How to Change Text in a Flat Image

You can change text in a JPG, PNG, screenshot, or other flat image, but you cannot recover the original text layer from the pixels alone. The practical job is reconstruction: remove or cover the old lettering, recreate the missing background, and add new live text that matches closely enough for the intended use.

The right method depends on the background, the font, and how exact the result must be. A simple label on a solid color can take a minute. Text over a face, product label, chart, or textured photograph may require careful retouching or a rebuild from the source.

First, make sure the source file is really gone

Before editing pixels, look for the original Canva, Figma, PowerPoint, Photoshop, Illustrator, or Moda file. Check shared drives, email attachments, export folders, version history, and the person who made the asset.

An original design file is better because it preserves the actual wording, font, spacing, layers, and background. OCR can recognize characters and software can estimate a typeface, but neither recreates information that was discarded during export.

If the original cannot be found, keep an untouched copy of the flat image. Work on a duplicate so you can compare the repair and start again if necessary.

Choose the method based on what sits behind the text

Situation	Best starting method	Main limitation
Solid color or simple gradient	Cover old text, then add a new text box	Font and spacing still need matching
Repeating or uncomplicated texture	Clone, heal, or use content-aware fill, then add text	Retouching can repeat obvious patterns
Detailed photograph	Select the lettering and reconstruct the background	Generated or sampled detail may be inaccurate
Designed graphic with several labels and shapes	Rebuild the needed area in a layout tool, or try layer decomposition	Automatic decomposition is an estimate, not recovered source data
Legal, financial, medical, or evidentiary image	Return to the authoritative source	A visually plausible repair is not a trustworthy record

Method 1: cover simple text and add a new text layer

For text on a flat color, do not overcomplicate the repair.

Sample the background color near the text.
Draw a shape over the old wording. Make it large enough to cover antialiasing and shadows around the letters.
Add the replacement as a real text box above that shape.
Match the typeface, weight, size, line height, letter spacing, alignment, and color.
Zoom out to the size at which the image will actually be used. A tiny mismatch that is obvious at 800% may disappear in a social thumbnail, while poor spacing can remain obvious at any size.

This method works in layout tools such as Canva, Figma, PowerPoint, or Moda, as well as image editors. It keeps the replacement editable even though the rest of the image remains flat.

If your main problem is placing new copy clearly rather than matching an old design, use the broader guide to putting text on a photo.

Method 2: remove the old text, then rebuild the background

When the lettering crosses a photograph or texture, covering it with a solid rectangle will look wrong. Remove the text first, then add the replacement.

In Photoshop, select the old lettering with a small margin and use Content-Aware Fill. Adobe lets you refine the sampling area and preview the fill before applying it. Clone and healing tools offer more manual control when the automatic result repeats a pattern or softens an edge.

Canva's Magic Eraser and other object-removal tools follow a similar basic idea: mark what should disappear, then let the software reconstruct the gap. Generative tools can also fill a selected area from a prompt.

Inspect the result closely. Background reconstruction invents or copies pixels. It may bend straight lines, repeat texture, alter a face, or change a product detail. If the covered area contains information that must remain exact, locate the original background or rebuild the region manually.

Method 3: use OCR as a reference, not as recovered typography

OCR can save retyping when an image contains a lot of copy. It identifies likely characters and sometimes text regions, but its output is plain recognized text. It does not know the original font file, kerning, line breaks, effects, or layer structure.

Use OCR to extract the wording, proofread every character, then create a new text layer. Check numbers, punctuation, accented characters, and similar-looking characters such as `0` and `O`. For a chart or price, compare against the authoritative data rather than trusting recognition.

Matching the appearance is a separate job. Photoshop's Match Font can suggest fonts based on selected text, and browser font-identification services can provide leads. A suggestion is not proof of the original font. If the typeface is licensed or unavailable, use an approved substitute and expect spacing to change.

Method 4: reconstruct editable pieces from a designed image

Some tools try to separate a flat design into text, shapes, and background regions. This can be useful for posters, ads, slides, and social graphics with clean boundaries. Treat the result as a starting point.

Check every recovered word, font, color, edge, and layer. A decomposition tool may split one object into several pieces, merge overlapping elements, or approximate text and shapes. It cannot prove what the original file contained.

Moda can help when the goal is to turn a flat designed asset into a canvas you can continue laying out. That is different from detailed photo retouching, where Photoshop, Affinity, GIMP, or another dedicated image editor is usually the better tool. It is also different from recovering the lost original. The output is a reconstruction that needs review.

For a wider comparison of editing tools, see Photoshop alternatives by editing task.

How to make the replacement look intentional

Start with the text box, not the font name. Match the width, number of lines, alignment, and position first. Then refine the typeface, weight, and spacing.

Look for effects around the original lettering:

A shadow or glow can leave a visible halo after the letters are removed.
Compressed JPEG edges can make perfectly sharp replacement text look pasted on. A very small amount of matching softness may be appropriate.
Text drawn in perspective needs a transform, not only a different font size.
Curved, warped, outlined, or textured display lettering may be faster to rebuild as artwork than as ordinary live text.

Export once at the final dimensions and inspect that file. Scaling or compression can change thin type and small details.

When you should not change the text

Only edit images you own or have permission to modify. Do not alter someone else's work to imply endorsement, copy a competitor's ad, remove attribution, or disguise the origin of a design.

Do not rewrite screenshots of conversations, receipts, financial statements, medical records, news images, certificates, dashboards, or other records in a way that could deceive someone. For a mockup or training example, label it clearly as a recreation or sample. For privacy, redact sensitive information visibly or replace the entire example with synthetic data instead of making a false record look authentic.

Watermarks and rights-management information are not ordinary unwanted text. Obtain a licensed clean copy from the owner rather than removing them.

The honest expectation

A flat image can be repaired, but it cannot be unflattened with perfect certainty. Simple backgrounds and common typefaces often produce convincing results. Complex scenes, unusual lettering, and files that must remain authoritative call for the original source or a deliberate rebuild.

The safest workflow is straightforward: preserve the original image, remove the old text without damaging meaningful detail, add a new editable text layer, and compare the final export at its real use size.