AI Applied: Covering AI News, Interviews and Tools - ChatGPT, Midjourney, Gemini, OpenAI, Anthropic AI Voice Cloning: Scams, Protections, and Truecaller
9 snips
Mar 24, 2026 A dive into how seconds of audio can produce convincing AI voice clones and why that matters for scams. A live demo shows voice synthesis risks and how video deepfakes can amplify them. Practical protections are discussed, from call-backs and secret code words to Truecaller’s AI-powered hang-up features. The conversation also touches on generational vulnerability and real-world scam examples.
AI Snips
Chapters
Transcript
Episode notes
Host Spends Big On Voice Tech And Gets Grants
- Jaeden spends heavily on 11 Labs to clone his own voice for projects and received startup credits from them.
- He mentions being on a $1,300/month tier and receiving $8,000 in Eleven Labs tokens across two grants.
Voice Cloning Works With Just Seconds Of Audio
- Voice cloning is trivially achievable with only a few seconds of audio, making virtually anyone vulnerable.
- Jaeden demonstrates Quen 3 TTS clones voices in ~3 seconds and warns open-source models lack enforcement or consent checks.
Responsible Services Versus Unregulated Open Models
- Responsible providers like Eleven Labs enforce verification steps for cloning, but open-source alternatives bypass those controls.
- Jaeden contrasts 11 Labs' multi-sentence live-read checks with Quen 3 TTS APIs that accept short clips without authorization.