Add Java API for non-streaming ASR (#807)

.github/workflows/run-java-test.yaml

@@ -92,13 +92,9 @@ jobs:
             make -j4
             ls -lh lib
 
-      - name:  Run java test
+      - name:  Run java test (Streaming ASR)
         shell: bash
         run: |
-          export CMAKE_CXX_COMPILER_LAUNCHER=ccache
-          export PATH="/usr/lib/ccache:/usr/local/opt/ccache/libexec:$PATH"
-          cmake --version
-
           cd ./java-api-examples
           ./run-streaming-decode-file-ctc.sh
           # Delete model files to save space
@@ -109,3 +105,16 @@ jobs:
 
           ./run-streaming-decode-file-transducer.sh
           rm -rf sherpa-onnx-streaming-*
+
+      - name:  Run java test (Non-Streaming ASR)
+        shell: bash
+        run: |
+          cd ./java-api-examples
+          ./run-non-streaming-decode-file-paraformer.sh
+          rm -rf sherpa-onnx-paraformer-zh-*
+
+          ./run-non-streaming-decode-file-transducer.sh
+          rm -rf sherpa-onnx-zipformer-*
+
+          ./run-non-streaming-decode-file-whisper.sh
+          rm -rf sherpa-onnx-whisper-*

java-api-examples/.gitignore

@@ -1,3 +1,4 @@
 lib
 hs_err*
 !run-streaming*.sh
+!run-non-streaming*.sh

java-api-examples/NonStreamingDecodeFileParaformer.java

@@ -0,0 +1,49 @@
+// Copyright 2024 Xiaomi Corporation
+
+// This file shows how to use an offline paraformer, i.e., non-streaming paraformer,
+// to decode files.
+import com.k2fsa.sherpa.onnx.*;
+
+public class NonStreamingDecodeFileTransducer {
+  public static void main(String[] args) {
+    // please refer to
+    // https://k2-fsa.github.io/sherpa/onnx/pretrained_models/offline-paraformer/paraformer-models.html#csukuangfj-sherpa-onnx-paraformer-zh-2023-03-28-chinese-english
+    // to download model files
+    String model = "./sherpa-onnx-paraformer-zh-2023-03-28/model.int8.onnx";
+    String tokens = "./sherpa-onnx-paraformer-zh-2023-03-28/tokens.txt";
+
+    String waveFilename = "./sherpa-onnx-paraformer-zh-2023-03-28/test_wavs/3-sichuan.wav";
+
+    WaveReader reader = new WaveReader(waveFilename);
+
+    OfflineParaformerModelConfig paraformer =
+        OfflineParaformerModelConfig.builder().setModel(model).build();
+
+    OfflineModelConfig modelConfig =
+        OfflineModelConfig.builder()
+            .setParaformer(paraformer)
+            .setTokens(tokens)
+            .setNumThreads(1)
+            .setDebug(true)
+            .build();
+
+    OfflineRecognizerConfig config =
+        OfflineRecognizerConfig.builder()
+            .setOfflineModelConfig(modelConfig)
+            .setDecodingMethod("greedy_search")
+            .build();
+
+    OfflineRecognizer recognizer = new OfflineRecognizer(config);
+    OfflineStream stream = recognizer.createStream();
+    stream.acceptWaveform(reader.getSamples(), reader.getSampleRate());
+
+    recognizer.decode(stream);
+
+    String text = recognizer.getResult(stream).getText();
+
+    System.out.printf("filename:%s\nresult:%s\n", waveFilename, text);
+
+    stream.release();
+    recognizer.release();
+  }
+}

