[Feat]新增支持多语言的TTS

[Feat]新增支持多语言的TTS

app/proguard-rules.pro

@@ -426,6 +426,6 @@
 -keep class com.squareup.haha.guava.collect.*{*;}
 -keep class **.zego.**{*;}
 
-
-
+#-----科大讯飞语音合成-----
+-keep class com.iflytek.**{*;}
 

config.gradle

@@ -142,6 +142,7 @@ ext {
             ttszhi                       : "com.mogo.tts:tts-zhi:${TTS_ZHI_VERSION}",
             ttspad                       : "com.mogo.tts:tts-pad:${TTS_ZHI_VERSION}",
             ttsnoop                      : "com.mogo.tts:tts-noop:${TTS_NOOP_VERSION}",
+            ttsiflytek                   : "com.mogo.tts:tts-iflytek :${TTS_IFLYTEK_VERSION}",
 
             //========================= 网约车 Maven 版本管理 =========================
             mogooch                      : "com.mogo.och:och:${MOGO_OCH_VERSION}",

foudations/mogo-commons/src/main/java/com/mogo/commons/voice/AIAssist.java

@@ -6,6 +6,7 @@ import com.mogo.eagle.core.utilcode.mogo.logger.CallerLogger;
 import com.mogo.eagle.core.utilcode.util.ToastUtils;
 import com.mogo.tts.base.IMogoTTS;
 import com.mogo.tts.base.IMogoTTSCallback;
+import com.mogo.tts.base.MultiLangTtsEntity;
 import com.mogo.tts.base.PreemptType;
 
 /**
@@ -52,7 +53,9 @@ public class AIAssist {
     private AIAssist(Context context) {
         try {
             // 暂时换成反射，解决死锁问题
+            // TODO:("支持切换思必驰和科大讯飞")
             Class<?> clazz = Class.forName("com.mogo.tts.pad.PadTTS");
+//            Class<?> clazz = Class.forName("com.mogo.tts.iflytek.IFlyTekTts");
             mTTS = (IMogoTTS) clazz.getConstructor().newInstance();
             mTTS.init(context);
 //            mTTS = (IMogoTTS) ARouter.getInstance().build(MogoTTSConstants.API_PATH).navigation(context.getApplicationContext());
@@ -122,6 +125,18 @@ public class AIAssist {
         }
     }
 
+    /**
+     * 支持多语言的Tts
+     * @param ttsEntity: 多语言Entity
+     * @param level: 等级由低到高为0、1、2、3，分别对应p3、p2、p1、p0
+     * @param callback
+     */
+    public void speakMultiLangTTSWithLevel(MultiLangTtsEntity ttsEntity, int level, IMogoTTSCallback callback) {
+        if (mTTS != null) {
+            mTTS.speakMultiLangTTSWithLevel(ttsEntity, level, callback);
+        }
+    }
+
     /**
      * 语音播报
      *

gradle.properties

@@ -123,6 +123,7 @@ TTS_DI_VERSION=2.1.16.10
 TTS_ZHI_VERSION=2.1.16.10
 TTS_PAD_VERSION=2.1.16.10
 TTS_NOOP_VERSION=2.1.16.10
+TTS_IFLYTEK_VERSION=2.1.16.10
 # 自研地图
 MAP_CUSTOM_VERSION=2.1.16.10
 

settings.gradle

@@ -50,6 +50,7 @@ include ':libraries:mogo-obu'
 // 语音
 include ':tts:tts-base'
 include ':tts:tts-pad'
+include ':tts:tts-iflytek'
 
 // 测试DEBUG
 include ':test:crashreport'

tts/tts-base/src/main/java/com/mogo/tts/base/IMogoTTS.java

@@ -64,6 +64,12 @@ interface IMogoTTS extends IProvider {
      */
     void speakTTSVoice( String tts, PreemptType type, IMogoTTSCallback callback );
 
