https://github.com/bytedance/MegaTTS3


- models
- 
For the VAE encoder model, which enables custom voice cloning without .npy files    



```

└─wavvae
        config.yaml
        
        decoder.ckpt
        
        model_only_last.ckpt

```
- script
- 
python generate_npy.py --input_wav assets/xxx.wav --output_npy assets/xxx.npy