java-api-examples/NonStreamingDecodeFileTransducer.java

@@ -0,0 +1,57 @@
+// Copyright 2024 Xiaomi Corporation
+
+// This file shows how to use an offline transducer, i.e., non-streaming transducer,
+// to decode files.
+import com.k2fsa.sherpa.onnx.*;
+
+public class NonStreamingDecodeFileTransducer {
+  public static void main(String[] args) {
+    // please refer to
+    // https://k2-fsa.github.io/sherpa/onnx/pretrained_models/offline-transducer/zipformer-transducer-models.html#sherpa-onnx-zipformer-gigaspeech-2023-12-12-english
+    // to download model files
+    String encoder =
+        "./sherpa-onnx-zipformer-gigaspeech-2023-12-12/encoder-epoch-30-avg-1.int8.onnx";
+    String decoder = "./sherpa-onnx-zipformer-gigaspeech-2023-12-12/decoder-epoch-30-avg-1.onnx";
+    String joiner = "./sherpa-onnx-zipformer-gigaspeech-2023-12-12/joiner-epoch-30-avg-1.onnx";
+    String tokens = "./sherpa-onnx-zipformer-gigaspeech-2023-12-12/tokens.txt";
+
+    String waveFilename =
+        "./sherpa-onnx-zipformer-gigaspeech-2023-12-12/test_wavs/1089-134686-0001.wav";
+
+    WaveReader reader = new WaveReader(waveFilename);
+
+    OfflineTransducerModelConfig transducer =
+        OfflineTransducerModelConfig.builder()
+            .setEncoder(encoder)
+            .setDecoder(decoder)
+            .setJoiner(joiner)
+            .build();
+
+    OfflineModelConfig modelConfig =
+        OfflineModelConfig.builder()
+            .setTransducer(transducer)
+            .setTokens(tokens)
+            .setNumThreads(1)
+            .setDebug(true)
+            .build();
+
+    OfflineRecognizerConfig config =
+        OfflineRecognizerConfig.builder()
+            .setOfflineModelConfig(modelConfig)
+            .setDecodingMethod("greedy_search")
+            .build();
+
+    OfflineRecognizer recognizer = new OfflineRecognizer(config);
+    OfflineStream stream = recognizer.createStream();
+    stream.acceptWaveform(reader.getSamples(), reader.getSampleRate());
+
+    recognizer.decode(stream);
+
+    String text = recognizer.getResult(stream).getText();
+
+    System.out.printf("filename:%s\nresult:%s\n", waveFilename, text);
+
+    stream.release();
+    recognizer.release();
+  }
+}

java-api-examples/NonStreamingDecodeFileWhisper.java

@@ -0,0 +1,50 @@
+// Copyright 2024 Xiaomi Corporation
+
+// This file shows how to use an offline whisper, i.e., non-streaming whisper,
+// to decode files.
+import com.k2fsa.sherpa.onnx.*;
+
+public class NonStreamingDecodeFileWhisper {
+  public static void main(String[] args) {
+    // please refer to
+    // https://k2-fsa.github.io/sherpa/onnx/pretrained_models/whisper/tiny.en.html
+    // to download model files
+    String encoder = "./sherpa-onnx-whisper-tiny.en/tiny.en-encoder.int8.onnx";
+    String decoder = "./sherpa-onnx-whisper-tiny.en/tiny.en-decoder.int8.onnx";
+    String tokens = "./sherpa-onnx-whisper-tiny.en/tiny.en-tokens.txt";
+
+    String waveFilename = "./sherpa-onnx-whisper-tiny.en/test_wavs/1.wav";
+
+    WaveReader reader = new WaveReader(waveFilename);
+
+    OfflineWhisperModelConfig whisper =
+        OfflineWhisperModelConfig.builder().setEncoder(encoder).setDecoder(decoder).build();
+
+    OfflineModelConfig modelConfig =
+        OfflineModelConfig.builder()
+            .setWhisper(whisper)
+            .setTokens(tokens)
+            .setNumThreads(1)
+            .setDebug(true)
+            .build();
+
+    OfflineRecognizerConfig config =
+        OfflineRecognizerConfig.builder()
+            .setOfflineModelConfig(modelConfig)
+            .setDecodingMethod("greedy_search")
+            .build();
+
+    OfflineRecognizer recognizer = new OfflineRecognizer(config);
+    OfflineStream stream = recognizer.createStream();
+    stream.acceptWaveform(reader.getSamples(), reader.getSampleRate());
+
+    recognizer.decode(stream);
+
+    String text = recognizer.getResult(stream).getText();
+
+    System.out.printf("filename:%s\nresult:%s\n", waveFilename, text);
+
+    stream.release();
+    recognizer.release();
+  }
+}

java-api-examples/README.md

@@ -4,8 +4,18 @@ This directory contains examples for the JAVA API of sherpa-onnx.
 
