- npm: to build docker
- hugging face account: download pretrained model
- docker: to run web-ui
-
Download pretrained all models from here and copy them to docker_trainer/pretrained.
-
Download pretrained_v2 all models from here and copy them to docker_trainer/pretrained_v2.
-
Download hubert_base.pt from here. Copy it to docker_trainer/hubert.
-
Run
npm run build:docker:trainer
-
Put training data into trainer/raw-data
-
Launch docker image
bash start_trainer.sh
-
In docker container, run
cp -r logs_org/mute logs/
-
In docker container, run
python infer-web.py
-
Access the web-ui by the browser. Open tab for training.
- http://
<address>
:7865/
- http://
-
Enter the speaker's name. Foler name includes the training data is
/rvc/raw-data
. -
Run the training as usual.
-
The weights is in trainer/weights.
.index
is in trainer/logs// -
Run with VC_Client!
Select v2 like as the below image.