Workflow guide
Music that breathes around speech instead of fighting it
Ducking lowers background music whenever someone speaks and brings it back up in pauses. Done well it is inaudible — viewers simply understand every word while feeling the energy. Done badly the music pumps, distracts, or buries the voice. Here is the level-and-fade method behind clean mixes.
From the Mellius product team · Updated September 28, 2026
Conspiracy theorists, you may hate this
AI Director
›
Just type what you need
- ›Remove all silent parts
- ›Add bold kinetic captions
- ›Only have Bill Ackman speaking
- ›Cut to highlights only
- ›Add cinematic transitions
- ›Match cuts to the beat
The short answer
Set dialogue peaking near minus 6 dB and music 8 to 12 dB lower underneath, dip the music with a 10 to 30 millisecond attack and 200 to 400 millisecond release, fade music up in pauses, then check the full mix on phone speakers and headphones.
Every word intelligible
Dialogue cuts through on laptop and phone speakers, not just studio headphones.
Inaudible movement
Listeners feel energy shifts without noticing the volume riding up and down.
Music keeps its role
The track still adds emotion in gaps rather than sounding muted throughout.
Set the level relationship first
Static levels solve most ducking problems before any automation. Start with dialogue peaking around −6 dB and the music bed sitting 8–12 dB below the voice during speech — roughly −16 to −20 dB for typical beds. Choose instrumental or lyric-sparse sections under narration, since lyrics compete with speech at any volume, and high-pass the music near 100 Hz so bass does not muddy the voice.
Carve one pocket for the voice with a gentle 2–4 dB EQ dip in the music around 1–4 kHz rather than turning the whole track down further. In Premiere this is the Essential Sound panel's Ducking controls; in Resolve it is sidechain compression on the Fairlight page; in Mellius you can ask for the music to dip under dialogue and then nudge any section on the timeline. Loudness targets like −16 LUFS for stereo web delivery are the final check, not the starting move.
- Dialogue near −6 dB peaks; music 8–12 dB under during speech.
- Use instrumental sections under narration, save vocals for gaps.
- High-pass music near 100 Hz to protect voice clarity.
- Aim web masters near −16 LUFS stereo, −19 LUFS mono speech.
Shape attack, release, and fades
Attack between 10–30 ms catches line entrances without clipping first syllables; release between 200–400 ms lets music swell back naturally instead of snapping up between words. The classic pumping mistake is a release under 100 ms, which makes the bed visibly breathe with every syllable. Longer releases around 500 ms suit sparse narration with generous pauses.
Automate the edges manually where it counts: fade music up over half a second in long pauses and montages, then pull it back down just before speech resumes. Always QC the mix on phone speakers, where bass disappears and midrange crowding is exposed, plus one headphone pass for clicks at edit points. If a line still fights the track, change the music section before reaching for more volume.
- Attack 10–30 ms, release 200–400 ms as a starting point.
- Fade beds up in pauses; never snap levels between words.
- Check mixes on phone speakers and headphones.
- Swap the music section before over-processing levels.
A practical sequence
Put it into practice
- 01
Balance static levels
Seat dialogue near −6 dB with music 8–12 dB beneath and lyrics cleared from speech zones.
- 02
Dial ducking movement
Set a fast attack and medium release so dips are felt but never heard pumping.
- 03
Ride the gaps
Fade music up in pauses and montages, then verify on small speakers.
Frequently asked questions
How much should music dip under voice?
Typically 8–12 dB below dialogue works for background beds, with the voice peaking near −6 dB. Dense or lyrical tracks may need more room or a different section entirely. Judge on phone speakers: if you strain to catch words, the bed is still too hot.
Why does my ducking sound like pumping?
Pumping comes from a release time that is too fast, so the music jumps up between syllables. Lengthen release to 200–400 ms and reduce the dip depth slightly. If it persists, the track's own compression may be fighting you — try a sparser section.
Should I use sidechain compression or volume keyframes?
Sidechain compression or auto-ducking handles long narration quickly and consistently. Manual keyframes win for short videos with few speech sections or precise musical hits. Many mixers run auto-ducking first, then hand-tune the entrances that matter most.
Continue reading
Primary sources
Product capabilities change. These official references support the comparisons above.
Describe the edit. Keep the timeline.
Mellius turns conversational direction into cuts, captions, graphics, motion, color, and sound you can still inspect and change.
Start editing free