SayaSS / vits-finetuning

Fine-Tuning your VITS model using a pre-trained model
MIT License
546 stars 86 forks source link

Where should the dataset file be placed? #1

Closed Camille1534 closed 1 year ago

SayaSS commented 1 year ago

Depends on your settings

path/to/XXX.wav|speaker id|transcript
python preprocess.py --filelists path/to/filelist_train.txt path/to/filelist_val.txt

path/to/XXX.wav, path/to/filelist_val.txt path/to/filelist_train.txt Both relative and absolute paths are acceptable.

Here is an example.

filelists/miyu_train.txt:

wav/ba/miyu/533522.wav|10|ブ、ブルー……アーカイブ……
wav/ba/miyu/978431.wav|10|SRT特殊学園、ラビット小隊の霞沢ミユ、です…… あの、もう帰……っちゃ、だめですよね……
wav/ba/miyu/589002.wav|10|お、お疲れ様です、先生……
......

filelists/miyu_val.txt:

wav/ba/miyu/916145.wav|10|ど、どしてこんなことに…
wav/ba/miyu/139528.wav|10|え、援護します
......
└── vits-finetuning/
    ├── filelists/
    │   ├── miyu_train.txt
    │   └── miyu_val.txt
    └── wav/
        └── ba/
            └── miyu/
                ├── 533522.wav
                ├── 978431.wav
                ├── 589002.wav
                └── 916145.wav
                └── 139528.wav
                └── ......