Shot2Text

Picture to Text Converter — Free, Copy the Words Out

Picture to text, whatever kind of picture it is. Photos, screenshots, saved images, exports from an app — if there are words in the picture, this pulls them out as plain text you can copy and edit.


Only the language you pick is downloaded.

What is a picture to text converter?

People say picture when they mean a screenshot, a photo, a downloaded image, or a frame grabbed from a video. Those four things behave very differently when you try to read text out of them, so it is worth knowing which one you have.

A screenshot is software-drawn: clean edges, flat colour, no noise. It reads almost perfectly. A photo has perspective, shadows and sensor noise, and needs correction before it will read well. A downloaded image has usually been recompressed once or twice on its way to you, which softens the letters. A video frame is a photo with motion blur on top.

This converter handles all four by looking at the image rather than trusting the file extension: it flattens transparency onto white, stretches contrast on washed-out images, and upscales anything too small to read. You do not need to tell it what kind of picture you have.

Which kinds of picture read well?

  • Screenshots — excellent. The cleanest case. Crop to the text and it will read almost perfectly, including code and numbers.
  • Phone photos — good, if shot straight. Contrast correction handles the washed-out grey. Tilt and shadow are what ruin it, not the format.
  • Downloaded images — fair to good. Depends how many times it was recompressed. A PNG from a website reads well; a JPEG forwarded five times does not.
  • Video frames — variable. Pause on the frame and screenshot it rather than photographing the screen. A screenshot of a paused video is far cleaner.
  • Memes and subtitle images — often good. Bold text with a heavy outline reads well, which is exactly how subtitles are drawn.
  • Scanned images — depends on dpi. 300 dpi reads very well. 100 dpi usually does not, because each character is only a few pixels tall.

Why won't my picture read?

In rough order of how often it happens.

  • Text is a small part of a large picture The reader looks at the whole image, and most of it is not text. Small letters get averaged away. Crop to the region with the words. Cropping is the single biggest improvement you can make.
  • Stylised or decorative fonts The reader was trained on common text faces. Script fonts, heavy display faces and all-caps logos are guesses. Try anyway — bold simple shapes often survive. Expect errors on very decorative lettering.
  • Text with an outline or drop shadow The outline looks like an extra stroke and gets read as a second character. Crop tighter so the outline is smaller relative to the letters, or accept the extra characters and delete them.
  • Very low resolution Below roughly 12 pixels per character there is nothing but noise to work with. Find a larger version of the image. Upscaling cannot invent detail that was never saved.
  • Text on a patterned background The reader cannot separate letters from texture when both are mid-tones. Crop to the text block, or export a version with a plain background.
  • Multi-column layouts The reader follows lines across the page, so it merges the ends of two columns into one line. Crop one column at a time and convert each separately.

What do you get back?

Plain text in reading order with line breaks. No font information, no colours, no images — just the words, ready to paste into a document, a spreadsheet or a search box.

Multi-column layouts come out interleaved, because the reader works line by line across the full width. If the picture has columns, convert them one at a time.

Everything stays in the page. There is no download, no history and no account — you copy what you need and move on.

Is my picture uploaded anywhere?

The picture is decoded and read inside your browser tab. It is not sent to a server, not queued, and not stored.

Pictures people want to read often contain more than they remember: a notification banner, a name in a corner, a second window behind the one they meant to capture. Keeping the reading local means none of that travels.

Questions

What picture formats does it accept?
JPG, PNG, WebP, GIF and BMP — essentially anything a browser can display. HEIC from an iPhone usually works in Safari; if it does not, set your camera to 'Most Compatible' or export the photo as JPG.
Why does the text come out in the wrong order?
Almost always a multi-column layout. The reader works line by line across the full width of the image, so it reads the end of the left column and the start of the right column as one line. Crop one column at a time and convert each separately.
Can it read text from an image with a patterned background?
Poorly. When the letters and the background sit at similar brightness, the reader cannot tell them apart and fills in noise. Cropping tightly around the text block helps more than any other single change.
Does it work on memes and subtitle images?
Usually yes, and better than you would expect. Subtitle text is deliberately bold with a heavy outline, which is close to ideal for a reader. The outline sometimes gets picked up as extra characters — delete them afterwards.
Can it read a picture of a screen?
It can, but a screenshot is always better. Photographing a screen adds moiré patterns, reflections and a colour cast, all of which a screenshot simply does not have. If you can capture the screen instead of photographing it, do that.
Is the picture uploaded to a server?
No. The reading happens entirely in your browser tab. Nothing is transmitted and nothing is stored. You can verify with your browser's network panel: no request carries your image.
Can it read several pictures at once?
Not on this page — one at a time. Doing them in sequence is fast, because the English text data stays cached after the first conversion.
Does it support languages other than English?
18 languages, chosen from the dropdown above the tool. Only the selected pack is downloaded, so this costs nothing until you need it.