Voice Conversion with Diverse Intonation using Conditional Variational Auto-Encoder
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866913796453302272 |
|---|---|
| author | Suh, Soobin Ahn, Dabi Park, Heewoong Park, Jonghun |
| author_facet | Suh, Soobin Ahn, Dabi Park, Heewoong Park, Jonghun |
| contents | Voice conversion is a task of synthesizing an utterance with target speaker's voice while maintaining linguistic information of the source utterance. While a speaker can produce varying utterances from a single script with different intonations, conventional voice conversion models were limited to producing only one result per source input. To overcome this limitation, we propose a novel approach for voice conversion with diverse intonations using conditional variational autoencoder (CVAE). Experiments have shown that the speaker's style feature can be mapped into a latent space with Gaussian distribution. We have also been able to convert voices with more diverse intonation by making the posterior of the latent space more complex with inverse autoregressive flow (IAF). As a result, the converted voice not only has a diversity of intonations, but also has better sound quality than the model without CVAE. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_12005 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Voice Conversion with Diverse Intonation using Conditional Variational Auto-Encoder Suh, Soobin Ahn, Dabi Park, Heewoong Park, Jonghun Sound Machine Learning Audio and Speech Processing Voice conversion is a task of synthesizing an utterance with target speaker's voice while maintaining linguistic information of the source utterance. While a speaker can produce varying utterances from a single script with different intonations, conventional voice conversion models were limited to producing only one result per source input. To overcome this limitation, we propose a novel approach for voice conversion with diverse intonations using conditional variational autoencoder (CVAE). Experiments have shown that the speaker's style feature can be mapped into a latent space with Gaussian distribution. We have also been able to convert voices with more diverse intonation by making the posterior of the latent space more complex with inverse autoregressive flow (IAF). As a result, the converted voice not only has a diversity of intonations, but also has better sound quality than the model without CVAE. |
| title | Voice Conversion with Diverse Intonation using Conditional Variational Auto-Encoder |
| topic | Sound Machine Learning Audio and Speech Processing |
| url | https://arxiv.org/abs/2504.12005 |