+    void speakMultiLangTTSWithLevel(
+            MultiLangTtsEntity ttsEntity,
+            int level,
+            IMogoTTSCallback callback
+    );
+
     void stopSpeakTts(String text);
 
     void stopTts();

tts/tts-base/src/main/java/com/mogo/tts/base/LanguageType.kt

@@ -0,0 +1,7 @@
+package com.mogo.tts.base
+
+enum class LanguageType(val langName: String) {
+    CHINESE("chinese"),
+    ENGLISH("english"),
+    KOREAN("korean")
+}

tts/tts-base/src/main/java/com/mogo/tts/base/MultiLangTtsEntity.kt

@@ -0,0 +1,42 @@
+package com.mogo.tts.base
+
+data class MultiLangTtsEntity(
+    private var ttsList: List<LangTtsEntity>
+) {
+    private val stringBuffer by lazy {
+        StringBuffer()
+    }
+
+    private var ttsIndex = 0
+
+    fun ttsNext(): LangTtsEntity? {
+        return if (ttsIndex in ttsList.indices) {
+            ttsList[ttsIndex++]
+        } else {
+            null
+        }
+    }
+
+    override fun toString(): String {
+        return stringBuffer.let {
+            it.setLength(0)
+            ttsList.forEachIndexed { index, langTtsEntity ->
+                if (index != ttsList.size - 1) {
+                    it.append("${langTtsEntity};")
+                } else {
+                    it.append(langTtsEntity)
+                }
+            }
+            it.toString()
+        }
+    }
+}
+
+data class LangTtsEntity(
+    var ttsContent: String,
+    var language: LanguageType
+) {
+    override fun toString(): String {
+        return "ttsContent is:$ttsContent,language is:${language.langName}"
+    }
+}

tts/tts-iflytek/.gitignore

@@ -0,0 +1 @@
+/build

tts/tts-iflytek/build.gradle

@@ -0,0 +1,62 @@
+plugins {
+    id 'com.android.library'
+    id 'kotlin-android'
+    id 'kotlin-android-extensions'
+    id 'kotlin-kapt'
+    id 'com.alibaba.arouter'
+}
+
+android {
+    compileSdkVersion rootProject.ext.android.compileSdkVersion
+    // buildToolsVersion rootProject.ext.android.buildToolsVersion
+    defaultConfig {
+        minSdkVersion rootProject.ext.android.minSdkVersion
+        targetSdkVersion rootProject.ext.android.targetSdkVersion
+        versionCode Integer.valueOf(VERSION_CODE)
+        versionName getValueFromRootProperties("${project.name.replace("-", "_").toUpperCase()}_VERSION")
+
+        testInstrumentationRunner "androidx.test.runner.AndroidJUnitRunner"
+        consumerProguardFiles "consumer-rules.pro"
+
+        //ARouter apt 参数
+        kapt {
+            useBuildCache = false
+            arguments {
+                arg("AROUTER_MODULE_NAME", project.getName())
+            }
+        }
+    }
+
+    buildTypes {
+        release {
+            minifyEnabled false
+            proguardFiles getDefaultProguardFile('proguard-android-optimize.txt'), 'proguard-rules.pro'
+        }
+    }
+
+    compileOptions {
+        sourceCompatibility JavaVersion.VERSION_1_8
+        targetCompatibility JavaVersion.VERSION_1_8
+    }
+}
+
+dependencies {
+    implementation fileTree(dir: "libs", include: ["*.jar"])
+    implementation rootProject.ext.dependencies.androidxappcompat
+
+    implementation rootProject.ext.dependencies.arouter
+    kapt rootProject.ext.dependencies.aroutercompiler
+
+    api rootProject.ext.dependencies.aiassist
+    api rootProject.ext.dependencies.aiassistReplace
+
+    if (Boolean.valueOf(USE_MAVEN_PACKAGE)) {
+        implementation rootProject.ext.dependencies.ttsbase
+        implementation rootProject.ext.dependencies.mogo_core_utils
+    } else {
+        implementation project(":tts:tts-base")
+        implementation project(':core:mogo-core-utils')
+    }
+}
+
+apply from: new File(rootProject.rootDir, "gradle/upload.gradle").toString()

