Development of WebSocket is over.

UtopiaXC · Dec 2, 2022 · af55b7e · af55b7e
1 parent ca833db
commit af55b7e
Show file tree

Hide file tree

Showing 11 changed files with 258 additions and 44 deletions.
diff --git a/app/src/main/java/com/utopiaxc/utopiatts/enums/SettingsEnum.java b/app/src/main/java/com/utopiaxc/utopiatts/enums/SettingsEnum.java
@@ -1,14 +1,15 @@
 package com.utopiaxc.utopiatts.enums;
 
 import com.utopiaxc.utopiatts.tts.enums.Driver;
+import com.utopiaxc.utopiatts.tts.enums.OutputFormat;
 
 public enum SettingsEnum {
 
     THEME("THEME", "theme", "auto"),
     FIRST_BOOT("FIRST_BOOT", "first_boot", true),
     AZURE_REGION("AZURE_REGION", "azure_region", ""),
     AZURE_TOKEN("AZURE_TOKEN", "azure_token", "NULL"),
-    OUTPUT_FORMAT("OUTPUT_FORMAT", "output_format", ""),
+    OUTPUT_FORMAT("OUTPUT_FORMAT", "output_format", OutputFormat.OGG_48K_HZ_16_BIT_MONO_OPUS.getName()),
     ACTOR("ACTOR", "actor", ""),
     ROLE("ROLE", "role", ""),
     STYLE("STYLE", "style", ""),

diff --git a/app/src/main/java/com/utopiaxc/utopiatts/fragments/FragmentTtsSettings.java b/app/src/main/java/com/utopiaxc/utopiatts/fragments/FragmentTtsSettings.java
@@ -4,6 +4,7 @@
 import android.content.SharedPreferences;
 import android.os.Bundle;
 import android.text.InputType;
+import android.util.Log;
 
 import androidx.preference.EditTextPreference;
 import androidx.preference.ListPreference;
@@ -98,5 +99,29 @@ public void onCreatePreferences(Bundle savedInstanceState, String rootKey) {
             }
             return true;
         });
+
+        ListPreference listFormat = findPreference(SettingsEnum.OUTPUT_FORMAT.getKey());
+        assert listFormat != null;
+        listFormat.setOnPreferenceChangeListener((preference, newValue) -> {
+            new AlertDialog.Builder(requireActivity()).setTitle(R.string.warning)
+                    .setMessage(R.string.warning_of_output_format)
+                    .setPositiveButton(R.string.confirm, null)
+                    .create()
+                    .show();
+            return true;
+        });
+
+        EditTextPreference editTextPreferenceToken = findPreference(SettingsEnum.AZURE_TOKEN.getKey());
+        assert editTextPreferenceToken != null;
+        editTextPreferenceToken.setEnabled(Driver.AZURE_SDK.getId().equals(
+                sharedPreferences.getString(SettingsEnum.TTS_DRIVER.getKey(),
+                String.valueOf(SettingsEnum.TTS_DRIVER.getDefaultValue()))));
+
+        ListPreference listDriver = findPreference(SettingsEnum.TTS_DRIVER.getKey());
+        assert listDriver != null;
+        listDriver.setOnPreferenceChangeListener((preference, newValue) -> {
+            editTextPreferenceToken.setEnabled(Driver.AZURE_SDK.getId().equals(newValue));
+            return true;
+        });
     }
 }
diff --git a/app/src/main/java/com/utopiaxc/utopiatts/tts/WsTts.java b/app/src/main/java/com/utopiaxc/utopiatts/tts/WsTts.java
@@ -4,15 +4,11 @@
 
 import android.content.Context;
 import android.content.SharedPreferences;
-import android.media.AudioFormat;
 import android.media.MediaCodec;
 import android.media.MediaExtractor;
 import android.media.MediaFormat;
-import android.os.Build;
 import android.os.SystemClock;
 import android.speech.tts.SynthesisCallback;
-import android.speech.tts.SynthesisRequest;
-import android.speech.tts.TextToSpeech;
 import android.text.TextUtils;
 import android.util.Log;
 
