Guide
How to cut B-roll to a voiceover
The search is “cut B-roll to voiceover,” “sync footage to VO,” “edit screen recording to script.” The craft is: the voice is the clock, the pictures cover the lines. Doing it by hand means a bin, subclips, and ripple edits. A speech-to-timeline pass can lay a first sequence from files you already have. It cannot invent a shot you did not shoot. Montage is that assembly in Tolik; Premiere is the same job with more control.
The job in one sentence
You have a voice (or you will). You have clips: product, city, UI, hands. Each sentence should land on a shot that makes sense, not a random Ken Burns of file 01. If the spine is your face, this is the wrong article — use jump cuts.
How a human editor does it
Transcribe or listen, split the VO on sentence boundaries, park selects on a layer, trim to the line, add handles for breathing. Screen recordings are B-roll. You still mix, you still swap the SKU that landed on the wrong sentence.
How a speech layout approximates it
Recognize the VO, clean filler and silence if you want density, place uploaded clips against remaining lines, caption the words. You then swap the warehouse wide onto line four because the model cannot know that. Product video from footage is this job with screens and SKUs named out loud.
No files, no cut. Prompt-to-video (Runway) is a different lie: pretty, not your product.
What to shoot
More coverage than you think: wides, inserts, cursor-visible UI, the box, hands. Garbage-in still applies. If every clip is the same 40-second pan, nothing intelligent can save it.
When talking-head is cheaper
If 80% of the video should be a person and B-roll is garnish, shoot the person. Montage is the inverse: picture is the show, voice is the guide.
| Montage / B-roll job | Talking-head job |
|---|---|
| UI, hands, places | Person in frame as the cut |
| Voice as the clock | Takes joined, jump cuts |
| Swap shots on lines | Swap takes of one setup |
Questions.
Do I have to pre-cut the VO?
No. You can lay picture to a long take. Pre-cutting helps only if you already know the script points.
Can I use screen recordings?
Yes. They are B-roll.
Will software generate missing footage?
A layout tool will not. A generative video tool will invent pixels that are not your product.
VO already locked in another NLE?
Export the mix, then layout, or stay in the NLE. Ping-pong only if you like pain.
Is this Descript’s script editor?
Descript is a transcript room. This job is a picture timeline. vs Descript.
Music bed?
Lay it after picture is honest, or the VO ducking will fight every clip swap.
Talking-head takes and B-roll together?
Pick a spine. Face or picture. Two exports if you need both.
Captions?
From the VO after cleanup, not from an old SRT. Auto captions.