Background Music Under AI Voiceover: Ducking and Levels

Updated September 3, 2026

TL;DR: put the music 12–18 dB under the voice, let it duck automatically while the narrator speaks, fade it in and out, and master the mix — not the voice — to platform loudness. LeapFun's Studio export does all of that with a track you upload: default bed −15 dB, sidechain ducking on, 1 s fade-in, 2.5 s fade-out, then the Standard master.

Levels

Speech intelligibility drops fast once music is within 10 dB of the voice. Start the bed at −15 dB relative to the track (LeapFun's default) and go quieter for dense narration or busy music; −30 to −6 dB is the useful range. If the voice sounds quiet, lower the music — never raise the voice into clipping.

Ducking

Sidechain ducking lowers the music while the voice is present and lets it recover in the gaps. It's what makes a bed feel produced rather than pasted. On LeapFun it's a toggle on export; in an NLE it's a compressor on the music bus keyed from the voice track.

Fades and loops

Fade in over about a second so the bed doesn't thump.
Loop the track to cover the whole narration and fade out over 2–3 seconds at the end; LeapFun stops the music when the voice ends.
Pick a track without a strong vocal or a busy top end — it fights consonants.

Sourcing music

Use royalty-free libraries you have rights to: YouTube's audio library, Creative Commons catalogues with attribution, or licensed stock. LeapFun does not generate or distribute music; you upload your own file, and the licence is yours to honour.

Master the mix, not the voice

Normalise after mixing so the final file — voice plus bed — lands at −14 LUFS. LeapFun mixes first and masters second for exactly this reason. If you mix in an editor, apply loudness normalisation on the export, not on the voice track.

Doing it in Studio

Upload a trackAny common audio format; it's stored with the project.
Set the bed and duckingChoose the level (default −15 dB) and leave ducking on.
ExportThe mix is rendered, faded, mastered with the Standard preset and delivered with subtitles. If the mix fails for any reason, you get the clean voice and the export says so.

Frequently asked questions

What level should music be under a voiceover?

12–18 dB below the voice for most narration; quieter for dense or fast speech.

Does LeapFun provide music?

No. Upload your own royalty-free or licensed track; the export mixes it under the voice.

Can I use ducking for a podcast intro?

Yes — put the intro music on its own track in your editor with the voice as the sidechain, or export the mix from Studio.

Why is my mix quieter than the voice alone?

Normalisation measures the whole mix. A loud bed pushes the voice down to hit the target; lower the bed and re-export.