tts/tts-iflytek/gradle.properties

@@ -0,0 +1,3 @@
+GROUP=com.mogo.tts
+POM_ARTIFACT_ID=tts-iflytek
+VERSION_CODE=1

tts/tts-iflytek/proguard-rules.pro

@@ -0,0 +1,21 @@
+# Add project specific ProGuard rules here.
+# You can control the set of applied configuration files using the
+# proguardFiles setting in build.gradle.
+#
+# For more details, see
+#   http://developer.android.com/guide/developing/tools/proguard.html
+
+# If your project uses WebView with JS, uncomment the following
+# and specify the fully qualified class name to the JavaScript interface
+# class:
+#-keepclassmembers class fqcn.of.javascript.interface.for.webview {
+#   public *;
+#}
+
+# Uncomment this to preserve the line number information for
+# debugging stack traces.
+#-keepattributes SourceFile,LineNumberTable
+
+# If you keep the line number information, uncomment this to
+# hide the original source file name.
+#-renamesourcefileattribute SourceFile

tts/tts-iflytek/src/main/AndroidManifest.xml

@@ -0,0 +1,4 @@
+<manifest xmlns:android="http://schemas.android.com/apk/res/android"
+    package="com.mogo.tts.iflytek">
+
+</manifest>

tts/tts-iflytek/src/main/java/com/mogo/tts/iflytek/IFlyTekTts.kt

@@ -0,0 +1,375 @@
+package com.mogo.tts.iflytek
+
+import android.content.Context
+import android.os.Bundle
+import android.os.Looper
+import android.util.Pair
+import androidx.annotation.Keep
+import com.iflytek.cloud.*
+import com.mogo.eagle.core.utilcode.mogo.logger.CallerLogger
+import com.mogo.eagle.core.utilcode.mogo.logger.CallerLogger.d
+import com.mogo.eagle.core.utilcode.mogo.logger.CallerLogger.i
+import com.mogo.eagle.core.utilcode.util.UiThreadHandler
+import com.mogo.tts.base.*
+import java.util.*
+
+@Keep
+class IFlyTekTts : IMogoTTS, InitListener {
+    companion object {
+        const val TAG = "IFlyTekTts"
+    }
+
+    private var context: Context? = null
+
+    private var ttsEngine: SpeechSynthesizer? = null
+    private var engineType = SpeechConstant.TYPE_CLOUD
+
+    private var voicer: String = "xiaoyan"
+    private var voicerEntries: Array<String>? = null
+
+    // 等级由低到高为0、1、2、3，默认为-1表示没有正在tts的
+    private var curTtsLevel = -1
+
+    // 由于主动打断不会有回调事件，所以主动打断时清掉map中被打断的text和callback
+    private var curTtsContent = ""
+    private var curTtsEntity: MultiLangTtsEntity? = null
+    private val linkedList = LinkedList<Pair<MultiLangTtsEntity, Int>>()
+
+    // 上一次语音合成的语言
+    private var lastLang = LanguageType.CHINESE
+
+    private val speakVoiceMap by lazy {
+        HashMap<String, IMogoTTSCallback>()
+    }
+
+    override fun init(context: Context) {
+        this.context = context
+
+        // 应用程序入口处调用，避免手机内存过小，杀死后台进程后通过历史intent进入Activity造成SpeechUtility对象为null
+        // 如在Application中调用初始化，需要在Mainifest中注册该Applicaiton
+        // 注意：此接口在非主进程调用会返回null对象，如需在非主进程使用语音功能，请增加参数：SpeechConstant.FORCE_LOGIN+"=true"
+        // 参数间使用半角“,”分隔。
+        // 设置你申请的应用appid,请勿在'='与appid之间添加空格及空转义符
+        // 注意： appid 必须和下载的SDK保持一致，否则会出现10407错误
+        SpeechUtility.createUtility(context, "appid=0c498b42")
+        ttsEngine = SpeechSynthesizer.createSynthesizer(context, this)
+        voicerEntries = context.resources?.getStringArray(R.array.voicer_cloud_values)
+    }
+
+    override fun initTts(sn: String?) {}
+
+    override fun onInit(code: Int) {
+        if (code != ErrorCode.SUCCESS) {
+            CallerLogger.e(
+                TAG,
+                "初始化失败，错误码为：$code,请点击网址https://www.xfyun.cn/document/error-code查询解决方案"
+            )
+        }
+    }
+
+    override fun release() {
+        ttsEngine?.stopSpeaking()
+        ttsEngine?.destroy()
+    }
+
+    override fun flush() {
+    }
+
+    override fun speakTTSVoice(tts: String?) {
+
+    }
+
+    override fun speakTTSVoice(tts: String?, callback: IMogoTTSCallback?) {
+    }
+
+    override fun speakTTSVoice(tts: String?, type: PreemptType?, callback: IMogoTTSCallback?) {
+    }
+
+    override fun speakTTSVoiceWithLevel(tts: String?, level: Int) {
+    }
+
+    override fun speakTTSVoiceWithLevel(tts: String?, level: Int, callBack: IMogoTTSCallback?) {
+    }
+
+    override fun speakMultiLangTTSWithLevel(
+        ttsEntity: MultiLangTtsEntity,
+        level: Int,
+        callBack: IMogoTTSCallback?
+    ) {
+        if (callBack != null) {
+            speakVoiceMap[ttsEntity.toString()] = callBack
+        }
+        speakMultiLangTTSWithLevel(ttsEntity, level)
