Why is it called b-roll?

It refers to the second reel of film. On 16mm, a cement splice showed in the projected picture, so editors split the shots across two reels, odd-numbered shots on the A roll and even-numbered shots on the B roll, with black leader filling the gaps, and printed both onto one copy. Every splice landed on black instead of the image. The B is a reel number, not a grade: it never meant second-rate footage.

What is the difference between A-roll and B-roll?

A-roll is the footage that carries the content: the interview or the piece to camera, recorded with sync sound. B-roll is everything cut in over or around it. Remove the A-roll and the video no longer says anything; remove the B-roll and it still says everything, just less watchably.

How much b-roll do I need?

For a talking-head video, one to three purposeful cutaways a minute is a common guide, at two to six seconds each, so a five-minute video might use 5 to 15 shots. Shoot several times more than you use, because shots get lost to focus, shake and framing.

How long should a b-roll shot be?

Two to six seconds over continuing narration. Shorter than about 1.5 seconds reads as a flash; longer than about eight seconds over someone still talking competes with the words. When filming, hold each shot for 10 seconds or more so you have clean handles.

Does adding b-roll make a video longer?

Only if you splice it in. A cover replaces the picture for a few seconds while the audio keeps running, so the runtime doesn't change. A splice inserts the clip as its own scene, which pauses the speech and adds the clip's length.

Can AI add b-roll automatically?

Yes. Valmera's AI agent reads the transcript of your edit, picks the moments that need a cutaway, finds real footage (your clips, stock video and photos, real footage or photos of what you name, recorded web pages), checks frames of each clip and places it on the words while your voice keeps playing. It doesn't generate footage, and you can name the moments that must stay on your face.

What Is B-Roll?

B-roll is supplementary footage cut in over or between the main footage of a video, which is called the A-roll. While a speaker talks, b-roll shows what they are describing, hides the cuts made in their take and keeps the picture moving; in its most common form only the picture changes, so the speaker's voice continues underneath. The name comes from 16mm film editing, where shots were split across two reels, the A roll and the B roll.

Types of B-Roll

B-roll is a role in the edit, not a kind of footage: anything cut in over the A-roll counts. The common types, with examples:

Type	What it shows
Cutaways to the subject	The thing the speaker names: the product, the city, the tool. The most common b-roll in talking-head video.
Insert (detail) shots	A close-up from inside the same scene: hands on the keyboard, the dial being turned, the contract being signed.
Establishing shots	Where we are: the building, the street, the room before the interview starts.
Process and action shots	Someone doing what is being described: cooking, coding, assembling, training.
Screen recordings	The app, the dashboard, the website. The b-roll of tutorials and software demos.
Photos and graphics	Stills with a slow push or pan (the Ken Burns effect), documents, charts and maps.
Archival and news footage	Historical events and named people, usually licensed from an archive or news agency.
Stock footage	Generic visuals licensed from a library. Useful, but usually the furthest from your own camera’s look.

A-Roll vs B-Roll

	A-roll	B-roll
What it is	The footage that carries the content: the interview, the piece to camera	Footage cut in over or around the A-roll
Sound	Recorded with sync sound; its audio carries the story	Usually used without its own sound; the A-roll audio plays underneath
Remove it and…	The video no longer says anything	The video still says everything, just less watchably
Example	A founder explaining the product to camera	The product in use, the team at work, the dashboard

Cutaway vs Insert, Cover vs Splice

Two distinctions get blurred under the word b-roll. The first is film grammar. A cutaway leaves the scene for something else, such as the city the speaker is describing. An insert shot stays inside the scene at a different size or angle, such as the hands on the keyboard she’s sitting at. An insert preserves the sense of continuous time; a cutaway resets it, which lets you remove more of the A-roll underneath.

The second is the edit operation, and it decides whether your video gets longer. A cover replaces only the picture for a few seconds; a splice adds the clip as its own scene.

	Cover (picture only)	Splice (adds a scene)
Runtime	Unchanged	Grows by the clip’s length
What the viewer hears	The A-roll audio, uninterrupted	The clip’s own scene; the speech pauses
Typical use	Showing what was just named; hiding a cut	A clip between sections, or a beat between sentences
Typical length	2–6 seconds	2–8 seconds, longer if the clip is the point

How Long Should B-Roll Be?

Two to six seconds per shot when it covers continuing speech. Under about 1.5 seconds the viewer registers a flash rather than an image; past about eight seconds over someone still talking, the shot starts competing with the words. For density, one to three purposeful cutaways a minute is a common guide for talking-head video, and covering much more than a third of the runtime starts to read as a montage.

When you shoot, hold each shot for 10 seconds or more so you have clean handles at both ends, and film several times more than you plan to use. With no b-roll at all, a punch-in on the A-roll hides a cut using footage you already have.

Where to Get B-Roll

Shoot it yourself. Same-shoot footage matches the A-roll’s white balance, exposure and lens by default, which is why it beats stock even when stock is prettier. Shoot sequences rather than single shots (hands, face, wide, over the shoulder, one unusual angle), and shoot at double your timeline’s frame rate, such as 60 fps for a 30 fps edit, if you want real slow motion.
Stock libraries. Pexels and Pixabay offer free footage under their own licenses; paid libraries sell licenses by use. Either way, read the license before commercial use, especially for recognizable people, brands and private property.
Archives and Creative Commons. Wikimedia Commons, Flickr, museums and public archives hold photos and footage of real people, places and events. Licenses are set per item: public domain, or Creative Commons terms that can combine credit required (BY), non-commercial only (NC) and no derivatives (ND). Credit exactly as the license asks.
Screen recordings. For software, record your own screen or capture the website itself; you own the result.
Footage from the web. A YouTube video or a news clip belongs to someone. Downloading it is not a license, so get permission or license it before you publish.

Common B-Roll Mistakes

Wallpaper. A generic timelapse under a paragraph about hiring motivates nothing. If you can’t say in one clause why this image is over these words, cut it.
Landing late. Cut away on the word that names the thing, not a second after it. Word-level timestamps make that precise.
Covering the payoff. Never cover a punchline, a reveal or a line delivered straight to camera; when the face is the content, the face is the shot.
Mismatched color. Stock shot on another camera in another room announces the cut. Grade it toward the A-roll or use your own footage.
Fixing only the picture. B-roll hides the visual jump cut, not a clipped breath or a room-tone change underneath. Fix the audio join separately, often with a J-cut or L-cut.

How Valmera Adds B-Roll

Valmera is an agentic AI video editor: you describe the edit and an AI agent performs it. Ask for b-roll and the agent reads the transcript of your current edit, picks the moments that need a cutaway, and sources real footage for them: your own clips first, then stock video and photos, real footage or photos of the people and places you name, or a recording of a web page. It checks frames of each downloaded clip, places it on the words as a cover so your voice and captions keep running, and tells you what it used and where it came from. The step-by-step version, with prompts to paste, is on the AI b-roll generator page.

Limits

Valmera doesn’t generate b-roll; it has no text-to-video or AI image generation, so every shot is real footage, a photo or a screen capture. Overlays hold their position and don’t track a moving subject. Spliced clips aren’t captioned, and a cover clip’s own audio doesn’t play.