SWAra ROOPantaran (SWAROOP) : A Tool for Transcription, Translation and Speech Synthesis
Amresh Kumar, T M Dhipu, R Rajesh · 2023
The advancements in the domain of Natural Language Processing has resulted in the latest Large Language Models (LLM) that power popular conversational systems like ChatGPT, Alexa etc. SWAROOP (SWAra ROOPantaran) is an LLM-based tool that performs speech recognition, translation and speech synthesis task, either separately or in combination of two or more tasks together with noisy speech data. One of the significant aspects of this tool is that it works in offline environment with human acceptable accuracy and can be easily deployed on local low resource compute hardware.