Remove Watermark from Video

Same-file test: the mark cleared, the line did not

We made two new six-second landscape videos on 25 October 2026, then selected each local file in this tool, set the repair rectangle, exported and downloaded the WebM. These are owned synthetic fixtures with DEMO lettering and a timer, not downloaded platform videos. Both have a moving gold object away from the mark and an audio tone. One background is flat; the other has a diagonal line running behind the lettering.

Each source is 480 by 270 pixels, contains 144 video frames at 24 frames per second, and carries a 600 Hz audio tone. We used one rectangle at X 190 and Y 82, width 108 and height 58, active from zero to six seconds. The second zone stayed disabled. These settings belong to this fixture; they are not a universal location for a watermark.

The plain-background export measured 140,493 bytes. We extracted frames at one and four seconds from the file we actually downloaded. They retained the 480 by 270 frame and showed the selected lettering replaced by a plain patch. Keeping the composition matters when the mark is inside the image rather than in a border you can discard. It still does not mean that the covered pixels have been recovered.

The diagonal-line result measured 139,251 bytes, and we kept it too. The lettering became less visible, but the line bent into a smeared rectangular patch. That is an appearance failure even though an export was produced. A clean sampling border is not enough to rebuild a feature that crosses the selected area. If your mark crosses a face, fine text, a wire or another useful edge, this example is a reason to stop and inspect rather than accept a cleaner-looking thumbnail.

Both downloaded files passed a complete decoding check. Each decoded audio stream contained six seconds of nonzero signal. That confirms an audible signal in these samples, not preservation of an original compressed soundtrack or a guarantee of exact timing for every clip. The browser records a new WebM. Check the saved file's sound and the beginning and end of the video, especially when synchronization matters.

The plain and line checks used the same rectangle and workflow, so changing the background was enough to change whether the result was acceptable. We have not turned that pair into a success rate, a ranking against other tools or a claim about all video watermarks. The source clips and the failed output remain available below so you can examine the limitation yourself.

Repeat with the plain source clip and downloaded plain result, then compare the line source with its saved failure. Retain your own original before making a decision. Our earlier offline algorithm images are not used here as substitutes for these newly downloaded browser results.