 # Usage
 
+## Streaming Speech recognition
+
 ```
 ./run-streaming-decode-file-ctc.sh
 ./run-streaming-decode-file-paraformer.sh
 ./run-streaming-decode-file-transducer.sh
 ```
+
+## Non-Streaming Speech recognition
+
+```bash
+./run-non-streaming-decode-file-paraformer.sh
+./run-non-streaming-decode-file-transducer.sh
+./run-non-streaming-decode-file-whisper.sh
+```

java-api-examples/StreamingDecodeFileCtc.java

@@ -1,4 +1,3 @@
-// Copyright 2022-2023 by zhaoming
 // Copyright 2024 Xiaomi Corporation
 
 // This file shows how to use an online CTC model, i.e., streaming CTC model,
@@ -16,8 +15,6 @@ public class StreamingDecodeFileCtc {
     String waveFilename = "./sherpa-onnx-streaming-zipformer-ctc-small-2024-03-18/test_wavs/8k.wav";
 
     WaveReader reader = new WaveReader(waveFilename);
-    System.out.println(reader.getSampleRate());
-    System.out.println(reader.getSamples().length);
 
     OnlineZipformer2CtcModelConfig ctc =
         OnlineZipformer2CtcModelConfig.builder().setModel(model).build();

java-api-examples/StreamingDecodeFileParaformer.java

@@ -1,4 +1,3 @@
-// Copyright 2022-2023 by zhaoming
 // Copyright 2024 Xiaomi Corporation
 
 // This file shows how to use an online paraformer, i.e., streaming paraformer,
@@ -16,8 +15,6 @@ public class StreamingDecodeFileParaformer {
     String waveFilename = "./sherpa-onnx-streaming-paraformer-bilingual-zh-en/test_wavs/2.wav";
 
     WaveReader reader = new WaveReader(waveFilename);
-    System.out.println(reader.getSampleRate());
-    System.out.println(reader.getSamples().length);
 
     OnlineParaformerModelConfig paraformer =
         OnlineParaformerModelConfig.builder().setEncoder(encoder).setDecoder(decoder).build();

java-api-examples/StreamingDecodeFileTransducer.java

@@ -22,8 +22,6 @@ public class StreamingDecodeFileTransducer {
         "./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/test_wavs/0.wav";
 
     WaveReader reader = new WaveReader(waveFilename);
-    System.out.println(reader.getSampleRate());
-    System.out.println(reader.getSamples().length);
 
     OnlineTransducerModelConfig transducer =
         OnlineTransducerModelConfig.builder()

java-api-examples/run-non-streaming-decode-file-paraformer.sh

@@ -0,0 +1,52 @@
+#!/usr/bin/env bash
+
+set -ex
+
+if [[ ! -f ../build/lib/libsherpa-onnx-jni.dylib  && ! -f ../build/lib/libsherpa-onnx-jni.so ]]; then
+  mkdir -p ../build
+  pushd ../build
+  cmake \
+    -DSHERPA_ONNX_ENABLE_PYTHON=OFF \
+    -DSHERPA_ONNX_ENABLE_TESTS=OFF \
+    -DSHERPA_ONNX_ENABLE_CHECK=OFF \
+    -DBUILD_SHARED_LIBS=ON \
+    -DSHERPA_ONNX_ENABLE_PORTAUDIO=OFF \
+    -DSHERPA_ONNX_ENABLE_JNI=ON \
+    ..
+
+  make -j4
+  ls -lh lib
+  popd
+fi
+
+if [ ! -f ../sherpa-onnx/java-api/build/sherpa-onnx.jar ]; then
+  pushd ../sherpa-onnx/java-api
+  make
+  popd
+fi
+
+if [[ ! -f ../build/lib/libsherpa-onnx-jni.dylib  && ! -f ../build/lib/libsherpa-onnx-jni.so ]]; then
+  cmake \
+    -DSHERPA_ONNX_ENABLE_PYTHON=OFF \
+    -DSHERPA_ONNX_ENABLE_TESTS=OFF \
+    -DSHERPA_ONNX_ENABLE_CHECK=OFF \
+    -DBUILD_SHARED_LIBS=ON \
+    -DSHERPA_ONNX_ENABLE_PORTAUDIO=OFF \
+    -DSHERPA_ONNX_ENABLE_JNI=ON \
+    ..
+
+  make -j4
+  ls -lh lib
+fi
+
+if [ ! -f ./sherpa-onnx-paraformer-zh-2023-03-28/tokens.txt ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-paraformer-zh-2023-03-28.tar.bz2
+
+  tar xvf sherpa-onnx-paraformer-zh-2023-03-28.tar.bz2
+  rm sherpa-onnx-paraformer-zh-2023-03-28.tar.bz2
+fi
+
+java \
+  -Djava.library.path=$PWD/../build/lib \
+  -cp ../sherpa-onnx/java-api/build/sherpa-onnx.jar \
+  NonStreamingDecodeFileParaformer.java

