TtsService: support detecting SSML documents

android/jni/jni/eSpeakService.cpp

@@ -314,7 +314,7 @@ JNICALL Java_com_reecedunn_espeak_SpeechSynthesis_nativeSetPitch(
 
 JNIEXPORT jboolean
 JNICALL Java_com_reecedunn_espeak_SpeechSynthesis_nativeSynthesize(
-    JNIEnv *env, jobject object, jstring text) {
+    JNIEnv *env, jobject object, jstring text, jboolean isSsml) {
   if (DEBUG) LOGV("%s", __FUNCTION__);
   native_data_t *nat = getNativeData(env, object);
   const char *c_text = text ? env->GetStringUTFChars(text, NULL) : NULL;
@@ -325,7 +325,8 @@ JNICALL Java_com_reecedunn_espeak_SpeechSynthesis_nativeSynthesize(
   espeak_SetSynthCallback(SynthCallback);
   const espeak_ERROR result = espeak_Synth(c_text, strlen(c_text), 0,  // position
                POS_CHARACTER, 0, // end position (0 means no end position)
-               espeakCHARS_UTF8, // text is UTF-8 encoded
+               isSsml ? espeakCHARS_UTF8 | espeakSSML // UTF-8 encoded SSML
+                      : espeakCHARS_UTF8,             // UTF-8 encoded text
                &unique_identifier, nat);
   espeak_Synchronize();
 

android/src/com/reecedunn/espeak/SpeechSynthesis.java

@@ -179,8 +179,8 @@ public class SpeechSynthesis {
         nativeSetPitch(pitch);
     }
 
-    public void synthesize(String text) {
-        nativeSynthesize(text);
+    public void synthesize(String text, boolean isSsml) {
+        nativeSynthesize(text, isSsml);
     }
 
     public void stop() {
@@ -257,7 +257,7 @@ public class SpeechSynthesis {
 
     private native final boolean nativeSetPitch(int pitch);
 
-    private native final boolean nativeSynthesize(String text);
+    private native final boolean nativeSynthesize(String text, boolean isSsml);
 
     private native final boolean nativeStop();
 

android/src/com/reecedunn/espeak/TtsService.java

@@ -192,7 +192,10 @@ public class TtsService extends TextToSpeechService {
             return;
         }
 
-        final String text = request.getText();
+        String text = request.getText();
+        if (text == null)
+            return;
+
         final int gender = getDefaultGender();
         final int rate = scaleRate(request.getSpeechRate());
         final int pitch = scalePitch(request.getPitch());
@@ -207,6 +210,13 @@ public class TtsService extends TextToSpeechService {
             }
         }
 
+        if (text.startsWith("<?xml"))
+        {
+            // eSpeak does not recognise/skip "<?...?>" preprocessing tags,
+            // so need to remove these before passing to synthesize.
+            text = text.substring(text.indexOf("?>") + 2).trim();
+        }
+
         mCallback = callback;
         mCallback.start(mEngine.getSampleRate(), mEngine.getAudioFormat(),
                 mEngine.getChannelCount());
@@ -214,7 +224,7 @@ public class TtsService extends TextToSpeechService {
         mEngine.setVoiceByProperties(null, mMatchingVoice.name, gender, 0, 0);
         mEngine.setRate(rate);
         mEngine.setPitch(pitch);
-        mEngine.synthesize(text);
+        mEngine.synthesize(text, text.startsWith("<speak"));
     }
 
     /**