About VideoTextRemove

VideoTextRemove is a small, focused tool: it removes text that has been rendered into video frames, and rebuilds the picture underneath. Nothing else. No timeline, no template library, no editor to learn.

Why we built it

Every existing option forced a bad trade. Crop the frame and lose composition. Blur the text and leave a visible scar. Re-shoot and lose the footage. Frame-level reconstruction is the only method that keeps both the framing and the resolution, and until recently it required a workstation and a research paper.

How it works

Text is located with a high-frame-rate OCR pass, masked, and then reconstructed using motion information from surrounding frames. That last part matters: a per-frame repair flickers, while a temporally aware one stays stable. We use a specialised subtitle-erase model rather than a generic inpainting model, because the failure modes are different.

What we will not do

We do not help remove watermarks or attribution from content you do not own. That rule is in the acceptable use policy and we enforce it. It costs us customers; it also keeps the service defensible for the people who use it legitimately.

Data

Uploaded files are deleted seven days after processing, or immediately when you delete a task. We do not train models on customer footage. Details are in the privacy policy.