Remove text from image areas that need a clean surface

If the caption exists as a text layer in the original design, deleting that layer is the more direct route. A flattened JPG or screenshot no longer contains a separate removable text object. The letters are part of the image, so clearing them requires reconstructing whatever surface they cover.

Use Remove Text from Image when you have the flattened picture and need a visual repair. This is distinct from translating text, changing a font or editing a document's wording. Keep those tasks in an editable design or document workflow when possible. An image repair gives you new pixels, not a recovered text box with its original settings.

A picture can contain an added caption, a product label and a sign in the background. "Remove the text" leaves the target unclear. Name the lettering by position, color and role: the small white caption at the bottom left. Explicitly preserve other words if their accuracy matters to the finished picture.

Include shadows and outlines that belong to the unwanted lettering. A bold caption may have a dark stroke or drop shadow outside the main characters. If those traces remain, the words can still be visible after the central fill. A Remove Text from Image request should describe the complete affected region without unnecessarily including a nearby logo or sign.

Upload JPG, PNG, WebP or HEIC up to 20MB. Choose the clearest available source; compression can spread colored fringes around letter edges.

1. 1

   Select the source and target

   Confirm the active image version, identify the unwanted text and use a selection if several pieces of lettering are close together.

2. 2

   Describe what should replace it

   Name the surface visible around the letters, such as pale plaster, blue sky or woven fabric. Mention any line or pattern that should continue through the repaired area.

3. 3

   Inspect the old strokes

   View the area at full size, including the spaces between letters and any former outline or shadow. Then compare all nearby text that should have remained unchanged.

Illustrative prompt, not a tested output: "Remove the white caption in the bottom-left corner and its dark drop shadow. Continue the pale wall texture and its soft lighting. Keep the red sign above the caption and all other lettering unchanged." This brief distinguishes the unwanted overlay from text that belongs in the photograph.

For this Remove Text from Image example, start by looking at the letter-shaped edges, then inspect the larger wall patch. The text can disappear while leaving a faint rectangular brightness change. If a wall seam crosses the caption, follow it through the repair. A smooth patch can still be wrong when the surrounding wall has visible grain or structure.

Text on a flat sky covers different information from text across a face or a woven pattern. The letters may be small, but the detail behind them can be important.

Follow where the original strokes were, especially rounded counters inside letters and thin punctuation. Compression or outlined fonts can leave pale or dark traces beyond the obvious character shape. During lettering cleanup, these remnants may be easier to see at full size than in the overall composition.

Next check whether the repair has introduced another pattern. A short caption over brickwork can be replaced by plausible bricks whose rows do not align. Text across a sleeve can leave fabric with a different weave. Compare neighboring material and the underlying perspective rather than accepting a patch because it no longer contains readable words.

Finally inspect retained lettering character by character. A nearby label may look unchanged from a distance while a small digit or logo stroke has drifted. Use an editable original or a clean reference when that text is a product specification, identifying mark or other detail that must be exact.

A flattened image does not preserve the exact background underneath an opaque caption. Remove Text from Image generates a plausible replacement from context. If the caption covered a serial number or a person's eye, the result cannot establish the missing original detail. Keep the unedited image and use an unobstructed source where factual accuracy matters.

Only modify material you have the right to change. Removing your own obsolete caption is a different use from obscuring someone else's ownership or attribution. This page describes a visual cleanup workflow; it does not grant permission to reuse an image or remove a mark that a license requires. For your own designs, keeping the editable source avoids this problem on the next revision.

One generated repair costs 10 credits at 1K, 20 at 2K or 30 at 4K. Auto normally starts at 1K. Internal retries do not add another charge, while a new variation you submit is a separate image task. Reviewing or selecting an existing version is free. Before another Remove Text from Image request, identify a concrete issue such as a remaining shadow or a broken pattern.

Export the accepted result as PNG, JPG or WebP. Open that file and inspect the repaired region again, especially if compression makes faint traces more visible. Use free basic resizing or cropping for delivery dimensions. If you need to add a replacement caption, do that in a workflow with editable text so you can change it later without generating another background repair.