Research Paper on LiveTranscribe AI: Intelligent Video Calls With Real- Time Voice Subtitles

Read full text Summarise with VIRAT
Authors: S.V Sonekar, Madhuri Kar, Pravin Shegmwar, Pranjal Indulkar, Mansi Gotmare
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

In an increasingly interconnected global society, the necessity for seamless communication across linguistic boundaries has become paramount for professional and social collaboration. While traditional video conferencing tools have revolutionized remote interaction, the persistent barrier of diverse native languages continues to limit the effectiveness of realtime cross-border dialogue. This research addresses the critical problem of real-time linguistic interpretation within digital communication platforms, where current solutions often suffer from significant processing latency, high computational costs, or a lack of intuitive integration. The primary objective of this study is to design and implement a decentralized, AI-powered video calling application capable of providing instantaneous, synchronized transcription and translation between English and regional languages such as Hindi. The methodology adopts a robust Peerto-Peer architecture leveraging WebRTC protocols to facilitate low-latency media streaming while bypassing intensive central server dependency. A specialized dualhandshake signaling logic was developed to ensure persistent two-way data channels, allowing for the concurrent delivery of media and linguistic data across heterogeneous operating systems. The technological framework utilizes a Python-based backend developed with FastAPI for efficient request handling, complemented by a JavaScript frontend employing the PeerJS library for decentralized connectivity. Real-time speech-to- conversion is executed via the Web Speech API, while linguistic processing is powered by advanced neural machine translation models and large language model frameworks. Experimental results demonstrate that the application successfully achieves sub-second latency for subtitle synchronization and maintains high transcription accuracy across various network conditions. The practical significance of this work lies in its capacity to democratize international communication, enabling users to engage in natural dialogue using their native tongues in business, educational, and social cons. By automating the translation process within a lightweight web interface, this system provides a scalable solution for eliminating global language barriers in real-time digital environments.

Fields of Study & Subjects

Speech Recognition and Synthesis Translation Studies and Practices Mobile and Web Applications

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
1 January 2026
Language
EN