The client sent a hero image with the old tagline printed across it. Or the AI mockup baked a headline into every product shot. Or the brand deck PDF has captions on every page. Removing that text — cleanly, without the model redrawing the subject behind it — used to be a Photoshop generative-fill loop. Here's how to do the whole set in one pass.

The 60-second answer

Drop the source image or mockup in.
Draw a box around every image that has text to remove.
Toggle Remove text on each box.
Submit once. Every box gets a clean version with the text erased and the background under it reconstructed — logos and wordmarks preserved.

Why removing text from an image is harder than it looks

Naive "erase the text" tools do one of two bad things: they leave a smudge where the letters were, or they redraw the whole subject from scratch and the fish in the photo suddenly has three eyes. The problem is that the model has to do two jobs at once — identify what is text vs. what is a logo or hand-lettered mark, and then reconstruct only the pixels behind the removed text without altering anything else.

Good text removal has to:

Erase headlines, subheads, captions, labels, taglines, and body copy.
Preserve logos, wordmarks, and decorative lettering pixel-for-pixel.
Rebuild the background behind the text — sky, texture, product surface — so it looks continuous.
Leave every other pixel of the image untouched. Same composition, same colors, same subject.

Why batch matters here specifically

Text-removal is rarely one image. It's a page of feature cards that all have baked-in labels. Or twelve product shots with the old campaign line on each. Or a 15-page brand PDF where every page has a caption you don't want.

Generative-fill-per-image at 30 seconds a piece is a 6-minute job on 12 images that expands to a 45-minute job with reviewing, re-prompting when it redraws the wrong thing, saving, and renaming. Batch collapses that to a single submit.

The stacked cleanup pattern

Removing text is almost never the only thing you want done. The powerful move is to stack it with the other cleanups on the same box:

Remove text + Remove overlay — kill the headline AND the dark wash it sat on.
Remove text + Remove background + Transparent PNG — strip the caption off a product shot and cut it out for a hero.
Remove text + Upscale + Sharpen — clean the label off a small mockup image and blow it up to hero size.

All of it in one pass per box, all boxes in one submit.

What gets preserved (and what gets erased)

Erased: headlines, subheads, taglines, captions, labels, section titles, body copy, any standalone text.
Preserved: standalone logos and wordmarks, decorative brand graphics that combine an icon with custom lettering, hand-drawn script logos. Custom lettering is intentional design, not distortion.

If you also want the logo gone, drop Remove background in the same box — that removes everything that isn't the primary subject.

Common questions

Will it remove text embedded in a t-shirt or a sign in the photo?

Yes, but be intentional — if the text is part of the subject (a shop sign, a product label), removing it also asks the model to reconstruct that surface. Preview the result and decide whether the reconstruction reads right.

What about watermarks?

Arturo's Acceptable Use Policy expressly prohibits using the service to remove watermarks or rights-management data. Don't do it.

Does batch text removal cost more?

No. Each box counts as one asset. Twenty boxes with text-removal is twenty assets whether you run them serially or in one submit.

Batch clean text off your images

First 25 assets are free — no signup required. Drop your mockup or image set in, box the ones with text to remove, submit once.

Try free — no signup See pricing

Remove Text From an Image — And Do It on 20 Images at Once