sherpa-onnx/java-api-examples/NonStreamingTtsVitsDePhonemize.java

// Copyright 2026 Xiaomi Corporation

// This file shows how to use a piper VITS German TTS model to convert text to
// speech.
//
// The text is phonemized outside of sherpa-onnx with piper-phonemize
// (https://github.com/csukuangfj/piper-phonemize) and the resulting phoneme
// codepoints are passed via GenerationConfig.phonemeCodepoints. No lexicon
// file and no espeak-ng-data are needed.
// There is no German lexicon, so this model is phonemize-only.
import com.github.csukuangfj.piper.phonemize.PiperPhonemize;
import com.k2fsa.sherpa.onnx.*;

public class NonStreamingTtsVitsDePhonemize {
  public static void main(String[] args) {
    // please visit
    // https://github.com/k2-fsa/sherpa-onnx/releases/tag/tts-models
    // to download model files
    String model = "./vits-piper-de_DE-glados-low/de_DE-glados-low.onnx";
    String tokens = "./vits-piper-de_DE-glados-low/tokens.txt";
    String text =
                    "Alles hat ein Ende, nur die Wurst hat zwei."
                        + "";

    OfflineTtsVitsModelConfig vitsModelConfig =
        OfflineTtsVitsModelConfig.builder()
            .setModel(model)
            .setTokens(tokens)
            .build();

    OfflineTtsModelConfig modelConfig =
        OfflineTtsModelConfig.builder()
            .setVits(vitsModelConfig)
            .setNumThreads(1)
            .setDebug(true)
            .build();
    OfflineTtsConfig config = OfflineTtsConfig.builder().setModel(modelConfig).build();
    OfflineTts tts = new OfflineTts(config);

    System.out.printf("piper-phonemize version: %s\n", PiperPhonemize.getVersionStr());
    // initialize() extracts the espeak-ng-data embedded in piper-phonemize.
    // That data only feeds the piper phonemizer; sherpa-onnx itself does not
    // use espeak-ng anymore.
    PiperPhonemize.initialize();

    // The espeak-ng voice for the phonemizer is taken from the model metadata.
    // Fall back to de if the model has no language information.
    String voice = tts.getLang();
    if (voice == null || voice.isEmpty()) {
      System.out.println("Warning: the model has no language information. Falling back to de.");
      voice = "de";
    }
    System.out.printf("Phonemizer language: %s\n", voice);

    int[][] phonemeCodepoints;
    try (PiperPhonemize.Result result = PiperPhonemize.text(text, voice)) {
      if (result == null) {
        throw new RuntimeException("Failed to phonemize the text");
      }
      phonemeCodepoints = new int[result.getNumSentences()][];
      for (int i = 0; i < result.getNumSentences(); i++) {
        phonemeCodepoints[i] = result.getPhonemes(i);
      }
    }
    System.out.printf("Number of phonemized sentences: %d\n", phonemeCodepoints.length);

    int sid = 0;
    float speed = 1.0f;
    GenerationConfig genConfig = new GenerationConfig();
    genConfig.setSid(sid);
    genConfig.setSpeed(speed);
    genConfig.setSilenceScale(config.getSilenceScale());
    genConfig.setPhonemeCodepoints(phonemeCodepoints);

    long start = System.currentTimeMillis();
    // The text has already been phonemized; pass an empty string here.
    GeneratedAudio audio = tts.generateWithConfig("", genConfig);
    long stop = System.currentTimeMillis();

    float timeElapsedSeconds = (stop - start) / 1000.0f;

    float audioDuration = audio.getSamples().length / (float) audio.getSampleRate();
    float realTimeFactor = timeElapsedSeconds / audioDuration;

    String waveFilename = "tts-vits-de-phonemize.wav";
    audio.save(waveFilename);
    System.out.printf("-- elapsed : %.3f seconds\n", timeElapsedSeconds);
    System.out.printf("-- audio duration: %.3f seconds\n", audioDuration);
    System.out.printf("-- real-time factor (RTF): %.3f\n", realTimeFactor);
    System.out.printf("-- text: %s\n", text);
    System.out.printf("-- Saved to %s\n", waveFilename);

    tts.release();
  }
}