Recent advancements in artificial intelligence and high-performance computing have revolutionized various facets of human interaction. Speech synthesis (text-to-speech) and automatic speech recognition (speech-to-text) technologies have become ubiquitous in smartphones, smart homes, websites, and automotive systems, transforming how we communicate with technology. This paper resembles a showcase of the efficacy of end-to-end text-to-speech (TTS) models trained specifically on Arabic data. The study presents an in-depth analysis of various TTS system architectures mainly Tacotron2 and FastSpeech2, highlighting their performance in generating high-quality synthesized speech from written Arabic text. Through a beginner’s lens, the article sheds light on the exciting domain of audio deep learning, providing valuable insights for researchers and practitioners interested in TTS technology. By demonstrating the capabilities and potential challenges of employing Arabic data in TTS model training, this paper contributes to the advancement of multilingual speech synthesis and inspires further exploration in the field of audio deep learning.