How to Add B-Roll to Talking-Head Videos Automatically

Put a short cutaway, 1.5 to 3 seconds, on the words that describe something the viewer can picture, and go back to your face in between. To do it automatically, use an editor that reads your transcript, picks those moments and finds the footage: in StoryCut, choose an edit style with B-roll, press Generate in the B-roll card (or run Auto Edit), and swap any insert you don’t like.

StoryCut’s B-roll is on Pro and up. It uses free stock videos first and makes an AI image only when no clip fits.

Disclosure: StoryCut publishes this guide and makes the tool it describes.

A talking head is one shot of a person speaking. It’s the easiest video to make and the hardest to keep watching, because nothing on screen changes. B-roll is any footage that isn’t the speaker: a shot of what you’re describing, a product, a place, a gesture.

It does three jobs: it shows what you say ("we posted every single day" over a phone on a tripod), it resets attention with a change of picture, and it hides cuts, since a jump cut in your face disappears under a cutaway.

It also has a cost: every second of B-roll is a second without your face, and your face is why people follow you. That is why the timing rules below matter more than the footage.

• Not in the first 3 seconds. The opening is where viewers decide to stay, and they decide on a face and a hook.
• On a word you can picture. Start the insert on the word it illustrates: "ring light", "ninety days", "the needle moved". Abstract sentences ("it changed my mindset") are better left on your face.
• Short. 1.5 to 3 seconds per insert. Longer, and it feels like stock footage instead of an edit.
• With room between. At least a few seconds of you between two inserts. StoryCut never puts two closer than 2 seconds, and aims for about 5.
• As often as your format needs. A calm explainer might use one a minute, a fast Short three or more. In StoryCut, the edit style sets it: a few (about one a minute) or some (about three a minute).
• Illustrate, don’t translate. Saying "phone" doesn’t mean you need a phone on screen every time. Pick the moments that add something.

The three ways StoryCut lays out an insert on a vertical video.

A full-screen cutaway on "every single day", then a card on "moved the needle", with the editor’s B-roll card beside the video.

StoryCut searches free stock videos on Pixabay, two or three searches per moment, and keeps clips that are long enough for the insert and sharp enough for their place on screen. An AI looks at a frame of each candidate and picks the best fit, or rejects them all. The editor credits each clip ("Video by … from Pixabay"), as Pixabay’s terms ask.

When no stock clip fits a moment, StoryCut makes an AI image for it. AI images never show a face or text: a stranger’s face full screen in your video is the last thing you want.

AI clips get B-roll too, on Pro and up: when your edit style has B-roll and is applied to a run, the three best clips get full-screen cutaways.

Any editor works (CapCut, Premiere Pro, DaVinci Resolve):

• Read your transcript and mark 1 to 3 picturable words per minute.
• Search free stock sites such as Pixabay or Pexels for each one, and download vertical clips if you post Shorts.
• Put each clip on the track above your video, starting on its word, 1.5 to 3 seconds long, muted.
• Scale it to cover the frame (no see-through overlays), and keep captions above it.
• Watch it back once without sound: if you can’t tell what you’re talking about from the pictures, the B-roll is too abstract.

Most of the time goes into searching for footage, which is the part worth automating.