LLM-based AI-Powered Offline Portable Transcriptor (OPT) : SURYA-TAC: A Tactical Speech-to-Speech Translation System

Effective intelligence gathering along the Middle Sector of the Line of Actual Control (LAC), which encompasses Himachal Pradesh, Uttarakhand, and the Nepal border, faces persistent challenges due to language barriers, including the use of Mandarin Chinese with military-specific jargon by opposing forces, and the prevalence of low-resource regional languages such as Kinnauri, Bhoti, Kumaoni, Garhwali, and Nepali. This paper proposes the design of an AI-powered offline portable transcription (OPT) based on a large language model (LLM) capable of real-time speech-to-speech (S2S) translation during patrol operations. The system operates entirely offline, using edge-deployable models for Automatic Speech Recognition (ASR), Neural Machine Translation (NMT) and Text-to-Speech (TTS) processing. Conceptually, the system, codenamed SURYA-TAC (Surya Tactical Translator), draws inspiration from the Indian Army’s Surya Command and is designed for disconnected tactical environments. This work lays the foundation for future development by focusing on system architecture, language dataset preparation, optimization strategies and projected operational performance.