Remove text from video

Your video plays through a canvas in your browser. On every frame, each box you drew is repaired by one of four methods:

• Fill from edges: each hidden pixel becomes a distance-weighted blend of the pixels just outside the box, the same idea as FFmpeg’s delogo filter. It works best on small marks over simple backgrounds.
• Blur and pixelate hide a mark without pretending it was never there.
• Solid paints a flat colour, which is useful for letterbox captions.

The frames are re-recorded with the original audio into a new MP4 (or WebM on browsers that can’t record MP4). Nothing is uploaded, so rendering takes about as long as the video plays. Keep the tab open until it finishes.

If the captions can be switched off in your player, they are a separate subtitle track, not part of the picture. Remove the track instead (for example, export without subtitles). This tool is for text that is “burned in”, meaning drawn into the frames themselves.

• Frame-by-frame generative AI on video is not available here. It costs dollars per clip and takes minutes, so this tool uses fast deterministic fills that run on your device.
• Long 4K files are scaled to 1080p while rendering to keep phones responsive.
• Marks that move continuously need several boxes with time ranges. A mark that jumps between two positions needs one box per position.