foreword
This is a real-time voice changer based on AI algorithm. If you don’t understand AI, it’s okay, just use the one-click installation package I provided.
Link: https://pan.baidu.com/s/1f3X6JdBVOgeTNPf0B3CRKg
Extraction code: k5v2
Voice changer installation and use
There are two voice changers, both based on RVC. We use the enhanced version, but the basic version is used for training. The enhanced version is developed by a Japanese.
train
If you want to train the voice you want, or you can look at your own voice, if you don’t want to train, there are ready-made models, don’t worry
- First we come to the source code interface
https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI
- There are instructions in the drop-down, you can use the packaged files directly, that is, what is the red box
Click on the red box to download
- After decompression, click go-web.bat, you can train your own voice, the above go-realtime-gui.bat file is the built-in voice changer, but the effect is too bad, don’t worry about it,
- Go to training page
- Change the name of the experiment to whatever you want
- Sound data (clean human voice, no noise), it is best to split it, one per minute
- Choose the V2 version, optimized on the basis of V1
- The number of training rounds can be 50 or 100
- The batch_size of each graphics card is the default, if the gpu can't move it down
- One key training
- After completion, the files are all in VC-fix1.53b\RVC-beta-v2-0528\logs\peiyin
- Index files and weights are inside
use
https://huggingface.co/wok000/vcclient000/tree/main
- directly download the latest
- We also need to download a Voicemeeter virtual driver
Baidu search for the Chinese cracked version, there are many, of course, installation packages are also provided, all in the package file
- Click Install and it will be installed automatically
- Then open the sound control panel, right-click to disable the first and third virtual drivers in the playback, the default device is still your own earphones or speakers do not need to move
- Same as recording, right click to disable 1, 3, 2 to the default device (right click properties, select listen and tick), others do not change
- Open the virtual drive just installed in the lower left corner, click the red box
- Choose your own device for input and output, I choose earphones, this kt is my earphones
- Download the decompressed file on behalf of Kai, click the voice changer to run locally, if it is a file downloaded by yourself, the name is start_http.bat,
- The first run will download the configuration file, and it will start after downloading. It comes with four models in the models folder.
Select RVC, enter
After the start, the voice will change when speaking, and it’s almost like this. If you have any questions, please ask in the comment area!