This repository contains the complete code for "WAKE: Watermarking Audio with Key Enrichment."

Environment Setup:

To set up the environment, please use the provided environment.yml file:

conda env create -f environment.yml

Pretrained Models:

Due to the size limitations of the uploaded materials, we have not included the model's checkpoint files. Upon the acceptance of our paper, we will promptly release the ckpt files and make all the code publicly available.

Training Script:

The training script can be found in script/train.py. You can easily replace the audio dataset with your own for training. However, please note that you must first convert the audio sample rate to 16,000 and ensure the length is 1 second.

Testing Script:

You can choose your own audio or use the provided test.wav file for testing. To test both single and double watermarking, run the following command:

python test.py