He spent an hour mixing his video on good headphones. The speech was clear, the music sat in its place, the room tone gave a sense of somewhere. He exported, published, and felt pleased.
The first comment he got: "the music is too loud, I can't hear you."
He opened the video on his phone's speaker with no headphones and heard what everyone else heard: music filling everything, and words disappearing under it in every second sentence.
He had not mixed it wrong. He had mixed it in the wrong place.
Your headphones lie to you
Headphones isolate you from the room, present low frequencies strongly, and place the sound in the middle of your head. So the music feels balanced to you and the speech feels clear.
Most of your viewers are listening on a small phone speaker: no low end, in a room with noise, while doing something else. What you hear as rich, they hear as thin and vague.
The rule: mix on what you like, check on what people use.
Sound is three layers, not one
The most common mistake is treating a video as one piece of audio. It is three:
Speech. The ruling layer. Everything else serves it or gets lowered for it.
Music. It carries feeling and rhythm. Its place is always under the speech, and it rises when nobody is talking.
Ambience. The market, the wind, the clatter of a machine, the hum of a café. This is the most forgotten layer, and the cheapest way to give a video a sense of place. Without it the video is dry, as if recorded inside a box.
The numbers that work
Speech: keep peaks under -6 dB and never let them touch zero. Aim for an average near -14 LUFS on the finished file, which is what most publishing platforms use; push it louder and the platform turns it down itself, sometimes in a way that does it no favours.
Music: 15 to 20 dB below the speech. That will feel far too quiet while you are mixing, and that is normal: music you plainly notice is music that is too loud.
Ambience: lower than the music, 20 to 25 dB below the speech. Its presence is felt rather than heard.
These are starting numbers, not rulings. Set them, then listen and decide.
The order: speech first, always
Follow this order without changing it:
1. Even out the speech across the whole video. If you filmed across days or with different microphones, one clip will be louder than another. Get them all to one level before adding anything. Most software has a command to normalise clip loudness; use it, then check by ear.
2. Clean what can be cleaned. Remove hum and steady noise; the limits of that are in fixing audio in the edit: what can be saved and what can't.
3. Add ambience. Before the music, because it is part of the scene rather than decoration on top of it.
4. Add music last. And lower it until you think it is too quiet, then lower it a little more.
Ducking: the trick that saves you
Most editing software can lower the music automatically when speech starts and raise it when speech stops. The name differs between programs, but almost all of them have it, and some do it in one click.
Turn it on, then check four moments by hand: the first line of speech, the end of a long sentence, a long pause, and the end of the video. Those four places are where the trick exposes itself if it was set carelessly.
If your software cannot do it, draw the music dips yourself at every sentence. Dull work, but it makes a large difference.
Three mistakes you hear in most videos
Music that never stops. A track running from the first second to the last tires the ear and loses its effect. Bring it in where you need it, and silence it at the important moments; deliberate silence is stronger than any track.
Dead silence. When you cut a piece of speech out and leave the gap completely empty, the ear hears an alarming hole. The fix is a thin layer of the room's own tone under the whole video. Record ten seconds of silence in every place you film; they are the most useful ten seconds of the shoot day.
Clipping. When audio hits the ceiling it turns into a crackle that cannot be repaired. Watch the meter: if it turns red, you are already late.
Choosing the music before setting its level
Before any of the numbers, the right music is music that does not compete with the human voice in the same frequency range. A track with vocals or a sharp lead instrument fights your speech even when you lower it, because it occupies the same space in the ear.
So choose tracks that are quiet in the mid-range, and save the vocals for the moments with no speech. Where to get music safely and legally is its own subject, covered in music in your videos: where do you get it safely.
A three-minute check before export
Listen to the video three times, each time differently:
On your phone speaker, holding it the way people hold it. If a single word disappears, the music is too loud.
On headphones, for the low end and hidden noise.
On computer speakers at low volume, while looking at something else. That reveals whatever irritates the ear without your knowing why.
And if the original audio was bad at the shoot, no mix will rescue it. The cause is usually the room rather than the microphone, which we went through in the room before the mic. Clean audio at the source is half the edit.