About ClearMark AI

ClearMark AI is a free AI video watermark remover that runs entirely in your web browser. This page explains who builds it, why it exists, how it works and where its limits are.

Who builds it

ClearMark AI is built and maintained by Ranjith R, an independent web builder based in Bengaluru, India, and the creator of the TryImager suite of image and video tools. There is no larger company behind it. You can see his work on GitHub and reach him at hello@tryimager.com.

Why it exists

Most online watermark removers ask you to upload your video to a server, create an account, and then pay with credits or accept a new watermark. For a personal clip, an unreleased project or client footage under NDA, sending the file to someone else's computer is a real cost. ClearMark AI was built to show that clean watermark reconstruction can happen on your own device: no upload, no sign-up, no daily limit and no watermark.

How it works, in plain language

When you choose a video, your browser reads and decodes it into memory using the WebCodecs API. Nothing is sent anywhere. When you draw a box around a watermark, the tool first checks neighbouring motion-aligned frames to recover any real background pixels that were uncovered before or after the current frame, and blends them with an unsharp mask and local texture synthesis so edges stay crisp.

For static marks or pixels that never become uncovered, the tool uses LaMa, a neural network trained to reconstruct realistic background structure and texture inside masked regions. The model file is downloaded once into your browser's cache when you run the tool and reused from then on. It runs on your graphics chip through WebGPU where supported, or on the processor through WebAssembly otherwise, and the finished video is re-encoded locally with the original audio preserved.

What it can and cannot do

It can remove static or fading watermarks, corner logos, timestamps and text overlays, reconstruct the covered area without leaving a blurry patch, and export an MP4 with original resolution, frame rate and audio preserved whenever your browser's encoder supports that combination. It supports up to four independent selection boxes with custom time ranges.

It cannot recover exact original detail that was permanently hidden behind a large opaque overlay in every frame, and the AI reconstructs plausible background texture rather than the exact original scene. It also cannot process video codecs your browser cannot decode (such as some HEVC or ProRes files) without converting them to MP4 (H.264) or WebM first, and slower or low-memory devices will take longer on high-resolution clips.

Ready to try it? Open the free video watermark remover, or read the privacy policy.