A few seconds of audio is now enough to clone a voice. The tools are public, cheap, and better every month.

Hearing is no longer believing.

Fraud calls that sound like your CEO. Fabricated statements that sound like a candidate. Synthetic evidence that sounds like proof.

We study how synthetic speech differs from human speech: the artifacts, the statistics, the traces generation leaves behind. And we build detectors that hold up outside the lab.

For platforms, that means a signal they can act on. For journalists and courts, a way to interrogate a recording. For research, open benchmarks and honest error bars.

What can be synthesized can be studied.
What can be studied can be detected.

We publish what we learn.