Hacker News new | ask | show | jobs
by arkobel 110 days ago
The lack of parallel accent data makes this fundamentally unsupervised. Curious if this leans more on latent disentanglement than direct supervision.