How to turn JPG images into text (and why the result is sometimes messy)

You've got a JPG with text in it. Maybe it's a photo of a receipt, a scan of a contract page, or a screenshot your colleague saved as .jpg because that's what the email client did. You want the words out of it, and you don't want to retype them.

That's the job of OCR — optical character recognition — and it's the whole trick behind every image to text converter. This article is about the JPG part specifically, because JPG brings one quirk to the table that PNG, TIFF, and the others don't.

What "jpg to text" actually does

OCR software looks at the shapes of the letters in the image and converts them into plain, editable text. It doesn't understand the document. It doesn't know a receipt from a poem. It just recognizes character shapes and passes them back to you in order.

That's why the output can read like a robot transcribed it — because, well, a robot did. The words are usually all there. The layout, the bold headings, the nice formatting? Those don't survive. You get a string of text, and that's the deal.

Why JPG specifically can be tricky

JPG is a lossy format. Every time a photo is saved as a JPG, the encoder throws away detail it thinks you won't notice — usually fine for looking at a photo on a screen, less fine for a computer trying to trace the exact outline of a letter.

At normal quality settings the effect is small. At low quality, you get visible artifacts around text: little smudges, halos, and blocky noise that sit exactly where the OCR engine needs a clean edge. A letter with fuzzy edges can be read as the wrong letter, or split into two, or not read at all.

So the rule of thumb: JPG is fine for converting to text, as long as the JPG itself is good. Screenshots should really be PNG or WEBP, but that's a separate complaint below.

The three things that mess up results

In my experience watching people use the converter, these three causes cover most failed or messy conversions.

1. Not enough resolution

OCR needs pixels. A 400-pixel-wide photo of a full page of text means each letter gets maybe 15 pixels to work with — not enough for a computer to distinguish an "e" from a "c". More resolution almost always means better recognition, up to a point of diminishing returns.

2. The photo was taken at an angle

Photograph a document from a corner and the letters bend into perspective. The engine can cope with a small angle. A steep angle turns clean text into a guessing game. The flatter and more head-on the photo, the better.

3. Glare, shadows, and compression noise

Glossy paper under a lamp, a window behind your shoulder, or a JPG saved at "web quality" — all of these add noise around the letters. Noise is what OCR mistakes are made of.

How to set up your JPG so the text comes out clean

None of this takes special equipment. It takes ten seconds of attention before you hit the shutter.

• Shoot straight-on. Hold the phone parallel to the paper, not from an angle. The built-in camera grid helps here.
• Use the light. Daylight or a lamp shining on the paper, not behind you. No window glare on glossy pages.
• Get close. Fill the frame with the text. Cropping a smaller area to the converter often works better than a wide shot of the whole desk.
• Keep the JPG quality high. When saving or exporting, use high quality (usually 90%+) if you get the choice.
• For screenshots, prefer PNG. A screenshot of UI text is flat colors on solid backgrounds, and PNG keeps those edges razor sharp. JPG screenshots work, but they're the worst-case input for OCR.
• If scanning, aim for 300 DPI. That's the standard for documents and it's plenty for OCR. More than that rarely helps.

When "good enough" is actually good enough

Here's the honest part. If you need the text for search, for quotes, for pasting into notes, or for reference — a conversion with a few stray characters wrong is usually fine. You'll read the result and the errors will be obvious.

If you're turning the result into a legal document or publishing it, proofread against the original. No converter I know is perfect, and the ones that claim to be are lying.

There's a whole separate article about why OCR accuracy varies between images and the fixes that genuinely help. If you've read this far, that one's worth your time too.

The short version

JPG to text works. The format isn't the problem — image quality is. Flat, straight, well-lit, reasonably high-resolution JPGs convert well. Everything else converts less well. And if the file is a screenshot, save it as PNG next time and thank yourself later.

Want to test it? The converter on the homepage takes JPG, PNG, WEBP, and about thirty other formats, up to ten files at once. Files are deleted from the server right after the task finishes.