Sound

Narrate over the cut, and score it.

Narration recorded against a rough cut never fits the finished one. Lumae records it against the assembled timeline while that timeline plays, so what you said at 0:32 stays at 0:32, and the mix under it gets out of the way on its own.

Duck ramp: 0.25 s each side
Music default level: 50 %
Minimum piece: 0.1 s
Fades: capped at half of what plays
Per-take gain: yes, plus fade in and out
Overlap: never, on any lane

Record against the timeline, retake in place

Press record and the timeline plays while you talk. The take lands at the playhead, or after the take already covering it, and is cut to the room ahead of it if it runs into the next one. Takes never sit on top of each other, so there is never a hidden second voice under the one you can hear.

Any take can be re-recorded in place, trimmed by its edges, given fades, and levelled with its own gain, which is what evens out takes recorded on different days.

Ducking with real ramps

While you are speaking, the clip's own audio drops to the duck level and comes back, with 0.25-second ramps on both sides. Overlapping narrated spans are merged first, so a gap of a few frames between two takes does not produce an audible bounce.

The levels are written as back-to-back ramps, flat stretches included, because a mixer reads them as keyframes and interpolates between them. Spelling out every hold and every step is what keeps a muted clip silent instead of fading in across its own length.

A music track that stays under

Drop a song beneath a demo and it gets its own lane, with a level, fades, and its own ducking under narration. A new project starts music at half gain, because music sits under a demo rather than beside it.

A piece may run past the end of the video: Lumae cuts it there and lands the fade-out before the cut, which means music laid under a sequence keeps covering clips you add later instead of stopping where the video used to end.

One set of rules for every lane

Narration, music, captions, zooms and clips all move, trim, split and snap the same way. Drag an edge and it clamps to its neighbours, to the media it came from, and to the lane's own minimum. Drag an item and it snaps to the playhead, to clip boundaries, to its neighbours' edges and to the ends, with a guide line where it is held.

Waveforms are drawn from the audio itself, with peaks as a halo under the body, so a long song cut into a short video shows the same detail as a short take.

Questions

Can I record narration for just one section?

Yes. Position the playhead and record; the take starts there and stops when you do. Takes are independent, so a later one can be re-recorded without touching the rest.

Which microphone does it use?

Whichever you choose in the capture bar or in Inspector › Audio; it is one app-wide choice covering both screen recordings and narration. The device is remembered by name, so unplugging it falls back to the system default rather than silently switching to something else.

Does music duck under narration automatically?

Yes, with the same ramps the clip audio uses, and you set the level it ducks to. You can switch it off if the track is already quiet enough.

All features