If you film yourself talking to a camera for ten, twenty, thirty minutes, you already know the problem: a static face gets tiring to watch. The advice everyone gives is "add B-roll." The advice nobody gives is *how* — without burying yourself in stock libraries that have nothing to do with what you are actually saying.
I make long-form talking-head videos myself, so let me walk through how I add B-roll to a talking-head video now, and why I stopped reaching for stock footage first.
Why stock footage is the wrong default
Stock footage feels like the obvious answer. You are talking about "productivity," you drop in a clip of someone smiling at a laptop, done. But watch it back and something is off. The clip is generic. It does not show *your* point. It shows a stranger doing a vaguely related thing, and your audience knows it — that little "this is filler" feeling is exactly what makes people click away.
Stock has a real place (I will get to it), but as the *default* layer over a talking head it fails for one reason: your video is specific, and stock is generic. When you are explaining a number, a process, or a comparison, the right visual is that number, that process, that comparison — not a mood.
Start from what you are saying, not from a library
The shift that fixed this for me was to stop searching a library and start from the transcript. Every talking-head video is really a sequence of beats: here you state a stat, here you list three steps, here you quote someone, here you show a bit of code. Each of those beats has an obvious *right* visual.
- A statistic wants an animated number or a bar that grows.
- A process wants labeled steps appearing one by one.
- A command or snippet wants a terminal or code scene.
- A comparison wants two options side by side.
- A key line wants a clean lower-third or a pulled quote on screen.
None of those are stock clips. They are motion-design scenes — the kind of thing you would normally build in After Effects, one painful keyframe at a time. That is the real reason creators fall back on stock: motion graphics look great but take forever, so the fast-but-bland option wins by default.
The four kinds of B-roll, and when each one fits
Once you think in beats, choosing B-roll gets simple. There are really only four options for any given moment:
- Animated scenes — for ideas, data, steps, code, quotes. This is the workhorse for talking-head explainers and it is what stock can never do well.
- AI images — for a specific thing that does not exist as stock and is not data: a concept, a metaphor, a scene you could never film.
- Licensed stock — for real-world texture. A real place, a real product in the wild, an establishing shot. Real footage earns its keep when you genuinely need the real world on screen.
- Your own filmed B-roll — if you already shoot cinematic clips of your workspace or product, keep using them. Nothing replaces those.
The mistake is using one type for everything. The best talking-head cuts mix all four: open on a filmed or stock establishing shot, explain with animated charts, punctuate with a generated image, and let your own footage carry the moments that are truly yours.
Keep it consistent, or it looks like clip art
Here is the trap even motivated creators fall into: they finally add animated B-roll, but every scene uses a different font, a different blue, a different animation style. The result looks like a pile of templates, not a channel.
The fix is to define your look once — your colors, your fonts — and make every scene inherit it. When a chart in video twelve matches the chart in video two, your channel starts to feel like a brand instead of a folder of uploads. This is the single cheapest thing you can do to look more professional, and almost nobody does it because doing it by hand is tedious.
How I add B-roll now
This is the part where I will be straight with you about what I built, because it is exactly how I solve this problem for my own videos.
With Brollss, I upload my recording (or paste the script before I even shoot). It transcribes the whole thing word by word, then reads it beat by beat and generates the visual layer: animated motion-design scenes built from what I actually said, AI images where a concept needs one, and licensed stock only where real footage genuinely fits. Every scene renders in my brand colors and fonts automatically, so I never fight consistency.
Then I review. I swap a scene I do not love, reword a title, nudge timing on the timeline. It is a review pass measured in minutes, not an animation session measured in evenings. The goal is not *more* B-roll — it is the *right* B-roll on the right beat, without me opening After Effects or scrolling a stock site for the hundredth time.
That is the whole difference: instead of hunting for footage that sort of relates to your point, you generate footage that *is* your point.
Try it on your own talking head
If you make long-form talking-head content and the B-roll step is what keeps stalling your uploads, this is exactly the workflow I built Brollss around. We are in founding-access stage right now — I am onboarding early creators and reading every piece of feedback personally.
Request founding access and bring one recording. If the first pass does not make your talking head noticeably better to watch, I want to hear why.