Traditional Text-to-Speech (TTS) systems rely on studio-quality speech recorded in controlled settings.a Recently, an effort known as noisy-TTS training has emerged, aiming to utilize in-the-wild data. However, the lack of dedicated datasets has been a significant limitation. We introduce the TTS In the Wild (TITW) dataset, which is publicly available, created through a fully automated pipeline applied to the VoxCeleb1 dataset. It comprises two training sets: TITW-Hard, derived from the transcription, segmentation, and selection of raw VoxCeleb1 data, and TITW-Easy, which incorporates additional enhancement and data selection based on DNSMOS. State-of-the-art TTS models achieve over 3.0 UTMOS score with TITW-Easy, while TITW-Hard remains difficult showing UTMOS below 2.8.
@article{arxiv.2409.08711,
title = {Text-To-Speech Synthesis In The Wild},
author = {Jee-weon Jung and Wangyou Zhang and Soumi Maiti and Yihan Wu and Xin Wang and Ji-Hoon Kim and Yuta Matsunaga and Seyun Um and Jinchuan Tian and Hye-jin Shim and Nicholas Evans and Joon Son Chung and Shinnosuke Takamichi and Shinji Watanabe},
journal= {arXiv preprint arXiv:2409.08711},
year = {2025}
}