java-api-examples/run-non-streaming-decode-file-transducer.sh

@@ -0,0 +1,52 @@
+#!/usr/bin/env bash
+
+set -ex
+
+if [[ ! -f ../build/lib/libsherpa-onnx-jni.dylib  && ! -f ../build/lib/libsherpa-onnx-jni.so ]]; then
+  mkdir -p ../build
+  pushd ../build
+  cmake \
+    -DSHERPA_ONNX_ENABLE_PYTHON=OFF \
+    -DSHERPA_ONNX_ENABLE_TESTS=OFF \
+    -DSHERPA_ONNX_ENABLE_CHECK=OFF \
+    -DBUILD_SHARED_LIBS=ON \
+    -DSHERPA_ONNX_ENABLE_PORTAUDIO=OFF \
+    -DSHERPA_ONNX_ENABLE_JNI=ON \
+    ..
+
+  make -j4
+  ls -lh lib
+  popd
+fi
+
+if [ ! -f ../sherpa-onnx/java-api/build/sherpa-onnx.jar ]; then
+  pushd ../sherpa-onnx/java-api
+  make
+  popd
+fi
+
+if [[ ! -f ../build/lib/libsherpa-onnx-jni.dylib  && ! -f ../build/lib/libsherpa-onnx-jni.so ]]; then
+  cmake \
+    -DSHERPA_ONNX_ENABLE_PYTHON=OFF \
+    -DSHERPA_ONNX_ENABLE_TESTS=OFF \
+    -DSHERPA_ONNX_ENABLE_CHECK=OFF \
+    -DBUILD_SHARED_LIBS=ON \
+    -DSHERPA_ONNX_ENABLE_PORTAUDIO=OFF \
+    -DSHERPA_ONNX_ENABLE_JNI=ON \
+    ..
+
+  make -j4
+  ls -lh lib
+fi
+
+if [ ! -f ./sherpa-onnx-zipformer-gigaspeech-2023-12-12/tokens.txt ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-zipformer-gigaspeech-2023-12-12.tar.bz2
+
+  tar xvf sherpa-onnx-zipformer-gigaspeech-2023-12-12.tar.bz2
+  rm sherpa-onnx-zipformer-gigaspeech-2023-12-12.tar.bz2
+fi
+
+java \
+  -Djava.library.path=$PWD/../build/lib \
+  -cp ../sherpa-onnx/java-api/build/sherpa-onnx.jar \
+  NonStreamingDecodeFileTransducer.java

