Edit text in image online

In a JPG, PNG or WebP export, visible letters are part of the pixels, not a separate text layer. Enter the words you can see and their replacement; the editor redraws the requested text in the surrounding style.

Use the original design file when you have it and need exact typography. For an exported image, inspect the result at full size: complex backgrounds and very small letters can need another attempt.

Type complete words with their original capitalization. A replacement of similar length is easier to fit into an existing headline, button or label; for several different styles, enter each line separately.

The same workflow works for a finished poster, a product image or an app screenshot. Only upload images you own or are allowed to edit.

If a layered source document is available, use its editable text layer when exact font files, vector output or layout control matter. If only an image remains, the letters are pixels: the editor must redraw the words and reconstruct the area around them.

TextImageEdit replaces existing wording or translates visible text. It does not expose editable design layers; for a new caption or moving elements around the page, use a layout editor instead.

Enter the old text exactly, including punctuation and capitals, and keep the replacement close in length where space is tight. Add separate pairs for different lines; up to five can be sent with one image.

Zoom in on the new lettering and nearby details. AI output can differ in small or blurry text, complex backgrounds and unusually long replacements. Start again from the original if another attempt is needed.

An image text editor is useful when words are baked into a photograph, a flattened export or a screenshot. The upload has no live text layer to select, so the requested wording has to be redrawn. Choose the original layered design instead when it is available and you need to reposition elements or preserve exact vector typography.

For a simple typo, enter the whole word as it appears and the corrected word. For a new event date, include the entire date and punctuation. If the image contains several different changes, add a separate pair for each line and review the complete result before publishing.

Copy the source lettering carefully, including capitalization, punctuation and spaces. The words you provide help identify the intended edit. Avoid entering a partial fragment from the middle of a longer word when the whole word is visible: the model has more context when it sees the complete source and destination.

A replacement close in length is easier to fit into a narrow badge, button or single-line headline. A longer phrase can wrap or alter spacing; if exact layout matters, compare at the intended display size or use the source design. Up to five text pairs can be included in one submitted image.

Use Replace text when you know the old and new words. Each pair describes a specific change. Use Translate when the goal is to translate all visible text into a chosen target language. These modes solve different jobs: translation does not require you to supply each translated line, but it does need human review.

One submission processes one image. To make language versions in several languages, choose one language and run a separate edit for each version. Proper names, numbers and small text deserve particular attention in every downloaded image.

After the editor finishes, compare the changed wording and the surrounding objects at full size. Look for missing punctuation, spacing changes, color drift, soft edges and background artifacts. The model aims to preserve the original design, but it is a generative image edit rather than a pixel-locked text layer.

The downloaded file's format and dimensions may differ from the upload. Before printing or adding it to a site, verify the file extension and pixel dimensions against the requirements of the destination. Keep the original image so you can try a different edit without repeatedly editing a generated result.

A first Google sign-in grants one credit once per Google account. That credit can be used for one submitted image with replacement or translation. A card is not required for the first edit. Purchased credits are one-time packs; they do not start a recurring subscription.

Each image submission costs one credit, including an image with multiple replacement pairs. A failed job returns its reserved credit. A completed result that you decide to try again is another submission, so review the image and your instructions before starting.

A typo can survive into a PNG banner, a presentation graphic or a JPG flyer even when the source file is no longer close at hand. Identify the entire word containing the error and enter both the visible spelling and the desired spelling. Preserve punctuation and capitalization if they are part of the design.

Look beyond the corrected letters when the result arrives. An AI edit can change the spacing between characters or the pixels around the text. If the design is a regulated or legally significant document, do not use this workflow; the service's acceptable-use policy excludes official and financial records.

Dates and names are easy to overlook because only a few characters may change. Enter the complete date or complete name as the source text. If a poster contains the date in both a headline and a footer, specify the intended occurrences as separate pairs and check each occurrence in the result.

For promotional graphics, use the real offer approved for that campaign. Editing a number in the image does not change the product listing, booking information or checkout. Update those systems separately so that the visible graphic remains accurate.

A single image may combine a heavy title, a small body line and a handwritten label. Giving each line its own original and replacement pair makes the requested scope clearer than combining different styles into one long instruction. One uploaded image can include up to five pairs.

If a phrase is too long for the space available, consider a shorter approved version. The model may change spacing or wrap a long replacement. Where the shape and layout must be exact, edit the original text layer in the source application instead.

Read the edited text letter by letter. Then compare the surrounding picture: check the edges of a product, lines in a screenshot, a person's face or fine background details near the edit. A generated image may look correct at thumbnail size while a typo or artifact is visible when zoomed in.

Open the downloaded file in the destination workflow and verify its size and extension. A photo editor, web upload form or printer may have different requirements. Keep the source image and document the exact wording that was approved for the final version.

Start with the version of the poster or program that was actually approved, not a screenshot of a compressed preview. Write down the current date, venue and any other field that needs revision. Enter the old date and its replacement as one pair; add a second pair for a name or location if it also changed. The editor accepts up to five pairs for one image.

After the result completes, read the entire image. A date can also appear in small print or in a call to action, and an instruction that changes one occurrence does not automatically update related event systems. Compare the final graphic with the registration page, event time and venue before sharing it.

For a product promotion, start with an image you own or are authorized to edit. Enter the complete visible offer, including any number and symbol, and type the approved replacement. Keeping the full phrase together gives the model context about which badge or headline you intend to update.

A modified picture is not a product database update. Change the actual price and terms separately, and confirm that the product, variant and promotion match what the customer will receive. Review the image at both full size and its final listing size.

For an app screenshot, choose a clear image of your own product or a permitted mockup. Give each label its own text pair so a button label is not confused with a nearby heading. A PNG capture is often useful for interface text because it can preserve sharp edges.

Check the output against the live interface when the screenshot is used as documentation or a store asset. Inspect icons and status bars as well as the requested wording. An image editor changes the screenshot, not the application or its actual behavior.

If the wrong text changes, compare the source pair with the image. Similar phrases can appear more than once; spelling, punctuation, capitalization and spacing help make the request clearer. For a poster, area selection provides an additional coordinate instruction, but it does not act as a locked pixel mask.

If a completed edit does not look right, start again from the original image with a shorter replacement or a clearer instruction. A completed retry uses another credit; a failed job returns its reserved credit. For exact typography, the source design remains the best reference.