# Training Multi-Speaker Neural Text-to-Speech Systems using Speaker-Imbalanced Speech Corpora

arXiv: https://arxiv.org/abs/1904.00771 · HTML: https://mlsanity.com/paper/1904.00771/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 1 Apr 2019.

## Built on by (most cited first, 8 of 8)

- [A Survey on Neural Speech Synthesis](https://mlsanity.com/paper/2106.15561/) (2021)
- [Combining speakers of multiple languages to improve quality of neural voices](https://mlsanity.com/paper/2108.07737/) (2021)
- [Custom Data Augmentation for low resource ASR using Bark and Retrieval-Based Voice Conversion](https://mlsanity.com/paper/2311.14836/) (2023)
- [Speech Rhythm-Based Speaker Embeddings Extraction from Phonemes and Phoneme Duration for Multi-Speaker Speech Synthesis](https://mlsanity.com/paper/2402.07085/) (2024)
- [A multilingual training strategy for low resource Text to Speech](https://mlsanity.com/paper/2409.01217/) (2024)
- [Building a Luganda Text-to-Speech Model From Crowdsourced Data](https://mlsanity.com/paper/2405.10211/) (2024)
- [Preliminary study on using vector quantization latent spaces for TTS/VC systems with consistent performance](https://mlsanity.com/paper/2106.13479/) (2021)
- [AI based Presentation Creator With Customized Audio Content Delivery](https://mlsanity.com/paper/2106.14213/) (2021)

## Abstract (arXiv)

When the available data of a target speaker is insufficient to train a high quality speaker-dependent neural text-to-speech (TTS) system, we can combine data from multiple speakers and train a multi-speaker TTS model instead. Many studies have shown that neural multi-speaker TTS model trained with a small amount data from multiple speakers combined can generate synthetic speech with better quality and stability than a speaker-dependent one. However when the amount of data from each speaker is highly unbalanced, the best approach to make use of the excessive data remains unknown. Our experiments showed that simply combining all available data from every speaker to train a multi-speaker model produces better than or at least similar performance to its speaker-dependent counterpart. Moreover by using an ensemble multi-speaker model, in which each subsystem is trained on a subset of available data, we can further improve the quality of the synthetic speech especially for underrepresented speakers whose training data is limited.