java-api-examples/run-non-streaming-decode-file-whisper.sh

@@ -0,0 +1,52 @@
+#!/usr/bin/env bash
+
+set -ex
+
+if [[ ! -f ../build/lib/libsherpa-onnx-jni.dylib  && ! -f ../build/lib/libsherpa-onnx-jni.so ]]; then
+  mkdir -p ../build
+  pushd ../build
+  cmake \
+    -DSHERPA_ONNX_ENABLE_PYTHON=OFF \
+    -DSHERPA_ONNX_ENABLE_TESTS=OFF \
+    -DSHERPA_ONNX_ENABLE_CHECK=OFF \
+    -DBUILD_SHARED_LIBS=ON \
+    -DSHERPA_ONNX_ENABLE_PORTAUDIO=OFF \
+    -DSHERPA_ONNX_ENABLE_JNI=ON \
+    ..
+
+  make -j4
+  ls -lh lib
+  popd
+fi
+
+if [ ! -f ../sherpa-onnx/java-api/build/sherpa-onnx.jar ]; then
+  pushd ../sherpa-onnx/java-api
+  make
+  popd
+fi
+
+if [[ ! -f ../build/lib/libsherpa-onnx-jni.dylib  && ! -f ../build/lib/libsherpa-onnx-jni.so ]]; then
+  cmake \
+    -DSHERPA_ONNX_ENABLE_PYTHON=OFF \
+    -DSHERPA_ONNX_ENABLE_TESTS=OFF \
+    -DSHERPA_ONNX_ENABLE_CHECK=OFF \
+    -DBUILD_SHARED_LIBS=ON \
+    -DSHERPA_ONNX_ENABLE_PORTAUDIO=OFF \
+    -DSHERPA_ONNX_ENABLE_JNI=ON \
+    ..
+
+  make -j4
+  ls -lh lib
+fi
+
+if [ ! -f ./sherpa-onnx-whisper-tiny.en/tiny.en-tokens.txt ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-whisper-tiny.en.tar.bz2
+
+  tar xvf sherpa-onnx-whisper-tiny.en.tar.bz2
+  rm sherpa-onnx-whisper-tiny.en.tar.bz2
+fi
+
+java \
+  -Djava.library.path=$PWD/../build/lib \
+  -cp ../sherpa-onnx/java-api/build/sherpa-onnx.jar \
+  NonStreamingDecodeFileWhisper.java

sherpa-onnx/java-api/Makefile

@@ -19,6 +19,15 @@ java_files += OnlineRecognizerConfig.java
 java_files += OnlineRecognizerResult.java
 java_files += OnlineRecognizer.java
 
+java_files += OfflineTransducerModelConfig.java
+java_files += OfflineParaformerModelConfig.java
+java_files += OfflineWhisperModelConfig.java
+java_files += OfflineModelConfig.java
+java_files += OfflineRecognizerConfig.java
+java_files += OfflineRecognizerResult.java
+java_files += OfflineStream.java
+java_files += OfflineRecognizer.java
+
 class_files := $(java_files:%.java=%.class)
 
 java_files := $(addprefix src/$(package_dir)/,$(java_files))

sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OfflineModelConfig.java

@@ -0,0 +1,117 @@
+// Copyright 2024 Xiaomi Corporation
+package com.k2fsa.sherpa.onnx;
+
+public class OfflineModelConfig {
+    private final OfflineTransducerModelConfig transducer;
+    private final OfflineParaformerModelConfig paraformer;
+    private final OfflineWhisperModelConfig whisper;
+    private final String tokens;
+    private final int numThreads;
+    private final boolean debug;
+    private final String provider;
+
+    private final String modelType;
+
+    private OfflineModelConfig(Builder builder) {
+        this.transducer = builder.transducer;
+        this.paraformer = builder.paraformer;
+        this.whisper = builder.whisper;
+        this.tokens = builder.tokens;
+        this.numThreads = builder.numThreads;
+        this.debug = builder.debug;
+        this.provider = builder.provider;
+        this.modelType = builder.modelType;
+    }
+
+    public static Builder builder() {
+        return new Builder();
+    }
+
+    public OfflineParaformerModelConfig getParaformer() {
+        return paraformer;
+    }
+
+    public OfflineTransducerModelConfig getTransducer() {
+        return transducer;
+    }
+
+    public OfflineWhisperModelConfig getZipformer2Ctc() {
+        return whisper;
+    }
+
+    public String getTokens() {
+        return tokens;
+    }
+
+    public int getNumThreads() {
+        return numThreads;
+    }
+
+    public boolean getDebug() {
+        return debug;
+    }
+
+    public String getProvider() {
+        return provider;
+    }
+
+    public String getModelType() {
+        return modelType;
+    }
+
+
+    public static class Builder {
+        private OfflineParaformerModelConfig paraformer = OfflineParaformerModelConfig.builder().build();
+        private OfflineTransducerModelConfig transducer = OfflineTransducerModelConfig.builder().build();
+        private OfflineWhisperModelConfig whisper = OfflineWhisperModelConfig.builder().build();
+        private String tokens = "";
+        private int numThreads = 1;
+        private boolean debug = true;
+        private String provider = "cpu";
+        private String modelType = "";
+
+        public OfflineModelConfig build() {
+            return new OfflineModelConfig(this);
+        }
+
+        public Builder setTransducer(OfflineTransducerModelConfig transducer) {
+            this.transducer = transducer;
+            return this;
+        }
+
+        public Builder setParaformer(OfflineParaformerModelConfig paraformer) {
+            this.paraformer = paraformer;
+            return this;
+        }
+
+        public Builder setWhisper(OfflineWhisperModelConfig whisper) {
+            this.whisper = whisper;
+            return this;
+        }
+
+        public Builder setTokens(String tokens) {
+            this.tokens = tokens;
+            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;
+        }
+
+        public Builder setModelType(String modelType) {
+            this.modelType = modelType;
+            return this;
+        }
+    }
+}

sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OfflineParaformerModelConfig.java

@@ -0,0 +1,32 @@
+// Copyright 2024 Xiaomi Corporation
+package com.k2fsa.sherpa.onnx;
+
+public class OfflineParaformerModelConfig {
+    private final String model;
+
+    private OfflineParaformerModelConfig(Builder builder) {
+        this.model = builder.model;
+    }
+
+    public static Builder builder() {
+        return new Builder();
+    }
+
+    public String getModel() {
+        return model;
+    }
+
+
+    public static class Builder {
+        private String model = "";
+
+        public OfflineParaformerModelConfig build() {
+            return new OfflineParaformerModelConfig(this);
+        }
+
+        public Builder setModel(String model) {
+            this.model = model;
+            return this;
+        }
+    }
+}

sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OfflineRecognizer.java

@@ -0,0 +1,55 @@
+// Copyright 2024 Xiaomi Corporation
+package com.k2fsa.sherpa.onnx;
+
+public class OfflineRecognizer {
+    static {
+        System.loadLibrary("sherpa-onnx-jni");
+    }
+
+    private long ptr = 0; // this is the asr engine ptrss
+
+    public OfflineRecognizer(OfflineRecognizerConfig config) {
+        ptr = newFromFile(config);
+    }
+
+    public void decode(OfflineStream s) {
+        decode(ptr, s.getPtr());
+    }
+
+    public OfflineStream createStream() {
+        long p = createStream(ptr);
+        return new OfflineStream(p);
+    }
+
+    @Override
+    protected void finalize() throws Throwable {
+        release();
+    }
+
+    // You'd better call it manually if it is not used anymore
+    public void release() {
+        if (this.ptr == 0) {
+            return;
+        }
+        delete(this.ptr);
+        this.ptr = 0;
+    }
+
+    public OfflineRecognizerResult getResult(OfflineStream s) {
+        Object[] arr = getResult(s.getPtr());
+        String text = (String) arr[0];
+        String[] tokens = (String[]) arr[1];
+        float[] timestamps = (float[]) arr[2];
+        return new OfflineRecognizerResult(text, tokens, timestamps);
+    }
+
+    private native void delete(long ptr);
+
+    private native long newFromFile(OfflineRecognizerConfig config);
+
+    private native long createStream(long ptr);
+
+    private native void decode(long ptr, long streamPtr);
+
+    private native Object[] getResult(long streamPtr);
+}

sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OfflineRecognizerConfig.java

