NouveauVoice framework enhances voice anonymization with diverse pseudo-speakers

Researchers have developed NouveauVoice, a new framework designed to generate diverse pseudo-speakers for voice anonymization. This system utilizes a Hierarchical Deep Variational Autoencoder (NVAE) and can be integrated with existing architectures like FACodec and CosyVoice2. Evaluations indicate that NouveauVoice significantly enhances speaker diversity and identity concealment, achieving an Equal Error Rate (EER) of over 38% against speaker verification attacks, while maintaining a balance between anonymity, diversity, and speech utility. AI

IMPACT Improves privacy in voice applications by generating more diverse and convincing anonymized voices.

RANK_REASON Academic paper detailing a new method for voice anonymization. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• CosyVoice2
• FACodec
• Hierarchical Deep Variational Autoencoder
• Maximum Mean Discrepancy
• NouveauVoice
• VoicePrivacy Challenge

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →