Image to text — free OCR in English and Arabic
An image to text converter reads the words in a photo or screenshot and turns them into editable text you can copy anywhere. Carbide's Image to Text tool does exactly that with free OCR in English and Arabic — and it runs entirely in your browser, so your image is never uploaded to a server.
That combination is rare. Almost every OCR site that ranks for this search sends your picture to its own servers first; most also charge, watermark or cap you after a few files. Here is how to extract text in three steps, how to get cleaner results, and where OCR honestly struggles.
Copy text from any photo or screenshot in three steps
The whole job takes under a minute in Image to Text:
- Drop your image in — or tap to choose one. PNG, JPG and WEBP all work, whether it is a phone photo, a screenshot or a scan.
- Pick the language: English, Arabic, or both (the default). Choosing the right one noticeably improves accuracy.
- Press "Extract text" and watch the progress bar. The recognized text appears below the image, ready to copy with one tap.
What result to expect
On the first run the page loads the recognition engine and language data — that one-time wait is the engine, not your image, moving over the network. After that, recognition happens on your device and you get plain, editable text that mirrors what is in the picture, line by line.
From there it is normal text like any other: paste it into an email, a document or a chat, keep it in the online notepad so it auto-saves in your browser, or run it through the character counter if you need a word count before reusing it. A crisp screenshot of typed text usually comes out nearly perfect; a slightly blurry photo of a printed page comes out very usable with a few corrections. If the tool finds no text at all, it says so — try a clearer or higher-contrast image rather than re-running the same one.
OCR that runs in your browser — contracts and IDs never uploaded
This is the part that separates this tool from every big OCR site: the recognition runs inside your browser. Your image is opened, read and discarded on your own device. It is never sent to a server, so there is nothing to be retained, leaked or quietly added to a training set.
Think about what people actually OCR: contracts, ID cards, bank statements, prescriptions, exam papers. The popular converters process all of that server-side — you are trusting a stranger's retention policy with your most sensitive documents just to copy a paragraph. With client-side OCR that trust is not required. It is the same privacy model as Carbide's PDF tools: because no server does the work, there is also no reason for sign-ups, watermarks or daily file limits.
Arabic OCR that actually works
Arabic is where most free OCR tools quietly fail. The script is cursive — letters connect and change shape by position — it runs right-to-left, and dots distinguish otherwise identical letters. Engines tuned only for Latin text return garbage for it, and many OCR sites treat Arabic as an afterthought or a paid add-on.
Image to Text ships Arabic recognition as a first-class option, free. Select Arabic for an Arabic-only document, or keep the default both mode for mixed content like invoices, forms and screenshots that switch between Arabic and English mid-line. For predominantly Arabic pages, picking Arabic alone is usually more accurate than the mixed mode. The extract text from image guide covers the language options in more detail.
Get better results: lighting, straightening and contrast
OCR accuracy is mostly decided before recognition starts — by the picture itself. A few habits make a big difference:
- Light the page evenly and avoid glare; a shadow across a line often costs you that line.
- Shoot straight-on and fill the frame with the text. Tilted or keystoned pages recognize worse.
- Prefer sharp over close: a slightly farther, focused shot beats a close, blurry one.
- Crop away backgrounds, straighten the page and push the contrast in the image editor before running OCR — dark text on a clean light background is the ideal input.
What OCR can't do well — honest expectations
OCR is built for printed, typed text — and on that it is genuinely good. Handwriting is a different problem: expect rough or partial results for cursive notes, and treat anything it does recover as a starting draft. Heavily stylized fonts, decorative Arabic calligraphy, very small type and low-resolution images also drag accuracy down.
Layout is the other limitation. OCR extracts the words, not the design — multi-column pages and tables come out as flowing text, so complex layouts need manual re-formatting after extraction. And always proofread numbers: a misread digit in an IBAN, a phone number or an invoice total is the most expensive kind of OCR error. For anything you will act on, compare the output against the original image before you use it.
From paper or a scanned PDF to editable text
If your text is trapped in a scanned PDF instead of an image, chain two tools: convert the pages to pictures with PDF to Image, then run each page through Image to Text. Both steps stay in your browser, so the document remains private end to end — the same no-upload workflow as the rest of the PDF toolkit.
Starting from paper? Photograph the page flat under even light and OCR the photo. For a cleaner capture, the document scanner in the Carbide app — coming soon for Android and iOS — auto-detects page edges and straightens the shot for you; the scanner apps post covers that whole camera-side workflow.
Frequently asked questions
Is my image uploaded when I extract text from it?
No. Image to Text runs OCR inside your browser — the only thing downloaded is the recognition engine itself, once. Your image is processed on your own device and never sent to any server, which is why it is safe for IDs, contracts and bank documents.
How do I copy text from a picture?
Open Image to Text, drop the picture in, pick the language and press "Extract text". The recognized text appears under the image with a copy button — one tap puts it on your clipboard, ready to paste anywhere.
Does it read Arabic text?
Yes — Arabic recognition is built in and free, not a paid add-on. Choose Arabic for Arabic-only documents or the default both mode for mixed Arabic-English content. That is still rare among free OCR tools.
Can OCR read handwriting?
Only roughly. OCR engines are trained on printed text, so neat block handwriting may partially work but cursive notes usually will not. For handwriting, treat the output as a draft to correct — or retype short notes directly into the notepad.
Is the image to text converter free? Are there limits?
It is completely free: no sign-up, no watermark and no daily file cap. Because the work happens on your device rather than a metered server, there is no reason to limit you — the practical limit is just your browser's memory on very large images.
Copying text from a picture should not cost you an upload, an account or your document's privacy. Drop a photo or screenshot into Image to Text, pick English, Arabic or both, and paste the result wherever it needs to go — and if the source is a scan, PDF to Image gets it OCR-ready first.