Video Watermark Remover erases station logos, corner badges or small logos that sit in a fixed spot in the picture. It is the “Remove watermark” tab of the audio/video workbench and is built on ffmpeg's delogo filter: for each rectangle you draw, the pixels just outside its four edges are interpolated inwards to paint over what is inside. That is what defines its range — small watermarks on a fairly smooth background come out well, while a large watermark, or one sitting on complex texture, leaves a blurred patch behind.

The panel allows up to 8 rectangles, and each one can switch on “Limit time range” on its own, so a logo that only appears in the first 5 seconds is only processed for those 5 seconds. delogo requires a rectangle that does not touch the edge of the frame, so the tool automatically shrinks each box inwards by at least 1 pixel, and a box smaller than 4×4 is ignored.

1. 01Add a video and select it— switch to “Remove watermark” and wait for the panel to read the video dimensions.
2. 02Box the watermark— drag a rectangle on the preview frame and draw as many as you need (8 at most); click a box to make it current, drag a corner to fine-tune it, and each box has a delete button in its top-right corner. The “Region list” below shows the pixel size and coordinates of each one.
3. 03Limit the time if needed— when a region only appears for part of the video, turn on “Limit time range” on that row and fill in the from / to times (mm:ss or seconds).
4. 04Remove and download— click “Remove watermark (N regions)” and download <original-name>_no-watermark.mp4 when it finishes (MOV / MKV / M4V keep their extension, other containers come out as .mkv).

Where the algorithm stops

— delogo fills a box by interpolating from its edges, so it suits small, fixed-position watermarks only; semi-transparent large text, or a diagonal watermark across the whole frame, can hardly be removed cleanly

Fixed position

— a rectangle does not move during its active time; a watermark that moves has to be approximated with several regions, each with its own time range

How many, how big

— 8 regions at most; boxes are shrunk to avoid the frame edge and need at least 4×4 pixels, so a watermark right against the edge leaves a 1-pixel border behind

Re-encoding

— the whole clip is re-encoded as H.264 CRF 20, taking time in proportion to its length; AV1 sources cannot be decoded; the audio track is copied

What it does not do

— it does not pixelate faces or licence plates (the “Pixelate a region” feature in the same panel explains the better fit for that) and it does not add watermarks (use Add Watermark); only process video you have the rights to, because removing someone else's copyright mark may be unlawful

The video is processed locally in the browser with ffmpeg.wasm and is not uploaded; the first use downloads about 31 MB of engine files from a CDN (the site's asset domain or jsDelivr) and that request contains no video content.