Your Audio Is Fine. Your Mix Is Not.
Answers why does my video sound bad
Almost everyone who sends a film back with "the audio is bad" has listened on a phone, and almost everyone who receives that note goes looking at the recording. It is rarely the recording. It is that the mix was never checked where the audience actually listens.
Six things, in the order they matter.
One: the scene has no floor
Every real space has a low bed: air conditioning, traffic, the hum of a room. A professional-sounding track has one running underneath the dialogue at all times, usually the room tone recorded on the day.
Without it, every cut between two shots of the same conversation is audible, because the silence between lines changes character. Add a loop of room tone across the whole scene at a low level and the cuts stop announcing themselves. This is the highest return for the least work in the entire mix.
Two: dialogue was levelled, not ridden
Levelling means setting every clip to the same number. Riding means moving the level through each line so the quiet words and the loud ones both land.
Levelling is what a compressor does badly on its own. Riding is what a person does. Words at the end of a sentence go soft in performance, and that is where the meaning usually is.
Three: no frequency work
Three moves, all narrow:
- a high-pass at 80–100 Hz on every voice track, to stop the room rumble competing with the music
- a gentle dip around 200–400 Hz, which is where dialogue goes muddy
- a small lift around 3–5 kHz, which is where intelligibility lives
Narrow, not enthusiastic. A wide smile curve is how a voice ends up sounding like a telephone.
Four: the music is at the wrong level
Music under dialogue should sit well below it, and the mistake is almost always that it is too loud because you have heard the track forty times and know what to listen for. Drop it four decibels below where it feels right. Then check on a phone.
The other half of this: the music should move. Down under dialogue, up in the gaps. A static bed under a whole film is a wall, and the ear stops hearing it, which means it stops doing any work.
Five: nothing is ducking
Even with the bed low, dialogue needs the music to give way when it starts. A sidechain or a volume curve doing 3–6 dB of reduction on the music whenever there is speech is the difference between "I can hear them" and "I am working to hear them".
Six: it was never checked on a phone
This is the one that causes the note. A mix that sounds wide and balanced on studio monitors collapses on a phone speaker, because the speaker cannot reproduce anything below about 400 Hz and everything under there simply is not heard. A voice whose body lives at 150 Hz arrives as a thin, brittle version of itself.
So: put the phone speaker in the chain before the client does. Play it through one, mono if you can. Everything that matters should survive. What disappears was probably decoration.
The order to do it in
Room tone, then dialogue ride, then the three filters, then the music level, then the ducking, then listen on the worst speaker you own. Two hours on a ten-minute film, and it is the two hours clients hear.