@@ -37,7 +33,6 @@
 
 import java.io.IOException;
 import java.nio.ByteBuffer;
-import java.nio.ByteOrder;
 import java.nio.charset.StandardCharsets;
 import java.util.Date;
 import java.util.Objects;
@@ -63,6 +58,7 @@ public class WsTts implements Tts {
     private final Buffer mData = new Buffer();
     private MediaCodec mMediaCodec;
     private String mOldMime;
+    private Ssml mSsml;
     private SynthesisCallback mCallback;
     private final WebSocketListener mWebSocketListener = new WebSocketListener() {
         @Override
@@ -87,16 +83,13 @@ public void onFailure(@NotNull WebSocket webSocket, @NotNull Throwable t, @Nulla
             super.onFailure(webSocket, t, response);
             mWebSocket = null;
             mWebSocketState = WebSocketState.OFFLINE;
-            Log.e(TAG, "onFailure" + t.getMessage(), t);
-            if (mIsSynthesizing) {
-                mWebSocket = getOrCreateWs();
-            }
+            Log.e(TAG, "onFailure, throwable = \n",t);
+            getOrCreateWs().send(mSsml.toStringForWs());
         }
 
         @Override
         public void onMessage(@NotNull WebSocket webSocket, @NotNull String text) {
             super.onMessage(webSocket, text);
-            //Log.v(TAG, "onMessage" + text);
             final String endTag = "turn.end";
             final String startTag = "turn.start";
             int endIndex = text.lastIndexOf(endTag);
@@ -121,15 +114,12 @@ public void onMessage(@NotNull WebSocket webSocket, @NotNull ByteString bytes) {
             final String audioTag = "Path:audio\r\n";
             final String startTag = "Content-Type:";
             final String endTag = "\r\nX-StreamId";
-
             int audioIndex = bytes.lastIndexOf(audioTag.getBytes(StandardCharsets.UTF_8)) + audioTag.length();
             int startIndex = bytes.lastIndexOf(startTag.getBytes(StandardCharsets.UTF_8)) + startTag.length();
             int endIndex = bytes.lastIndexOf(endTag.getBytes(StandardCharsets.UTF_8));
             if (audioIndex != -1) {
-
                 try {
                     String temp = bytes.substring(startIndex, endIndex).utf8();
-                    Log.v(TAG, "当前Mime:" + temp);
                     String mCurrentMime;
                     if (temp.startsWith("audio")) {
                         mCurrentMime = temp;
@@ -139,12 +129,10 @@ public void onMessage(@NotNull WebSocket webSocket, @NotNull ByteString bytes) {
                     if (!mOutputFormat.needDecode()) {
                         if ("audio/x-wav".equals(mCurrentMime) && bytes.lastIndexOf("RIFF".getBytes(StandardCharsets.UTF_8)) != -1) {
                             audioIndex += 44;
-                            Log.v(TAG, "移除WAV文件头");
                         }
                     }
                     mData.write(bytes.substring(audioIndex));
                 } catch (Exception e) {
-                    Log.e(TAG, "onMessage Error:", e);
                     mIsSynthesizing = false;
                 }
             }
@@ -199,12 +187,17 @@ public boolean doSpeak(String text, int pitch, int rate,SynthesisCallback synthe
             }
         }
         mIsSynthesizing = false;
-        return false;
+        return true;
     }
 
     @Override
     public void stopSpeak() {
-
+        if (mWebSocket != null) {
+            Objects.requireNonNull(mWebSocket).close(1000, "closed by call onStop");
+            mWebSocket = null;
+        }
+        mIsSynthesizing = false;
+        mData.clear();
     }
 
     @Override
@@ -237,24 +230,41 @@ public synchronized void sendText(String text, int pitch, int rate) {
         int styleDegree = mSharedPreferences.getInt(SettingsEnum.STYLE_DEGREE.getKey(),
                 (Integer) SettingsEnum.STYLE_DEGREE.getDefaultValue());
 
-        Ssml ssml = new Ssml(text, actor.getId(), pitch,
+        mSsml = new Ssml(text, actor.getId(), pitch,
                 rate, role.getId(), style.getId(), styleDegree);
 
