Research map / Label-efficient and robust learning / Recognition and long-tail
VoiceMixer: Adversarial Voice Style Mixup
Sang-Hoon Lee, Ji-Hoon Kim, Hyunseung Chung, Seong-Whan Lee
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add contrastive losses Lpos and Lneg (ablation) | VoiceMixer · VCTK zero-shot | Similarity: 42.3% -> 63.7% | Improved |
| add negative contrastive loss Lneg (ablation) | VoiceMixer · VCTK zero-shot | Naturalness MOS: 3.18 -> 3.72 | Improved |
| add adversarial speaker classification Ladvsc (ablation) | VoiceMixer · VCTK zero-shot | Similarity: 41.3% -> 63.7% | Improved |
| add adversarial speaker classification Ladvsc (ablation) | VoiceMixer · VCTK zero-shot | FDSD: 1.217 -> 2.416 | Hurt |
| use similarity-based bottleneck instead of fixed-length (tau=16) (ablation) | VoiceMixer · VCTK zero-shot | Naturalness MOS: 3.35 -> 3.72 | Improved |
| use similarity-based bottleneck instead of fixed-length (tau=32) (ablation) | VoiceMixer · VCTK zero-shot | ASV EER: 20.7% -> 18.5% | Improved |
| add GAN training (ablation) | VoiceMixer · VCTK zero-shot | FDSD: 2.666 -> 2.416 | Improved |
| use disentangled content/style discriminator instead of single discriminator (ablation) | VoiceMixer · VCTK zero-shot | FDSD: 3.502 -> 2.416 | Improved |
| add repulsive style loss (ablation) | VoiceMixer · VCTK zero-shot | Similarity: 55.0% -> 63.7% | Improved |
| add repulsive style loss (ablation) | VoiceMixer · VCTK zero-shot | FDSD: 2.173 -> 2.416 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- AUTOVC (τ=32) — AutoVC: Zero-Shot Voice Style Transfer with Only Autoencoder Loss (ICML 2019) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DDDM-VC: Decoupled Denoising Diffusion Models with Disentangled Representation and Prior Mixup for Verified Robust Voice Conversion (AAAI 2024 · similarity 0.88)
- SelfVC: Voice Conversion With Iterative Refinement using Self Transformations (ICML 2024 · similarity 0.86)
- Seeing Your Speech Style: A Novel Zero-Shot Identity-Disentanglement Face-based Voice Conversion (AAAI 2025 · similarity 0.85)
- Vevo: Controllable Zero-Shot Voice Imitation with Self-Supervised Disentanglement (ICLR 2025 · similarity 0.85)
- Improving Generalization for AI-Synthesized Voice Detection (AAAI 2025 · similarity 0.85)
- StyleSinger: Style Transfer for Out-of-Domain Singing Voice Synthesis (AAAI 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
