Pre-requisites

1. Python == 3.8
2. Clone this repository.
3. Install python requirements. Please refer requirements.txt
4. Download a 48k dataset, such as genshin or VCTK.

Training

python train.py \
--config config_v1_16k_to_48k.json \
--input_wavs_dir VCTK-Corpus/wav48/,genshin --checkpoint_path exp/v1_16k_to_48k/

To train 24k_to_48k, replace config_v1_16k_to_48k.json with config_v1_24k_to_48k.json.
Checkpoints and copy of the configuration file are saved in checkpoint_path directory by default.
You can change the path by adding --checkpoint_path option.

SR model sample theory

The hifigan means hiftnet here.

SR results

Dir of gen_from_wav is the generated wavs, which sound good, and better than hifigan-sr.

• origin 16k mel-spectrum
• generated 48k mel-spectrum

Pretrained Model

1. The pretrained models provided is in "exp/v1_16k_to_48k/g_bst", trained with StarRail_Datasets and VCTK.
2. For i don't have GPU resources, a kind person(@Lucy) train config_v1_16k_to_48k version and trained stop at 300k, maybe training to 800k is better.

Inference from wav file

1. Make test_files directory and copy wav files into the directory.
2. Run the following command.

   # in inference.py, change the 'cp_path' param to your checkpoint dir.
   python inference.py
   

Generated wav files are saved in generated_files by default.
You can change the path by adding --output_dir option.

Reference

Our repository is heavily based on yl4579 's HiFTNet.