Editable text
Read and correct the recognized text in the result area, then copy it or download a plain TXT file.
Free online OCR for images and scanned PDFs
Upload one PDF or image up to 20 MB. OCR.now extracts editable text, keeps the source visible for review, reports recognition diagnostics, and preserves page boundaries for PDFs of up to 25 pages.
Recognized text
Check names, dates, totals, identifiers, and instructions against the source.
Know what you will get before you upload
The current tool returns reviewable text, not a rebuilt office document. The source, result, and diagnostic signals stay close enough to support a real correction pass.
Reviewable JSON
Download the text with its page count, recognized-word count, and average engine confidence for an inspectable handoff.
Source-led checking
An uploaded image remains visible in your browser so names, digits, punctuation, and uncertain words can be compared with the source.
PDF page boundaries
Multi-page PDF results are labeled by page. This keeps page identity clear even though the output is plain text rather than a new PDF.
| Need | What works now | Important boundary |
|---|---|---|
| Upload | One PDF, JPG, JPEG, PNG, WebP, BMP, or TIFF file | 20 MB per request; the first 25 PDF pages are processed |
| Recognition | Printed text using a language available in the selector | Handwriting, weak scans, mixed scripts, and complex layouts need closer review |
| Export | Copyable text plus TXT and JSON downloads | No DOCX, XLSX, searchable PDF, or layout-preserving export |
| Document structure | Page labels and recognition diagnostics | Tables, form fields, columns, and reading order are not certified automatically |
Start with the document you have
Use the Image to Text tool for photos and scans, or the PDF to Text tool for image-only pages. Both keep the source visible so uncertainty and the next review action remain clear.
- No account
- Source-aware review
- Copy-ready output
One engine, focused document workflows
Each tool page exists only where the source, layout, verification method, or required output materially changes the work.
Image to Text
Turn a document photo or scan into text while keeping the source close for checking.
PDF to Text
Read image-only PDF pages without losing which page produced each block of text.
JPG to Text
Work from the original camera file and catch blur, glare, perspective, and JPEG damage.
PNG to Text
Preserve sharp screenshot and scan pixels while checking transparency and panel order.
WebP to Text
Read a WebP directly and decide whether the available web asset still holds enough detail.
Screenshot to Text
Copy hard-to-select interface text after cropping notifications and private context.
Table to Markdown
Recognize table content, then rebuild headers, cells, blanks, and row relationships deliberately.
Formula Image to LaTeX
Use OCR as a reference while transcribing mathematical structure and checking a rendered result.
Bilingual OCR
Test mixed scripts without translating names or silently dropping minority-language regions.
Long PDF OCR
Divide a long scan into traceable batches and keep weak pages in a correction queue.
Searchable PDF Maker
Prepare checked recognition text before adding a hidden layer in a compatible PDF editor.
Handwriting to Text
Create a cautious draft from clear block handwriting and preserve unresolved words honestly.
Invoice and Receipt OCR
Extract financial text while keeping labels, rows, currencies, and arithmetic under review.
Form to JSON
Recognize form labels and answers before mapping blank, checked, and unreadable states into a schema.
OCR Document Compare
Prepare two reviewed text versions without confusing OCR disagreement with a real edit.
Chart Image to Data
Recover visible chart labels and values while separating source readings from estimates.
OCR Quality Checker
Compare controlled source changes using visible corrections, word counts, and engine confidence.
Business Card to vCard
Recognize card text before assigning person, company, role, and contact fields.
Image Translator
Extract words trapped inside an image as editable text in the language shown.
When OCR output looks wrong, fix the earliest failure
A different converter cannot restore evidence that the camera or scan already removed. Match the visible symptom to the next useful action.
Letters merge or disappear
Blur, glare, compression, or tiny type has removed character detail. Use the original file or retake the source with a controlled accuracy checklist.
Words are right but the order is wrong
Columns, sidebars, or text boxes are being flattened into one stream. Apply the multi-column reading-order method before reusing the text.
Digits and similar letters swap
Characters such as O and 0 or I and 1 can look alike. Check totals, dates, units, and identifiers with the numeric OCR validation guide.
One script works and another fails
The selected language may not match all text on the page. Use the bilingual OCR workflow to separate scripts, or use image to text extraction for a source-faithful editable transcript.
Confidence is a review signal, not a promise.
