diff --git a/CHANGELOG.md b/CHANGELOG.md
index d2af96e..20d4315 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,3 +1,5 @@
 - automatic download and installation model
 - batch change language
 - model list reorder
+- engine cache size
+- number of threads per engine 
\ No newline at end of file
diff --git a/app/src/main/java/com/k2fsa/sherpa/onnx/Tts.kt b/app/src/main/java/com/k2fsa/sherpa/onnx/Tts.kt
index 5e6100c..6c1146f 100644
--- a/app/src/main/java/com/k2fsa/sherpa/onnx/Tts.kt
+++ b/app/src/main/java/com/k2fsa/sherpa/onnx/Tts.kt
@@ -159,7 +159,7 @@ fun getOfflineTtsConfig(
                 model = "$modelDir/$modelName",
                 lexicon = "$modelDir/$lexicon",
                 tokens = "$modelDir/tokens.txt",
-                dataDir = "$dataDir"
+                dataDir = dataDir
             ),
             numThreads = 2,
             debug = true,
diff --git a/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/conf/TtsConfig.kt b/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/conf/TtsConfig.kt
index e674846..1a825c5 100644
--- a/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/conf/TtsConfig.kt
+++ b/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/conf/TtsConfig.kt
@@ -16,7 +16,19 @@ object TtsConfig {
     val timeoutDestruction = mutableDataSaverStateOf(
         dataSaverInterface = dataSaverPref,
         key = "timeoutDestruction",
-        initialValue = 0
+        initialValue = 3
+    )
+
+    val cacheSize = mutableDataSaverStateOf(
+        dataSaverInterface = dataSaverPref,
+        key = "cacheSize",
+        initialValue = 3
+    )
+
+    val threadNum = mutableDataSaverStateOf(
+        dataSaverInterface = dataSaverPref,
+        key = "threadNum",
+        initialValue = 2
     )
 
 }
\ No newline at end of file
diff --git a/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/synthesizer/ModelManager.kt b/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/synthesizer/ModelManager.kt
index f854c88..61d7c95 100644
--- a/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/synthesizer/ModelManager.kt
+++ b/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/synthesizer/ModelManager.kt
@@ -135,7 +135,6 @@ object ModelManager {
                     tokens = format(tokens),
                     dataDir = format(dataDir),
                 ),
-                numThreads = 2,
                 debug = true,
                 provider = "cpu",
             ),
diff --git a/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/synthesizer/SynthesizerCache.kt b/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/synthesizer/SynthesizerCache.kt
index 4094633..84e745c 100644
--- a/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/synthesizer/SynthesizerCache.kt
+++ b/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/synthesizer/SynthesizerCache.kt
@@ -6,6 +6,8 @@ import java.util.concurrent.DelayQueue
 import java.util.concurrent.Delayed
 import java.util.concurrent.Executors
 import java.util.concurrent.TimeUnit
+import kotlin.math.max
+import kotlin.math.min
 
 
 interface ImplCache {
@@ -13,10 +15,13 @@ interface ImplCache {
 }
 
 
-class SynthesizerCache {
+internal class SynthesizerCache {
     companion object {
         private val delayTime: Int
-            get() = 1000 * 60 * TtsConfig.timeoutDestruction.value
+            get() = 1000 * 60 * min(1, TtsConfig.timeoutDestruction.value)
+
+        private val maxCacheSize: Int
+            get() = max(1, TtsConfig.cacheSize.value)
     }
 
     private val delayQueue = DelayQueue<DelayedDestroyTask>()
@@ -44,15 +49,31 @@ class SynthesizerCache {
         }
     }
 
+    private fun limitSize() {
+        if (queueMap.size > maxCacheSize) {
+            val oldestEntry =
+                queueMap.entries.minByOrNull { it.value.getDelay(TimeUnit.MILLISECONDS) }
+            oldestEntry?.let {
+                queueMap.remove(it.key)
+                delayQueue.remove(it.value)
+            }
+        }
+    }
+
+    @Synchronized
     fun cache(id: String, obj: ImplCache) {
-        val task =
-            DelayedDestroyTask(delayTime = delayTime, id, obj)
+        limitSize()
+
+        val task = DelayedDestroyTask(delayTime = delayTime, id, obj)
         delayQueue.add(task)
         queueMap[id] = task
         ensureTaskRunning()
     }
 