+    }
+
+    override fun stopSpeakTts(text: String?) {
+        stopTts()
+    }
+
+    override fun stopTts() {
+        curTtsEntity?.let {
+            val string = it.toString()
+            if (speakVoiceMap.containsKey(string)) {
+                speakVoiceMap.remove(string)
+            }
+            curTtsEntity = null
+        }
+        curTtsContent = ""
+        curTtsLevel = -1
+        ttsEngine?.stopSpeaking()
+    }
+
+    override fun speakQAndACmd(tts: String?, callback: IMogoTTSCallback?) {
+    }
+
+    override fun speakQAndACmd(
+        tts: String?,
+        okWords: Array<out String>?,
+        cancelWords: Array<out String>?,
+        callback: IMogoTTSCallback?
+    ) {
+    }
+
+    override fun registerUnWakeupCommand(
+        cmd: String?,
+        cmdWords: Array<out String>?,
+        callback: IMogoTTSCallback?
+    ) {
+    }
+
+    override fun unregisterUnWakeupCommand(cmd: String?) {
+    }
+
+    override fun unregisterUnWakeupCommand(cmd: String?, callback: IMogoTTSCallback?) {
+    }
+
+    override fun startAIAssist(context: Context?) {
+    }
+
+    override fun startAIAssist(context: Context?, status: Int) {
+    }
+
+    override fun breakOffSpeak() {
+    }
+
+    // 降序插入Tts(目前Level0、1可排队)
+    private fun insertTts(ttsEntity: MultiLangTtsEntity, level: Int) {
+        var index = -1
+        for (i in linkedList.indices.reversed()) {
+            val nodeLevel = linkedList[i].second
+            // 只有高优先级才插入到前面，等于的情况下是插到后面
+            if (level > nodeLevel) {
+                index = i
+            }
+        }
+        if (index >= 0) {
+            linkedList.add(index, Pair(ttsEntity, level))
+        } else {
+            linkedList.addLast(Pair(ttsEntity, level))
+        }
+        for (ttsPair in linkedList) {
+            d(TAG, "tts文本为：" + ttsPair.first + ",level为：" + ttsPair.second)
+        }
+        d(TAG, "===================")
+    }
+
+    private fun speakMultiLangTTSWithLevel(ttsEntity: MultiLangTtsEntity, ttsLevel: Int) {
+        ttsEngine?.let {
+            if (ttsLevel == curTtsLevel) {
+                // 对应p3、p2级别的排队
+                if (ttsLevel == 0 || ttsLevel == 1) {
+                    d(TAG, "===================")
+                    d(TAG, "插入消息：$ttsEntity,level为：$ttsLevel")
+                    insertTts(ttsEntity, ttsLevel)
+                    return
+                } else {
+                    // 打断并合成新的
+                    stopTts()
+                    d(TAG, "非Level1同级别打断!")
+                }
+            } else {
+                // 将要TTS的比现在正在TTS的优先级高
+                if (ttsLevel > curTtsLevel) {
+                    if (curTtsLevel >= 0) {
+                        // 打断并合成高优先级的
+                        stopTts()
+                    }
+                    d(TAG, "高优先级打断低级别的!")
+                } else {
+                    if (ttsLevel == 0 || ttsLevel == 1) {
+                        d(TAG, "===================")
+                        d(TAG, "插入消息：$ttsEntity,level为：$ttsLevel")
+                        insertTts(ttsEntity, ttsLevel)
+                    }
+                    return
+                }
+            }
+            curTtsLevel = ttsLevel
+            curTtsEntity = ttsEntity
+            // 合成并播放
+            d(TAG, "tts准备合成：$ttsEntity,curTtsLevel为：$curTtsLevel")
+            startSpeak(ttsEntity.ttsNext())
+        }
+    }
+
+    private fun startSpeak(langTtsEntity: LangTtsEntity?) {
+        langTtsEntity?.let {
+            if (it.language != lastLang) {
+                updateVoicer(it.language)
+                updateTtsParam()
+                lastLang = it.language
+            }
+            ttsEngine?.startSpeaking(it.ttsContent, ttsListener)
+        }
+    }
+
+    private fun handleCompleteEvent() {
+        curTtsLevel = -1
+        if (curTtsEntity != null) {// 多语言
+            val langTtsEntity = curTtsEntity!!.ttsNext()
+            if (langTtsEntity != null) {
+                speakVoiceMap[curTtsEntity.toString()]?.onSpeakEnd(curTtsEntity.toString())
+                ttsNextLanguage(langTtsEntity)
+            } else {
+                speakVoiceMap.remove(curTtsEntity.toString())?.onSpeakEnd(curTtsEntity.toString())
+                curTtsEntity = null
+                ttsNextMultiLangEntity()
+            }
+        } else {// 单语言
