Image + Video Editor

Auto-detect the watermark.
Restore the background.

Upload authorized media and CleanMark will suggest the watermark area automatically. Review or correct the red mask, then reconstruct only that selected area while keeping the remaining image unchanged.

Open the editor See how it works

Runs on your own Flask server Automatic + manual masking Images & video, both supported

1Upload

2Select area

3Download

Drop images or one video here

Select one file, or up to 20 images · maximum 250 MB total

JPGPNGWEBPMP4AVIMOVMKV

Uploaded file

Preview

Review the automatic red mask. Use Brush or Eraser to correct it before processing.

Restoration settings

Automatic detection Ready

The app now uses a dominant-background detector for faint transparent logos and a generic text/logo detector as fallback. Review the red character-shaped mask before processing.

Detection sensitivity75%

Increase it when a faint watermark is missed. Lower it when too much of the image is selected.

Detection padding2 px

Run automatic detection immediately after upload.

Brush size18 px

OpenCV radius2 px

Used by OpenCV modes. Keep it small to preserve nearby details.

Mask expansion0 px

Start at 0. Increase only when a thin watermark outline remains.

Restoration engine

First AI use downloads the LaMa model. Unselected pixels are preserved exactly in the decoded image.

For video, Adaptive Video Mode checks whether the reviewed watermark is actually visible in each frame. Clean frames are preserved, short intro/outro marks use the nearest clean frame to avoid smears, and fixed full-video marks use the optimized FFmpeg/OpenCV ROI path. Original audio is preserved.

Important: the Box tool replaces the whole rectangle. When a watermark crosses a product, face, or object, use the Brush and paint only over the watermark characters.

I own this media or have permission to edit and remove this visible mark.

Restoring your media

Preparing your file...

OpenCV inpainting0%

Why CleanMark

Built to detect first, then touch only the reviewed pixels

Automatic detection prepares the selection, while the editor keeps you in control of every pixel that will be reconstructed.

Automatic detection + editing

CleanMark suggests the watermark mask immediately after upload. You can refine that red mask with the Brush, Eraser, Box and Undo tools.

Two restoration engines

Choose AI LaMa for the best quality around objects and detail, or a fast local OpenCV mode for quick, precise fills.

Images and video

Works on JPG, PNG and WEBP images as well as MP4, AVI, MOV, MKV and WEBM video. Adaptive frame detection avoids processing frames where the watermark is no longer visible.

Clean-frame protection

Images keep unselected pixels exactly and save as lossless PNG. For video, watermark-free frames are detected and preserved instead of being inpainted unnecessarily.

Runs on your own Flask server

Uploads, masking and restoration all happen through your local Flask backend, not a third-party service.

Undo and fine control

Adjust brush size, OpenCV radius and mask expansion, and undo or clear a selection at any point before processing.

Process

Three steps from upload to download

The same steps you see inside the editor above, explained.

Upload your file

Drop in one image/video or up to 20 images, with a total upload limit of 250 MB. CleanMark reads its dimensions and, for video, its duration.

Review the detected mask

The watermark is suggested automatically in red. Use the Brush or Eraser to correct the mask; the Box tool still replaces the entire rectangle you draw.

Download the result

Pick an engine and confirm you're authorized to edit the media. CleanMark reconstructs the masked area and hands you the file.

FAQ

Good to know before you start

Only media you own or have explicit permission to modify. You'll be asked to confirm this in the editor before any file is processed.

Use the Brush for anything precise — it only reconstructs the pixels under your stroke. The Box tool replaces the whole rectangle you draw, so it should only be used when the entire boxed area may be replaced, such as a watermark that doesn't cross a product, face or other object.

Yes. The selected position is checked frame by frame. Frames without the watermark are preserved, while frames containing it are restored. The position should remain fixed while the watermark is visible.

For images, all unselected pixels come from the original decoded image and the result is saved as lossless PNG. For video, clean frames are preserved; a very short watermarked intro or outro may use the nearest clean frame to prevent a visible blur or white patch.

Images: JPG, PNG, WEBP. Video: MP4, AVI, MOV, MKV, WEBM. Uploads are limited to 250 MB.

Have a mark to remove?

Upload one file or a batch of images, review the automatically detected masks, and let CleanMark reconstruct only those areas.

Open the editor