Skip to content
Boostlix

·3 min read·Guides

How to Add Background Music Under a Voice Without Drowning It (Audio Ducking)

How loud background music should be under speech, what audio ducking is, how to duck music automatically in Premiere Pro and other editors, which part of a song to use, and how to avoid muted or claimed clips on TikTok, Shorts and Reels.

To add background music under a voice, keep the music well below the speech (as a starting point, roughly 15 to 25 dB quieter while someone talks) and let it come up a little in the pauses. That automatic lowering is called ducking. Premiere Pro does it in the Essential Sound panel (tag the music and dialogue, then Ducking and Generate Keyframes), and AI clipping tools such as Boostlix Clipper duck your own music automatically and start it at the track's strongest part. Use music you have the rights to, or the platform's own sound library, or the clip can be muted or claimed.

We make Boostlix Clipper, which finds the moments worth clipping in a long video and exports them with captions. Free trial, no card needed.

Music makes a talking clip feel finished, and it is also the easiest way to ruin one. Too loud and the viewer misses the words; too quiet and it is noise. The fix is to keep the music under the speech and let it breathe in the pauses, which editors call ducking.

How loud the music should be

There is no single correct number, because songs and voices differ, but these starting points work for most short-form clips:

MomentMusic level
While someone speaksroughly 15 to 25 dB below the voice
In pauses and at the starta few dB louder than under speech
Fade in and outabout half a second in, a second out

Then listen on a phone speaker, not studio headphones. If any word is harder to follow with the music on, the music is too loud.

What ducking does

Ducking watches the voice track and lowers the music while there is speech, then raises it again in the gaps. Without it, you either keep the music very quiet all the time (and it vanishes) or you adjust the volume by hand around every sentence.

Two settings matter:

  • How much it drops (the reduction): more for fast talkers and busy music.
  • How quickly it comes back (release): too fast and the music "pumps" between words; a few hundred milliseconds sounds natural.

How to duck music in common editors

Premiere Pro: open the Essential Sound panel, tag the voice clips as Dialogue and the song as Music, select the music, turn on Ducking, choose what it ducks against (Dialogue), set the sensitivity, reduction and fade, and click Generate Keyframes. Premiere adds volume keyframes you can adjust by hand. Adobe's steps: Automatically duck audio.

DaVinci Resolve and CapCut: both can lower music under speech, with the option names and availability changing between versions. Without an automatic option, add volume keyframes on the music track around each sentence, or simply keep the music low throughout.

Use the best part of the song

A 30-second clip that starts a song at 0:00 usually plays the intro and never reaches the part people know. Start the music at the chorus or the drop, where the energy comes in, and trim so it lasts as long as the clip, with a short fade at each end. If the clip is longer than the section, loop it at a natural point.

Music rights: avoid muted and claimed clips

  • Popular songs added in your editor are the main cause of muted audio and copyright claims.
  • Royalty-free music: check that the licence covers social media and monetised video.
  • The platform's sound library: adding a sound inside TikTok, Instagram or YouTube is often the safest route, but business and brand accounts usually get a smaller, commercial-use library.
  • Your own music, or music you commissioned, is the cleanest option.

Doing it automatically

Boostlix Clipper has music in its brand kit. Upload your own tracks (MP3, M4A, WAV, OGG or FLAC), choose one, and every clip gets it:

  • Starts at the best part: Clipper measures the track's loudness every half second and starts the music where it is strongest for that clip's length, at the chorus or drop rather than the intro, or from the beginning if you prefer.
  • Ducked under speech: the music drops automatically whenever someone talks and comes back up in the pauses.
  • Looped, faded and level-controlled: it loops for the whole clip, fades in and out, and you choose quiet, medium or loud.
  • Works with jump cuts: when pauses are cut out, the music still plays smoothly across the cuts.

Your logo and brand colour live in the same kit, and the style is remembered for every new clip. As with any tool, use tracks you have the rights to.

For cutting dead air before you add music, see how to remove silence and pauses from a video.

Questions this raises

How loud should background music be under talking?

Quiet enough that every word is clear on a phone speaker. A common starting point is music about 15 to 25 dB below the voice while someone speaks. Then listen on a phone, which is where most people hear it, and lower the music if any word is hard to follow.

What is audio ducking?

Automatically lowering one sound (usually music) whenever another (usually speech) is playing, and bringing it back up in the pauses. It keeps the voice clear without making the music disappear completely.

Can I use copyrighted music in my Reels or Shorts?

Not safely. A popular song added in your editor can get the clip muted, blocked or claimed. Use music you have licensed, royalty-free music whose licence covers social video, or the sounds offered inside each app, which come with their own rules for business accounts.

Which part of a song should I use?

Usually the chorus or the drop, not the intro. A short clip rarely reaches the best part of a song if it starts from 0:00. Start where the energy comes in, and fade in and out so it does not start or stop abruptly.

Should music play during the whole clip?

For talking clips, a low bed under the whole clip works well, louder only in pauses. For reaction and gaming clips, the original sound is often the point; leave music out or keep it very low.

Related reading