This thesis presents the research effort and its results regarding the design and implementation of a speech synthesis system based on time-domain techniques. More specifically, this thesis, apart from a detailed literature review, focuses on the following specific points: • The study of prosody and its modeling for the Text-to-Speech system. The proposed mechanism provides a novel algorithm for modeling, producing and applying of prosodic curves onto the synthetic speech signal, in a manner that allows the preservation of microprosody and diversity in prosodic patterns. • The design and development of primary material for the database of a voice synthesizer, which is based on recordings of a native speaker. The method proposed is an innovative one for the design of the textual corpus, taking into account the specific characteristics of the voice synthesizer and ensuring maximum coverage of different acoustic and linguist phenomena. • The different subsystems of the speech synthesizer, which are responsible for natural language processing, text normalization and phonetic transcription of the text. New methods have been proposed for the phonetic transcription of the Greek language, as well as the conversion from Greeklish to Greek. • Extra focus was given in the customization and adaptation of the speech synthesis system for optimal performance in the framework of accessibility, providing the ground for an optimized TTS system as an assistive tool. The results of this research effort have led to the realization of the highest quality TTS system for Greek language, providing at the same time the basis for creating a TTS system for the Bulgarian language of similarly high quality and naturalness.