Sources & licenses
AI Text Scanner credits & licenses
The AI Text Scanner was built for LoschaZ. The text recognition runs in your browser with open-source software and open language models:
What we built
The workspace, adding images by drag and drop, file picker, camera and clipboard, the previews with rotation and area selection, the image preparation (contrast boost, dark background inversion, enlarging), the line joining and hyphen repair, the confidence summary, the text export and the packaging of the engine on this website were written for LoschaZ.
Components & licenses
- tesseract.js 7.0.0 – Apache License 2.0 (license). Its browser files include small helpers named in their notices (main file, worker): regenerator-runtime (MIT), buffer (MIT), ieee754 (BSD 3-Clause) and zlib.js (MIT). Served from this website.
- tesseract.js-core 7.0.0, the Tesseract OCR engine compiled to WebAssembly – Apache License 2.0 (license). Tesseract is used in the modified version described in the tesseract.js-core README. The engine files also contain the image libraries Tesseract is built with: Leptonica (BSD 2-Clause style), libpng (libpng license), zlib (zlib license), libjpeg (IJG license), libtiff (libtiff license), giflib (MIT) and libwebp (BSD 3-Clause). Served from this website.
- Language models for English, German, French and Spanish from tessdata_best (integer version) – Apache License 2.0, packaged by the tesseract.js project as @tesseract.js-data 1.0.0 (MIT) (notice). Served from this website.
Our changes
tesseract.js, the engine and the language models are used unchanged. We only removed the source map comments at the end of tesseract.esm.min.js and worker.min.js, because the source maps are not published.