How the image to text converter works

The short version

You add files. The text gets extracted. Your files get deleted. Everything else on this page is just the details behind those three sentences.

Step by step

1. You add the files

Drag and drop up to 10 images or PDFs onto the upload box, or paste one straight from your clipboard. Each file sits in a queue with its own status, so a batch of nine screenshots and one photo all get processed without you babysitting them.

2. You mask anything sensitive (optional)

Before anything is uploaded, you can open the mask tool on any file. It's a small canvas editor: brush over a name, box out a number, or blur a whole region. The image gets replaced with the masked version, and that's the only version the server ever sees. It's your file, and the tool respects that.

3. The file travels over an encrypted connection

The upload runs over TLS, the same encryption your bank and your email use in transit. Whatever happens next, the file is not sitting on the wire in a form anyone can read.

4. OCR extracts the text

The recognition engine looks at the shapes of the characters — printed text, not handwriting — and turns them into a string of plain text. For multi-page PDFs, that means each page is handled separately and the results are stitched together in order. Things like text direction (both left-to-right and right-to-left) and common languages are handled automatically or picked from the language dropdown.

5. Your files are deleted

This is the part we want to be clear about: files are deleted from the server the moment the task finishes. There's no archive, no cache to warm up, no copy tucked away for training. The extracted text is returned to you, and the source file is gone.

6. You copy or download the result

Each finished file shows its extracted text. Copy it to your clipboard or download it as a .txt file. If a file didn't convert well, the error under it will usually tell you why.

What the converter is good at

In everyday use, the sweet spots are:

• Printed documents, scans, and photos of pages where text is flat and legible
• Screenshots — UI text, error messages, code, chat quotes
• Receipts, invoices, and forms with clean type
• Scanned PDFs that need to be searchable
• Text in 100+ languages, including Arabic, Hebrew, Chinese, Japanese, Korean, Hindi, and Thai

Where it struggles

We'd be lying if we said the result is always perfect. These cause trouble, in roughly this order:

• Handwriting. Printed text is the engine's home turf. Handwritten notes are hit or miss, and heavily cursive handwriting is mostly a miss.
• Angle and perspective. A photo taken from a corner instead of straight-on bends the letters. The more off-axis, the more errors.
• Glare and shadows. Glossy paper, windows behind you, or a hand casting a shadow can split letters into pieces.
• Very low resolution. A screenshot saved as a 400-pixel-wide thumbnail doesn't leave the engine enough pixels to work with.

There's a practical guide with fixes for each of these in the OCR accuracy article.

Practical limits

• Up to 10 files per batch
• Up to 20 MB per file
• Images in 30+ formats, plus PDF (single and multi-page)
• No account, no sign-up, no install

Why the privacy part matters

Encrypted transfer means your file can't be read in transit. Deletion after processing means there's nothing left for us — or anyone else — to look at later. The mask tool means you don't even have to trust that: you can remove sensitive content before the file ever leaves your machine.

If you want the formal version, the privacy page spells out the same thing in slightly more legal-sounding words.