5. Opus — the modern default

Opus is a modern, royalty-free codec specified in RFC 6716. It uses a hybrid of SILK (for speech) and CELT (for music) and adapts dynamically to bitrate, sample rate, and content type.

Pros

• Best-in-class audio quality at any given bitrate, especially in adverse network conditions
• Adaptive bitrate from 6 kbps to 510 kbps and sample rates from 8 to 48 kHz, all in one codec
• Built-in forward error correction (FEC) and packet loss concealment (PLC)
• Royalty-free, open standard
• Mandatory in WebRTC, so widely deployed in browser-based and modern softphone implementations

Cons

• Not supported by older SIP phones — though this gap is closing rapidly
• Cannot bridge to PSTN without transcoding
• Higher CPU cost than simpler codecs, particularly for the wideband modes
• fmtp parameter complexity — many configuration options can lead to interop issues if not aligned

When to use it

WebRTC, modern UCaaS platforms, browser-based softphones, and any IP-only deployment where both sides support it. Opus's adaptive nature makes it especially good for unpredictable network conditions like mobile or remote work scenarios.

6. Choosing between them

The decision tree:

If the call touches PSTN at any point

Use G.711. Avoiding transcoding preserves quality and reduces processing cost. PSTN endpoints expect G.711.

If both sides are IP-only and support G.722

Use G.722. Same bandwidth as G.711, much better audio. Requires PBX/SBC awareness so PSTN-bound calls fall back to G.711.

If bandwidth is severely constrained

G.729 if both sides require it for legacy reasons; Opus at low bitrate (8-12 kbps) if both sides support it. Opus produces noticeably better quality than G.729 at the same bitrate.

If the call is browser-based or WebRTC

Opus. Mandatory in WebRTC. Universal browser support.

If quality matters more than bandwidth

Opus. With good bandwidth, Opus at 32-48 kbps produces audio quality indistinguishable from a same-room conversation.

Multi-codec offers

Most production deployments offer multiple codecs in priority order (e.g., Opus, G.722, PCMU, G.729). The negotiation picks the best mutually supported option. This handles interop with diverse endpoints automatically.

Frequently asked questions

What is the best VoIP codec?

There is no single best codec — the right choice depends on the deployment. Opus is the highest-quality option for any IP-to-IP scenario where both sides support it, especially over unreliable networks. G.711 is universal and required for PSTN bridging without transcoding. G.722 doubles audio quality at the same bandwidth as G.711 for IP-only calls. G.729 is the legacy low-bandwidth option, increasingly being replaced by Opus.

Why does G.711 use so much bandwidth?

G.711 uses 64 kbps because it does not compress audio — it samples at 8 kHz and uses 8 bits per sample with logarithmic encoding (mu-law for PCMU or A-law for PCMA). The lack of compression means negligible CPU cost and zero quality loss, but every call uses ~80-100 kbps per direction including RTP/UDP/IP overhead. Codecs like G.729 (8 kbps) and Opus (adaptive 6-510 kbps) compress to use less bandwidth at the cost of CPU and some quality.

Should I use Opus or G.729 for low-bandwidth calls?

Opus, if both sides support it. At equivalent low bitrates (8-12 kbps), Opus produces noticeably better audio quality than G.729. Opus also handles packet loss better through built-in forward error correction. G.729 remains relevant only for legacy interop where Opus is not available. For new deployments needing low bandwidth, Opus is the better choice.