Voiceover (VO)
Narration laid over the picture instead of recorded in sync with it, whether or not you ever see the speaker on screen.
See it
What it is
Voiceover is narration laid over the picture rather than recorded in sync with it, the opposite of sync sound, where the voice and the image were captured together. The invisible narrator is the common case, but visibility is not the test: a presenter you see on camera or a character in the scene can narrate over the cut too, and it is still voiceover. It gets recorded either to picture (the reader watches the cut and times the read to it) or wild (read free, then cut to fit). For explainers, ads, and product demos the VO usually comes first and the edit is built to the voice.
Write it for the ear. Short sentences, one idea each, second person, contractions, no clause a person would run out of breath inside. Conversational pace lands around 150 words per minute, but that is raw reading speed with nothing in between. Pauses, emphasis, and breathing all cost seconds, so a 60-second spot wants roughly 130 to 145 words, not a flat 150. Write it, read it aloud against a stopwatch, and cut until it fits before anyone books a booth.
Gotcha: the fixes people skip. Record 30 seconds of room tone in the same session so edits between takes do not click, leave headroom rather than tracking hot, and stop deleting every breath, because a totally breathless read sounds like a robot reading a menu. Keep the raw take too: a client who wants one word changed will need a matching mic and room, and that is a lot easier when nothing was baked in.
Ask AI for it
Write a 60-second voiceover script for [product/topic]. Target roughly 130 to 145 spoken words, not a flat 150, and budget explicit pause time between beats so the timed read still finishes under 60 seconds at a conversational pace. Conversational second person, sentences under 15 words, one idea per sentence, no jargon a newcomer would trip on. Open with the problem in the first 5 seconds, land one benefit per beat, close on a single call to action. Mark pauses with a line break, mark emphasis in bold, and add a bracketed timecode every 10 seconds so the editor can cut to it. Report the word count and the timed duration of your own read-through.