-        try {
-            boolean success = getOrCreateWs().send(ssml.toStringForWs());
-            if (!success && mIsSynthesizing) {
-                getOrCreateWs().send(ssml.toStringForWs());
-            }
-        } catch (Exception e) {
-            getOrCreateWs();
-            while (mWebSocket == null) {
+        while (mIsSynthesizing){
+            Log.w(TAG,"try sendText");
+            try {
+                if (getOrCreateWs().send(mSsml.toStringForWs())){
+                    break;
+                }
+            }catch (Exception e){
                 try {
                     this.wait(500);
                 } catch (Exception ignored) {
                 }
+                Log.w(TAG,"Retry sendText");
             }
-            getOrCreateWs().send(ssml.toStringForWs());
         }
+//        try {
+//            while(!(getOrCreateWs().send(ssml.toStringForWs()))&&mIsSynthesizing){
+//                Log.w(TAG,"Retry sendText");
+//            }
+//            boolean success = getOrCreateWs().send(ssml.toStringForWs());
+//            if (!success && mIsSynthesizing) {
+//                getOrCreateWs().send(ssml.toStringForWs());
+//            }
+//        } catch (Exception e) {
+//            getOrCreateWs();
+//            while (mWebSocket == null) {
+//                try {
+//                    this.wait(500);
+//                } catch (Exception ignored) {
+//                }
+//            }
+//            getOrCreateWs().send(ssml.toStringForWs());
+//        }
     }
 
     private synchronized void sendConfig(@NonNull WebSocket ws) {
@@ -265,7 +275,7 @@ private synchronized void sendConfig(@NonNull WebSocket ws) {
                 "{\"synthesis\":" +
                 "{\"audio\":" +
                 "{\"metadataoptions\":" +
-                "{\"sentenceBoundaryEnabled\":\"false\",\"wordBoundaryEnabled\":\"true\"" +
+                "{\"sentenceBoundaryEnabled\":\"true\",\"wordBoundaryEnabled\":\"true\"" +
                 "},\"outputFormat\":\"" + mOutputFormat.getValue() + "\"}}}}";
         ws.send(msg);
     }
@@ -387,7 +397,6 @@ private synchronized void doDecode(@NonNull ByteString data) {
     private synchronized void doUnDecode(@NonNull ByteString data) {
         mIsSynthesizing = true;
         int length = data.toByteArray().length;
-        //最大BufferSize
         final int maxBufferSize = mCallback.getMaxBufferSize();
         int offset = 0;
         while (offset < length && mIsSynthesizing) {

diff --git a/app/src/main/java/com/utopiaxc/utopiatts/tts/enums/OutputFormat.java b/app/src/main/java/com/utopiaxc/utopiatts/tts/enums/OutputFormat.java
@@ -247,6 +247,6 @@ public static OutputFormat getOutputFormat(String name) {
                 return outputFormat;
             }
         }
-        return RAW_48K_HZ_16_BIT_MONO_PCM;
+        return OGG_48K_HZ_16_BIT_MONO_OPUS;
     }
 }
diff --git a/app/src/main/java/com/utopiaxc/utopiatts/tts/utils/CommonTool.java b/app/src/main/java/com/utopiaxc/utopiatts/tts/utils/CommonTool.java
@@ -180,6 +180,16 @@ public static String getTime() {
         return sdf.format(date);
     }
 
+    /**
+     * 获取时间戳
+     *
+     * @return String time
+     */
+    public static String getTime(long timestamp) {
+        Date date = new Date(timestamp);
+        return sdf.format(date);
+    }
+
 
     public static String localeToEmoji(Locale locale) {
         String countryCode = locale.getCountry();

diff --git a/app/src/main/java/com/utopiaxc/utopiatts/tts/utils/Ssml.java b/app/src/main/java/com/utopiaxc/utopiatts/tts/utils/Ssml.java
@@ -4,24 +4,27 @@
 
 import androidx.annotation.NonNull;
 
