零存储危机隐私保护：您上传的证据影像仅在内存中进行即时神经分析，完成后立即彻底销毁。无日志、无云端备份。
WhatsApp Voice Note Romance Catfish: Spotting AI-Synthesized Voice Memos
Detecting synthetic text-to-speech audio notes and voice changers used by pig butchering syndicates.
使用 Sealed Rose 验证媒体真实性
实时神经网络检测 · 免费 · 无需注册 · 零存储 · 内存瞬时处理
When victims on dating apps become suspicious of fake profiles, their first demand is: "Send me a voice note to prove you are real." In the past, this stopped romance syndicates in their tracks. Today, scammers use AI voice cloning (ElevenLabs, Bark, Speechify) to type any custom message into a prompt box and generate a lifelike, flirty 5-second voice note within seconds. This guide teaches you how to identify the subtle audio artifacts that expose AI-synthesized romance audio.
关键警讯与破绽特征
1Generic, Scripted Voice Notes Without Answering Real-Time Questions
The voice note says pleasantries like "I miss you" or "have a great day", but never addresses complex multi-part questions you asked in the preceding text.
2Complete Absence of Ambient Background Noise
The audio sounds as if recorded in a soundproof studio booth without car noise, wind, keyboard clicks, or room echo.
3Unwillingness to Have a Live Conversational Voice Call
They send multiple recorded voice memos throughout the day, but immediately reject live phone or video calls, claiming anxiety or busy meetings.
Neural Vocoder Artifacts in Compressed Voice Notes
- Robotic Flat Pitch Contours: Text-to-speech models often fail to capture subtle dynamic pitch modulation (prosody), resulting in a slight melodic uniformity across sentences.
- High-Frequency Sibilance Distortion: The "s", "sh", and "f" sounds exhibit a metallic, staticky timbre when compressed by WhatsApp’s Opus voice codec at low bitrates.
- Pre-Recorded Script Libraries: Syndicates maintain shared Google Drive folders containing hundreds of categorized voice notes ("good morning", "eating lunch", "going to gym") recorded by hired female operatives.
防范措施与应对步骤
1. Ask a Hyper-Specific Verbal Challenge
Ask them to send a voice note saying a completely bizarre sentence: "The purple elephant ordered tacos on a Tuesday." Pre-recorded audio libraries cannot answer this, and typing it into an AI tool takes them several minutes.
2. Analyze the Audio File in Sealed Rose Audio Detector
Save the voice memo file (.opus or .ogg) and upload it to Sealed Rose Audio Detector. Our model inspects audio spectral roll-off and synthetic pitch contours.
3. Cross-Check the Sender’s Phone Number
Run their phone number through Sealed Rose Phone Lookup to determine if the number is a VoIP burner or linked to overseas fraud rings.
常见问题
Why do scammers send voice notes instead of calling live?▼
Generating an asynchronous voice note from text takes 10 seconds in a browser. Live bidirectional voice conversation requires expensive low-latency voice-to-voice hardware that frequently glitches during fast dialogue.
Can free tools fake any accent?▼
Yes. Modern neural TTS tools feature hundreds of preset voices spanning French, Russian, British, Latin American, and East Asian accents.
支持 23 种语言：