About

Mask the logo on your own video with Box, Brush and Eraser, and a neural inpainting model repaints that area on every frame. It suits the usual case: a corner logo from an AI generator like Veo, Sora, Kling or Runway that stays put. Choose Fast or the stronger Best model (a one-time download, then cached). Because each frame is filled on its own, the repaired patch can shimmer slightly between frames on moving backgrounds - it is cleanest where the area behind the logo is fairly steady.

100% in your browser - no upload, no account needed. The video is re-encoded at high bitrate(visually lossless, not untouched), and this removes the visible logo only. Invisible watermarking such as Google's SynthID is written into the pixels themselves: it survives this tool, and no browser tool can remove it. C2PA Content Credentials are a separate thing entirely - they live in the file structure rather than the picture, and the Metadata Remover strips those. Use this on footage you created or have the rights to.

Frequently Asked Questions

About removing a visible watermark from your own video, in the browser.

No. The whole pipeline - decoding, masking, inpainting, encoding and muxing - runs inside your browser using WebCodecs. Your file never leaves your device, there is no server involved, and no account is needed. That also means large files stay fast: nothing has to be uploaded or downloaded.

Yes, automatically, in the same pass. The audio track is copied across as encoded packets - it is never decoded or re-encoded - so it comes out bit-for-bit identical to the original with no quality loss. You drop in one file and get one file back with its sound intact. There is no second step and no separate audio file to reattach.

It is visually lossless, re-encoded at high bitrate - not untouched. Repainting pixels means the video has to be encoded again, and every re-encode costs something in theory. At the bitrate used here the difference is not visible in normal viewing, but we will not claim "full quality", because that would not be true.

Paint slightly beyond the watermark rather than hugging its edge - the anti-aliased fringe is part of the logo and leaving it behind is what produces a faint ghost. Keep the mask tight otherwise: the smaller the region, the less has to be reconstructed. Then use Preview to check a few frames before exporting the whole clip. A neural AI model fills the masked area on every frame; pick Fast for speed or Best for a stronger model (a one-time download, then cached). Because each frame is filled independently, the patch is cleanest where the background behind the logo is fairly steady, and can shimmer slightly on strongly moving backgrounds.

Chrome or Edge is the reliable choice. The tool needs WebCodecs (VideoDecoder and VideoEncoder) plus OffscreenCanvas. Safari has shipped WebCodecs since 16.4 but with gaps that break this pipeline, and Firefox support is still partial. The tool checks your browser up front and tells you plainly if something is missing rather than failing halfway through an export.

Up to 4K resolution and up to 100 MB per file. Processing happens frame by frame with only one frame in memory at a time, so large clips stay within reach, but a 4K export is genuinely heavy work and will take a while - the progress bar reflects real frames processed.

No. The mask is drawn once and stays fixed for the whole clip, which fits the normal case: a generator logo sits in a corner and does not move. There is no object tracking, so a watermark that drifts across the frame is out of scope - you would need to mask the whole area it travels through.