Audio

Audio for live streaming: codecs, loudness and A/V sync

Viewers forgive a soft picture far sooner than missing, jumping or out-of-sync sound. A practical look at the audio side of live delivery.

Key takeaways

  • Broadcast sources often carry MP2 or AC-3; most streaming players and devices are happiest with AAC.
  • Loudness should be consistent across channels — EBU R128 and ATSC A/85 are the common references.
  • Audio/video sync depends entirely on timestamps; broken audio timestamps cause drift and freezes.
  • Silent output is a fault. Detect audio loss automatically, just like video loss.

Audio codecs you will meet

Live sources reflect where they come from. Satellite and terrestrial contribution feeds often carry MPEG-1 Layer II (MP2) or Dolby Digital (AC-3); newer HD services increasingly use AAC or Dolby Digital Plus (E-AC-3). On the delivery side, the picture is simpler: AAC is supported by virtually every browser, phone, smart TV and streaming device.

CodecWhere it appearsChannelsPlayer support
AAC (LC, HE-AAC)Modern HD services, streamingStereo to multichannelNear universal
MPEG-1 Layer II (MP2)Legacy broadcast, radioMono / stereoGood on set-top boxes, uneven elsewhere
AC-3 (Dolby Digital)Broadcast, film contentUp to 5.1Often needs device or receiver support
E-AC-3 (Dolby Digital Plus)Newer HD and UHD servicesUp to 7.1Growing, still device-dependent

Because audio encoding is inexpensive compared with video, converting audio to AAC while passing the video through untouched is a common and cost-effective way to maximise compatibility without touching picture quality.

Stereo, multichannel and multiple tracks

A single channel may carry several audio tracks: different languages, audio description, or a 5.1 mix alongside stereo. Decide deliberately which tracks to keep. Passing every track increases bandwidth and can confuse simpler players; dropping tracks may remove a language some subscribers need. Explicit track mapping per channel gives control over exactly which video, audio and subtitle streams reach the output.

Where a stereo version is required for broad device support, a downmix from 5.1 is a reasonable fallback, but a native stereo mix from the source is preferable when it exists.

Loudness and consistency

Nothing makes a service feel less professional than reaching for the remote every time the channel changes. Loudness standards address this by measuring perceived loudness over time rather than peak level:

  • EBU R128 (Europe) targets an integrated loudness of −23 LUFS with a true-peak ceiling.
  • ATSC A/85 (North America) targets −24 LKFS, the same measurement under a different name.

Ideally, broadcasters deliver feeds that already comply. When they don't, loudness can be corrected during processing, at the cost of decoding and re-encoding the audio. Whatever you choose, apply it consistently across the line-up so that switching channels doesn't change volume noticeably.

Timestamps and A/V sync

In an MPEG transport stream, audio and video are separate elementary streams. Each frame carries a presentation timestamp (PTS) that tells the player when to show or play it, and video frames may also carry a decoding timestamp (DTS) because of frame reordering. The player keeps the two streams aligned purely by comparing these timestamps.

Sync problems therefore almost always come from timestamps, not from the media itself:

  • Discontinuities after a source reconnect or encoder restart, where timestamps jump.
  • Non-monotonic audio timestamps in some legacy MP2 and MP3 feeds — often invisible live, but causing drift or freezes when viewers seek in catch-up or time-shift.
  • Different start points for audio and video after a cut, leaving a constant offset.

Resampling audio against its timestamps while re-encoding to AAC corrects most of these cases. For recordings in particular, storing audio as AAC with corrected timing makes the archive behave reliably on every player.

Detecting audio loss

A channel can be technically "running" while delivering silence or no audio track at all — for example when an upstream feed drops audio, or when audio only appears after the channel has started. Monitoring should treat audio loss as a fault in the same way as video loss or frozen frames: detect it, alert the operations team and, where appropriate, restart or switch to a backup source.

The detection must be careful not to flag channels that never had audio, and not to loop endlessly restarting a source that is genuinely silent. Limited retries followed by an alert is a sensible pattern.

Radio channels

Radio stations delivered alongside television are audio-only streams with their own expectations: low bitrate, very long listening sessions and frequent playback on mobile devices in the background. Keeping radio as a separate channel type allows separate packages, simpler players and correct ordering in playlists, while sharing the same ingest, monitoring and failover as television.

How SaurusRex handles it: per-channel AAC audio normalisation keeps output compatible with every player while video is passed through, custom track mapping selects exactly which tracks reach viewers, and health checks detect audio loss and restart the channel automatically, with sensible limits for channels that never had sound. Recordings store audio as AAC by default to keep catch-up in sync, and radio channels are managed alongside TV. Explore the platform.

Let's talk

See SaurusRex running on your own network.

Tell us about your channels, sites and subscribers. We'll walk you through a live demo and propose an architecture that fits.