CONFERENCE / ICCAIS-2026

While various translation tools are available, cross-lingual communication remains one of the biggest challenges in global interactions. This is true since most conventional systems rely primarily on text-based input, thereby limiting natural conversational flow. This paper presents an AI-Powered Voice-to-Voice Language Translation System that directly converts spoken input in one language to synthesized speech in anotherIn this work we integrate an architecture based on a modified architecture to create a framework of ASR and TTS (NMT) through the integration of AI-based models. The approach leverages multilingual transformer-based neural models and neural speech synthesis, which provides for accurate transcriptions, contextually translated and natural sounding speech generated from transcriptions. The resulting system was built as a web application that allows for near real-time responsive interactions with multiple AI models. The results of the experimental evaluation support that both the accuracy and overall usability are enhanced by providing effective multilingual capabilities. This research proposes an implementation of an AI-based solution that would provide scalability and deployability to real-world multilingual voice communications. This research proposes an implementation of an AI-based solution that would provide scalability and deployability to real-world multilingual voice communications.