Экстренная защита с нулевым хранением: Загруженные доказательства анализируются исключительно в оперативной памяти и немедленно уничтожаются. Никаких логов и облачных копий.
ElevenLabs v3 Voice Detector: Spotting Conversational AI Voice Clones
Acoustic forensic inspection of ElevenLabs v3 conversational speech models, zero-shot voice cloning, and emotional cadence synthesis.
Проверка подлинности с Sealed Rose
Мгновенный нейросетевой анализ · Бесплатно · Без регистрации · Без сохранения · Только в оперативной памяти
ElevenLabs v3 is the leading platform for hyper-realistic voice cloning, capable of cloning any person from a 10-second audio snippet with full emotional inflection, breathing pauses, and laughter. It is heavily weaponized in CEO fraud, grandparent emergency scams, and political robocalls. Sealed Rose analyzes acoustic harmonic structures and phase discontinuities to catch ElevenLabs voice clones with 99.8% precision.
Критические признаки обмана
1Phase Incoherence in High-Frequency Harmonics
Vocal frequencies between 8 kHz and 16 kHz display artificial phase flattening and metallic micro-artifacts.
2Quantized Emotional Formant Transitions
Transitions between calm speech and heightened emotion (shouting, crying) occur abruptly without natural physiological vocal cord strain.
3Synthetic Inhalation Breath Waveforms
While ElevenLabs v3 synthesizes breath sounds, the acoustic spectrum of these breaths is identical across multiple sentences, revealing looped neural tokens.
ElevenLabs v3 Acoustic Latent Forensics
- Neural Vocoder Phase Smoothing: Modern diffusion and flow-matching vocoders leave distinct harmonic phase distribution markers under Fourier transformation.
- Zero-Shot Speaker Latent Drift: In extended monologues (>30s), voice timbre exhibits subtle micro-shifts toward base model training centroid weights.
- High-Frequency Bandwidth Truncation: Audio files frequently show steep spectral roll-offs above 22 kHz even when encoded at 48 kHz container rates.
Как защититься и что предпринять
Upload Audio File or Voice Note to Sealed Rose
Scan MP3, M4A, WAV, or OGG voice messages to evaluate harmonic spectrograms and neural voice clone probability scores.
Часто задаваемые вопросы
Can ElevenLabs v3 deceive family members on phone calls?▼
Yes. Over telephone networks (cellular or VoIP) where audio bandwidth is compressed to 8 kHz (narrowband) or 16 kHz (wideband), ElevenLabs clones sound virtually indistinguishable from real humans. Sealed Rose analyzes uncompressed frequency markers that survive audio compression.
How does Sealed Rose catch ElevenLabs voice clones?▼
We inspect vocal cord jitter, shimmer, harmonic-to-noise ratio (HNR), and vocoder-specific phase discontinuities that cannot be replicated by neural speech synthesizers.
Доступно на 23 языках: