1. Choose music that leaves room
- No vocals. Singing competes directly with speech. Use instrumental versions.
- Steady, not busy. Pads, soft piano and gentle rhythms sit under a voice; drum fills and lead melodies fight it.
- Licensed for your use. Check the licence covers where you're publishing, including monetized video.
2. Set the levels
Get the voice sounding right on its own first, then bring the music up underneath until you can just feel it. While someone is speaking, music typically sits around 15 to 20 dB below the voice; between lines, and in intros and outros, it can come up. If you're unsure, it's too loud: listeners forgive quiet music, not a voice they can't follow.
3. Duck the music under speech
Ducking lowers the music automatically whenever the voice is speaking and brings it back up in the gaps. Most editors have it:
- Audacity: put the music on the track above the voice, select the music, and use Effect › Volume and Compression › Auto Duck.
- Video editors: Premiere Pro's Essential Sound panel can duck music under dialogue, and DaVinci Resolve and Final Cut Pro have equivalents; or draw volume changes by hand.
- DAWs: a compressor on the music, side-chained from the voice.
Because abcVoiceover gives you one file per line, gaps between lines are clean silence, and ducking responds precisely to each one.
4. Make room with EQ
- On the voice: a high-pass filter around 80 Hz removes rumble a voice doesn't need.
- On the music: a gentle dip of a few dB somewhere between about 1 and 4 kHz, where speech is clearest, lets the voice cut through without the music getting quieter overall.
- Small moves. If you can hear the EQ, it's too much.
5. Fades and edges
- Fade music in over a second or two, and out over two to five at the end.
- Let the music swell in a pause between sections, then duck back before the next line.
- End the voice before the music, not after: a voice left alone after a cut-off sounds like a mistake.
6. Check, then set the loudness
- Listen on headphones, laptop speakers and a phone. Phones are where buried voices get lost.
- Mix down to one file.
- Normalize to the platform's target, such as -16 LUFS for a podcast or about -14 LUFS for YouTube. See loudness standards.
Common mistakes
- Music mixed on studio monitors at high volume, which sounds fine there and buries the voice on a phone.
- Compressing the voice so hard it sounds squashed. AI voice takes are already even; they need little compression.
- Music that changes key or mood in the middle of a sentence. Edit the music to the script's sections.