Remove text from image

Yes. Upload the photo and the remover erases the lettering already in it and rebuilds the background. If you need to remove text without removing the background, it reconstructs the pixels behind the words rather than deleting the whole surface.

Yes. The same remove-text-from-image workflow works with photos and other JPG, PNG or WebP images.

Upload the JPG, PNG, or WebP and the remover clears the lettering it finds, then rebuilds what sat behind it. To erase text from an image, you do not need to type the words or drag a brush: the surface is cleaned in one pass, so check first that you want every word gone.

Yes — that is exactly what this page does. Delete text from an image when the words should be gone and nothing should take their place. If you want different words there instead, use the main editor: it replaces text rather than erasing it.

It removes the lettering it detects word by word, not line by line. There is no field for naming one word to keep, so if only part of the text should go, crop the image first or make a copy for each half.

Yes. An AI model reads the lettering in your upload, clears it, and reconstructs the background pixels behind it. That reconstruction is generated, not restored from a hidden original, so inspect the cleared area and the texture around it before you publish the result.

No. Those tools remove objects. This page is for erasing text from an image. Use an object remover if the thing you want gone is not lettering.

Choose a brush-based tool when you need to mark one area and leave the rest alone. This page removes every word it detects.

Use the main editor. Replacing text and removing text are different tasks, so they live on different URLs.

JPG, PNG or WebP, 10 MB or smaller. There is nothing to type: the model reads the lettering from the picture itself.

It will attempt to, and that is exactly why you should not rely on it here. Handwriting varies stroke by stroke, so detection is less predictable than with set type, and thin pen lines frequently survive a pass. Signatures are also usually the part of a document that must not be altered at all — read the acceptable-use policy before uploading one, because a cleaned file can still be a forged file.

The preview appears when processing finishes. Review the result before downloading it.

Yes, for lawful work you own or are authorized to edit, once a result exists. Do not erase text on documents, receipts or safety labels.

IDs, passports, tickets, receipts, invoices, bank statements, contracts, certificates, and food or medicine labels. Cleaning those is outside this tool.

No. This page removes all lettering it detects. Crop a copy to isolate an area, or choose a tool with a selection brush.

Thin or low-contrast lettering can leave faint strokes. Try a sharper image or crop closer to the words, then review at the final display size.

No. Text removal reconstructs the covered area from nearby pixels; it cannot recover the original background. Check patterns and edges after editing.

Only text-shaped marks are handled, and only on your own material. Removing another photographer's or a stock agency's watermark is not a supported use, and it is the kind of request that gets accounts closed.

No. The letters and the shadow they cast are both part of the area being rebuilt, so removing text from an embossed sign removes the relief as well and the patch can read as flat.

The patch is generated from the surrounding pixels, so it inherits their softness. On a sharp flat background it matches closely; on a photo with grain or a gradient it can read as a soft patch. Compare at 100 percent and try again if it is obvious.