metadata

license: mit
datasets:
  - fixie-ai/librispeech_asr
language:
  - en
base_model:
  - facebook/wav2vec2-base
pipeline_tag: voice-activity-detection

Voice Detection AI - Real vs AI Audio Classifier

Model Overview

This model is a fine-tuned Wav2Vec2-based audio classifier capable of distinguishing between real human voices and AI-generated voices. It has been trained on a dataset containing samples from various TTS models and real human audio recordings.

Model Details

Architecture: Wav2Vec2ForSequenceClassification
Fine-tuned on: Custom dataset with real and AI-generated audio
Classes:
1. Real Human Voice
2. AI-generated (e.g., Melgan, DiffWave, etc.)
Input Requirements:
- Audio format: .wav, .mp3, etc.
- Sample rate: 16kHz
- Max duration: 10 seconds (longer audios are truncated, shorter ones are padded)

Performance

Validation Accuracy: 99.8%
Robustness: Successfully classifies across multiple AI-generation models.
Limitations: Struggles with certain unseen AI-generation models (e.g., ElevenLabs).

How to Use

1. Install Dependencies

Make sure you have transformers and torch installed:

pip install transformers torch torchaudio