Indian languages represent a highly multilingual and low-resource speech ecosystem, where the scarcity of high-quality parallel speech corpora significantly limits the development of speech-to-speech translation systems. Most existing approaches rely on cascaded pipelines that combine automatic speech recognition (ASR), machine translation (MT), and text-to-speech synthesis (TTS). While effective, these cascaded systems often suffer from cumulative error propagation, increased latency, and higher computational complexity, particularly in low-resource Indian languages. To address these challenges, my doctoral work proposes a novel sequence-to-sequence direct speech translation framework capable of translating speech from one Indian language to another without relying on intermediate text representations. Recent advances in deep learning, however, indicate that direct speech translation architectures can surpass conventional cascaded systems in both efficiency and translation quality, motivating the design of our fully end-to-end solution. We aim to release an initial dataset comprising at least 120,000 real speech samples within a 6–12 month timeframe.

Speech-to-Speech Translation (S2ST) focuses on generating spoken output in a target language directly from spoken input in a source language. Despite progress in S2ST modeling, low-resource Indic languages remain poorly supported, primarily because large-scale parallel speech corpora are unavailable. We present UrHiOdSynth, a three-language parallel S2ST dataset containing approximately 75 hours of speech across Urdu, Hindi, and Odia. The corpus consists of 10,735 aligned sentence triplets, with an average utterance length of 8.45 seconds. To our knowledge, UrHiOdSynth represents the largest multi-domain resource offering aligned speech and text for S2ST in this language context. Beyond speech-to-speech translation, the dataset supports tasks such as automatic speech recognition, speech-to-text translation, text-to-speech synthesis, and machine translation. This flexibility enables the training of unified multilingual models, particularly for low-resource Indic languages.