+            speakVoiceMap.remove(curTtsContent)?.onSpeakEnd(curTtsContent)
+            curTtsContent = ""
+            ttsNextMultiLangEntity()
+        }
+    }
+
+    private fun handleErrorEvent(error: SpeechError) {
+        curTtsLevel = -1
+        if (curTtsEntity != null) {
+            speakVoiceMap.remove(curTtsEntity.toString())
+                ?.onSpeakError(curTtsEntity.toString(), error.message)
+        } else {
+            speakVoiceMap.remove(curTtsContent)?.onSpeakError(curTtsContent, error.message)
+        }
+        curTtsEntity = null
+        curTtsContent = ""
+    }
+
+    private val ttsListener = object : SynthesizerListener {
+        override fun onSpeakBegin() {
+
+        }
+
+        override fun onBufferProgress(p0: Int, p1: Int, p2: Int, p3: String?) {
+        }
+
+        override fun onSpeakPaused() {
+        }
+
+        override fun onSpeakResumed() {
+        }
+
+        override fun onSpeakProgress(p0: Int, p1: Int, p2: Int) {
+        }
+
+        override fun onEvent(p0: Int, p1: Int, p2: Int, p3: Bundle?) {
+
+        }
+
+        /**
+         * error - 错误信息，若为null，则没有出现错误。
+         * 在音频播放完成，或会话出现错误时，将回调此函数。
+         * 若应用主动调用SpeechSynthesizer.stopSpeaking()停止会话，则不会回调此函数。
+         */
+        override fun onCompleted(error: SpeechError?) {
+            if (Thread.currentThread() == Looper.getMainLooper().thread) {
+                if (error == null) {// 播放完成
+                    handleCompleteEvent()
+                } else {// 播放出错
+                    handleErrorEvent(error)
+                }
+            } else {
+                UiThreadHandler.post {
+                    if (error == null) {
+                        handleCompleteEvent()
+                    } else {
+                        handleErrorEvent(error)
+                    }
+                }
+            }
+        }
+    }
+
+    /**
+     * 语音合成下一个MultiLangTtsEntity
+     */
+    private fun ttsNextMultiLangEntity() {
+        if (!linkedList.isEmpty()) {
+            val ttsPair = linkedList.removeFirst()
+            i(TAG, "排队播放的下一条文本为：" + ttsPair.first + ",级别为：" + ttsPair.second)
+            curTtsLevel = ttsPair.second
+            startSpeak(ttsPair.first.ttsNext())
+        } else {
+            i(TAG, "队列为空")
+        }
+    }
+
+    /**
+     * 语音合成下一个MultiLangTtsEntity中的语言
+     */
+    private fun ttsNextLanguage(langTtsEntity: LangTtsEntity) {
+        startSpeak(langTtsEntity)
+    }
+
+    private fun updateVoicer(language: LanguageType) {
+        voicerEntries?.let {
+            voicer = when (language) {
+                LanguageType.ENGLISH -> it[5]
+                LanguageType.KOREAN -> it[6]
+                else -> it[0]
+            }
+        }
+    }
+
+    private fun updateTtsParam() {
+        // 清空参数
+        ttsEngine?.setParameter(SpeechConstant.PARAMS, null)
+        // 根据合成引擎设置相应参数
+        if (engineType == SpeechConstant.TYPE_CLOUD) {
+            ttsEngine?.setParameter(SpeechConstant.ENGINE_TYPE, SpeechConstant.TYPE_CLOUD)
+            // 支持实时音频返回，仅在 synthesizeToUri 条件下支持
+            ttsEngine?.setParameter(SpeechConstant.TTS_DATA_NOTIFY, "1")
+            //	ttsEngine?.setParameter(SpeechConstant.TTS_BUFFER_TIME,"1");
+            // 设置在线合成发音人
+            ttsEngine?.setParameter(SpeechConstant.VOICE_NAME, voicer)
+            //设置合成语速
+            ttsEngine?.setParameter(
+                SpeechConstant.SPEED, "50"
+            )
+            //设置合成音调
+            ttsEngine?.setParameter(
+                SpeechConstant.PITCH, "50"
+            )
+            //设置合成音量
+            ttsEngine?.setParameter(
+                SpeechConstant.VOLUME, "50"
+            )
+        } else {
+            ttsEngine?.setParameter(SpeechConstant.ENGINE_TYPE, SpeechConstant.TYPE_LOCAL)
+            ttsEngine?.setParameter(SpeechConstant.VOICE_NAME, "")
+        }
+
+        //设置播放器音频流类型
+        ttsEngine?.setParameter(
+            SpeechConstant.STREAM_TYPE, "3"
+        )
+        // 设置播放合成音频打断音乐播放，默认为true
+        ttsEngine?.setParameter(SpeechConstant.KEY_REQUEST_FOCUS, "false")
+
+        // 设置音频保存路径，保存音频格式支持pcm、wav，设置路径为sd卡请注意WRITE_EXTERNAL_STORAGE权限
+        ttsEngine?.setParameter(SpeechConstant.AUDIO_FORMAT, "pcm")
+        ttsEngine?.setParameter(
+            SpeechConstant.TTS_AUDIO_PATH,
+            context?.getExternalFilesDir("msc")?.absolutePath + "/tts.pcm"
+        )
+    }
+}

