How to keep music under a voiceover

The loudest problem in a voiceover mix is usually the arrangement, not the level. A busy melody in the same range as a voice competes with it at any volume. So the brief starts with what the music should leave out.

• No lead melody while someone talks. Pads, soft chords and a light pulse sit under words; a hook does not.
• Steady energy. Sudden builds and drops pull attention from the sentence that lands on them.
• Soft, low or high instruments. Felt piano, warm pads, soft plucks and light percussion leave the middle of the range, where speech lives, mostly free.

When you generate music for a video in Sonilo, the Preserve speech switch decides what happens to the voices already in the clip. It starts off.

• On: Sonilo separates the dialogue and narration from the clip's audio and keeps them under the new music. The music is generated with that voice track as part of its input.
• Off: the finished video has only the new music. The dialogue in the clip will not be in it.

For any video with a voiceover, turn it on before you generate. It costs no extra credits. On a phone the same switch is called Keep speech & vocals.

Ducking lowers one sound while another plays: here, the music while a voice speaks. In the Studio result, Auto Duck sits on the Music track and is on by default whenever speech was found in the mix. It lowers the music while someone is speaking and lets it back up in between, and the exported video keeps it.

Two things to know:

• Auto Duck follows the Original track, the kept voices. If Original is muted, there is nothing to duck under.
• Sound effects never duck. A whoosh or an impact stays at its own level.

The Music track also has its own level. 0 dB is the level the music was matched to, and the slider moves it 12 dB either way. A common starting point is to leave it at 0 with Auto Duck on, then lower it a few dB if the words still feel crowded.

Listen once at a low volume and once on a phone speaker. If you can follow every sentence on a phone at low volume, the mix will hold up almost everywhere.

1. Play the busiest spoken section and the quietest one.
2. If a word disappears, lower the Music track or ask for a sparser direction and generate again.
3. Check the gaps between sentences: the music should come back up there, not stay buried.
4. Export the video (MP4): it combines every track you can hear, with the music level and ducking you set.

Sonilo can write the voice too. In Video translation, switch to Voiceover, write the script one line per cue, and choose whether each line keeps its natural length or is fitted to the clip. The voiceover replaces the clip's audio track, so make it first, then give the voiced video to Video to Music with Preserve speech on. Doing it in that order means the music is generated around the final words.

More in the tutorial and podcast groups of the prompt library.