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/

Global: Zendy Plus annual

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

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

In addition to conveying the linguistic content from source speech toconverted speech, maintaining the speaking style of source speech also plays animportant role in the voice conversion (VC) task, which is essential in manyscenarios with highly expressive source speech, such as dubbing and dataaugmentation. Previous work generally took explicit prosodic features orfixed-length style embedding extracted from source speech to model the speakingstyle of source speech, which is insufficient to achieve comprehensive stylemodeling and target speaker timbre preservation. Inspired by the style'smulti-scale nature of human speech, a multi-scale style modeling method for theVC task, referred to as MSM-VC, is proposed in this paper. MSM-VC models thespeaking style of source speech from different levels. To effectively conveythe speaking style and meanwhile prevent timbre leakage from source speech toconverted speech, each level's style is modeled by specific representation.Specifically, prosodic features, pre-trained ASR model's bottleneck features,and features extracted by a model trained with a self-supervised strategy areadopted to model the frame, local, and global-level styles, respectively.Besides, to balance the performance of source style modeling and target speakertimbre preservation, an explicit constraint module consisting of a pre-trainedspeech emotion recognition model and a speaker classifier is introduced toMSM-VC. This explicit constraint module also makes it possible to simulate thestyle transfer inference process during the training to improve thedisentanglement ability and alleviate the mismatch between training andinference. Experiments performed on the highly expressive speech corpusdemonstrate that MSM-VC is superior to the state-of-the-art VC methods formodeling source speech style while maintaining good speech quality and speakersimilarity.

MSM-VC: High-fidelity Source Style Transfer for Non-Parallel Voice Conversion by Multi-scale Style Modeling