Six Level Balancing Mistakes That Make Good Dialogue Sound Uneven
Most uneven mixes come from a handful of habits, not bad recordings. Here are the mistakes we see most in explainers and interview films, and how to correct each one.
Daniel Okafor · · 6 minutes to read
Fixing loudness by ear alone
The first mistake is trusting one listen at one volume. After an hour on a timeline, your ears adapt, and a segment that sounded balanced at a loud monitoring level can sound thin when a viewer plays it quietly on a laptop. Producers then push the fader up again and again, and the whole programme drifts louder without ever becoming more even.
A steadier habit is to pick one monitoring level at the start of the session and mark it on the dial or note it down. Return to it whenever you judge a transition. Check the same passage once at a lower level and once through small speakers, because those two checks reveal most imbalances before a client does.
Meters are there to support your ears, not replace them. Use a loudness reading to confirm what you hear, and treat any large disagreement between the two as a prompt to listen again rather than a verdict.
- Fix one monitoring level and keep it
- Check quietly and on small speakers
- Use meters to confirm, not to decide
Treating every speaker as if they were the same voice
In an interview film, the presenter and the guest were rarely recorded in the same room, on the same microphone, at the same distance. Applying one gain value across both tracks leaves one voice sitting forward and the other slightly behind, and the listener hears this as strain rather than as a mixing fault.
Balance each speaker against the others first, using a typical spoken passage rather than the loudest laugh or the quietest aside. Aim for voices that feel equally close, not identical in measured level, since a deeper voice often needs a little more energy to feel as present as a lighter one.
Do this before touching music or effects. If the speakers are not settled with each other, every later decision about the bed underneath them rests on a moving base.
- Balance speakers against each other before anything else
- Judge on typical speech, not peaks
- Allow for differences in voice weight
Letting music win the argument
A music bed that sounds lovely on its own often fights the narration once the two are played together. The usual mistake is to set the music first, then lower the voice to suit it, which leaves the spoken words doing the compromising when they are the reason the segment exists.
Set the voice where it belongs, then bring the music in underneath until it supports the words without competing for the same space. If the track still masks consonants, a gentle dip in the frequencies the voice relies on usually helps more than simply turning the music down further.
Remember that the gap between voice and music should change with purpose. A title sequence can carry more music than a spoken explanation of a process, and both can be correct in the same programme.
- Set the voice first, then fit the music beneath it
- Carve space in the music rather than only lowering it
- Let the gap change with the purpose of the scene
Over-compressing to hide variation
When a speaker moves between a murmur and an emphatic phrase, heavy compression feels like a quick solution. It brings the quiet words up and the loud words down, but it also lifts breaths, room tone and small mouth noises, and it flattens the natural emphasis that makes spoken delivery believable.
A lighter approach is to ride the level by hand or with clip gain on the few phrases that genuinely sit too low, then use compression only to tidy what remains. The edit stays closer to how the person spoke, and the cleanup done earlier is not undone by the processing that follows.
If you find yourself needing a strong ratio on every track, go back a step. The cause is often an uneven recording or an edit that was never levelled phrase by phrase, and no amount of compression repairs that kindly.
- Ride clip gain on the phrases that need it
- Use compression to tidy, not to rescue
- Listen for lifted breaths after processing
Ignoring the jumps between segments
Each segment can be balanced well and the programme can still feel uneven, because the join between a studio voiceover and an on-location interview is where a viewer notices a step in loudness. Pre-recorded programme segments are especially prone to this, as they are often finished separately and assembled later.
Play every transition on its own, a few seconds before and after the cut, and compare the perceived loudness rather than the numbers alone. Room tone also shifts at these points, so a level that matches can still sound like a jump if the background changes abruptly.
Where a gap is unavoidable, a short crossfade or a brief bed of matching room tone smooths the change. Keep a list of the transitions you have checked, so none is left to chance on the final listen.
- Audition each join separately
- Match room tone as well as level
- Keep a checklist of transitions
Skipping a final check on the finished timeline
The last mistake is declaring the balance done because individual tracks look tidy. Balance is a property of the whole programme heard from start to end, and small changes made during revisions can quietly undo earlier decisions in places you are no longer looking at.
Before sending a review version, play the full programme once without stopping, at your fixed monitoring level, and note anything that makes you reach for the volume control. Those moments are the ones a viewer will also feel, even if they cannot name why.
After feedback, repeat the same pass instead of only checking the lines that changed. It takes a little time, and it is the habit that keeps a mix consistent across versions rather than slowly wandering.
- Listen to the whole programme without stopping
- Repeat the full pass after every revision
Uneven audio rarely comes from one dramatic error. It builds from small habits: judging at one volume, treating voices alike, letting music lead, leaning on compression, ignoring joins and skipping the final listen. Correct them one at a time, keep a fixed monitoring level, and give each transition its own attention. The result is a programme where nobody has to reach for the volume control, which is the quiet sign that the balance is working.
Daniel Okafor
Instructor, Sync and ReviewDaniel works with picture editors on synchronisation and review workflows. He guides learners through feedback rounds and revision scope.