English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

Disentangling content and speaking style information is essential forzero-shot non-parallel voice conversion (VC). Our previous study investigated anovel framework with disentangled sequential variational autoencoder (DSVAE) asthe backbone for information decomposition. We have demonstrated thatsimultaneous disentangling content embedding and speaker embedding from oneutterance is feasible for zero-shot VC. In this study, we continue thedirection by raising one concern about the prior distribution of content branchin the DSVAE baseline. We find the random initialized prior distribution willforce the content embedding to reduce the phonetic-structure information duringthe learning process, which is not a desired property. Here, we seek to achievea better content embedding with more phonetic information preserved. We proposeconditional DSVAE, a new model that enables content bias as a condition to theprior modeling and reshapes the content embedding sampled from the posteriordistribution. In our experiment on the VCTK dataset, we demonstrate thatcontent embeddings derived from the conditional DSVAE overcome the randomnessand achieve a much better phoneme classification accuracy, a stabilizedvocalization and a better zero-shot VC performance compared with thecompetitive DSVAE baseline.

Towards Improved Zero-shot Voice Conversion with Conditional DSVAE