HOW IT WORKS
Image to text in three steps
- Choose one clear image, up to 20 MB and 20 million pixels.
- Rotate upright, select a language and press Recognize text.
- Compare text with the preview, then copy or download TXT.
ON-DEVICE PROCESSING
Image to text: local Vietnamese and English OCR
Choose one clear PNG/JPG/WebP image, rotate it upright and select Vietnamese or English. VietPDF runs OCR in your browser and produces copyable text or TXT; verify accents and numbers against the source.
Source image
PNG, JPG, WebP · up to 20 MBYour image stays on this device. The recognition engine downloads on first use.
Recognized text
Check accents and numbers before usePRACTICAL GUIDE
What to know before you image to text
OCR turns visible character shapes into text rather than reading an existing text layer. This tool reuses VietPDF’s Tesseract.js engine, keeps the source preview available and lets you copy or download plain text without an account.
Printed notes
Extract text from a clear photo of printed material, then verify accents.
One region
Crop outside the tool to recognize only the needed region and reduce noise and memory.
Text preparation
Create editable plain text from material you are entitled to process.
HOW THE PDF CHANGES
Technical limits to understand
Upright, clear images
Straight, well-lit print on a simple background is easier to check. Rotation applies before OCR without modifying the original.
Language models
Choose Vietnamese, English or both. Models download when needed, so first use requires connectivity; this is not a fully bundled offline tool.
Memory bounds
20 MB and 20 million pixels are ceilings, not RAM guarantees. Decoded bitmaps and canvases use more memory than compressed files. Resize on low-memory devices.
TROUBLESHOOTING
If the result is not what you expected
No text found?
Try a clearer upright image, check the language and wait for models to download. Handwriting is not guaranteed.
Broken reading order?
Multiple columns can confuse OCR. Crop and recognize each region separately; review before cleaning line breaks.
FAQ
Frequently asked questions
Is OCR perfectly accurate?
No. An image saying “Quantity: 15” may produce a wrong character or number. Check important fields, names and amounts manually.
Is my image uploaded?
No image upload occurs. Your browser downloads OCR engines and language models through the existing OCR infrastructure, then recognizes locally. Recognized text is never sent to analytics.
Can I cancel or rotate the image?
Yes. Select 0/90/180/270 before running, or press Cancel to stop. Changing the image, angle or language clears the previous result to prevent confusion.