How to Remove Text from an Image
A FEW SMALL MOVES
A cleaner image in three steps.
1. Find and choose
Upload a JPG, PNG, or still WebP, then choose Find text. Select only the words you want to remove. The detector starts with nothing selected.
2. Refine the selection
Use Brush for missed letters, Erase to shrink the removal mask, and Area for a rectangle. Protect paints a blue area that the repair will preserve. Green marks what will be removed; blue marks what stays. Zoom in for small details, then choose Pan to move around the image.
3. Repair and inspect
Quick repair works best on smooth backgrounds. AI repair downloads an approximately 28 MB model on first use, which can take several minutes on a slow connection. It may help with textures; it can still produce visible artifacts. Drag the comparison slider before you download. Changing a selection returns you to the original so the next repair uses the new mask.
PNG keeps transparency and the original decoded pixels outside the mask. JPG flattens transparency to white and re-encodes the whole image. Detection may miss rotated, small, decorative, or low-contrast text.
Working on a batch
Add up to 10 images, with a 40 megapixel total limit. Find text in the queue, apply optional keyword suggestions, then review each image in the editor and save it back. Words to keep add protected areas around matching detected text. You can explicitly approve the current selections for the batch, review each image first, or use Scan & erase matches to automatically repair only your entered keyword matches. An empty library never removes anything, and images with no matching removal area are skipped without a fake result. Repair is serial; cancel preserves completed results. Retry an individual failed image, and download successful results individually or as a ZIP. Nothing is selected for you by default.
Keyboard: V select, B brush, E erase, R area, P protect, H pan. Ctrl/⌘ Z undo; Ctrl/⌘ Shift Z redo. Animation, SVG, and GIF are not supported.