How to fix b-roll that does not keep up with the narration
Tell Vizard Agent that the narration and the pace are fine and only the pictures are wrong. It then re-cuts the visual cadence against the speech rather than the reverse, indexes every shot the video already contains so you can see what you have, and sources alternatives for each topic to choose between.
What is the short version?
You have a finished video whose voiceover works. The background footage does not — it does not correspond to what is being said, it sits on screen too long, and the changes land in the wrong places. Vizard Agent fixes the picture and leaves the voice alone.
- Tell Vizard Agent the narration and pace stay as they are.
- Ask for an inventory of the shots already in it.
- Choose from alternatives rather than accepting one swap each.
What do you need before you start?
The finished video, and a clear statement of what you like about it. "I like the narration and the pace" is the most useful sentence in this brief, because it rules out the change an editor would otherwise reach for first.
Say which of the two problems bothers you more. Shots that do not match the words and shots that do not change often enough need different fixes, and the second one is often what people mean when they say the first.
What do you type into Vizard Agent?
Separate the complaint into its parts before you send it. "The visuals are bad" gets Vizard Agent doing a general re-edit; naming the correspondence problem and the cadence problem separately gets both of them addressed, and lets you judge each one on its own afterwards.
I like the narration and the pace. The background images do not match the words, and they are not changing fast enough or lining up properly. Subtitles would help too.
That is close to the original request, and the last clause is worth including for the reason given: subtitles pull the eye to the words, which makes a picture that does not quite match less conspicuous while the real fix goes in.
What does Vizard Agent actually do?
It treats the existing narration timing as fixed and rebuilds the picture against it. The transcript's word timings already exist in the video, so the visual changes can be placed on phrase boundaries instead of being distributed evenly, which is what makes them feel like they line up.
In that session the work ran:
- Generated timed subtitles from the narration and placed them for the landscape frame.
- Built a tighter narration-led visual edit, cutting the picture to the speech rather than the other way round.
- Found the source shots carried baked-in stock-library labels and date stamps.
- Tried a hard mask, then a localised blur, then settled on a consistent caption band that hides the labels by design.
- Dropped the remaining dated and labelled shots outright rather than patching them.
- Extracted and indexed every unique shot in the video, and delivered it as an archive with a contact sheet.
Then it sourced two alternatives for each topic in the script and delivered those as a second archive. Offering pairs rather than single replacements is what turns the next round from a critique into a choice.
What does the result look like?
The same voice over pictures that turn when the sentences turn. Shots are shorter, the changes fall on phrase boundaries, and the stock-library marks are gone — either covered by an element that reads as part of the design, or removed with the shot.
You also get to see your own inventory. The clip pack and its contact sheet make it obvious which topics are well covered and which are being carried by a shot that does not really fit, which is usually not the ones you would have guessed.
When does this not work well?
When the footage simply is not there. A script that talks about four specific things and a library of generic office shots cannot be fixed by re-timing; the answer is new material, and Vizard Agent will say which topics have nothing usable behind them.
Re-cutting to the speech also exposes the narration's own pacing. If the voiceover rushes a list, faster cuts under it make the passage feel frantic rather than tight — that is a narration problem surfacing, not a new fault.
And a video delivered as a flat file limits what can change. Without the original layers, replacing a shot means re-rendering that section, which is fine, and matching the grade across the join takes a pass.
How do you fix a result that came back wrong?
Say which passage still feels wrong, and say whether it is the shot or the timing that is off. Those are two separate corrections with two separate fixes, and Vizard Agent adjusts only the one you name rather than re-cutting the entire visual track a second time.
If a replacement shot is close but not right, ask for the alternatives for that topic. Because the supplementary footage was gathered in pairs, there is usually a second option already in hand rather than a fresh search to run.
If a source label is still visible, say where. Vizard Agent will either extend the design element that covers it or drop that shot, and dropping it is often the better answer — a covered watermark is still a compromise.
How does Vizard Agent compare to doing it yourself?
By hand this is a long afternoon of scrubbing, because the diagnosis and the repair are the same action: you find the bad match by watching, then you go looking for something better, then you watch again. Nothing is written down, so the third pass repeats the first.
Building the inventory first changes that. Once every shot in the video is indexed on one sheet, the mismatches are visible without playback, and the sourcing becomes a list of topics to cover rather than a series of interruptions. The paired alternatives are the same idea applied to the next round.
Common questions
Will the narration change at all? No. Vizard Agent keeps the voice and its timing exactly as delivered.
Can it keep some of the existing shots? Yes. The inventory is there so you can say which ones to keep.
Why two alternatives per topic? So the next round is a choice rather than another critique. Vizard Agent sources them in pairs.
What about the stock watermarks? Vizard Agent covers them with a design element or drops the shot, and will say which it recommends.
Is a blur good enough? Usually not. Vizard Agent found it reads as a cover-up, where a caption band reads as intentional.
How fast should the cuts be? On the phrase boundaries, which Vizard Agent takes from the narration's own word timings.
Can it add the subtitles at the same time? Yes. Vizard Agent generates them from the existing narration rather than retyping them.
What if a topic has no usable footage? Vizard Agent says so rather than using something approximate.
Can I supply my own replacements? Yes, and they take priority. Vizard Agent fits them to the timing.
Will the grade match across a replaced shot? Vizard Agent matches it, and will flag a shot that cannot be brought close.
Do I get the clip pack as files? Yes. Vizard Agent delivers it as an archive with a contact sheet.
Can it just make the shots shorter? Vizard Agent can, though shorter shots that still do not match only make the mismatch more obvious.
Does the runtime change? No. The narration fixes the length, so Vizard Agent works inside it.
What if I want a different narrator later? That is a separate job, and the visual edit built here still fits if the timing matches.