The accuracy of OCR software varies from program to program. Research has been going on in this area for many years, so modern text recognition software is already delivering much better results than it used to. However, free tools usually do not achieve the same hit rate as high-priced professional solutions. A basic judgment is difficult, however, because the source material also plays a major role. While most programs achieve good results with printed black letters in Latin script on a white background, deviations from this ideal template are much more difficult to identify.
East Asian characters, for example, pose major problems even for professional OCR software due to their fine but significant lines. Logos, graphics, special characters, small letters or blurred copies also pose a major challenge to OCR software. Spelling errors in the source material are also a hurdle, since many programs recognize not only individual letters, but entire words.
The greatest variations, even within individual OCR software, occur when reading handwritten text. If the document was written in block letters, the results are better than for a hastily written note in cursive. Overall, OCR image to text converters do not offer one hundred percent certainty and it is best to always check for accuracy.