Website of the project in voicecamo.cs.columbia.edu
Please install the necessary requirements with: pip install -r requirements.txt
We work with the LibriSpeech dataset for this project. To run our code, you will need to download their wav files and transcriptions. You can do this by running the data/librispeech.py file. This will generate the manifest json files. If this step was done correctly, the LibriSpeech dataset will be a folder in the data folder as such: data/LibriSpeech_dataset. This folder will contain test_clean, train, and val folders. Within each of these folders you should have a txt and a wav folder.
The path to this directory has to be introduced in the config file.
The pretrained models reported in our paper can be found by emailing the author at mac2500@columbia.edu, at which point we will grant you access to our google drive link with the checkpoints.
Extract the models under the /path/to/your/checkpoints directory you introduce in
the checkpoint_dir argument in the config file.
To run inference, please just use the command: "python run_inference.py"
To train your own modle, please run the command: "python run_train.py"