Choose an image and Photopip finds the text on your device. Over a smooth backdrop such as a wall, sky or gradient, the gap is filled from its edges with no model; over busier areas an inpainting model fills it. Download the result in the original format, or keep editing: touch up the fill with the Remove tool or set new words with the Type tool.

1. Click Choose an image, or drop a screenshot, poster or photo on the page.
2. Wait while the text is detected and filled. The first run downloads about 33 MB of models (text detection and inpainting); after that an image takes a few seconds.
3. Click Download JPG, PNG or WEBP. The format follows the original, and "-no-text" is added to the file name.
4. To remove only some of the words, close the card with ✕ at its top right, drag the Rectangular Marquee (M) around them and click Remove Text in the Properties panel. The fill stays inside the selection.

How is the gap filled?

Over a smooth backdrop (wall, sky, gradient) Photopip interpolates from the edges of the hole, no model, so the fill is clean. Over people, fabric or patterns the inpainting model (MI-GAN, 28 MB) works on a 512-pixel crop, so a headline across a busy background comes back a little soft; paint over leftovers with the Remove tool (J).

Does it work on screenshots, posters, memes and photos of signs?

Printed Chinese and English are detected reliably. Handwriting, heavily stylized lettering, vertical text and text under strong perspective are often missed, and detection runs at 1280 pixels on the long side, so tiny text in a large image may be skipped. Select those by hand and click Remove Text, or paint them with the Remove tool.

Can I change the words instead of deleting them?

Yes. Close the card (✕), drag the Rectangular Marquee (M) around the words and click Edit Text in Image in the task bar that appears next to the selection, then type the new words: the old ones are erased and the new ones are set in the closest of three font styles on your computer (sans-serif, serif or kai). To turn every line into editable text instead, ask the AI Assistant (Ctrl+K / Command+K) to "turn the text into editable layers": each line, up to 60, becomes a text layer at the original position, angle, size and color. The original font is not copied; refine it in the Character panel.

Is my image uploaded?

No. Detection, recognition and filling run in the browser, and the notice after detection says so. If the inpainting model can't load, the fill falls back to a Content-Aware patch from nearby pixels; if no text is found, Photopip tells you; paint logos and graphic watermarks with the Remove tool (J). Use it on your own images.

Updated 2026-10-21

More toolsRemove objectsRemove peopleProduct photo on whiteRemove backgroundRaw adjustmentsPassport photoBlur backgroundResize image

Choose an image and Photopip finds the text on your device. Over a smooth backdrop such as a wall, sky or gradient, the gap is filled from its edges with no model; over busier areas an inpainting model fills it. Download the result in the original format, or keep editing: touch up the fill with the Remove tool or set new words with the Type tool.

1. Click Choose an image, or drop a screenshot, poster or photo on the page.
2. Wait while the text is detected and filled. The first run downloads about 33 MB of models (text detection and inpainting); after that an image takes a few seconds.
3. Click Download JPG, PNG or WEBP. The format follows the original, and "-no-text" is added to the file name.
4. To remove only some of the words, close the card with ✕ at its top right, drag the Rectangular Marquee (M) around them and click Remove Text in the Properties panel. The fill stays inside the selection.

How is the gap filled?

Over a smooth backdrop (wall, sky, gradient) Photopip interpolates from the edges of the hole, no model, so the fill is clean. Over people, fabric or patterns the inpainting model (MI-GAN, 28 MB) works on a 512-pixel crop, so a headline across a busy background comes back a little soft; paint over leftovers with the Remove tool (J).

Does it work on screenshots, posters, memes and photos of signs?

Printed Chinese and English are detected reliably. Handwriting, heavily stylized lettering, vertical text and text under strong perspective are often missed, and detection runs at 1280 pixels on the long side, so tiny text in a large image may be skipped. Select those by hand and click Remove Text, or paint them with the Remove tool.

Can I change the words instead of deleting them?

Yes. Close the card (✕), drag the Rectangular Marquee (M) around the words and click Edit Text in Image in the task bar that appears next to the selection, then type the new words: the old ones are erased and the new ones are set in the closest of three font styles on your computer (sans-serif, serif or kai). To turn every line into editable text instead, ask the AI Assistant (Ctrl+K / Command+K) to "turn the text into editable layers": each line, up to 60, becomes a text layer at the original position, angle, size and color. The original font is not copied; refine it in the Character panel.

Is my image uploaded?

No. Detection, recognition and filling run in the browser, and the notice after detection says so. If the inpainting model can't load, the fill falls back to a Content-Aware patch from nearby pixels; if no text is found, Photopip tells you; paint logos and graphic watermarks with the Remove tool (J). Use it on your own images.

Updated 2026-10-21

More toolsRemove objectsRemove peopleProduct photo on whiteRemove backgroundRaw adjustmentsPassport photoBlur backgroundResize image