VAE
https://paperswithcode.com/paper/rave-a-variational-autoencoder-for-fast-and-1
RAVE 2021.11 并不能直接用于VC,可能需要微调
https://github.com/kamepong/ACVAE-VC
ACVAE-VC 2018.8的论文,代码最后一次维护3年前,可能有点古老了
https://paperswithcode.com/paper/non-parallel-voice-conversion-with-cyclic
CycleVAE-VC 2019.7
https://github.com/rupschy/specGenerationVAE
specGenerationVAE
https://github.com/vsimkus/vae-voice-conversion
Diffusion
https://github.com/huawei-noah/Speech-Backbones/tree/main/DiffVC
❤DiffVC ICLR-2022 这个很经典,似乎是第一个将diffusion用在VC的模型
https://paperswithcode.com/paper/dddm-vc-decoupled-denoising-diffusion-models
DDDM-VC 2023.5
https://paperswithcode.com/paper/duta-vc-a-duration-aware-typical-to-atypical
duta-VC 2023.6 代码基于DiffVC
https://paperswithcode.com/paper/diff-hiervc-diffusion-based-hierarchical
Diff-HierVC 2023.11 最新的模型,代码上个月还在维护,有train.py但readme没说怎么训练,感觉复现有难度
https://github.com/lmnt-com/diffwave/tree/master
diffwave 功能与wavenet近似,是通用波形生成器。已经将api植入到了pip中,训练较为容易,效果似乎有待观察