When You Only Need the Sound

What you actually save

Measured on our own service on 2 August 2026 — one video, one server, so treat them as proportions rather than promises — the wider size picture.

The interesting row is the last one. Audio is comparable to the 360p file and a third of the 1080p one, and unlike 360p it loses nothing you were going to use.

The proportion holds better than the absolute numbers do. Audio is usually somewhere between a quarter and a third of a 1080p file, because a music or speech track sits around 128 kbps while the picture eats the rest — which is why the saving grows with duration rather than staying flat.

Two hours of a conference talk is the case where this stops being an optimisation and becomes the difference between a file that fits on the phone and one that does not.

The cases where audio is the right answer

• Interviews, podcasts and talks. A recorded conversation is a recorded conversation; the video is two people at a table.
• Music. The point is the track, and a static image or a performance you will not watch again is the rest of the file.
• Lectures you are revising from. Unless the slides matter, and if they do, take the video once and the audio for repeat listening — the study case.
• Anything you will listen to while doing something else. Commuting, walking, cooking — the screen is off regardless.
• A metered or slow connection. A third of the bytes is a third of the wait — where every megabyte counts.

The common thread is not the content type but your intention. If you cannot picture yourself looking at the screen, the screen is overhead.

There is a second thread worth naming: repetition. Anything you will play more than once is usually something you listen to on the later plays, even if you watched it the first time — which makes audio the right format for the copy you keep.

When audio only is the wrong call

The failure mode is discovering afterwards that the picture carried information.

Anything demonstrated rather than described. A tutorial where the presenter says ‘click here’ is useless as sound, and a surprising number of explanations are less verbal than they feel while you are watching them.

Slides, diagrams and code. Even a talk that is mostly speech usually has three minutes that only exist on screen.

Anything you might reuse. A clip you may want later should be taken as video, because you cannot add the picture back — the editing case.

Anything that might disappear. If this is your only copy of something, take the complete thing while it exists.

The test that works: ask what you would do if the audio turned out to be useless. If the answer is ‘download it again’, audio-only is a safe experiment. If the answer is ‘the video is gone by then’, it is not.

M4A or MP3

Two audio options and the difference is smaller than the internet suggests.

M4A is the platform's own track, copied out unchanged. Nothing is re-encoded, so it is exactly the audio the video had, and it is usually the better technical choice.

MP3 is converted, which means one generation of loss, and in exchange it plays on everything — car stereos, older players, cheap devices that have never heard of M4A. That compatibility is a real feature, not a legacy quirk.

Practical rule: M4A unless you know the device you are targeting, MP3 when you do and it is fussy — what the bitrate numbers mean.

Neither improves the source. A recording that was compressed to 128 kbps by the platform stays a 128 kbps recording, whichever container you put it in — why the ceiling is upstream.

The bitrate question people ask next — 128 against 320 — is mostly answered by the source. Converting a 128 kbps platform track to a 320 kbps MP3 produces a larger file containing exactly the same information, which is the most common way people waste storage on audio.

Where the source is genuinely high quality — a music platform serving a good stream — the higher setting is worth taking, and the difference is audible on decent headphones and inaudible on phone speakers.

Battery, storage and the things nobody measures

Size is the obvious saving and not the only one.

Some of the rest matter more day to day than the megabytes do, and none of them show up in a comparison table.

Decoding. Playing video keeps a decoder and a screen busy; playing audio keeps neither. On a long journey the difference in battery is the difference between arriving with a phone and arriving without one.

Backup and sync. A library of audio files moves between devices and into backups without thought. A library of videos is the thing that fills the cloud allowance — keeping a library manageable.

Attention. Less measurable and probably the biggest one: an audio file is something you can listen to while doing something else, and a video is not.

Playback control. Podcast and music apps do things video players generally do not — remembering position across a whole library, queueing, playing with the screen locked, and surviving a phone call without losing your place. Putting a talk into that ecosystem as audio is often a better experience than the video ever was.

Transfer time. A third of the bytes is a third of the wait, and on a hotel or mobile connection that is the difference between downloading three talks before a flight and downloading one.

Taking both

1. Take the video once if there is any chance the picture matters.
2. Take the audio as well for the listening copy — it is small enough that the duplication costs almost nothing.
3. Name them so you can tell them apart before you have forty files with similar names.
4. Delete the video once you know you never opened it. Most of the time you will not, and that is the useful signal for next time.

On our free tier this costs one of your ten daily downloads for the video and, for the platform's own audio track, nothing at all — the M4A option is a direct link rather than an assembly, so it is not counted.

Which makes the both-copies habit cheaper than it sounds. The expensive half is the video you were taking anyway, and the listening copy rides along for free — why the two routes differ.

One caveat on the free M4A: it exists only where the platform stores audio separately. Sites that serve a single finished file have no separate track to link to, so there the audio route is the MP3 conversion and it does count. That is a property of the platform rather than a choice of ours.

If you are unsure which case you are in, the quality list answers it: an ‘Audio only’ row with a size next to it means the separate track exists.