Add Java API for text-to-speech (#811)

This commit is contained in:
Fangjun Kuang
2024-04-26 09:26:39 +08:00
committed by GitHub
parent fa2429920f
commit 15772d2150
15 changed files with 692 additions and 0 deletions

View File

@@ -0,0 +1,32 @@
// Copyright 2024 Xiaomi Corporation
package com.k2fsa.sherpa.onnx;
public class GeneratedAudio {
static {
System.loadLibrary("sherpa-onnx-jni");
}
private final float[] samples;
private final int sampleRate;
public GeneratedAudio(float[] samples, int sampleRate) {
this.samples = samples;
this.sampleRate = sampleRate;
}
public int getSampleRate() {
return sampleRate;
}
public float[] getSamples() {
return samples;
}
// return true if saved successfully.
public boolean save(String filename) {
return saveImpl(filename, samples, sampleRate);
}
private native boolean saveImpl(String filename, float[] samples, int sampleRate);
}

View File

@@ -0,0 +1,53 @@
// Copyright 2024 Xiaomi Corporation
package com.k2fsa.sherpa.onnx;
public class OfflineTts {
static {
System.loadLibrary("sherpa-onnx-jni");
}
private long ptr = 0; // this is the asr engine ptrss
public OfflineTts(OfflineTtsConfig config) {
ptr = newFromFile(config);
}
public GeneratedAudio generate(String text) {
return generate(text, 0, 1.0f);
}
public GeneratedAudio generate(String text, int sid) {
return generate(text, sid, 1.0f);
}
public GeneratedAudio generate(String text, int sid, float speed) {
Object[] arr = generateImpl(ptr, text, sid, speed);
float[] samples = (float[]) arr[0];
int sampleRate = (int) arr[1];
return new GeneratedAudio(samples, sampleRate);
}
@Override
protected void finalize() throws Throwable {
release();
}
public void release() {
if (this.ptr == 0) {
return;
}
delete(this.ptr);
this.ptr = 0;
}
private native void delete(long ptr);
private native int getSampleRate(long ptr);
private native int getNumSpeakers(long ptr);
private native Object[] generateImpl(long ptr, String text, int sid, float speed);
private native long newFromFile(OfflineTtsConfig config);
}

View File

@@ -0,0 +1,68 @@
// Copyright 2024 Xiaomi Corporation
package com.k2fsa.sherpa.onnx;
public class OfflineTtsConfig {
private final OfflineTtsModelConfig model;
private final String ruleFsts;
private final String ruleFars;
private final int maxNumSentences;
private OfflineTtsConfig(Builder builder) {
this.model = builder.model;
this.ruleFsts = builder.ruleFsts;
this.ruleFars = builder.ruleFars;
this.maxNumSentences = builder.maxNumSentences;
}
public static Builder builder() {
return new Builder();
}
public OfflineTtsModelConfig getModel() {
return model;
}
public String getRuleFsts() {
return ruleFsts;
}
public String getRuleFars() {
return ruleFars;
}
public int getMaxNumSentences() {
return maxNumSentences;
}
public static class Builder {
private OfflineTtsModelConfig model = OfflineTtsModelConfig.builder().build();
private String ruleFsts = "";
private String ruleFars = "";
private int maxNumSentences = 1;
public OfflineTtsConfig build() {
return new OfflineTtsConfig(this);
}
public Builder setModel(OfflineTtsModelConfig model) {
this.model = model;
return this;
}
public Builder setRuleFsts(String ruleFsts) {
this.ruleFsts = ruleFsts;
return this;
}
public Builder setRuleFars(String ruleFars) {
this.ruleFars = ruleFars;
return this;
}
public Builder setMaxNumSentences(int maxNumSentences) {
this.maxNumSentences = maxNumSentences;
return this;
}
}
}

View File

@@ -0,0 +1,56 @@
// Copyright 2024 Xiaomi Corporation
package com.k2fsa.sherpa.onnx;
public class OfflineTtsModelConfig {
private final OfflineTtsVitsModelConfig vits;
private final int numThreads;
private final boolean debug;
private final String provider;
private OfflineTtsModelConfig(Builder builder) {
this.vits = builder.vits;
this.numThreads = builder.numThreads;
this.debug = builder.debug;
this.provider = builder.provider;
}
public static Builder builder() {
return new Builder();
}
public OfflineTtsVitsModelConfig getVits() {
return vits;
}
public static class Builder {
private OfflineTtsVitsModelConfig vits = OfflineTtsVitsModelConfig.builder().build();
private int numThreads = 1;
private boolean debug = true;
private String provider = "cpu";
public OfflineTtsModelConfig build() {
return new OfflineTtsModelConfig(this);
}
public Builder setVits(OfflineTtsVitsModelConfig vits) {
this.vits = vits;
return this;
}
public Builder setNumThreads(int numThreads) {
this.numThreads = numThreads;
return this;
}
public Builder setDebug(boolean debug) {
this.debug = debug;
return this;
}
public Builder setProvider(String provider) {
this.provider = provider;
return this;
}
}
}

View File

@@ -0,0 +1,116 @@
// Copyright 2024 Xiaomi Corporation
package com.k2fsa.sherpa.onnx;
public class OfflineTtsVitsModelConfig {
private final String model;
private final String lexicon;
private final String tokens;
private final String dataDir;
private final String dictDir;
private final float noiseScale;
private final float noiseScaleW;
private final float lengthScale;
private OfflineTtsVitsModelConfig(Builder builder) {
this.model = builder.model;
this.lexicon = builder.lexicon;
this.tokens = builder.tokens;
this.dataDir = builder.dataDir;
this.dictDir = builder.dictDir;
this.noiseScale = builder.noiseScale;
this.noiseScaleW = builder.noiseScaleW;
this.lengthScale = builder.lengthScale;
}
public static Builder builder() {
return new Builder();
}
public String getModel() {
return model;
}
public String getLexicon() {
return lexicon;
}
public String getTokens() {
return tokens;
}
public String getDataDir() {
return dataDir;
}
public String getDictDir() {
return dictDir;
}
public float getLengthScale() {
return lengthScale;
}
public float getNoiseScale() {
return noiseScale;
}
public float getNoiseScaleW() {
return noiseScaleW;
}
public static class Builder {
private String model;
private String lexicon = "";
private String tokens;
private String dataDir = "";
private String dictDir = "";
private float noiseScale = 0.667f;
private float noiseScaleW = 0.8f;
private float lengthScale = 1.0f;
public OfflineTtsVitsModelConfig build() {
return new OfflineTtsVitsModelConfig(this);
}
public Builder setModel(String model) {
this.model = model;
return this;
}
public Builder setTokens(String tokens) {
this.tokens = tokens;
return this;
}
public Builder setLexicon(String lexicon) {
this.lexicon = lexicon;
return this;
}
public Builder setDataDir(String dataDir) {
this.dataDir = dataDir;
return this;
}
public Builder setDictDir(String dictDir) {
this.dictDir = dictDir;
return this;
}
public Builder setNoiseScale(float noiseScale) {
this.noiseScale = noiseScale;
return this;
}
public Builder setNoiseScaleW(float noiseScaleW) {
this.noiseScaleW = noiseScaleW;
return this;
}
public Builder setLengthScale(float lengthScale) {
this.lengthScale = lengthScale;
return this;
}
}
}

View File

@@ -1,4 +1,5 @@
// Copyright 2024 Xiaomi Corporation
package com.k2fsa.sherpa.onnx;
public class WaveReader {