The Blizzard Challenge 2013

By Simon King, Vasilis Karaiskos. 2013

126 citations recorded by the source corpus.

The Blizzard Challenge 2013 was the ninth annual Blizzard Challenge which was once again organised by the University of Edinburgh with advice from the other members of the Blizzard Challenge committee – Prof. Keiichi Tokuda and Prof. Alan Black – joined this year by Dr. Kishore Prahallad who organised the Indian languages tasks. This summary paper only covers the English tasks; a separate paper is available which summarises those Indian language tasks. For the English tasks, a large single-speaker English corpus was used, comprising around 300 hours of audio from professionally-produced audiobooks.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

1. Xu Tan
2. Yu Zhang
3. Wei-Ning Hsu
4. Ye Jia
5. Yoshua Bengio
6. Tao Qin
7. Ron J. Weiss
8. Dong Yu

Citations in the corpus

• Guided-TTS: A Diffusion Model for Text-to-Speech via Classifier Guidance (2021, methodology): “For the other case which makes use of only the untranscribed target speaker speech, we use LJSpeech, Hi-Fi TTS (Bakhturina et al., 2021), and Blizzard 2013 (King & Karaiskos, 2013).”
• Hierarchical Generative Modeling for Controllable Speech Synthesis (2018, methodology): “A single speaker US English audiobook dataset of 147 hours, recorded by professional speaker, Catherine Byers, from the 2013 Blizzard Challenge (King & Karaiskos, 2013) is used for training.”
• Sequential Neural Models with Stochastic Layers (2016, methodology): “Blizzard is a dataset of 300 hours of English, spoken by a single female speaker.”
• A study of speaker adaptation for DNN-based speech synthesis (2015, background): “Due to the robust performance of speaker adaptation, a major advantage of SPSS over unit-selection speech synthesis is its flexibility in changing speaker characteristics, speaking styles and emotions [1], and there has been a significant improvement in naturalness in recent years [2].”
• Fully-Hierarchical Fine-Grained Prosody Modeling For Interpretable Speech Synthesis (2020, methodology): “The proposed models are evaluated on the LibriTTS multi-speaker audiobook dataset [35] and the Blizzard Challenge 2013 single-speaker audiobook dataset [36].”
• Z-Forcing: Training Stochastic Recurrent Networks (2017, methodology): “We show that our model can achieve state-of-the-art results on two speech modeling datasets: Blizzard (King and Karaiskos, 2013) and TIMIT raw audio datasets (also used in Chung et al.”
• Hi-Fi Multi-Speaker English TTS Dataset (2021, background): “The largest high-quality English dataset for speech synthesis was provided by The Voice Factory as part of the Blizzard Challenge 2013 [12].”
• Robust universal neural vocoding (2018, methodology): “Finally we also evaluated highly expressive children audiobook speech extracted from Blizzard2013 development set [23], which was out of domain in speaker, speaking style and recording conditions.”
• Universal MelGAN: A Robust Neural Vocoder for High-Fidelity Waveform Generation in Multiple Domains (2020, unknown): “1[23] Seen: single speaker E2 LibriTTS(train-clean-360)[24] Seen: multi-speaker E3 LibriTTS(test-clean)[24] Unseen: speaker E4 BlizzardChallenge2013[25] Unseen: emotion Multiple language dataset Index Name Remarks M1 CSS10[26] Unseen: language”
• The SIWIS French Speech Synthesis Database ? Design and recording of a high quality French database for speech synthesis (2017, unknown): “In English, one of the recent Blizzard Challenge datasets provided a high quality somewhat expressive speech corpus [Karaiskos et al., 2008].”
• Speech Enhancement Using Bayesian Wavenet (2017, background): “The clean training set contains a total of 9700 utterances (19 hours) from audio books played by a female speaker [25].”