By employing hand gestures, individuals with hearing or speech disabilities are able to communicate their ideas without the need for spoken words. The work envisaged in this paper is to create an affordable yet real-time sign language to speech conversion system for the Tamil speaking regions. The system makes use of a web camera to capture hand gestures which are processed according to MediaPipe and then classified using Convolutional Neural Networks. The recognized gestures are converted to human voice in Tamil using gTTS API for Tamil language, thus providing instant voice output in Tamil language. The proposed system is highly beneficial for the hearing-impaired population in the Tamil-speaking areas and its accuracy stands remarkable at 92%. The system has been developed enough for 12 alphabets of the Tamil language which can as well be updated in the future to cover all the 247 alphabets.