Effective Zero-Shot Multi-Speaker Text-to-Speech Technique Using Information Perturbation and a Speaker Encoder
Chae-Woon Bang, Chanjun Chun, Christoph M Friedrich · 2023
Read Effective Zero-Shot Multi-Speaker Text-to-Speech Technique Using Information Perturbation and a Speaker Encoder by Chae-Woon Bang, Chanjun Chun, Christoph M Friedrich (2023) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-10708733.txt.
Description
Chae-Woon Bang and co-authors improve a speech system so it can copy a new voice it has never been trained on. Their base model builds speech from text with a diffusion method and handles many voices, but it could not handle an unseen voice at all.
The new method takes speaker facts from a sample clip, using a model first trained to tell speakers apart. A trick that scrambles some speaker facts helps the system learn a wider range of voices. In tests it matched or beat rival systems for known voices and did about as well for new ones, based on scores for likeness and quality.
Catalog record
| CREATOR | Chae-Woon Bang, Chanjun Chun, Christoph M Friedrich |
|---|---|
| YEAR | 2023 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |