Reproducing Whisper-Style Training Using an Open-Source Toolkit and Publicly Available Data

TL;DR

This paper introduces an open-source toolkit to reproduce Whisper-style speech model training using publicly available data, enabling broader research and improvements in speech recognition and translation.

Contribution

It presents the Open Whisper-style Speech Model (OWSM), supporting more translation directions and improved training efficiency, with publicly released resources for transparency and reproducibility.

Findings

OWSM supports additional translation directions.

OWSM can be trained more efficiently.

All scripts and models will be publicly released.

Abstract

Pre-training speech models on large volumes of data has achieved remarkable success. OpenAI Whisper is a multilingual multitask model trained on 680k hours of supervised speech data. It generalizes well to various speech recognition and translation benchmarks even in a zero-shot setup. However, the full pipeline for developing such models (from data collection to training) is not publicly accessible, which makes it difficult for researchers to further improve its performance and address training-related issues such as efficiency, robustness, fairness, and bias. This work presents an Open Whisper-style Speech Model (OWSM), which reproduces Whisper-style training using an open-source toolkit and publicly available data. OWSM even supports more translation directions and can be more efficient to train. We will publicly release all scripts used for data preparation, training, inference, and scoring as well as pre-trained models and training logs to promote open science.