Free OCR by Tabaix
How Tabaix OCR works
Tabaix OCR turns pictures of text into text you can copy, edit and search. There is no sign-up, no upload and no install. You open the tool, add your files, and the reading happens on your own device, in your browser.
Step 1 — Add your files
Drag JPG, PNG or WebP images, a PDF, or a DOCX file into the upload area — or use Select Folder to load a whole folder at once. The files are read by your browser only; they are never sent to a server. You can see every file in the list before you run anything, and remove any file you do not want processed.
Step 2 — Pick the language
Auto Detect handles Arabic, Urdu and English on its own. If your page is in one language only, picking it manually usually gives cleaner results — the engine then expects that script and makes fewer guesses. This matters most for Urdu and Arabic, where connected letters and right-to-left text need the right script model. You can also choose the output language for the extracted text.
Step 3 — Run OCR
Press Run Master Process. The tool uses Tesseract.js, an open-source OCR engine, running locally in your browser. It reads each page, finds the words, and shows the extracted text next to a preview of your file so you can check it. For difficult pages — faint print, unusual fonts, dense Urdu or Arabic — you can switch on Cloud Vision mode in Settings with your own Google Cloud Vision API key for a second, cloud-based reading.
Step 4 — Copy or export
Copy the text straight from the results table, or download it. Export formats include TXT, JSON, Markdown, HTML, PDF, DOCX and XLSX. If you processed a batch, turn on “Wrap exports in ZIP” to download everything in one archive that keeps your original file and folder names.
Tips for the best results
- Use clear, straight, well-lit photos. Blurry or tilted pages are the most common cause of errors.
- Crop out unrelated background before uploading — the engine reads only what is in the frame.
- Pick the language manually when a page is in a single script.
- For handwriting, expect mistakes: OCR engines are built for printed text, and handwriting varies too much to read reliably.
- Always proofread important text — names, numbers and dates — before using it anywhere official.
Try it now
The tool is free and ready on the homepage. Try it with an Urdu page, an Arabic document, a scanned PDF or a whole folder at once.