How do you lower background music under dialogue?
Short answer
To lower background music under dialogue, keep voice and music on separate tracks, set a useful voice level, reduce the music bed, and apply ducking during speech. Rushilo can use measured speech intervals or a caption track to write volume keys on the music. Review the fades and final mix; ducking does not separate music already mixed into a voice recording.
If the music masks the first word of every sentence, turning up the finished video also turns up the music. Fix the relationship between the tracks. Dialogue should be understandable before you add a music bed, and music should remain useful when it rises between lines.
Automatic ducking creates level changes around speech. Adobe documents amount, fade duration and generated keyframes in its Premiere ducking guide. Rushilo provides its own local ducking and loudness tools, described in the release notes.
Prepare the voice and music separately
- 1Separate the sources. Keep the voice and music on separate tracks. If both are already mixed into one file, this workflow cannot independently turn down its music.
- 2Listen to the voice alone. Check clipped words, noise and large differences between speakers. A loudness adjustment does not restore distorted speech or balance every sentence.
- 3Set a useful voice level. Measure the voice track, then choose a loudness target that suits the delivery. Do this before deciding how loud the music should be.
- 4Set the music bed. Lower the track until the busiest musical passage is comfortable behind the voice. Ducking should refine that balance rather than rescue an overpowering starting level.
- 5Find the speech intervals. Use the voice track's speech measurements or an accurate caption track. Check the intervals against the current edit.
A transcript made before a large cut may no longer describe the current sequence. Refresh the timing used for ducking after you change the dialogue. Review long breaths and deliberate pauses so the music does not swell into a moment meant to remain quiet.
Apply ducking in Rushilo
Connect Claude Code, Codex or another supported client using the setup page. Ask it to inspect the project, identify the voice and music tracks, and save a named version before changing automation.
The agent can use rushilo_listen to measure the voice track and identify speech intervals. It applies audio.duck to the music tracks. Alternatively, the command can use a caption track as its timing reference. Choose the reference that matches the actual speech, rather than whichever track happens to exist.
Save a version called Before music ducking. Identify the dialogue and music tracks and report their current levels. Measure the dialogue track, then lower only the music during its speech intervals. Start with a 12 dB reduction and a 0.3-second ramp. Keep the dialogue unchanged. Tell me which existing music volume keys will be replaced, and let me review the opening, a short pause and the busiest section before export.The 12 dB reduction and 0.3-second ramp are Rushilo defaults, not a universal mixing rule. A sparse instrumental may need a gentler dip; a dense track with vocals may be a poor bed regardless of the setting.
Understand what regeneration replaces
Ducking writes visible volume keys on affected music clips and replaces their previous keys. Running it again does not stack another reduction on the old result. Save your version before generation, review the automated result, then make intentional manual adjustments afterward.
This also means a carefully drawn musical swell can be lost when you regenerate ducking. Apply the automated pass before the final hand-tuned mix, or be prepared to restore those intentional changes. Adobe makes a similar overwrite warning for its own generated keyframes.
Understand the numbers without guessing
A hypothetical voice track measured at -22 LUFS and adjusted toward -16 LUFS needs a 6 dB gain increase before other mix interactions. That calculation is target minus measured loudness: -16 - (-22) = 6. It is not a promise that the complete exported mix will measure -16 LUFS after music and processing are added.
A 12 dB duck reduces the music’s amplitude multiplier to about 0.25: 10 raised to (-12/20). It does not mean reducing a volume slider by 12 percentage points. Keep units explicit when instructing the agent.
Apple Podcasts recommends overall loudness around -16 LKFS with a 1 dB tolerance and a true-peak ceiling of -1 dBFS. That is delivery guidance for that destination, not a rule for every video. Rushilo reports loudness and sample peaks; its export limiter’s -1 dBFS ceiling is not a substitute for measuring true peaks when a delivery specification requires them.
Repair common ducking problems
First words are masked
- What to inspect
- The speech interval and the start of the fade
- A useful change
- Correct the interval or allow the level to fall earlier
Music pumps between words
- What to inspect
- Short gaps and rapid level changes
- A useful change
- Use smoother ramps or keep the bed lower through the sentence
Music still dominates
- What to inspect
- The baseline level and arrangement
- A useful change
- Lower the bed or choose less busy music before increasing the duck
Music never returns
- What to inspect
- Speech detection or overlong caption intervals
- A useful change
- Check the timing reference and remove incorrect intervals
The ending suddenly jumps
- What to inspect
- The last speech interval and the outro fade
- A useful change
- Shape the outro intentionally after the automatic pass
| What you hear | What to inspect | A useful change |
|---|---|---|
| First words are masked | The speech interval and the start of the fade | Correct the interval or allow the level to fall earlier |
| Music pumps between words | Short gaps and rapid level changes | Use smoother ramps or keep the bed lower through the sentence |
| Music still dominates | The baseline level and arrangement | Lower the bed or choose less busy music before increasing the duck |
| Music never returns | Speech detection or overlong caption intervals | Check the timing reference and remove incorrect intervals |
| The ending suddenly jumps | The last speech interval and the outro fade | Shape the outro intentionally after the automatic pass |
Compare voice only, music only and the combined mix at the same playback setting. Then listen on headphones and a small speaker. Low playback volume is a useful final test: if words disappear before the music does, keep working on the balance.
For an episode with separate video and audio deliverables, continue with the video podcast workflow. Keep a saved master before creating a teaser or a different mix.
Sources
- Rushilo release historyRecording, regions, reframing, local editing, sound tools and export, checked October 13, 2026.
- Adobe Premiere automatic duckingAmount, fade timing and regenerated-keyframe overwrite behavior. Updated January 7, 2026.
- Apple Podcasts audio requirementsDestination-specific formats, loudness, encoding and true-peak guidance, checked October 13, 2026.
Cite this page
Last verified: Oct 13, 2026
Patel, O. (2026, October 13). How to lower background music under dialogue. Rushilo. https://rushilo.com/guides/duck-music-under-dialogue
About the author
Founder, Rushilo
Om Patel is the founder of Rushilo, a desktop video editor driven by the user's own AI, and of BigIdeasDB, a research platform that reads real complaints and reviews to help founders validate products. He writes Rushilo's research articles with the same method: primary sources, dated numbers and the words real users write.
