The recognition of Arabic words presents considerable difficulties owing to the complex characteristics of the Arabic script, which encompasses letters positioned both above and below the baseline, hamzas, and dots. In order to address these intricacies, we provide a structured approach for transforming handwritten Arabic text into a digital format. We employ a hybrid deep learning technique that combines Convolutional Neural Networks (CNNs), Bidirectional Long Short-Term Memory (BLSTM), and Connectionist Temporal Classification (CTC). We collected datasets that cover a wide range of Arabic text variations. We have also created a pre-processing pipeline. Our methodology successfully achieved an accuracy rate of 99.52%. At the level of recognizing the letters of the word, with an accuracy of 98.36% at the level of the full word. In order to evaluate the effectiveness of our suggested method for recognizing handwritten text, we utilize two essential metrics: Word Error Rate (WER) and Character Error Rate (CER) to compare its performance. The experimental research demonstrates a WER of 1.64 % and a CER of 0.48%.