OCR recognition mistakes
Optical character recognition can omit text, insert text, change reading order, confuse punctuation, or substitute similar shapes. Common examples include O and 0, I and 1, S and 5, and missing decimal points.
The Tesseract.js OCR engine returns an automated result. No confidence score proves that a word, amount, date, identifier, or instruction is correct.
Image quality limitations
Blur, glare, shadows, low resolution, compression, curved pages, perspective, decorative fonts, small letters, mixed languages, and complex columns can reduce accuracy. Handwriting is generally less dependable than clear printed text.
Image text enhancement limitations
Crop, brightness, contrast, grayscale, threshold, sharpening, rotation, and inversion can make existing pixels easier to inspect. They cannot recreate characters that are absent, covered, censored, removed, or destroyed by severe blur.
Table reconstruction limitations
The table extractor proposes rows and columns from recognized word positions. It may merge columns, split a cell, miss headers, change reading order, or lose borders, formulas, colors, merged cells, and nested structures.
Review every cell, total, percentage, date, code, and amount before CSV or XLSX export. The image table extractor provides an editor because manual correction is expected.
Important and regulated decisions
Do not rely on unverified OCR output for medical, legal, financial, identity, compliance, employment, academic, emergency, or safety-critical decisions. Retain the original source and seek qualified review when the consequences of an error are significant.
Third-party resources and links
Picture2Txt may link to authoritative documentation for technical context. External websites control their own content, availability, and privacy practices.
User responsibility
Users are responsible for having permission to process the source material, reviewing the output, protecting sensitive information, and deciding whether the result is suitable for its intended use.