A voice from off-screen explains the scene or leads the story, tying short cuts together into one story.
Explainer ads, brand stories, montages where the place changes every cut, and scenes where it's hard to match a character's lip movements.
AI clips have trouble matching dialogue to lip movements, so leading the story with an off-screen voice instead of scenes where characters talk means fewer failures. Write the script first, record it or make it with voice synthesis, then decide the number and length of cuts to fit its length (see B-roll Overlay). Level every sentence to the same volume and lower the music with ducking (see Sidechain Ducking). Use the same voice across the whole video, and don't repeat in words what the picture already shows.
Montage of four quiet morning shots with no one speaking on screen: steam from a kettle, a hand opening curtains, a bike leaving a gate, sunlight on a desk, while a calm female narrator voice speaks off-screen, saying: Every day begins with one small habit. Soft room sound underneath, warm light, 3 seconds per shot.
Sunset Boulevard (1950)) — The main character, Joe Gillis, who is already dead, tells the story in voice-over
ElevenLabs — How to Make AI Ads with ElevenLabs: Audio First chapter