Indian Sign Language (ISL) is an essential means of communication for the deaf and non-verbal communities in India. Despite its importance, significant barriers remain in facilitating fluid interaction between these individuals and the broader public. Recent advances in deep learning have propelled the automation of ISL recognition, offering transformative potential for accessible communication. This review synthesizes recent innovations in deep learning for ISL recognition, focusing on models such as “convolutional neural networks (CNNs), recurrent neural networks (RNNs),” and transformer architectures. These models significantly improve the recognition and interpretation of ISL’s complex gestures, facial expressions, and body movements. Our proposed deep learning-based framework combines CNNs with the MediaPipe framework, enabling real-time ISL recognition through live video input. Recognized ISL gestures are seamlessly translated into text and subsequently into speech using text-to-speech (TTS) technology, facilitating two-way communication between deaf, non-verbal, and hearing individuals. This paper critically evaluates the effectiveness of these models across key parameters, including gesture segmentation, spatial–temporal dynamics, computational efficiency, and scalability for real-time applications. We emphasize the role of multi-modal feature integration, incorporating hand shapes, motion, and facial cues to accurately capture ISL’s nuanced, context-sensitive nature. Furthermore, the review addresses pressing challenges, such as limited annotated datasets, regional dialectal variations within ISL, and the need for culturally representative training data to enhance recognition accuracy. By conducting a comparative analysis, we identify optimal practices in model architecture, data augmentation, and transfer learning to strengthen ISL recognition systems. This paper outlines a roadmap for future research, aiming to create scalable, high-accuracy ISL recognition systems capable of empowering deaf and non-verbal individuals through real-time, accessible communication solutions.