@@ -0,0 +1,71 @@
+// Copyright 2024 Xiaomi Corporation
+package com.k2fsa.sherpa.onnx;
+
+public class OfflineRecognizerConfig {
+    private final FeatureConfig featConfig;
+    private final OfflineModelConfig modelConfig;
+    private final String decodingMethod;
+    private final int maxActivePaths;
+    private final String hotwordsFile;
+    private final float hotwordsScore;
+
+    private OfflineRecognizerConfig(Builder builder) {
+        this.featConfig = builder.featConfig;
+        this.modelConfig = builder.modelConfig;
+        this.decodingMethod = builder.decodingMethod;
+        this.maxActivePaths = builder.maxActivePaths;
+        this.hotwordsFile = builder.hotwordsFile;
+        this.hotwordsScore = builder.hotwordsScore;
+    }
+
+    public static Builder builder() {
+        return new Builder();
+    }
+
+    public OfflineModelConfig getModelConfig() {
+        return modelConfig;
+    }
+
+    public static class Builder {
+        private FeatureConfig featConfig = FeatureConfig.builder().build();
+        private OfflineModelConfig modelConfig = OfflineModelConfig.builder().build();
+        private String decodingMethod = "greedy_search";
+        private int maxActivePaths = 4;
+        private String hotwordsFile = "";
+        private float hotwordsScore = 1.5f;
+
+        public OfflineRecognizerConfig build() {
+            return new OfflineRecognizerConfig(this);
+        }
+
+        public Builder setFeatureConfig(FeatureConfig featConfig) {
+            this.featConfig = featConfig;
+            return this;
+        }
+
+        public Builder setOfflineModelConfig(OfflineModelConfig modelConfig) {
+            this.modelConfig = modelConfig;
+            return this;
+        }
+
+        public Builder setDecodingMethod(String decodingMethod) {
+            this.decodingMethod = decodingMethod;
+            return this;
+        }
+
+        public Builder setMaxActivePaths(int maxActivePaths) {
+            this.maxActivePaths = maxActivePaths;
+            return this;
+        }
+
+        public Builder setHotwordsFile(String hotwordsFile) {
+            this.hotwordsFile = hotwordsFile;
+            return this;
+        }
+
+        public Builder setHotwordsScore(float hotwordsScore) {
+            this.hotwordsScore = hotwordsScore;
+            return this;
+        }
+    }
+}

sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OfflineRecognizerResult.java

@@ -0,0 +1,26 @@
+// Copyright 2024 Xiaomi Corporation
+package com.k2fsa.sherpa.onnx;
+
+public class OfflineRecognizerResult {
+    private final String text;
+    private final String[] tokens;
+    private final float[] timestamps;
+
+    public OfflineRecognizerResult(String text, String[] tokens, float[] timestamps) {
+        this.text = text;
+        this.tokens = tokens;
+        this.timestamps = timestamps;
+    }
+
+    public String getText() {
+        return text;
+    }
+
+    public String[] getTokens() {
+        return tokens;
+    }
+
+    public float[] getTimestamps() {
+        return timestamps;
+    }
+}

sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OfflineStream.java

@@ -0,0 +1,49 @@
+// Copyright 2024 Xiaomi Corporation
+package com.k2fsa.sherpa.onnx;
+
+public class OfflineStream {
+    static {
+        System.loadLibrary("sherpa-onnx-jni");
+    }
+
+    private long ptr = 0;
+
+    public OfflineStream() {
+        this.ptr = 0;
+    }
+
+    public OfflineStream(long ptr) {
+        this.ptr = ptr;
+    }
+
+    public long getPtr() {
+        return ptr;
+    }
+
+    public void setPtr(long ptr) {
+        this.ptr = ptr;
+    }
+
+    public void acceptWaveform(float[] samples, int sampleRate) {
+        acceptWaveform(this.ptr, samples, sampleRate);
+    }
+
+    public void release() {
+        // stream object must be release after used
+        if (this.ptr == 0) {
+            return;
+        }
+        delete(this.ptr);
+        this.ptr = 0;
+    }
+
+    @Override
+    protected void finalize() throws Throwable {
+        release();
+        super.finalize();
+    }
+
+    private native void acceptWaveform(long ptr, float[] samples, int sampleRate);
+
+    private native void delete(long ptr);
+}

sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OfflineTransducerModelConfig.java