tts/tts-iflytek/src/main/res/values/strings.xml

@@ -0,0 +1,17 @@
+<?xml version="1.0" encoding="utf-8"?>
+<resources>
+    <string-array name="voicer_cloud_values">
+        <!-- 汉语发音人 -->
+        <item>xiaoyan</item>
+        <item>aisjiuxu</item>
+        <item>aisxping</item>
+        <item>aisjinger</item>
+        <item>aisbabyxu</item>
+
+        <!-- 英语发音人 -->
+        <item>x2_enus_catherine</item>
+
+        <!-- 韩语发音人 -->
+        <item>zhimin</item>
+    </string-array>
+</resources>

tts/tts-pad/src/main/java/com/mogo/tts/pad/PadTTS.java

@@ -29,6 +29,7 @@ import com.mogo.eagle.core.utilcode.util.ThreadUtils;
 import com.mogo.eagle.core.utilcode.util.ToastUtils;
 import com.mogo.tts.base.IMogoTTS;
 import com.mogo.tts.base.IMogoTTSCallback;
+import com.mogo.tts.base.MultiLangTtsEntity;
 import com.mogo.tts.base.PreemptType;
 import com.zhidao.auto.platform.voice.VoiceClient;
 import com.zhidao.voicesdk.MogoVoiceManager;
@@ -505,6 +506,11 @@ public class PadTTS implements IMogoTTS, VoiceClient.VoiceCmdCallBack, OnTtsList
         }
     }
 
+    @Override
+    public void speakMultiLangTTSWithLevel(MultiLangTtsEntity ttsEntity, int level, IMogoTTSCallback callback) {
+
+    }
+
     /**
      * 问答类型语音注册:默认确认和取消
      *