+import java.util.List;
+
 public class Ssml {
     private static final String TAG = "Ssml";
     private final String mActor;
     private final int mPitch;
     private final int mRate;
     private final String mStyle;
     private final int mStyleDegree;
-    private final String mText;
+    private final StringBuilder mText;
     private final String mRole;
 
     public Ssml(String text, String actor, int pitch, int rate,String role, String style, int styleDegree) {
-        this.mText = text;
-        this.mActor = actor;
-        this.mPitch = (pitch/4)-25;
-        this.mRate = (rate/4)-25;
-        this.mRole =role;
-        this.mStyle = style;
-        this.mStyleDegree = styleDegree;
+        mText = new StringBuilder(text);
+        mActor = actor;
+        mPitch = (pitch/4)-25;
+        mRate = (rate/4)-25;
+        mRole =role;
+        mStyle = style;
+        mStyleDegree = styleDegree;
+        handleContent();
     }
 
 
@@ -53,14 +56,15 @@ public String toString() {
     }
 
     public String toStringForWs(){
+        long timestamp=System.currentTimeMillis();
         StringBuilder sb = new StringBuilder()
                 .append("Path:ssml\r\n")
-                .append("X-RequestId:").append(CommonTool.getMD5String(mText + "" + System.currentTimeMillis()))
+                .append("X-RequestId:").append(CommonTool.getMD5String(mText + "" + timestamp))
                 .append("\r\n")
                 .append("X-Timestamp:")
-                .append(CommonTool.getTime()).append("Z\r\n")
+                .append(CommonTool.getTime(timestamp)).append("Z\r\n")
                 .append("Content-Type:application/ssml+xml\r\n\r\n");
-        sb.append("<speak xmlns=\"http://www.w3.org/2001/10/synthesis\" xmlns:mstts=\"http://www.w3.org/2001/mstts\" xmlns:emo=\"http://www.w3.org/2009/10/emotionml\" version=\"1.0\" xml:lang=\"en-US\">");
+        sb.append("<speak xmlns=\"http://www.w3.org/2001/10/synthesis\" xmlns:mstts=\"http://www.w3.org/2001/mstts\" xmlns:emo=\"http://www.w3.org/2009/10/emotionml\" version=\"1.0\" xml:lang=\"zh-CN\">");
         sb.append("<voice name=\"").append(mActor).append("\">");
         if (!"".equals(mStyle)||!"".equals(mRole)) {
             sb.append("<mstts:express-as ");
@@ -82,4 +86,14 @@ public String toStringForWs(){
         Log.d(TAG, "toString = " + sb);
         return sb.toString();
     }
+
+    private void handleContent() {
+        CommonTool.replace(mText, "\n", " ");
+        CommonTool.Trim(mText);
+        CommonTool.replace(mText, "&", "&amp;");
+        CommonTool.replace(mText, "\"", "&quot;");
+        CommonTool.replace(mText, "'", "&apos;");
+        CommonTool.replace(mText, ">", "&lt;");
+        CommonTool.replace(mText, "<", "&gt;");
+    }
 }
diff --git a/app/src/main/res/values-zh/arrays.xml b/app/src/main/res/values-zh/arrays.xml
@@ -90,4 +90,49 @@
         <item>轻松叙事</item>
         <item>纪录片解说</item>
     </string-array>
+
+    <string-array name="tts_output_format_entries">
+        <item>AUDIO_16K_HZ_128K_BIT_RATE_MONO_MP3 *</item>
