VoxENES 2026: Benchmarking Generalization of Speech Spoofing Detectors Against LLM-Era TTS and Voice Conversion
VoxENES 2026 is a bilingual (English and Spanish) benchmark of 53,628 audio samples generated using 10 contemporary speech synthesis methods and evaluated under 10 standardized post-processing conditions. Eight pretrained detectors were benchmarked without fine-tuning; the best model achieved 28.98% EER overall, while most performed near or below random chance.
发展脉络
- 首次出现VoxENES 2026: Benchmarking Generalization of Speech Spoofing Detectors Against LLM-Era TTS and Voice ConversionarXiv cs.AI
- 当前判断The benchmark reveals a temporal generalization gap in spoofing detection, suggesting that deployed systems may be vulnerable to modern synthetic speech. Next signal: adoption of VoxENES 2026 as a standard evaluation set by industry or regulatory bodies.Agent Pulse · 分析
VoxENES 2026 is a bilingual benchmark of 53,628 audio samples from 10 modern TTS/VC methods, tested under 10 post-processing conditions. Eight pretrained detectors showed substantial performance degradation, with the best achieving 28.98% EER and most near random chance.
Current speech spoofing detectors rely on brittle artifacts that fail to generalize to LLM-era TTS and VC, as evidenced by near-random performance on VoxENES 2026. Next signal: development of detectors trained on modern synthetic speech or using artifact-agnostic features.
The benchmark reveals a temporal generalization gap in spoofing detection, suggesting that deployed systems may be vulnerable to modern synthetic speech. Next signal: adoption of VoxENES 2026 as a standard evaluation set by industry or regulatory bodies.
Robust spoofing detectors are critical for voice authentication systems in banking, call centers, and security. VoxENES 2026 highlights the need for updated detection models, creating opportunities for companies offering anti-spoofing solutions. Next signal: a commercial spoofing detection product claiming VoxENES 2026 performance.
VoxENES 2026 provides a testbed for developing robust countermeasures. Future work may focus on fine-tuning detectors on modern generators or designing artifact-invariant features. Next signal: publication of a detector achieving <10% EER on VoxENES 2026.