[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[AUDITORY] Launching the Blizzard Challenge 2025 - Synthesizing Bildts



**Apologies for cross posting**

Dear colleagues,
We are delighted to announce the 2025 edition of the Blizzard Challenge, which will be co-located with the Speech Synthesis Workshop 2025 (SSW13).
The theme of this edition of SSW is "Scaling down: sustainable synthesis for language diversity". In line with this theme, the challenge focuses on synthesizing speech for Bildts, a unique language variety from the Netherlands.
The information about the challenge is available (and will be updated) here: https://blogs.helsinki.fi/ssw13-2025/the-blizzard-challenge-2025/
About Bildts
Bildts (Indo-European > West Germanic) is spoken in Het Bildt, a region in the Dutch province of Friesland. With approximately 10,000 first and second language speakers, it represents a vibrant example of European linguistic diversity. The language variety has been systematically through grammatical descriptions, dictionaries with pronunciation information, literary works, media productions (weekly radio broadcasts, theater performances), and regular newspaper columns.
For the Blizzard Challenge 2025, we have curated a dataset of high-quality audio recordings with corresponding linguistic resources. This choice of Bildts aligns with our theme of "sustainable synthesis for language diversity" - it presents participants with the real-world challenge of developing synthesis capabilities for a well-documented but data-limited language variety, representative of the thousands of smaller languages that could benefit from speech technology. The challenge is composed of two tasks:
  1. the traditional hub which this year consists of creating a synthetic voice for Bildts using the data
  2. a Zeroshot TTS for Bildts for different speakers with reference audio files provided during the release of the test set
Data download
The training data is available here: https://doi.org/10.5281/zenodo.14792457
A pronunciation dictionary for Bildts will be made available to participants in February 2025. This will include phonetic transcriptions that can be used for training text-to-speech systems. Participants will be notified when this resource becomes available.
Timeline
This is a preliminary timeline and is subject to change.
  • Feb 032025 - challenge announcement and registration open
  • Feb 03 2025 - training dataset released
  • Mar 07 2025 - team registration closes
  • Mar 28 2025 - test sentences released to participants
  • Apr 04 2025 - deadline for participants to submit synthetic speech (23:59 AoE)
  • Apr 07 2025 - last date for payment of the entry fee (more details in February)
  • Apr 24 2025 - evaluation systems go live
  • Jun 20 2025 - end of the evaluation period
  • Jun 27 2025 - release of results
  • Jul 18 2025 - deadline to submit workshop papers (23:59 AoE)
  • Aug 01 2025 - notification of acceptance
  • Aug 08 2025 - camera ready
  • Aug 17-21 2025 - INTERSPEECH 2025, Rotterdam, Netherlands
  • Aug 24-26 2025 - SSW13, Leeuwarden, Netherlands
  • Aug 27 2025 - Blizzard Challenge 2025 Workshop, same location as SSW
Any questions?
Please contact blizzard-challenge-organisers@xxxxxxxxxxxxxxxx if you have any questions.

Please feel free to distribute this announcement.

The Blizzard Challenge 2025 Organising team