How to remove burnt-in captions so you can reuse old footage

Last updated 2026-08-25 · 7 min read

What is the short version?

Footage you already published usually has subtitles baked into it. Reusing it means either living with two sets of captions on screen or removing the old ones first, and removing them properly is a measurement job before it is an editing job.

1. Go to Vizard Agent and upload the old videos.
2. Say the burnt-in captions must be removed before anything is cut.
3. Say what the new video is meant to be.

What do you need before you start?

The old files and a plan for the new video. Vizard Agent finds where the captions sit in each file itself, which matters because three videos made at different times rarely put their subtitles in exactly the same place.

• The old videos. However many you want to reuse.
• The new brief. What the recut is for.
• The new caption style. Or the same one, cleanly rendered.
• The frame. Vertical, square or wide.
• The audio plan. Original sound, or a new voice over it.

What do you type into Vizard Agent?

Name the problem before naming the outcome. Vizard Agent will otherwise cut your footage and add captions on top of the ones already there, which looks exactly as bad as it sounds and is only discovered on playback.

Prompt

Variants worth knowing:

• A clean erase rather than a blur. When the band sits over detail.
• Consistent new captions. One style across footage from different years.
• Levels matched. Old files rarely share a loudness.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real recut built from three previously published videos. The stretch that matters runs from the fourteenth step: it found the caption bands, tested a blur, judged it insufficient, and tried a cleaner method.

1. Checks the client's past project files and reads the page extraction options.
2. Extracts images and facts from the product page for the claims it can use.
3. Downloads and probes the three source videos and transcribes all of them.
4. Extracts frame grids from each source and tiles them into review grids.
5. Reviews the frames from all three videos by eye.
6. Locates the existing caption bands and crops them at full resolution.
7. Inspects each source's caption band closely to see what is underneath.
8. Tests a blur to hide the old captions and reviews the result honestly.
9. Tests a cleaner erase method, retests it, and measures each source's speech loudness before cutting.

Steps 8 and 9 are the whole job. A blur is the obvious first attempt and it often leaves a smeared rectangle that draws more attention than the text did, so the test has to be looked at rather than assumed.

What does the result look like?

From the run this page is written from, probed on the delivered file: 1080x1920, H.264, 30fps, 37.85 seconds, AAC audio. Vertical, thirty-eight seconds, cut from three previously published videos with their old subtitles removed and one consistent new caption style running across all of them.

The join is what you notice, or rather do not. Three files shot at different times, with different loudness and different subtitle styles, end up reading as one video because all three of those differences were measured and corrected.

When does this not work well?

Removing text that is baked into the pixels means reconstructing whatever was behind it, and sometimes there is no good answer available. Vizard Agent shows you its own test of the result rather than promising something it cannot actually deliver.

• Captions over faces are hardest. There is detail to reconstruct, not just background.
• Moving backgrounds complicate it. A static patch fails when the shot pans.
• Outlined text bleeds wider. Strokes and shadows extend past the letters.
• Low resolution shows repairs. Compression artefacts survive the erase.
• Reuse rights still apply. Cleaning captions off footage does not make it yours.

How do you fix a result that came back wrong?

Point at the section that still looks wrong. Vizard Agent keeps the located caption bands, the full-resolution crops, both removal tests and the loudness measurements, so the method can be changed on one source without redoing any of the others.

• "The blur is obvious." Replaced with the cleaner erase on that file.
• "A caption still flashes." The band is re-located across the whole clip.
• "One clip is much louder." Re-levelled from the measurements already taken.

How does Vizard Agent compare to doing it yourself?

By hand this means masking the caption area in an editor, trying a blur, deciding it looks bad, trying a patch, and repeating that for every source file. Most people give up and crop the bottom off, which loses a third of a vertical frame.

Common questions

Can Vizard Agent remove any burnt-in text? Vizard Agent removes what can be reconstructed. Text sitting over a moving face is the hard case.

Is a blur good enough? Often not. Vizard Agent tests it and moves to a cleaner erase when the blur is visible.

Can it handle several videos at once? Yes. Vizard Agent locates the caption band separately in each file, since they rarely match.

Will the new captions look consistent? Yes. Vizard Agent applies one style across all the cleaned footage regardless of where it came from.

What about differing audio levels? Vizard Agent measures speech loudness per source and matches them before cutting.

Why not just crop the captions off? Because on a vertical video that costs a third of the frame, and the subject is usually near the bottom. Reconstructing the band keeps the composition you originally shot.

Does it work on old low-resolution files? Partly. Compression artefacts survive the repair, so Vizard Agent's result shows more on poor sources.

Can it reuse footage from a published video? Yes, if it is yours. Removing captions gives you no rights you did not already have.

Will Vizard Agent tell me if it cannot do it cleanly? Yes. It reviews its own test and reports when the result is not good enough.

How much footage can I recover this way? Whatever the caption band does not destroy. Vizard Agent reconstructs the strip behind the text, so most of a library that was published with subtitles becomes reusable again.

Should I re-caption in the same style as before? Usually not. Vizard Agent applies one consistent style across footage from several years, and matching an old style defeats the point of cleaning the originals in the first place.

Does Vizard Agent check the finished video? Yes. It reviews the cleaned footage and the new captions before delivery.