Does my image get uploaded to read the text?

No. Every other free image-to-text site sends your picture to a server to recognise it — and people photograph exactly the documents that should not be sent anywhere: IDs, prescriptions, payslips, contracts. Here the recognition engine is downloaded into your browser (a one-time fetch from this site, cached afterwards) and runs on your own device. The image never leaves your computer.

What kind of images does it work well on?

Anything where the text is reasonably large and in focus: a photo of a page, a screenshot, a scan, a whiteboard. It handles a slightly skewed, unevenly lit phone photo well. What it needs is resolution — the text has to be big enough in the image itself, which matters more than the file size.

What if it finds no text at all?

You get a message saying so, and no file. That is deliberate: when text is too small or too blurry, the recognition engine returns nothing rather than an error, so a tool that is not careful hands you an empty text file and calls it a success. We would rather tell you it did not work and why. Usually the fix is a higher-resolution image — the original file rather than a screenshot of it.

What does the confidence percentage mean?

It is the recognition engine's own reported confidence in what it read — not a measure of how correct the result actually is. We show it verbatim because it is a genuinely useful signal: in our testing it tracked quality closely, dropping sharply on images that produced poor results. But it is not an accuracy score, so we never call it one. Below 60% we show a warning telling you to check the text against the image.

Which languages can it read?

English, French, German, Spanish, Italian, Portuguese, and Dutch — pick the one your text is in before you start, since it changes how the engine reads the characters. Each language model is served from this site rather than a third-party CDN, so not even the language you chose is revealed to anyone else.

Which file types can I use?

JPG, PNG, WebP, GIF, and BMP. HEIC (the default on iPhones) and TIFF are not supported yet, and you will be told by name rather than left with a failure — on an iPhone you can share the photo as JPG, or use a screenshot of it if the text is large enough.

Not this tool — it takes an image. For a scanned PDF use OCR PDF, which does the same recognition across every page and can also give you back a searchable PDF. If your PDF already has selectable text, PDF to Text extracts it directly, which is faster and exact.

image

Image to Text

Read the text out of a JPG, PNG, WebP, GIF, or BMP and get it back as plain text you can copy or download. The image is never uploaded — text recognition runs entirely inside your browser, on your device, so a photo of an ID, a payslip, or a contract stays on your computer. Free, no account, no email.

Frequently asked questions

Does my image get uploaded to read the text?: No. Every other free image-to-text site sends your picture to a server to recognise it — and people photograph exactly the documents that should not be sent anywhere: IDs, prescriptions, payslips, contracts. Here the recognition engine is downloaded into your browser (a one-time fetch from this site, cached afterwards) and runs on your own device. The image never leaves your computer.
What kind of images does it work well on?: Anything where the text is reasonably large and in focus: a photo of a page, a screenshot, a scan, a whiteboard. It handles a slightly skewed, unevenly lit phone photo well. What it needs is resolution — the text has to be big enough in the image itself, which matters more than the file size.
What if it finds no text at all?: You get a message saying so, and no file. That is deliberate: when text is too small or too blurry, the recognition engine returns nothing rather than an error, so a tool that is not careful hands you an empty text file and calls it a success. We would rather tell you it did not work and why. Usually the fix is a higher-resolution image — the original file rather than a screenshot of it.
What does the confidence percentage mean?: It is the recognition engine's own reported confidence in what it read — not a measure of how correct the result actually is. We show it verbatim because it is a genuinely useful signal: in our testing it tracked quality closely, dropping sharply on images that produced poor results. But it is not an accuracy score, so we never call it one. Below 60% we show a warning telling you to check the text against the image.
Which languages can it read?: English, French, German, Spanish, Italian, Portuguese, and Dutch — pick the one your text is in before you start, since it changes how the engine reads the characters. Each language model is served from this site rather than a third-party CDN, so not even the language you chose is revealed to anyone else.
Which file types can I use?: JPG, PNG, WebP, GIF, and BMP. HEIC (the default on iPhones) and TIFF are not supported yet, and you will be told by name rather than left with a failure — on an iPhone you can share the photo as JPG, or use a screenshot of it if the text is large enough.
Can it read a PDF?: Not this tool — it takes an image. For a scanned PDF use OCR PDF, which does the same recognition across every page and can also give you back a searchable PDF. If your PDF already has selectable text, PDF to Text extracts it directly, which is faster and exact.

Frequently asked questions

Related tools