Skip to content

Nes2Net

A lightweight architecture for speech anti-spoofing

Visit official site

What Nes2Net is

The model is a lightweight nested architecture for speech anti-spoofing driven by foundation models, aimed at audio deepfake detection. Pretrained models and code are released on GitHub.

What you can do with it

  • Run simple inference on audio
  • Train the model on a custom dataset
  • Evaluate results with the provided tools

How to start

  • Clone the library locally
  • Install the required packages
  • Fetch pretrained weights and run the inference script

Pros & cons

✓ What we like

  • Conda and Pip environments are supported
  • Evaluation tools compute EER and related metrics

! What to watch out for

  • Pretrained weights must be downloaded to a set path
  • Environment setup uses Conda or Pip

FAQ

What is it designed for?

The material describes speech anti-spoofing and audio deepfake detection.

Where are models published?

The material states pretrained models and code are on GitHub.

Last reviewed: 2026-11-15

More AI audio tools tools

View all →

How we review