Local text recognition: PaddleOCR.js 0.4.2 (PaddlePaddle Authors, Apache-2.0), PP-OCRv6 tiny and en_PP-OCRv5 mobile recognition models (PaddlePaddle), OpenCV.js 4.10.0-release.1 (Apache-2.0), ONNX Runtime Web 1.24.3 (Microsoft, MIT), js-yaml 4.3.2 (MIT), Clipper 6.4.2 (Angus Johnson and Timo, Boost 1.0), JSBN (Tom Wu).

Modified OpenCV bindings: four dynamic Function wrappers replaced with equivalent static closures, to retain the site Content Security Policy. Inference uses the WASM-only ONNX build. The recognition models and recognition algorithms are unmodified. See BUILD.json and SOURCE.json.