How to Duck Music Under Voice: Levels, Ramps and -14 LUFS
To duck music under voice, lower the music while someone speaks and bring it back up in the gaps. Drop it far enough that every word is clear, often to a quarter of its level or lower. Ramp it down just before the voice starts and back up just after it ends, then level the whole video at the end. Done well, nobody notices the ducking: they just hear the voice clearly over a full-sounding track.
What does ducking do for a video?
Music and speech fight for the same space. When both play at full level, the voice loses, and viewers on a phone speaker lose words first. Ducking gives the voice the room while it talks and hands the music its energy back in the pauses. That’s why a well-mixed video feels louder and clearer at the same time.
How far should the music drop?
Enough that you never strain to hear a word. This is a starting point many editors use:
| Moment | Music level |
|---|---|
| Intro, no voice | Full, the energy of the video |
| While the voice speaks | Around 20 to 30% of full |
| Short pauses between lines | Stays down (don’t pump up and down) |
| Longer gaps, b-roll moments | Back up |
| The last seconds | Up, then fade out |
If the music has vocals of its own, drop it lower still. Two voices at once is the hardest mix to understand.
How fast should the ramp be?
A ramp of about a fifth of a second sounds natural: quick enough to clear the first word, slow enough not to jump. Start it just before the first word so the voice never begins on top of loud music. Release it just after the last word. Short gaps between sentences shouldn’t bring the music back. Treat lines close together as one stretch of speech, or the music will pump in and out.
What about the clips’ own sound?
Footage has its own audio: wind, room noise, people in the background. Under a voice over, lower each clip’s sound to a fraction of its level, or mute it, so it doesn’t compete. Keep it a little up when it carries the scene, like a crowd at an event or waves at a beach. Talking-head clips are different: there the speech is the clip’s own sound, so the music goes under it for the whole clip.
Where should the music come back up?
In the moments without speech: the intro before the first word, a b-roll stretch between sections, the end card. Those are the places the music carries the energy, so let it breathe there. If your video has a reveal or a punchline followed by a pause, bringing the music up in that pause makes the moment land. Then dip it again before the next line.
How loud should the final video be?
Level the finished mix, not each track. Several platforms normalize playback loudness, and around -14 LUFS integrated is a common target for streaming and social video. Leveling to that, with peaks under -1 dBTP, means your video isn’t quieter than the one before it in the feed. It also won’t get turned down hard. Check the platform’s current guidance if you need a precise figure.
How do you check the mix before you post?
Listen twice: once on a phone speaker at low volume, once on earbuds. The phone speaker shows whether the words survive; earbuds show whether the music pumps or clicks. If you have to turn the phone up to follow the voice, the music is too loud under it.
What music works under a voice?
Pick instrumental tracks that don’t crowd the middle frequencies where speech lives. Tracks with a strong lead melody or lyrics compete with the voice. A simple beat or a soft acoustic bed sits underneath much more easily. Loop a shorter track rather than switching songs mid-video.
How do you duck music with Cut Express?
You don’t adjust keyframes. You say it, and Claude writes it into the timeline:
- Mark the tracks. Your music track gets a duck level (for example 0.25, a quarter of its volume), and the voice tracks are marked as voice. The voice can be a recording you upload or a voice over Claude makes.
- Automatic windows. While any voice track plays, the music drops to the duck level with 0.2 s ramps. Voice lines that sit close together count as one stretch, so it doesn’t pump.
- The final level. The whole mix is measured and brought to about -14 LUFS with a limiter at -1 dBTP. A silent video stays silent.
A prompt that works: “Use my track as music, looped, ducked to 25% under the voice, fading out over the last two seconds.” Add sound effects on top and they sit between the two layers.
FAQ
Does ducking work with the clips’ own audio? Ducking reacts to voice tracks. For a talking-head clip, lower the music overall or put the speech on its own track.
Can the music come back up between sentences? Long gaps, yes. Short pauses stay ducked on purpose.
What if the voice is still hard to hear? Ask for a lower duck level, like 15%, and render the next version.
Do I need to master the audio separately? No. Leveling happens on every render.
See the features, the FAQ, or how captions carry the words for viewers with the sound off.
Get a clean mix on every video with Cut Express
Cut Express is the video editor you drive from Claude: music, voice and effects mixed for you. See the plans or the home page.