Voice AntiSpoofing with RawNet2
pip install -r ./requirements.txtFor this project Hydra is supported, so one can find corresponding configs for training and testing in the hw_as/configs folder.
python train.pyYou need to change the path to the model's checkpoint and list audios you want to test in a hw_as/configs/test.yaml
python test.pyBelow you can see audios with probabilities of a bonafide class. There are two audios generated by FastSpeech2, two audios synthesized by HiFiGAN, one bonafide audio and three audios from teachers.
- as_examples/tts_audio1.wav
tts_audio1.mp4
- as_examples/tts_audio2.wav
tts_audio2.mp4
- as_examples/nv_audio1.wav
nv_audio1.mp4
- as_examples/nv_audio2.wav
nv_audio2.mp4
- as_examples/bonafide_gf.wav
bonafide_gf.mp4
- as_examples/blinorot_audio1.wav
blinorot_audio1.mp4
- as_examples/blinorot_audio2.wav
blinorot_audio2.mp4
- as_examples/blinorot_audio3.wav