diff --git a/local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/AzureTtsHelper.kt b/local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/AzureTtsHelper.kt
index 05223bd8d..5b0770249 100644
--- a/local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/AzureTtsHelper.kt
+++ b/local_plugins/azure_speech/android/src/main/kotlin/com/yunqiinnovation/azure_speech/AzureTtsHelper.kt
@@ -14,6 +14,7 @@ import com.deep_voice.speech.tts.TtsEventType
 import kotlinx.coroutines.*
 import java.io.ByteArrayInputStream
 import java.io.InputStream
+import java.util.concurrent.atomic.AtomicInteger
 import com.yunqiinnovation.azure_speech.tools.SimpleAudioPlayer
 /**
  * Azure TTS Helper
@@ -51,6 +52,9 @@ class AzureTtsHelper(private val context: Context) : ITtsService {
     private val streamBuffer = StringBuilder()
     private var lastSpeakTime = 0L
 
+    // 待完成的合成任务计数：只有全部完成才触发 PLAYBACK_COMPLETED
+    private val pendingSpeakCount = AtomicInteger(0)
+
     private var enableVoiceFlocking = false
     private var speakerProfileId: String = ""
 
@@ -310,8 +314,11 @@ class AzureTtsHelper(private val context: Context) : ITtsService {
                 )
                 isSpeaking = false
                 notifyEvent(TtsEventType.SYNTHESIS_COMPLETED)
-                // Azure TTS 合成完成即播放完成
-                notifyEvent(TtsEventType.PLAYBACK_COMPLETED)
+                // 所有句子都播完后才触发 PLAYBACK_COMPLETED，避免多句流式合成时重复触发
+                if (pendingSpeakCount.decrementAndGet() <= 0) {
+                    pendingSpeakCount.set(0)
+                    notifyEvent(TtsEventType.PLAYBACK_COMPLETED)
+                }
             }
 
             // 合成取消事件
@@ -561,8 +568,9 @@ class AzureTtsHelper(private val context: Context) : ITtsService {
             val ssml = generateOptimizedSsml(text)
 
             isSpeaking = true
+            pendingSpeakCount.incrementAndGet()
 
-            FileLogger.d(TAG, "开始语音合成，文本长度: ${text.length}")
+            FileLogger.d(TAG, "开始语音合成，文本长度: ${text.length}，当前待完成任务: ${pendingSpeakCount.get()}")
 
             // 直接调用SpeakSsmlAsync，SDK内部已经是异步的
             synthesizer?.SpeakSsmlAsync(ssml)
@@ -700,6 +708,7 @@ class AzureTtsHelper(private val context: Context) : ITtsService {
             synthesizer?.StopSpeakingAsync()
             streamBuffer.clear()
             isSpeaking = false
+            pendingSpeakCount.set(0)
             recreateSynthesizer()
             return true
         } catch (e: Exception) {