How to extract text from an image
Drop a photo, scan or screenshot into the box above, or click Choose files. JPG, PNG, WebP, HEIC and AVIF all work.
Click Extract text. The first time, the recognition models (about 13 MB) are downloaded once and kept in your browser.
Copy the text straight from the page, or download it as a .txt file.
Why extract text from an image?
Retyping a page from a photo is slow and error-prone. Recognising it gives you text to paste into a document, an email or a form in seconds: a quote from a book, a table from a handout, the details on a receipt or a business card.
Screenshots and scans can't be searched or edited. Once the text is extracted, you can search it, translate it and fix it like any other text.
What recognises well, and what doesn't
Printed text works best: documents, book pages, slides, screenshots, receipts and signs. Each line is found first and then read, so a photo taken at a slight angle or a slide with a busy background still reads well. Very small, blurred or curved text lowers the accuracy, so check the result before you rely on it.
Handwriting usually isn't recognised reliably. Layout is kept only as lines: text at the same height, like an item and its price on a receipt, stays on one line, and columns and tables come out as plain text.
Is it safe?
Yes. Image to Text runs entirely in your browser. Your files are read and converted by your own device using built-in browser technology, so they are never uploaded to a server and nobody else can see them.
You can check this yourself: open your browser's developer tools, switch to the Network tab and convert a file. No upload requests are made.
Text recognition uses PaddleOCR, open-source AI models that find each line of text and read it, running in your browser. The models come from our own server; your images and the text found in them are never uploaded.
Files stay in this tab's memory only until you remove them or close the page.
Frequently Asked Questions
Which languages are supported?
Turkish, English and other languages written in the Latin alphabet, such as German, French and Spanish. There's nothing to choose: Turkish letters like ç, ğ, ı, İ, ö, ş and ü are read as they are, even in a text that mixes languages.
Does it read handwriting?
Not reliably. The models are trained on printed text. Neat block capitals sometimes work, but joined-up handwriting usually doesn't.
How can I get better results?
Photograph the page straight on, in good light and in focus, with the text filling most of the picture. A scan or a screenshot gives the best results of all.
Can I extract text from a PDF?
If the PDF has selectable text, open it and copy the text directly, or use PDF to Word. For a scanned PDF, save its pages as images with PDF to JPG first, then extract the text here.
How many images can I process at once?
Up to 500 images, each up to 50 MB. Each one gets its own text file, or download them all together as a ZIP.