PyTorch Implementation of WMCodec: End-to-End Neural Speech Codec with Deep Watermarking for Authenticity Verification
pip install -r requirement.txt
We are using the LibriTTS dataset.
Modify the parameter --input_training_file --input_validation_file --checkpoint_path in train.py
Modify the parameter --input_wavs_dir --output_dir --checkpoint_file in inference-at.py
Modify the config.json
The watermark configuration is in the watermark.py, defaulting to 4-digit base-16.
python train.py
python inference-at.py
This implementation uses parts of the code from the following Github repos: TraceableSpeech
If you find this code useful in your research, please consider citing:
@misc{zhou2024wmcodecendtoendneuralspeech,
title={WMCodec: End-to-End Neural Speech Codec with Deep Watermarking for Authenticity Verification},
author={Junzuo Zhou and Jiangyan Yi and Yong Ren and Jianhua Tao and Tao Wang and Chu Yuan Zhang},
year={2024},
eprint={2409.12121},
archivePrefix={arXiv},
primaryClass={cs.SD},
url={https://arxiv.org/abs/2409.12121},
}