A fluent paragraph can still contain the wrong name, decimal, date, or identifier. OCR.now reports an engine-level estimate and makes the high-consequence checks explicit.
Understand OCR confidence
A useful OCR result starts before recognition
These checks prevent wasted processing and make the final review smaller, faster, and safer.
- Check for native text first.If words can already be selected and copied accurately, use that text instead of OCR.
- Minimize the document.Crop unrelated pages and fields, then confirm you are authorized to process what remains.
- Use the best source.Choose the original scan or photo, correct rotation, and avoid screenshots of thumbnails.
- Define the risky fields.Decide which names, numbers, dates, totals, units, or instructions must match the source exactly.
A field guide for better OCR decisions
The OCR guide library connects capture quality, document structure, languages, archival quality control, privacy, and downstream verification across eight focused collections.
OCR basics
Understand the recognition pipeline before choosing a file format or trusting fluent output.
3 focused guides
Accuracy lab
Diagnose the earliest failed stage and test one source change on the same difficult passage.
3 focused guides
Document workflows
Choose an output by the structure and verification the next task actually needs.
3 focused guides
Privacy and safety
Decide whether an online upload is appropriate before convenience overrides document risk.
3 focused guides
Capture and preprocessing
Fix focus, glare, perspective, borders, and scan mode before changing recognition settings.
5 focused guides
Layout and structured data
Preserve columns, headers, fields, coordinates, and explicit blank states when position carries meaning.
5 focused guides
Languages and scripts
Map scripts, directions, names, and language changes before selecting recognition passes.
5 focused guides
Archives and quality control
Keep masters, derivatives, metrics, samples, and correction records explainable over time.
5 focused guides
Explore another part of the OCR library
This daily selection opens a different path through tools, practical guides, topic collections, and the standards behind OCR.now. Every public page appears here during each 7-day cycle.
OCR.now recognition methodology
Inspect how recognition, confidence, source review, and current output boundaries are defined.
JPG to Text
Work from the original camera file and catch blur, glare, perspective, and JPEG damage.
Document workflows
Choose an output by the structure and verification the next task actually needs.
How to Improve OCR Accuracy: A Controlled Checklist
Diagnose the first damaged stage, change one variable, and compare the same difficult passage again.
Editorial policy
Review the standards used to hold thin, repetitive, unsupported, or mismatched content.
PNG to Text
Preserve sharp screenshot and scan pixels while checking transparency and panel order.
Privacy and safety
Decide whether an online upload is appropriate before convenience overrides document risk.
Best Scan DPI for OCR: Choose Resolution by Character Size
Choose scan resolution from real character detail instead of treating a larger DPI number as proof.
Privacy policy
Understand temporary upload handling, minimization, and when online OCR is inappropriate.
What makes OCR.now different
Source beside output
Recognition is easier to trust when the page that produced it remains available for comparison.
Consequences before averages
Names, totals, identifiers, and instructions get explicit review guidance even when the page score is high.
Useful boundaries
We state when OCR is the wrong tool, when a native text layer is better, and when sensitive work belongs offline.
Quick answers before you upload
Know the input limits, output boundaries, and review work before sending a document.
What files can OCR.now read?
The live tool accepts PDF, JPG, JPEG, PNG, WebP, BMP, and TIFF files up to 20 MB. A PDF request processes up to 25 pages. For larger documents, use the long PDF review workflow to divide and check the work.
What can I download after OCR?
You can copy the recognized text or download TXT and JSON. The JSON includes the text and recognition diagnostics, not automatically inferred form fields. Compare choices in the OCR output format guide.
Can I paste an image or submit a URL?
No. The current form accepts one local file upload per request. Download an authorized source first, then use the image upload OCR workflow without placing private document details in a URL.
Does OCR.now create a searchable PDF?
Not yet. It returns text and diagnostics, but it does not write a hidden text layer into a new PDF. The searchable PDF preparation guide explains what to verify in a compatible PDF editor.
Does a high confidence score mean the text is correct?
No. Confidence is an engine signal, not proof. Always compare names, dates, totals, identifiers, and instructions with the source. See how to interpret OCR confidence.
Should I upload a private document?
Only when you are allowed to process it online and the possible harm is acceptable. Crop unrelated information first, review the upload and data privacy policy, and use an approved offline workflow when online processing is not appropriate.