@@ -0,0 +1,55 @@
+// Copyright 2024 Xiaomi Corporation
+package com.k2fsa.sherpa.onnx;
+
+public class OfflineTransducerModelConfig {
+    private final String encoder;
+    private final String decoder;
+    private final String joiner;
+
+    private OfflineTransducerModelConfig(Builder builder) {
+        this.encoder = builder.encoder;
+        this.decoder = builder.decoder;
+        this.joiner = builder.joiner;
+    }
+
+    public static Builder builder() {
+        return new Builder();
+    }
+
+    public String getEncoder() {
+        return encoder;
+    }
+
+    public String getDecoder() {
+        return decoder;
+    }
+
+    public String getJoiner() {
+        return joiner;
+    }
+
+    public static class Builder {
+        private String encoder = "";
+        private String decoder = "";
+        private String joiner = "";
+
+        public OfflineTransducerModelConfig build() {
+            return new OfflineTransducerModelConfig(this);
+        }
+
+        public Builder setEncoder(String encoder) {
+            this.encoder = encoder;
+            return this;
+        }
+
+        public Builder setDecoder(String decoder) {
+            this.decoder = decoder;
+            return this;
+        }
+
+        public Builder setJoiner(String joiner) {
+            this.joiner = joiner;
+            return this;
+        }
+    }
+}

sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OfflineWhisperModelConfig.java

@@ -0,0 +1,80 @@
+// Copyright 2024 Xiaomi Corporation
+package com.k2fsa.sherpa.onnx;
+
+public class OfflineWhisperModelConfig {
+    private final String encoder;
+    private final String decoder;
+    private final String language;
+    private final String task;
+    private final int tailPaddings;
+
+    private OfflineWhisperModelConfig(Builder builder) {
+        this.encoder = builder.encoder;
+        this.decoder = builder.decoder;
+        this.language = builder.language;
+        this.task = builder.task;
+        this.tailPaddings = builder.tailPaddings;
+    }
+
+    public static Builder builder() {
+        return new Builder();
+    }
+
+    public String getEncoder() {
+        return encoder;
+    }
+
+    public String getDecoder() {
+        return decoder;
+    }
+
+    public String getLanguage() {
+        return language;
+    }
+
+    public String getTask() {
+        return task;
+    }
+
+    public int getTailPaddings() {
+        return tailPaddings;
+    }
+
+    public static class Builder {
+        private String encoder = "";
+        private String decoder = "";
+        private String language = "en"; // used only with multilingual models
+        private String task = "transcribe"; // used only with multilingual models
+
+        private int tailPaddings = 1000; // number of frames to pad
+
+        public OfflineWhisperModelConfig build() {
+            return new OfflineWhisperModelConfig(this);
+        }
+
+        public Builder setEncoder(String encoder) {
+            this.encoder = encoder;
+            return this;
+        }
+
+        public Builder setDecoder(String decoder) {
+            this.decoder = decoder;
+            return this;
+        }
+
+        public Builder setLanguage(String language) {
+            this.language = language;
+            return this;
+        }
+
+        public Builder setTask(String task) {
+            this.task = task;
+            return this;
+        }
+
+        public Builder setTailPaddings(int tailPaddings) {
+            this.tailPaddings = tailPaddings;
+            return this;
+        }
+    }
+}

sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OnlineModelConfig.java

@@ -12,6 +12,7 @@ public class OnlineModelConfig {
     private final boolean debug;
     private final String provider;
     private final String modelType;
+
     private OnlineModelConfig(Builder builder) {
         this.transducer = builder.transducer;
         this.paraformer = builder.paraformer;

sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OnlineRecognizer.java

@@ -15,19 +15,6 @@ public class OnlineRecognizer {
         ptr = newFromFile(config);
     }
 
-    /*
-    public static float[] readWavFile(String fileName) {
-        // read data from the filename
-        Object[] wavdata = readWave(fileName);
-        Object data = wavdata[0]; // data[0] is float data, data[1] sample rate
-
-        float[] floatData = (float[]) data;
-
-        return floatData;
-    }
-    */
-
-
     public void decode(OnlineStream s) {
         decode(ptr, s.getPtr());
     }
@@ -55,7 +42,7 @@ public class OnlineRecognizer {
         release();
     }
 
-    // recognizer release, you'd better call it manually if not use anymore
+    // You'd better call it manually if it is not used anymore
     public void release() {
         if (this.ptr == 0) {
             return;

sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OnlineRecognizerConfig.java

@@ -12,6 +12,7 @@ public class OnlineRecognizerConfig {
     private final int maxActivePaths;
     private final String hotwordsFile;
     private final float hotwordsScore;
+
     private OnlineRecognizerConfig(Builder builder) {
         this.featConfig = builder.featConfig;
         this.modelConfig = builder.modelConfig;