Voice and audio
How Loud Should a Video Be?
Platforms turn loud uploads down automatically, so pushing the level gains nothing and costs dynamics. What to aim for, and why your video sounds quieter than everyone else's.
2 min read · Published 10 September 2026
The usual complaint is that a video sounds quiet next to everything else in the feed, and the usual response is to raise the level until it does not. That stopped working years ago, because the platforms now adjust playback loudness themselves.
YouTube, and most of the services that followed it, measure how loud an upload is overall and turn down anything above their target. Uploading louder does not make you louder; it makes the platform apply more reduction, and a heavily limited track sounds flat once that reduction lands.
Two different measurements
Peak level is the tallest single moment in the waveform. Loudness is how loud the whole thing feels over time, measured in LUFS. They are not the same, and only the second is what platforms act on.
This is why a track that never touches the top of the meter can still be turned down, and why a quiet recording with one loud door slam is not loud at all despite hitting the ceiling.
| Measurement | What it describes | What to aim for |
|---|---|---|
| True peak | The loudest instant | Stay at or below minus 1 dBTP |
| Integrated loudness | The average felt loudness | Around minus 14 LUFS for online video |
| Dialogue level | How present the voice is | Clear above music, always |
Minus 14 LUFS is a widely used target for online video rather than a rule written down in one place, and platforms revise their handling from time to time. Treat it as the sensible middle of the road: a video mixed around there is turned down very little on any major service.
Why yours sounds quiet anyway
Usually not the overall level at all. It is the relationship between the voice and everything else.
- Music sitting too close to the voice, so the words fight for room rather than sitting on top
- A recording made too far from the microphone, which brings the room up with the voice
- Wide dynamics, where quiet sentences are lost even though loud ones are fine
- Mono voice spread across both channels unevenly, so half the listeners hear less of it
Fixing any of these does more for perceived loudness than raising the master by three decibels, and none of them triggers the platform's reduction.
A practical order of operations
- Get the voice recording right at the source; nothing later recovers a distant, roomy take
- Set the voice so it sits comfortably, then bring music up underneath it rather than the reverse
- Even out the obvious extremes with gentle compression, not heavy limiting
- Check the integrated loudness lands near minus 14 LUFS
- Leave a decibel of headroom at the top so encoding does not clip
That last point matters more than it sounds. Compression to a delivery format can push peaks slightly higher than they were in the master, so a file that just touches zero can arrive clipped. A single decibel of space costs nothing audible and prevents it.
Generated narration is consistent by nature and usually needs less work than a recorded take: the level does not wander between sentences, so the remaining job is placing music underneath it rather than rescuing the voice.
Common questions
What loudness should I export a video at?
Around minus 14 LUFS integrated, with true peaks kept at or below minus 1 dBTP. Video mixed near that is turned down very little by the major platforms, which all normalise loudness on playback.
Will making my video louder help it compete in the feed?
No. Platforms measure the loudness of an upload and turn down anything above their target, so a louder file simply receives more reduction. What is lost is dynamics, because getting louder usually means limiting harder.
Why does my video sound quiet even at a normal level?
Usually the balance rather than the level. Music too close to the voice, a microphone too far away, or wide swings between quiet and loud sentences all make speech harder to hear without showing up as a low number on the meter.
What is the difference between peak and LUFS?
Peak is the single loudest instant; LUFS describes how loud the material feels across its whole length. Platforms normalise using LUFS, which is why a track that never clips can still be turned down.