People with hearing loss would greatly benefit from assistive technology if it used Audio Visual Speech Recognition (AVSR). 466 million individuals worldwide are deaf or partially deaf. They use lip reading to grasp what is being said since they are deaf or hearing impaired. Many students with hearing loss struggle because of a scarcity of qualified sign language facilitators and the hefty price tag on assistive technology. Using cutting-edge deep learning models, we've developed a new way for visual voice detection. In addition, the present VSR approaches have flaws that need to be corrected. Our new method merges audio and visual speech outputs as a result. An audio-visual speech recognition model that incorporates deep learning has been proposed to improve lip reading speed and accuracy. According to this study, the system's performance has been significantly improved, with word error rates of 6.59 % for ASR and 95 % for lip reading.