Audio Codecs Explained: AAC, Opus, MP3 and the Silent Video Problem

Almost every discussion of video quality is about the picture. Meanwhile the audio track is where a surprising share of playback failures actually happen, because the rules about which audio belongs in which container are less familiar and more often broken.

The four you will meet

AAC, the one to default to

If you are producing video for other people and have no specific reason to do otherwise, use AAC. It is efficient enough that nobody complains, it sits naturally inside MP4, and there is essentially no device that will not play it.

For a stereo track, 128 kbps is transparent for speech and fine for most music. 192 kbps is generous. Going beyond 256 for stereo is spending data on something almost nobody can hear.

Opus, better and fussier

Opus genuinely beats AAC at the same bitrate, especially at low rates. At 64 kbps it produces speech that AAC needs considerably more data to match, which is why voice and conferencing applications adopted it quickly.

The catch is where it lives. Opus was designed for WebM and Ogg. It can be placed inside an MP4 and the specification allows it, but support is patchy, and the result is a file that many players open happily and then play in total silence.

That specific combination, Opus audio inside an MP4 container, is one of the most common causes of a video with picture and no sound.

Why sound goes missing, in order of likelihood

1. There is no audio stream at all. The source separates picture and sound and only the picture arrived. Read the file: if no audio track is listed, no player setting will produce one.
2. The codec is unsupported in this container. Opus in MP4 is the usual example. The track exists and the player cannot decode it.
3. The channel layout is wrong. A 5.1 track played through a stereo device without downmixing can leave dialogue, which sits in the centre channel, entirely absent.
4. Something more ordinary. The player is muted, the wrong output device is selected, or the track is disabled. Worth ruling out before anything technical.

The difference between no audio stream and an undecodable one matters, because the fixes are unrelated. Missing means fetch the sound from the source. Undecodable means re-encode the track you already have. Reading the file settles it in seconds.

The centre channel trap

Surround audio puts dialogue in the centre channel by convention. A naive downmix to stereo that simply takes the front left and right channels drops the centre, and the result is music and effects at normal volume with the speech nearly gone.

People usually describe this as the volume being wrong or the mix being bad. It is neither. It is a downmix that discarded a channel it should have folded in.

Choosing, in one line each

• Distributing an MP4 to anyone at all: AAC, 128 to 192 kbps stereo.
• Web delivery in WebM: Opus, and enjoy the efficiency.
• Compatibility with something old and unknown: MP3, which nothing refuses.
• Surround for television or disc: AC-3, and check the downmix behaves.
• Archiving your own masters: keep the audio uncompressed or lossless, since it is a small fraction of the file size and every re-encode costs something.

Audio is usually a few percent of a video file's size. Being generous with it is one of the cheapest quality decisions available, and being clever with it is one of the least rewarding.