When music fights your voice-over: six symptoms, six fixes

This is masking, and it is the most common fault by a distance.

The track has energy between 500 Hz and 4 kHz. That is where speech lives, and where consonants sit at their quietest.

Lowering the music does not fix it, because the overlap is in frequency rather than volume.

The fix is a different track. Describe it as sparse, nothing in the midrange, low end and high percussion only.

That costs 50 credits and takes one attempt, against an afternoon of level adjustments that never quite work.

Pads, keys, guitars and synth leads are the usual culprits. Bass and hats are safe.

The reason this fault is so common is that a brief asking for warmth, fullness or energy returns exactly those instruments.

Describing absences feels wrong and produces the right result, which is why it takes most people two or three ads to adopt it.

Rushed means the tempo agrees with the read. A fast voice-over over a fast track leaves neither anywhere to sit.

The fix is to ask for slow, around 70 BPM, and let the disagreement create the space.

Nothing landing means the track is busy through the moments that should be quiet.

A hook needs a gap under it, and a track with continuous movement removes every gap you had.

Ask for almost no rhythm, or ambient pads and texture, when the script has beats that need to breathe.

Both faults are described into existence, and both are described back out for 50 credits.

If there is singing on the track and a voice-over on the ad, that is the fault. There is nothing to adjust.

A listener cannot follow two sets of words, so the sung line has to go. Leave the lyrics box empty and regenerate for 50 credits.

The wall symptom is different. The ad is fine on headphones and turns to mush on a phone.

A phone speaker reproduces almost nothing below roughly 500 Hz, so the safe half of the track disappears and only the competing half survives.

That is not a mixing skill problem, and headphones will keep telling you the ad is fine.

Judge every ad on a phone at half volume held at arm's length. It is the only listening test that matters.

Sometimes the music is simply too loud, and the level is the whole fault.

You can tell because lowering it fixes the problem completely. The music stays audible.

If lowering it helps a little and the ad still feels muddy, you have symptom one and not symptom six.

If you lower it until the problem goes away and the music has become inaudible, you also have symptom one.

The editor controls are turn music on, pick the track and set how loud it sits, so this is the only lever available.

That is by design. There is no ducking and no stem separation, which means the description is doing the work a mixing desk would.

Once you have run this diagnosis twice, you stop reaching for the level first, and the ads get clearer without anybody learning to mix.

That is the whole point of moving the control from a fader to a sentence.

Teams that run this regularly end up with music on about half their ads rather than all of them.

Their voice-overs get measurably clearer as a result. That is an odd outcome for a music module and the most useful one on offer here.

Run the ad muted, then voice only, then voice with the bed. Most people skip the middle step.

If voice only holds attention, the bed is a decision rather than a requirement. Skipping it costs nothing.

A dense read with no pauses has nowhere for music to live, whatever you brief.

A demonstration where the product makes a real sound is stronger with that sound than with a track over it.

The bed is a supporting decision. When it stops supporting anything, removing it is a fix rather than a compromise.

Four plain limits sit around all of this. No ducking, no stems, no editing after generation, and 120 seconds is the ceiling.

The track is generated for you rather than licensed from a catalogue. Confirm the usage terms that apply to your account before running it behind paid spend.