Balance the voice first, then lower music by phrases rather than chasing every syllable. A slower, intentional return often sounds more natural than aggressive automatic ducking.
The goal is intelligibility with a musical shape
Ducking lowers one sound while another needs attention. In a narration video, the music should make room for the words and recover during meaningful gaps. A mix that jumps up between every word can sound nervous, even when a meter says the voice is louder.
For a short edit, manual clip-volume keyframes are a dependable starting point. They let you anticipate the first word and shape each phrase. A sidechain compressor or dedicated ducking control can save time on a long program, but it needs sensible detector, attack and release behavior. Resolve’s available automation and ducking features vary by version and edition; this guide’s manual method works without depending on a particular premium effect.
Balance the dialogue before touching the music
Put dialogue and music on separate tracks, then mute the music. Listen through the entire voice passage at a fixed monitoring volume. Use clip gain or volume to bring unusually quiet or loud phrases into a reasonable range before asking compression to solve everything.
Fix distracting edits with short fades and remove irrelevant bumps where practical. Preserve enough room tone that the background does not vanish unnaturally between phrases. Watch the track and bus meters so neither clips. Peak level alone does not measure perceived loudness, and there is no single dialogue peak that fits every delivery target. If your final destination specifies integrated loudness and true-peak limits, keep those for the complete mix check rather than normalizing every sentence independently.
Choose two music levels by listening
Unmute the music and establish a speech level, where quiet words remain clear, and a gap level, where the music can carry a transition. As an illustrative starting point, you might lower the music 6 dB from its gap level under a phrase. A busy vocal song may need more room than sparse instrumental music.
Try this on a phone speaker as well as headphones. If a melody competes strongly with speech, lowering everything is not the only option: choose a less busy section, cut around the lyric, or use a different licensed track. Equalization can make space, but broad removal of the music’s midrange can sound hollow. Arrangement and level usually deserve attention before complex processing.
Draw a four-point dip around a phrase
- On the Edit page, expand the music track so its clip-volume line is easy to see. Enable the relevant keyframe controls if hidden.
- Add volume keyframes around one sentence. Depending on platform and version, the clip-volume line supports modifier-click keyframes; use the application’s keyframe controls if your shortcut differs.
- Place point A at the gap level a little before the voice begins. Place point B at the speech level by the first important consonant.
- Place point C near the end of the phrase at the speech level. Place point D after the phrase at the gap level.
- Listen at normal speed and lengthen either ramp when it feels abrupt.
Try a fade down over roughly 0.15–0.3 seconds and a return over 0.4–0.8 seconds as starting values. These are creative examples, not universal settings. A dramatic pause may want a slower return; rapid instructions may work better with the music held low across several sentences.
Avoid the three common automation mistakes
| Symptom | Likely cause | Adjustment to try |
|---|---|---|
| The first word is masked | The dip starts too late | Move the downward ramp earlier |
| Music breathes after every syllable | Too many dips or a short release | Shape whole phrases and bridge short gaps |
| Music never seems to return | Gap level is too low or recovery too long | Check musical gaps separately from speech |
If you later use automatic ducking, listen to the same passage against the manual version. Choose a threshold that responds to intended dialogue rather than room noise. Excessive gain reduction and a fast release encourage pumping. Do not stack manual dips, track automation and an automatic ducker without knowing their combined effect; you can accidentally suppress the music far more than intended.
Verify the whole mix outside the session
Listen from ten seconds before the voice begins through the end of the passage. A fade that sounds good in an isolated sentence may be awkward against the preceding beat. Check that edit points and narration changes still align after any timeline revisions; clip-bound automation and track automation can follow edits differently.
Export a short delivery file and listen in a normal player. Check a quiet sentence, a loud sentence, a musical gap and the ending. If speech disappears at low listening volume, revisit balance before adding more compression. Save a clean version of the music and dialogue tracks so you can rebuild the mix later, and measure the complete export against any actual destination loudness requirements.
Common questions
Is automatic ducking always better than keyframes?
No. Automation can save time, but intentional keyframes give direct control over timing and musical phrasing. Compare both on a representative section.
Should background music completely disappear under speech?
Usually it only needs to make enough room for intelligibility. The right amount depends on the music, voice, pacing and listening device.
Sources & further reading
Written for ToolWillow with original explanations and examples. Consult current product documentation for controls that vary by version.