+        <item>AUDIO_16K_HZ_16_BIT_32KBPS_MONO_OPUS *</item>
+        <item>AUDIO_16K_HZ_16KBPS_MONO_SIREN</item>
+        <item>AUDIO_16K_HZ_32K_BIT_RATE_MONO_MP3 *</item>
+        <item>AUDIO_16K_HZ_64K_BIT_RATE_MONO_MP3 *</item>
+        <item>AUDIO_24K_HZ_160K_BIT_RATE_MONO_MP3 *</item>
+        <item>AUDIO_24K_HZ_16_BIT_24KBPS_MONO_OPUS *</item>
+        <item>AUDIO_24K_HZ_16_BIT_48KBPS_MONO_OPUS *</item>
+        <item>AUDIO_24K_HZ_48K_BIT_RATE_MONO_MP3 *</item>
+        <item>AUDIO_24K_HZ_96K_BIT_RATE_MONO_MP3 *</item>
+        <item>AUDIO_48K_HZ_192K_BIT_RATE_MONO_MP3 *</item>
+        <item>AUDIO_48K_HZ_96K_BIT_RATE_MONO_MP3 *</item>
+        <item>OGG_16K_HZ_16_BIT_MONO_OPUS *</item>
+        <item>OGG_24K_HZ_16_BIT_MONO_OPUS *</item>
+        <item>OGG_48K_HZ_16_BIT_MONO_OPUS * 默认格式</item>
+        <item>RAW_16K_HZ_16_BIT_MONO_PCM</item>
+        <item>RAW_16K_HZ_16_BIT_MONO_TRUE_SILK</item>
+        <item>RAW_22050_HZ_16_BIT_MONO_PCM</item>
+        <item>RAW_24K_HZ_16_BIT_MONO_PCM</item>
+        <item>RAW_24K_HZ_16_BIT_MONO_TRUE_SILK</item>
+        <item>RAW_44100_HZ_16_BIT_MONO_PCM</item>
+        <item>RAW_48K_HZ_16_BIT_MONO_PCM</item>
+        <item>RAW_8K_HZ_16_BIT_MONO_PCM</item>
+        <item>RAW_8K_HZ_8_BIT_MONO_ALAW</item>
+        <item>RAW_8K_HZ_8_BIT_MONO_MULAW</item>
+        <item>RIFF_16K_HZ_16_BIT_MONO_PCM</item>
+        <item>RIFF_16K_HZ_16KBPS_MONO_SIREN</item>
+        <item>RIFF_22050_HZ_16_BIT_MONO_PCM</item>
+        <item>RIFF_24K_HZ_16_BIT_MONO_PCM</item>
+        <item>RIFF_44100_HZ_16_BIT_MONO_PCM</item>
+        <item>RIFF_48K_HZ_16_BIT_MONO_PCM</item>
+        <item>RIFF_8K_HZ_16_BIT_MONO_PCM</item>
+        <item>RIFF_8K_HZ_8_BIT_MONO_ALAW</item>
+        <item>RIFF_8K_HZ_8_BIT_MONO_MULAW</item>
+        <item>WEBM_16K_HZ_16_BIT_MONO_OPUS *</item>
+        <item>WEBM_24K_HZ_16_BIT_24KBPS_MONO_OPUS *</item>
+        <item>WEBM_24K_HZ_16_BIT_MONO_OPUS *</item>
+    </string-array>
+
+    <string-array name="tts_driver_entries">
+        <item>Azure SDK个人密钥</item>
+        <item>官网Demo WebSocket</item>
+    </string-array>
 </resources>
diff --git a/app/src/main/res/values-zh/strings.xml b/app/src/main/res/values-zh/strings.xml
@@ -23,7 +23,7 @@
     <string name="confirm">确认</string>
     <string name="tts_driver_title">TTS引擎驱动方式</string>
     <string name="tts_driver_title_sdk">Azure SDK</string>
-    <string name="tts_driver_title_ws">WebSocket</string>
+    <string name="tts_driver_title_ws">Demo WebSocket</string>
     <string name="warning">警告</string>
     <string name="warning_blank_token">请输入Azure TTS服务密钥。</string>
     <string name="warning_azure_token_not_checked">您的Azure信息未经检查，请先点击确认进行检查。</string>
@@ -52,4 +52,6 @@
     <string name="sorry">抱歉</string>
     <string name="tutorials_is_not_ready">教程还未编写完成。</string>
     <string name="tips_azure_token">此处输入从微软Azure中获取的TTS服务密钥，如果不理解，可以点击教程前往了解。</string>
+    <string name="output_format">音频流格式</string>
+    <string name="warning_of_output_format">"如果您的语音不卡顿，请不要修改本项，因为部分语音在WebSocket模式下存在部分无法解决的问题以加剧卡顿。\n如果需要修改，请首选带'*'的格式，因为它们是需要进行解码的格式，因此可能更加稳定。"</string>
 </resources>