Recognize printed English or Chinese text from a local PNG.
Your input
Outputs are cleared when the input changes.
Result
Enter your data or use the example, then run the tool.
Worked example
The generated sample reads ORCHARD 123. Use a clear, upright image and select the matching language model; inspect mistakes before reusing the text.
Method
Pinned Tesseract.js uses a dedicated, terminable OCR worker, fixed Tesseract.js-core 7.0.0 and official integer-quantized tessdata_fast models. Only same-origin code/model assets load; image pixels remain local. A confidence score is a model estimate, not proof of correctness.
Limits
Static PNG ≤5 MiB / 2 million pixels, no APNG/PDF/JPEG input. First use downloads a core variant of about 3.9 MB plus a language model of 1.6–2.0 MB. Fixed fast models trade accuracy for size/speed; handwriting, layout, tables and mixed scripts are not guaranteed. OCR ends after 45 seconds or cancellation, with no cloud OCR or remote input processing.
All input processing stays in this browser; only fixed same-origin runtime assets are fetched, with no input or logs sent to analytics.
Your inputs stay on this device. Only optional, consented page visits are measured. Input values, files, filenames and processing logs are not sent to Analytics.
When switching languages, text, settings and selected files are temporarily kept on this device. Transfers are removed after restoring; interrupted transfers are cleaned up on a later visit. Inputs and files are never uploaded.