Skip to main content
  • Pricing
AI audio

Audio Ducking: Lower Music Under Voiceover Without Pumping

By Oakgen TeamPublished 4 min read

Share
Audio Ducking: Lower Music Under Voiceover Without Pumping

Audio ducking lowers a background track while another sound, usually speech, is present. The music returns toward its original level when the speaker stops. Use it when narration and music alternate; keep a steady lower music level when almost the whole video is spoken. The useful result is clear words with unobtrusive transitions, rather than music that jumps up between syllables.

You can create narration and music with Oakgen, then mix the downloaded files in an audio editor. This guide uses Audacity's documented Auto Duck workflow. It does not assume that an Oakgen generation automatically mixes or ducks a soundtrack. The example timings below are illustrative, not a tested output.

Should you duck the music or just turn it down?

SituationStart withListen for
Speech runs through almost the entire editA steady lower music levelEvery word stays clear without audible level changes
A quiet introduction leads into narrationDucking or a deliberate volume fadeMusic settles before the opening words
Long silent gaps separate spoken sectionsDucking with a controlled returnMusic fills the gap without an abrupt surge
Short breaths separate phrasesHold the reduced level through the gapMusic does not swell on each breath
Music masks speech even when quietA sparser arrangement or different trackThe voice feels clear without extreme attenuation

If you are still choosing the track, use the product-demo music brief first. Ducking fixes changing levels. It cannot make every dense arrangement suitable for narration.

Prepare separate narration and music tracks

Finish the spoken edit before setting the ducking. Removing a sentence later changes where the background should rise and fall. Keep the voice and music separate until the mix is approved; a single file containing both gives the detector less useful control.

Listen to the narration alone. Correct a missing word, awkward pause, or clipped line before making the music follow it. If the voice sounds damaged even with the music muted, use the clipping and distortion diagnostic. Ducking is not a repair for distorted speech.

Set a comfortable, consistent voice level, then audition the music underneath it. Save an untouched copy of both sources. For generated assets, a filename such as demo-voice-approved is more useful than another anonymous download when several versions are open.

Set up Auto Duck in Audacity

The Audacity Auto Duck manual, read November 20, 2026, specifies a selected track to change and an unselected control track underneath it. For a straightforward two-track mix:

  1. Import the music and the approved voice recording into the same project.
  2. Align the voice with its intended start. Place the music track above the voice track.
  3. Select the music track or the region you want to reduce.
  4. Leave the voice track unselected. It must be the first unselected track below the selected music.
  5. Open Effect โ†’ Volume and Compression โ†’ Auto Duck.
  6. Adjust the parameters, apply the effect, and listen across an entire spoken section.

That selection order matters. Selecting both tracks does not express โ€œduck the music using this voice.โ€ Check the highlighted track before applying the effect. If the wrong audio changes, undo the effect and fix the selection instead of compensating with more processing.

Adjust the detection and transitions separately

Threshold decides what level on the control track triggers ducking. If quiet words fail to lower the music, inspect the voice level and detection threshold. If room noise triggers attenuation during silence, examine the source noise and threshold rather than increasing the duck amount.

Duck Amount controls the reduction. Start with enough attenuation to hear consonants, names, and numbers, then listen in the context of the full mix. A particular decibel value cannot account for every voice, musical arrangement, or listening device.

Fade Down and Fade Up shape the transitions. A very abrupt change can sound like a fault. A return that is too slow can leave the music low after the narration has finished. Audacity exposes inner and outer fade controls; its graph helps show how long the music stays fully reduced.

Maximum Pause helps prevent fast changes through gaps. A short breath should usually remain inside the spoken passage, rather than becoming an opportunity for a music swell. Adjust this while listening to real pauses in your recording.

Change one group of controls at a time. Otherwise you may fix a missed quiet word while introducing a new surge after every phrase.

Worked example: a short narrated demo

Imagine a fictional organizer demonstration. The music starts alone, a speaker explains the dividers, and the video closes with a silent end card. This is an editing plan, not a recommended preset.

MomentIntended music behaviorReview question
Opening product viewMusic audible without speechDoes the opening feel louder than the rest of the video?
First spoken sentenceMusic moves down smoothlyCan you hear the first word, including its opening consonant?
Breath between two clausesMusic stays downIs there an unwanted swell in the gap?
Final instructionMusic remains behind the voiceAre the product name and final verb clear?
End card after speechMusic returns or fades to finishDoes the return feel intentional rather than startling?

If a real product click matters, give it a separate listening check. Reducing music does not prove that the click is loud enough or accurately synchronized. Avoid letting an effect detector decide the creative importance of every sound.

Fix the common failure patterns

When music repeatedly rises between words, review the pause and fade settings. If the whole track barely changes, confirm the music selection, control-track placement, and threshold. If the music remains intrusive after attenuation, try an arrangement with fewer competing elements.

In Premiere's documented ducking workflow, clips receive audio types and ducking creates adjustable keyframes. That is a different interface from Audacity's selected-track/control-track setup. Follow the method for the editor you are using rather than mixing the two sets of steps.

Before delivery, listen to the exported mix on headphones and a phone speaker. Compare it with the voice-only version at a similar listening level. Check the opening word, quietest phrase, longest gap, and ending. Reopen the export to catch missing audio or a late music return that a timeline preview might conceal.

Where Oakgen fits

Use the Voice Generator for the narration and the Music Generator for a suitable bed when you need new source material. Approve the voice and musical arrangement before mixing. Reuse a good take when the problem is only the mix.

Create an Oakgen account to prepare those assets, and review current plans and credits before generating. Generation and external audio editing are separate steps; this workflow makes no promise of free generation or an automatic finished mix.

Sources

  1. Audacity Manual: Auto Duck, read 2026-11-20
  2. Adobe Premiere: Automatically duck audio, read 2026-11-20
  3. Oakgen AI Music Generator
  4. Oakgen AI Voice Generator
Share