Choose
Select, drop, or paste one JPG, PNG, or static WebP. The file contents and decoded dimensions are checked before any OCR resource loads.
Kivoza text tools · OCR
Read printed text from an image.
Recognize printed English or Simplified Chinese in this browser tab. Your image stays on your device, and the result remains editable.
Local processing
No upload
EN + 中文
Bring one image
Choose an image.
JPG, PNG, or static WebP. Recognition resources load only when you start.
Drop an image here
or browse your device. You can also paste an image.
One static image · 20 MB maximum · 40 megapixels maximum · printed text only
Editable output
Recognized text
A short path
Choose, recognize, refine.
Recognize
Pick English, Simplified Chinese, or both. The OCR worker and selected language data run locally in a Web Worker.
Refine
Edit the text, copy it to the clipboard, or download a plain TXT file. Recognition is a starting point for printed text, not a promise of perfect transcription.
Know the edges
Sized for one browser tab.
The limits keep image decoding and WebAssembly memory bounded. Large, low-quality, tilted, or visually complex images can take longer or produce less reliable text.
20 MBmaximum input size
40 MPmaximum decoded pixels
1 imageper recognition run
What is supported?
Static JPG, PNG, and WebP images with printed text. The browser must support Web Workers, WebAssembly, and the selected image decoder.
What is not promised?
This first version does not claim handwriting, table structure, receipt fields, PDF OCR, layout reconstruction, or 100% accuracy.
Why can text be wrong?
Blur, glare, tilt, small type, unusual fonts, mixed backgrounds, and low contrast can reduce recognition quality. Review and edit the result.
A real workflow
Turn a printed note into editable text.
note.png
The preview reports the file name, bytes, and decoded dimensions. The image is read by this tab and is never sent to an OCR API.
note.txt
The recognized text remains in an editable field. Correct names, punctuation, and line breaks before copying or downloading.
Frequently asked questions
Are my images uploaded?
No. Validation, preview decoding, OCR worker execution, language data loading, and TXT creation happen in this browser tab. The app has no image upload or remote OCR endpoint.
Why does the first recognition take longer?
The selected OCR worker and language data are loaded on the first run and may be cached by this browser. Later runs can reuse the cached resources. The page reports stages instead of pretending that a percentage equals recognition accuracy.
Can it read handwriting or a PDF?
No promise is made for handwriting, PDF OCR, tables, forms, or structured extraction. Use a clear printed image and review the editable result.
What happens if recognition fails?
The page keeps the image selected, reports the error, and offers Recognize again. You can change the language, replace the image, or clear the tool without uploading anything.