How long does it take for Qwen3-TTS voice clone to generate 2 hours of audio?

In January 2026, a businessman located in the Swiss canton of Schwyz became the victim of a sophisticated financial fraud scheme. The perpetrators successfully defrauded the individual of several million Swiss francs by utilizing manipulated audio to impersonate a trusted business partner during a series of phone calls. This incident highlights the growing risks associated with generative AI-enabled impersonation, as security experts warn that the technology is outpacing current detection and legal frameworks.

What happened

The attack relied on the use of advanced voice-cloning technology to deceive the target. By manipulating audio to mimic the specific voice of a business partner the victim trusted, the fraudsters were able to conduct a series of calls that successfully bypassed the target's suspicions. Research cited by experts, including studies from UC Berkeley, indicates that listeners frequently struggle to distinguish between real human voices and high-quality AI-generated replicas, often performing only slightly better than random guessing. The lack of robust technical mechanisms to verify consent in voice-cloning tools—whether commercial or self-hosted—has exacerbated the potential for such impersonation scams. As the technology becomes more accessible, the barrier to entry for bad actors attempting to conduct financial fraud continues to drop, leaving individuals and organizations increasingly vulnerable to high-fidelity audio deception.