Skip to content

libvits-ncnn is an ncnn implementation of the VITS library that enables cross-platform GPU-accelerated speech synthesis.🎙️💻

License

Notifications You must be signed in to change notification settings

Sg4Dylan/libvits-ncnn

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

8 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

libvits-ncnn 🎙️

This is an ncnn implementation of the VITS library that supports cross-platform GPU-accelerated speech synthesis.
The project is forked from weirdseed/Vits-Android-ncnn, Thanks to the original author for their contribution.

📖 Usage

  1. 🔍 Prepare dependencies

    Get the ncnn static library suitable for your runtime environment from its wiki or ncnn/releases.
    Place it in the root directory ncnn folder like this:

    ncnn
    ├─bin
    ├─include
    └─lib
    
  2. 🛠️ Compile the project

    a. libvits-ncnn
    Execute in the repo root directory:

    mkdir build && cd build  
    cmake .. -DCMAKE_CXX_COMPILER=/usr/bin/clang++ -DCMAKE_C_COMPILER=/usr/bin/clang  
    make
    

    After compilation, you can find libvits-ncnn.so in the build directory.

    b. vits-cli
    Enter the demo directory and execute:

    mkdir build && cd build  
    cmake .. -DCMAKE_CXX_COMPILER=/usr/bin/clang++ -DCMAKE_C_COMPILER=/usr/bin/clang  
    make
    

    After compilation, you can find vits-cli in the build directory.

  3. 🚀 Run the demo

    In the directory where vits-cli is located, prepare the dependencies required for running:

    a. Download the openjtalk dictionary file and unzip it;
    b. Download the VITS ncnn model, unzip the Atri part to the atri/ directory (for testing the monophonic model), and extract 365_epochs to the 365_epochs/ directory (for testing the multi-tone model).
    c. Download the VITS ncnn params (single/multi directories).

    At this time, the directory has:

    build
    ├─vits-cli
    ├─365_epochs
    ├─atri
    ├─multi
    ├─open_jtalk_dic_utf_8-1.11
    └─single
    

    Now execute ./vits-cli.

🔧 Main changes

  • JNI export interface changed to C++
  • C++ re-implemented preprocessing

📝 TODO

  • Support for Japanese monophonic model
  • Support for Japanese multi-tone model
  • Support for Chinese monophonic model
  • Support for Chinese multi-tone model

About

libvits-ncnn is an ncnn implementation of the VITS library that enables cross-platform GPU-accelerated speech synthesis.🎙️💻

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published