Add Java API for Matcha-TTS models. (#1673)
This commit is contained in:
@@ -0,0 +1,116 @@
|
||||
// Copyright 2025 Xiaomi Corporation
|
||||
|
||||
package com.k2fsa.sherpa.onnx;
|
||||
|
||||
public class OfflineTtsMatchaModelConfig {
|
||||
private final String acousticModel;
|
||||
private final String vocoder;
|
||||
private final String lexicon;
|
||||
private final String tokens;
|
||||
private final String dataDir;
|
||||
private final String dictDir;
|
||||
private final float noiseScale;
|
||||
private final float lengthScale;
|
||||
|
||||
private OfflineTtsMatchaModelConfig(Builder builder) {
|
||||
this.acousticModel = builder.acousticModel;
|
||||
this.vocoder = builder.vocoder;
|
||||
this.lexicon = builder.lexicon;
|
||||
this.tokens = builder.tokens;
|
||||
this.dataDir = builder.dataDir;
|
||||
this.dictDir = builder.dictDir;
|
||||
this.noiseScale = builder.noiseScale;
|
||||
this.lengthScale = builder.lengthScale;
|
||||
}
|
||||
|
||||
public static Builder builder() {
|
||||
return new Builder();
|
||||
}
|
||||
|
||||
public String getAcousticModel() {
|
||||
return acousticModel;
|
||||
}
|
||||
|
||||
public String getVocoder() {
|
||||
return vocoder;
|
||||
}
|
||||
|
||||
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 static class Builder {
|
||||
private String acousticModel = "";
|
||||
private String vocoder = "";
|
||||
private String lexicon = "";
|
||||
private String tokens = "";
|
||||
private String dataDir = "";
|
||||
private String dictDir = "";
|
||||
private float noiseScale = 1.0f;
|
||||
private float lengthScale = 1.0f;
|
||||
|
||||
public OfflineTtsMatchaModelConfig build() {
|
||||
return new OfflineTtsMatchaModelConfig(this);
|
||||
}
|
||||
|
||||
public Builder setAcousticModel(String acousticModel) {
|
||||
this.acousticModel = acousticModel;
|
||||
return this;
|
||||
}
|
||||
|
||||
public Builder setVocoder(String vocoder) {
|
||||
this.vocoder = vocoder;
|
||||
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 setLengthScale(float lengthScale) {
|
||||
this.lengthScale = lengthScale;
|
||||
return this;
|
||||
}
|
||||
}
|
||||
}
|
||||
@@ -4,12 +4,14 @@ package com.k2fsa.sherpa.onnx;
|
||||
|
||||
public class OfflineTtsModelConfig {
|
||||
private final OfflineTtsVitsModelConfig vits;
|
||||
private final OfflineTtsMatchaModelConfig matcha;
|
||||
private final int numThreads;
|
||||
private final boolean debug;
|
||||
private final String provider;
|
||||
|
||||
private OfflineTtsModelConfig(Builder builder) {
|
||||
this.vits = builder.vits;
|
||||
this.matcha = builder.matcha;
|
||||
this.numThreads = builder.numThreads;
|
||||
this.debug = builder.debug;
|
||||
this.provider = builder.provider;
|
||||
@@ -23,8 +25,13 @@ public class OfflineTtsModelConfig {
|
||||
return vits;
|
||||
}
|
||||
|
||||
public OfflineTtsMatchaModelConfig getMatcha() {
|
||||
return matcha;
|
||||
}
|
||||
|
||||
public static class Builder {
|
||||
private OfflineTtsVitsModelConfig vits = OfflineTtsVitsModelConfig.builder().build();
|
||||
private OfflineTtsMatchaModelConfig matcha = OfflineTtsMatchaModelConfig.builder().build();
|
||||
private int numThreads = 1;
|
||||
private boolean debug = true;
|
||||
private String provider = "cpu";
|
||||
@@ -38,6 +45,11 @@ public class OfflineTtsModelConfig {
|
||||
return this;
|
||||
}
|
||||
|
||||
public Builder setMatcha(OfflineTtsMatchaModelConfig matcha) {
|
||||
this.matcha = matcha;
|
||||
return this;
|
||||
}
|
||||
|
||||
public Builder setNumThreads(int numThreads) {
|
||||
this.numThreads = numThreads;
|
||||
return this;
|
||||
|
||||
@@ -60,9 +60,9 @@ public class OfflineTtsVitsModelConfig {
|
||||
}
|
||||
|
||||
public static class Builder {
|
||||
private String model;
|
||||
private String model = "";
|
||||
private String lexicon = "";
|
||||
private String tokens;
|
||||
private String tokens = "";
|
||||
private String dataDir = "";
|
||||
private String dictDir = "";
|
||||
private float noiseScale = 0.667f;
|
||||
|
||||
Reference in New Issue
Block a user