Singing Voice Conversion Challenge 2023 Official Samples
This website shows some singing voice conversion samples with two systems using the official SVCC 2023 dataset. We will be providing two source speakers during the evaluation phase (SM1 and SF1) and the goal will be to convert these to the target speakers (IDM1, IDF1, CDM1, CDF1).
Please note that for this year, the dataset will only be given to registered participants.
You can check out the registration form (free of charge!) and the challenge rules at the official SVCC 2023 website here.
Baseline System: DiffSVC Official Implementation
Paper: DiffSVC: A Diffusion Probabilistic Model for Singing Voice Conversion
Authors: Songxiang Liu, Yuewen Cao, Dan Su, Helen Meng
[ArXiv paper] [Official demo]
Notes: These samples were provided by the official author.
Task 1: In-domain conversion
Sample 1
Sample 2
Sample 3
Sample 4
SM1 to IDM1
SF1 to IDM1
SM1 to IDF1
SF1 to IDF1
Task 2: Cross-domain conversion
Sample 1
Sample 2
Sample 3
Sample 4
SM1 to CDM1
SF1 to CDM1
SM1 to CDF1
SF1 to CDF1
Starter Kit: FastSVC Reimplementation
Paper: FastSVC: Fast Cross-Domain Singing Voice Conversion with Feature-wise Linear Modulation
Authors: Songxiang Liu, Yuewen Cao, Na Hu, Dan Su, Helen Meng
[ArXiv paper] [Github repository] [Official demo]
Notes: Please note that there are some differences between this reimplementation and the official paper to accommodate 24kHz sampling. You can read more about the changes in the Github repository.