Gemini watermark remover
for video
Choose a Gemini or Veo clip and the watermark is found and cleaned frame by frame by a small AI model running on your own device. Free, no sign-up, and the video never leaves your computer.
Your browser can’t encode video (WebCodecs is missing). Please use the latest Chrome or Edge on a computer.
Drop a video here or browse
MP4, WebM or MOV · up to 300 MB · output is MP4 (H.264)
- No upload
- Local AI cleanup
- Audio kept
Keep this tab open while processing. A 8-second 1080p clip usually takes 20-90 seconds depending on your computer.
How video cleanup works
-
Detect
A dozen frames are sampled to find the exact size and position of the Gemini / Veo mark.
-
Reverse & polish
Each frame is decoded, the logo is reversed with alpha blending, then an FDnCNN model smooths the edges.
-
Re-encode
Frames are encoded to H.264 MP4 with your original audio track copied over when possible.
Free Gemini watermark remover for video
This page removes the visible Gemini and Veo watermark from video the same way the image tool does it for pictures, only many times over: every frame is decoded, cleaned and encoded again inside your browser. The clip is never uploaded, which is why there is no file-size fee, no queue and no copy of your video sitting on somebody’s server.
A dozen frames are sampled first to pin down the exact size and position of the mark. Each frame is then restored by reversing the blend, and a small AI denoise model smooths the area the logo covered. That last step matters for video and not for stills: compression distorts the watermark a little differently in every frame, so a purely mathematical reversal can leave flickering specks behind.
What to expect
- MP4, WebM and MOV clips up to 300 MB.
- Output is H.264 MP4 with your original audio track copied over when possible.
- Two models: the default polishes with AI, the second exports faster and lets you adjust the mark by hand.
- Desktop Chrome or Edge is required - Safari and Firefox cannot yet encode the result.
Still in beta
Video is newer than the image tool and harder to get perfect, because the watermark survives compression differently in every clip. If the detection is unsure, open Advanced and let it process anyway, or try the other model. Clips that were cropped or re-encoded before you got them are the least predictable.
For pictures rather than clips, use the Gemini watermark remover for images.
Questions
Model One or Model Two?
Model One samples a dozen frames to find the mark and polishes the edges with a small local AI model. Model Two finds the mark on a preview frame, lets you adjust strength, size and position with a zoomed before/after preview, and exports faster without the AI step. Try the other model if a trace remains.
Which browsers work?
Recent desktop Chrome and Edge work best (they support WebCodecs and WebGPU). Safari and Firefox may be unable to encode the output video.
Is my video uploaded?
No. Decoding, cleanup and encoding all happen in this tab. The AI model file (~2.6 MB) is downloaded once from this site.
Why is the result slightly softer under the logo?
Video compression distorts the watermark differently in every frame, so a light AI denoise is applied to the small logo area to hide leftovers.
Is the video watermark remover free?
Yes - free and unlimited, with no account and no watermark added to the export. Your computer does the decoding, cleanup and re-encoding, so there is no server cost to pass on and no upload limit to hit.
How long does a clip take?
Roughly 20-90 seconds for an 8-second 1080p clip, depending on your computer and whether WebGPU is available. Longer clips scale about linearly. Keep the tab open while it works.
Can I remove the watermark from an image instead?
Yes - the Gemini watermark remover for images handles PNG, JPG and WebP, one at a time or in a batch. It is faster than video, about a second per image.
The detection says “uncertain”. What now?
Open Advanced and enable “Process even if detection is uncertain”. If the clip was cropped or resized, results may vary.