This research proposes a fully client-side system for digitising and summarising handwritten notes that integrate optical character recognition (OCR) with advanced natural language generation models. Initial experimentation with a Convolutional Neural Network (CNN)– based handwritten character recognition system revealed significant challenges, including low recognition accuracy, strong dependence on dataset quality, and poor generalisation across diverse handwriting styles. A subsequent implementation using Gemini 2.5 Pro provided high-quality English summaries but failed to deliver equivalent performance for multilingual academic content. To overcome these issues, this study adopts a hybrid methodology combining Tesseract.js OCR for multilingual text extraction with Gemini 2.5 Flash for fast, context-aware, and language-flexible summarization. The system is able to take handwritten notes, convert them into clear and readable text, and produce well-structured academic summaries across different subjects and languages. Our tests show that it works reliably, adapts well to multiple languages, and is easy to use, making it a strong and practical tool for educational digitization.
