The Blizzard Challenge
Data and tools from previous Blizzard Challenges: 2008-2019
Various data and tools are available. Please carefully check the terms of use for each item - some items have restrictions on commercial use, for example.- Tools and listening test results
- Data sets
- 2011 - The data for the Blizzard Challenge 2011 tasks EH1 and ES1
- 2012 phase 1 - The data for Phase One of the Blizzard Challenge 2012
- 2012 phase 2 - The data and development set for Phase Two of the Blizzard Challenge 2012
- 2013 - The English audiobook data for the Blizzard Challenge 2013
- 2017 - The English audiobook data for the Blizzard Challenges 2017 & 2018 (a subset was used in 2016)
- 2019 - The Chinese data for the Blizzard Challenge 2019 (only available to registered participants)
- 2020 - The Chinese Mandarin and Shanghainese data for the Blizzard Challenge 2020 (only available to registered participants)
- 2021 - The Spanish data for the Blizzard Challenge 2021 (only available to registered participants)
- 2023 - The French data for the Blizzard Challenge 2023 (freely available on Zenodo)
- Software
- Models
- 2013 - The models used for the 2013-EH2 extension conducted in 2023. To access these models, you need to obtain a license for The English audiobook data for the Blizzard Challenge 2013 as listed above, then use the same credentials. (paper under review)
- 2008 - The British English 'roger' corpus
- 2009 - The British English 'roger' corpus
- 2010 - The British English 'rjs' corpus
- 2010 - The British English 'roger' corpus
- 2010 - The Mandarin Chinese corpus
Contact: Simon King.