PiperJNI

A JNI wrapper for piper a local neural text to speech system, so it can be used on Java projects.

Platform support

This library aims to support the following platforms:

Windows x86_64
Linux GLIBC x86_64/arm64 (built with debian focal, GLIBC version 2.31)
macOS x86_64/arm64 (built targeting v11.0)

The jar includes the piper-jni native library, which includes the piper source, and the shared libraries it depends on: espeak, piper-phonetize and onxxruntime.

Installation

The package is distributed through Maven Central.

You can also find the package's jar attached to each release.

Basic Example

A basic example available at the tests.

        ...
        String voiceModel = System.getenv("VOICE_MODEL");
        String voiceModelConfig = System.getenv("VOICE_MODEL_CONFIG");
        String textToSpeak = System.getenv("TEXT_TO_SPEAK");
        if(voiceModel == null || voiceModel.isBlank()) {
            throw new ConfigurationException("env var VOICE_MODEL is required");
        }
        if(voiceModelConfig == null || voiceModelConfig.isBlank()) {
            throw new ConfigurationException("env var VOICE_MODEL_CONFIG is required");
        }
        if(textToSpeak == null || textToSpeak.isBlank()) {
            throw new ConfigurationException("env var TEXT_TO_SPEAK is required");
        }
        try (var config = piper.createConfig()) {
            try (var voice = piper.loadVoice(config, Paths.get(voiceModel), Path.of(voiceModelConfig), 0)) {
                int sampleRate = voice.getSampleRate();
                piper.initialize(config, voice);
                short[] samples = piper.textToAudio(config, voice, textToSpeak);
                piper.terminate(config);
                createWAVFile(samples, sampleRate, Path.of("test.wav"));
            }
        }
        ...

Building the project from source.

You need to have Java and Cpp setup.

After cloning the project you need to init the piper submodule by running:

git submodule update --init

Build the piper-jni library:

# You need to set the correct resources folder as install prefix.
cmake -Bbuild -DCMAKE_INSTALL_PREFIX=src\main\resources\debian-amd64
cmake --build build --config Release
cmake --install build

Then you need to download a piper voice and its config.

Finally, you can run the project tests to confirm it works:

# Path to piper voice model
export VOICE_MODEL=/test-data/es_ES-sharvard-medium.onnx
# Path to piper voice model config
export VOICE_MODEL_CONFIG=/test-data/es_ES-sharvard-medium.onnx.json
# Text to speak
export TEXT_TO_SPEAK="Buenos días"
# tests will generate test.wav in the root dir.
mvn test

Extending the native api

If you want to add any missing piper functionality you need to:

Add the native method description in src/main/java/io/github/givimad/piperjni/PiperJNI.java.
Run the gen_header.sh script to regenerate the src/main/native/io_github_givimad_piperjni_PiperJNI.h header file.
Add the native method implementation in src/main/native/io_github_givimad_piperjni_PiperJNI.cpp.
Add a new test for it at src/test/java/io/github/givimad/piperjni/PiperJNITest.java.

GiviMAD / piper-jni