How to extract text from an image using OCR

What OCR is and how it works

OCR software analyses the shapes of letters and words in an image and converts them into editable, selectable text. Modern on-device OCR engines handle printed text in dozens of languages, and many cope reasonably well with neat handwriting. The key insight is that OCR is pattern recognition: the cleaner and higher-contrast the input image, the more patterns the engine can match confidently. Carbide's Image to Text tool runs the entire recognition process on your device — nothing leaves your phone — so it works offline and keeps your documents private.

Getting accurate results: lighting, focus and contrast

Three factors control OCR accuracy more than anything else. First, lighting: shoot in even, diffuse light (near a window on an overcast day is ideal) and avoid harsh shadows cutting across the text. Second, focus: tap the text area on your camera screen to lock focus before capturing. Third, contrast: dark ink on white paper scans best; if the background and text are similar in tone, adjust your phone's exposure slider before shooting. Holding the phone parallel to the page — not at an angle — also removes perspective distortion that can confuse the recogniser.

Scanning multi-page documents

When you need text from several pages, Carbide's Document Scanner is the right starting point. It detects the edges of each page automatically and applies a perspective correction and contrast boost so every page arrives as a clean, flat image. Work through the pages one at a time in good light, confirm the edge overlay looks right before capturing each page, then pass the result to Image to Text for recognition. Both tools are free, work completely offline, and never upload your files anywhere.

What to do with the extracted text

Once Image to Text has recognised your text, copy it directly to the clipboard and paste it into any app — a notes tool, a translation service, a spreadsheet, or an email draft. For longer documents, paste into a plain-text editor so you can search and correct any recognition errors before using the content. Common error patterns: 0 mistaken for O, 1 for l or I, rn for m. A quick read-through catches most of them. Because Carbide stores nothing and sends nothing, you can safely extract text from sensitive documents such as passports, contracts, or medical records.

Frequently asked questions

Does OCR work on handwritten text?

It depends on how neat the handwriting is. Printed block letters and neat cursive often recognise well; loose or heavily stylised handwriting is much harder for current engines. For best results with handwriting, shoot in strong, even light and keep the text as large in the frame as possible.

Why is my extracted text full of errors?

The most common causes are a blurry photo, uneven lighting that creates shadows, or low contrast between the text and background. Retake the image in better light, tap to focus on the text, and make sure the page fills most of the frame. Angled shots also introduce distortion — hold the camera directly above the page.

Does Carbide's Image to Text send my photos to a server?

No. Carbide runs entirely on your device. Your images are processed locally and never uploaded to any server. This means the tool also works with no internet connection, which is useful when you are travelling or have limited data.

Can I extract text from a PDF or a screenshot?

Image to Text works on any image file — screenshots, photos, and exported PDF page images all work. If you have a multi-page PDF, export each page as an image first, then run Image to Text on each one. Carbide's Document Scanner can also prepare physical pages for recognition by correcting perspective and boosting contrast automatically.