Last updated: October 2026
Edit text on image online free — fix typos without Photoshop
Short answer: upload any photo, click Scan words, then click any highlighted word and type the replacement. The tool erases the original word and repaints your text automatically matched to the original's size, color, bold/italic style and font family — then you download a full-resolution PNG. No signup, no upload, no watermark.
Everyone has been there: a screenshot with a typo, a meme with the wrong caption, a product photo with last year's price, a presentation slide with a misspelled name. The old fix was re-creating the whole graphic. This tool takes the other path: it finds the text that's already in the image with on-device OCR, lets you click any word, and paints your replacement back into the photo so it looks like it was always there.
How the automatic style matching works
Short answer: for each word the tool segments the actual glyph pixels from the background, measures the ink height for size, reads the dominant glyph color, compares stroke density against a clean render to detect bold, measures glyph slant to detect italic, and shape-matches the font family (serif, sans or mono). Every attribute can still be overridden by hand.
Matching "looks right" is five separate problems. Size comes from the true ink height of the original word — not the OCR box, which includes padding. Color is read from the glyph pixels themselves after thresholding the word box, so a colored word on a colored background doesn't get contaminated by the background. Bold is detected by comparing the word's ink density against the same word rendered in a regular weight — bold strokes pack measurably more ink at the same height. Italic is detected from the slant of the glyphs' vertical profile. Font family is shape-matched against built-in fonts and mapped to the closest one: a clean sans, an elegant serif, or a monospace.
One honest boundary: pixels don't carry font data the way a PDF's text layer does, so the match is approximate — excellent on plain printed text (receipts, screenshots, signs, documents), rougher on decorative typography. That's why every attribute stays manually adjustable: font, size from 50–200%, any color, bold and italic toggles, one click back to auto.
When it works best — and when it doesn't
Short answer: best on straight, high-contrast printed text with flat backgrounds — screenshots, memes, signs, labels. It struggles with skewed or tiny text, handwriting, and busy textured backgrounds, where the erase patch can show.
The text detector wants what OCR always wants: straight lines, sharp edges, good contrast. A meme caption, a UI screenshot, a photographed shop sign — all ideal. What breaks it: text shot at a steep angle, motion-blurred or tiny text, cursive handwriting (the engine is trained on print), and artistic warped text. On the repaint side, the erase fills the word's box with the sampled background color and feather-blends the edges — invisible on flat backgrounds, potentially a faint rectangle on gradients, textures or busy photos. If your image is the hard case, flat and frontal always wins.
Frequently asked questions
How do I edit text on an image online for free?
Drop your photo into the box above (or tap Load sample image), then click Scan words. Every detected word gets a clickable highlight — click one, type the replacement, and hit Apply. The tool erases the original word and paints your text back in, matched to the original style. Download the result as a PNG.
Will the replacement text match the original font?
It gets close automatically: the tool measures the original word's size, samples its exact color, detects bold vs regular and italic vs upright, and matches the font family (serif, sans-serif or monospace) against built-in fonts. It cannot clone the exact font file — pixels don't carry font data the way PDFs do — so on decorative or unusual typography the match is approximate. You can override font, size, color, bold and italic manually for any word.
What kind of images work best?
Straight, high-contrast printed text: screenshots, memes, signs, labels, receipts, presentation slides. The same limits every tool in this class has apply — heavily skewed or rotated text, tiny blurry text, handwriting, and text on busy textured backgrounds detect poorly or repaint visibly.
Can it edit handwriting?
No. The recognition engine is trained on printed text, so handwriting usually comes out garbled and can't be reliably boxed for editing. Neat block capitals sometimes work.
Is my photo uploaded to a server?
No. The OCR engine downloads to your browser once and runs entirely on your device — your image never leaves your device, is never sent to a server, and is never stored.
Which languages can it read?
English, with the OCR engine running locally in your browser. The first scan downloads the engine and language data once (a few MB), then it is cached and later scans start faster.
Can I change the font or color myself?
Yes. After clicking a word, the editor panel shows the auto-detected style and lets you override everything: font family, size (50–200%), color picker, and bold/italic toggles. Reset to auto anytime with one click.
What happens on textured or gradient backgrounds?
The tool fills the word's box with the sampled background color and feather-blends the edges, which is invisible on flat backgrounds. On textured, gradient or busy photographic backgrounds the patch can show as a faint rectangle — flat, high-contrast backgrounds give the cleanest results.
Does it work on my phone?
Yes. The upload uses your phone's native file picker, word boxes are tap targets, and the editor panel is fully usable on small screens. Large photos are processed at full resolution; the scan itself may take a little longer on older phones.
What file do I download?
A PNG at your image's full original resolution, with all edits baked in. Edits are non-destructive while you work — the original pixels are kept in memory, so you can re-edit or clear any word before downloading.
