Whispy: Adapting STT Whisper Models to Real-Time Environments

TL;DR

Whispy is a system that adapts Whisper models for real-time speech transcription, achieving low latency and high accuracy through architectural optimizations, enabling practical live speech analysis applications.

Contribution

This paper introduces Whispy, a novel system that enables real-time transcription with Whisper models by optimizing architecture for low latency and high accuracy.

Findings

Whispy maintains high transcription accuracy in real-time settings.

The system demonstrates robustness across diverse speech datasets.

Whispy achieves low computational cost suitable for practical deployment.

Abstract

Large general-purpose transformer models have recently become the mainstay in the realm of speech analysis. In particular, Whisper achieves state-of-the-art results in relevant tasks such as speech recognition, translation, language identification, and voice activity detection. However, Whisper models are not designed to be used in real-time conditions, and this limitation makes them unsuitable for a vast plethora of practical applications. In this paper, we introduce Whispy, a system intended to bring live capabilities to the Whisper pretrained models. As a result of a number of architectural optimisations, Whispy is able to consume live audio streams and generate high level, coherent voice transcriptions, while still maintaining a low computational cost. We evaluate the performance of our system on a large repository of publicly available speech datasets, investigating how the transcription mechanism introduced by Whispy impacts on the Whisper output. Experimental results show how Whispy excels in robustness, promptness, and accuracy.