Fact · Voice privacy and security
3seconds
A 2023 Microsoft research paper reported that its VALL-E model can synthesize personalized speech from only a 3-second enrolled recording of an unseen speaker.
Checked against its source on October 16, 2026
In the source's own words
“can be used to synthesize high-quality personalized speech with only a 3-second enrolled recording of an unseen speaker as an acoustic prompt”
Neural Codec Language Models are Zero-Shot Text to Speech Synthesizers, Chengyi Wang et al., arXiv (Microsoft Research), 2023, Abstract.
Open the original- What it measures
- The result is a research claim from 2023 and shows how little audio a clone can need.
- What it does not show
- Lab result by the system's authors. Quality varies with voice, recording and system.