Building a Luganda Text-to-Speech Model from Crowdsourced Data
Keywords: Text-to-Speech, Crowdsourced, Luganda, Speech, Text, Multiple Speakers
TL;DR: This paper shows that the quality of Luganda TTS can be significantly improved by training models on multiple speakers with closely matched intonations from crowdsourced data and applying advanced preprocessing techniques to enhance the data quality.
Abstract: Text-to-Speech (TTS) development for African languages such as Luganda is still limited, primarily due to the scarcity of high-quality, single-speaker recordings essential for training TTS models. Prior work has focused on utilizing the Crowdsourced Common Voice Luganda recordings of multiple female speakers aged between 20-49. Although the generated speech is intelligible, it is still of lower quality compared to their model trained on studio-grade recordings. This is due to the insufficient data preprocessing methods applied to improve the quality of the Common Voice recordings. Furthermore, speech convergence is more difficult to achieve due to varying intonations from multiple speakers, as well as background noise in the training samples. In this paper, we show that the quality of Luganda TTS from Common Voice can improve by training on multiple speakers of close intonation in addition to further preprocessing of the training data. Specifically, we selected six female speakers with close intonation determined by subjectively listening and comparing their voice recordings. In addition to trimming out silent portions from the beginning and end of the recordings, we applied a pre-trained speech enhancement model to reduce background noise and enhance audio quality. We also utilized a pretrained, non-intrusive, self-supervised Mean Opinion Score (MOS) estimation model to filter recordings with an estimated MOS over 3.5, indicating high perceived quality. Subjective MOS evaluations from nine native Luganda speakers demonstrate that our TTS model achieves a significantly better MOS of 3.55 compared to the reported 2.5 MOS of the existing model. Moreover, for a fair comparison, our model trained on six speakers outperforms models trained on a single-speaker (3.13 MOS) or two speakers (3.22 MOS). This showcases the effectiveness of compensating for the lack of data from one speaker with data from multiple speakers of close intonation to improve TTS quality.
Submission Number: 33
Loading