How to Extract Text from Images with OCR on Android

OCR — optical character recognition — reads the shapes in an image and returns the characters they represent. Ten years ago it needed a scanner and a desktop. Today it runs comfortably on a mid-range Android phone, offline, in under a second.

Things people actually use OCR for

• Pulling totals, dates and VAT numbers off receipts for expense claims
• Capturing a whiteboard after a meeting instead of copying it into a notebook
• Getting a WiFi password, serial number or IMEI off a sticker without squinting
• Copying a paragraph from a printed book or a paywalled screenshot into notes
• Turning a business card into a contact in seconds
• Extracting a reference number from a letter to paste into a portal
• Reading a foreign-language menu, then sending the text straight into translation

Using the OCR Scanner in Utility Toolbox

1. Open OCR Scanner and either capture with the camera or import an existing photo or screenshot.
2. Wait for the scan overlay to finish — recognition runs locally, so nothing uploads.
3. Review the detected blocks. Text is grouped by region, so you can copy just the line you need.
4. Copy a block, copy everything, or share the result into notes, email or a spreadsheet.
5. Send the text on to Translation or Text Utilities if you need it in another language or a different case.

How to get accurate OCR results

• Light evenly — glare and hard shadows destroy more accuracy than low resolution.
• Keep the text horizontal and the camera square to the page; skew is the biggest cause of garbled output.
• Get close enough that characters are clearly separated, then crop out background clutter.
• Flatten curled receipts and creased pages; a folded line often reads as a stray character.
• Prefer printed text — handwriting recognition is far less reliable, especially cursive.
• For long documents, capture page by page rather than one wide shot of two pages.

Why on-device OCR matters for privacy

The documents worth scanning are usually the sensitive ones: payslips, medical letters, IDs, invoices. Cloud OCR services send that image to a server, and their retention terms vary wildly. Utility Toolbox runs recognition locally on the phone, so the image and the extracted text stay on your device and the tool keeps working in airplane mode.

Where OCR still struggles

Expect imperfect results on stylised logos, dot-matrix printing, faded thermal receipts, low-contrast screenshots and dense tables where column alignment matters. Always check numbers — an OCR mistake in a bank account or invoice total is expensive. Treat OCR as a fast first draft you verify, not a source of truth.