Skip to content

SafeEar

Detect deepfake audio without using speech content

Visit official site

What SafeEar is

The framework is described as an audio deepfake detector that works without relying on speech content.

What you can do with it

  • Choose detection parameters in the interface
  • Use the built benchmark for evaluation

Pros & cons

✓ What we like

  • Multi-language speech content is protected
  • Tests on four benchmark datasets are described

! What to watch out for

  • Detection avoids using speech content
  • An audio sample must be uploaded

FAQ

What does it use for detection?

Acoustic information such as prosody and timbre is named.

What figures are listed?

An EER as low as 2.02 percent and WER above 93.93 percent are stated.

Last reviewed: 2026-11-15

More AI audio tools tools

View all →

How we review