Most B-roll makes a video worse. Not because it looks bad, but because it asks the viewer to process a new image that tells them nothing they did not already hear, and every one of those images is a small invitation to swipe.

So here is the short answer to “what B-roll should I film for TikTok”: film proof, not filler. Good B-roll shows the thing your words just claimed, at the moment you claim it. A number on a screen after you say the number. The burnt pan after you say “this is what happens if you skip step two”. The before photo when you say “before”. If a shot would still make sense muted, with the words removed, it is doing work. If it only exists because the timeline looked empty, cut it and stay on your face.

The rest of this post is about how to plan, shoot, and place B-roll so it holds attention instead of leaking it.

Why most B-roll loses viewers

Every cut is a reset. The viewer’s eye has to find the new subject, read the frame, and decide whether it is worth staying. We covered the cost of that reset in how fast you should cut: a cut pays for itself only when it brings new information.

Decorative B-roll fails that test by design. A laptop on a desk, a city at night, hands typing, a coffee being poured. These shots feel “produced”, but they carry zero information the voiceover has not already delivered. The viewer pays the reset cost and gets nothing back.

When we look across the informational videos in our library, the pattern is consistent. The ones that hold up tend to keep each shot short and get to their reveal early. The ones that sag lean on long, drawn-out shots that never earn their screen time. Generic illustrative montages, the kind that could sit under any script, do not separate the strong videos from the weak ones. What does separate them is whether the picture is carrying part of the argument.

B-roll as proof: the one rule

Think of every sentence in your script as a claim. B-roll is the evidence.

“I tested five of these.” Show the five, lined up.
“It took me three hours.” Show the clock, or the mess, or the tab count.
“Most people hold it like this.” Show the wrong grip, then yours.
“The results were not close.” Show both results side by side.

That is the whole idea. The shot does not decorate the sentence. It makes the sentence believable. A viewer who hears a claim and sees the proof a beat later has a reason to keep watching, because the video keeps paying off what it promises.

This also fixes a problem most creators do not notice: the claim without proof reads as an ad. A talking head that says “this changed everything” with nothing on screen to back it up sounds like every sponsored post the viewer has learned to skip. The same line with the result on screen sounds like a person showing you something.

The cut-on-the-claim rule

Timing matters as much as the shot itself. The strongest placement is simple:

Say the claim on camera. Your face and voice set it up.
Cut to the proof on the key word. Not before, not two seconds after. On the word that names the thing.
Hold only as long as it takes to read. A number needs a second. A before and after needs two. A process needs as long as the process takes to understand, and no longer.
Come back to your face for the next claim.

Cutting early spoils the line: the viewer sees the result before they know why it matters. Cutting late means they have already imagined it and your shot confirms instead of reveals. Holding too long turns proof into wallpaper.

If you want the B-roll to feel rhythmic rather than random, editing to the beat covers how to line cuts up with the audio so the switch feels intended.

Seven shots that keep working

These are the shot types that come up again and again among videos that hold attention in our library. None of them need a crew.

1. The result, first

If your video is a process, film the finished thing and put a flash of it early. “This is what we are making” gives the viewer a destination. Then every process shot is progress toward something they have already seen.

2. The before and after

Two states, same framing. A room, a face, a spreadsheet, a garden bed, a closet. Film the “before” before you start. Most creators forget and try to recreate it afterwards, and it never looks as honest.

3. The hands doing it

Top down, close, steady. Hands working on the thing the video is about carry more trust than any wide shot. The viewer is watching the skill, not the setting.

4. The screen, cropped tight

If the proof is on a phone or a laptop, do not film the whole desk. Crop to the one number, the one line, the one notification that matters. A full-screen dashboard is noise. A single circled figure is evidence.

5. The comparison

Two options, next to each other or in quick succession. Two products, two techniques, two outcomes. Comparison shots do the viewer’s thinking for them, which is why they hold attention.

6. The mistake

Show what goes wrong. The overcooked version, the bad grip, the cluttered frame. Viewers stay for mistakes because they recognise themselves in them, and the fix lands harder right after.

7. The texture close-up

The sizzle, the stitch, the pour, the click. Short, sensory, and tied to the thing you are talking about. This is the one “pretty” shot that earns its place, because it is still about the subject.

What to film when you have nothing to show

Some topics are abstract. Opinions, advice, stories. You are not making a product; you are making a point. Here the temptation to reach for stock footage is strongest, and it is still the wrong move.

Better options:

Stay on your face. A well-lit, well-framed talking head with good audio holds attention longer than a stock montage. The gear that moves retention is mostly about making that face shot easy to watch.
Show the artefact behind the story. The email, the receipt, the old photo, the notebook page. Anything real from the event you are describing.
Use text on screen as your B-roll. One key phrase, large, on a plain background. It is information, it is fast to read, and it works muted.
Change your own framing. A punch-in from a medium shot to a close-up on the important line is a visual change with no new subject to process.

If your whole format is faceless, the rule gets stricter, not looser. What works in faceless content goes deeper on why story and specificity beat generic footage every time.

Plan B-roll in the script, not the edit

Most bad B-roll is chosen in the edit, when the timeline has a gap and the fastest fix is whatever clip is in the camera roll. The fix is to plan it before you shoot.

Take your script and go line by line. Next to each claim, write one of three things:

Proof shot: the specific thing to film.
Face: the line lands best on you.
Text: a word or number on screen.

If a line has no proof and does not land on your face, that line is probably filler, and the script is telling you to cut it. This is the quiet benefit of planning B-roll: it edits the script as well as the footage.

Then shoot the proof list in one pass. Most of these shots take seconds each. Twenty minutes with a phone and a list will give you better B-roll than an hour of “getting some nice shots”.

The first three seconds are different

Everything above applies inside the video. The opening has its own rules.

If you open on B-roll, it has to be the most interesting frame you have, and it has to make sense muted. A lot of viewers see your first second with the sound off. An opening shot of your result, your mistake, or your before state can stop a scroll on its own. An opening shot of a coffee cup cannot.

We broke down how the opening frame and the first spoken line share that job in visual vs verbal hooks. The short version: if your first frame is B-roll, it is your hook, so pick it like one.

How Blossom helps

The hard part of B-roll is that you cannot judge it on your own footage. You know what every shot means, so every shot feels like it is doing something.

When you run a video through Blossom, the analysis reads it the way a viewer does. It scores the hook from 1 to 10 and explains why, maps where the pacing drags with timestamps, and hands you one fix plan: the single change most worth making first. If a stretch of decorative footage is where attention slips, that is what the plan points at. You can also browse the formats and hooks holding up in your category across our library of analyzed videos, so you can see which kinds of proof shots show up in your niche before you plan your own.

Scores are content-side estimates, not view predictions. Treat them as a guide to what to fix, not a promise of reach. Blossom is a paid tool with a trial; if you have a video where the talking was strong and the retention still dropped, start a trial and run that one first.

The short version

Good B-roll is proof, not filler. It shows what your words just claimed.
Cut to it on the key word, hold it only as long as it takes to read, then come back.
Generic montages cost the viewer a reset and give nothing back.
The shots that keep working: the result first, before and after, hands doing it, a tight screen crop, a comparison, the mistake, and a texture close-up tied to the subject.
With nothing to show, stay on your face, show the real artefact, or use on-screen text.
Plan B-roll next to each line of the script. Lines with no proof are often lines to cut.

The best B-roll is not the prettiest shot in the video. It is the one that makes the viewer believe the sentence they just heard.

B-Roll That Holds Attention: What to Film When You Are Not Talking