+    @Synchronized
     fun getById(id: String): ImplCache? {
+        limitSize()
+
         queueMap[id]?.let {
             if (it.getDelay(TimeUnit.MILLISECONDS) <= 1000 * 10) { // 小于10s便重置
                 it.reset()
diff --git a/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/synthesizer/SynthesizerManager.kt b/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/synthesizer/SynthesizerManager.kt
index 9a0d6bb..60af762 100644
--- a/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/synthesizer/SynthesizerManager.kt
+++ b/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/synthesizer/SynthesizerManager.kt
@@ -3,6 +3,8 @@ package com.k2fsa.sherpa.onnx.tts.engine.synthesizer
 import android.util.Log
 import com.k2fsa.sherpa.onnx.OfflineTts
 import com.k2fsa.sherpa.onnx.OfflineTtsConfig
+import com.k2fsa.sherpa.onnx.tts.engine.conf.TtsConfig
+import kotlin.math.max
 
 object SynthesizerManager {
     const val TAG = "SynthesizerManager"
@@ -15,9 +17,11 @@ object SynthesizerManager {
             Log.d(TAG, "getTTS (from cache): ${it.config}")
         }
 
-        return tts ?: OfflineTts(
-            config = cfg
-        ).run {
+        val model = cfg.model.copy(
+            numThreads = max(1, TtsConfig.threadNum.value)
+        )
+
+        return tts ?: OfflineTts(config = cfg.copy(model = model)).run {
             cacheTTS(cfg.model.vits.model, this)
             this
         }
diff --git a/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/ui/settings/SettingsScreen.kt b/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/ui/settings/SettingsScreen.kt
index 0815825..b8ee527 100644
--- a/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/ui/settings/SettingsScreen.kt
+++ b/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/ui/settings/SettingsScreen.kt
@@ -17,20 +17,41 @@ fun SettingsScreen() {
     Scaffold { paddingValues ->
         Column(Modifier.padding(paddingValues)) {
             DividerPreference {
-                Text(stringResource(id = R.string.engine_cache))
+                Text(stringResource(id = R.string.engine))
             }
 
             var timeout by remember { TtsConfig.timeoutDestruction }
             val timeoutStr = stringResource(id = R.string.minute_format, timeout)
             SliderPreference(
                 valueRange = 1f..60f,
-                steps = 60,
                 title = { Text(stringResource(R.string.timeout_destruction)) },
                 subTitle = { Text(stringResource(R.string.timeout_destruction_summary))},
                 value = timeout.toFloat(),
                 onValueChange = { timeout = it.toInt() },
                 label = timeoutStr
             )
+
+            var cacheSize by remember { TtsConfig.cacheSize }
+            val cacheSizeStr = cacheSize.toString()
+            SliderPreference(
+                valueRange = 1f..10f,
+                title = { Text(stringResource(R.string.cache_size)) },
+                subTitle = { Text(stringResource(R.string.cache_size_summary))},
+                value = cacheSize.toFloat(),
+                onValueChange = { cacheSize = it.toInt() },
+                label = cacheSizeStr
+            )
+
+            var threadNum by remember { TtsConfig.threadNum }
+            val threadNumStr = threadNum.toString()
+            SliderPreference(
+                valueRange = 1f..8f,
+                title = { Text(stringResource(R.string.thread_num)) },
+                subTitle = { Text(stringResource(R.string.thread_num_summary))},
+                value = threadNum.toFloat(),
+                onValueChange = { threadNum = it.toInt() },
+                label = threadNumStr
+            )
         }
     }
 }
diff --git a/app/src/main/res/values/strings.xml b/app/src/main/res/values/strings.xml
index eb3ed08..54cf0e0 100644
--- a/app/src/main/res/values/strings.xml
+++ b/app/src/main/res/values/strings.xml
@@ -19,7 +19,7 @@
     <string name="desc_seekbar_remove">Seekbar remove</string>
     <string name="desc_seekbar_add">Seekbar add</string>
     <string name="close">Close</string>
-    <string name="engine_cache">Engine cache</string>
+    <string name="engine">Engine</string>
     <string name="timeout_destruction">Timeout destruction</string>
     <string name="minute_format">%1$s min</string>
     <string name="timeout_destruction_summary">The model is automatically destroyed after it has not been used for a specified period of time to save running memory.</string>
@@ -53,4 +53,8 @@
     <string name="model_install_failed">Model installation failed ❌</string>
     <string name="model_installed">Model installed</string>
     <string name="timed_out">Timed out</string>
+    <string name="cache_size">Cache size</string>
+    <string name="cache_size_summary">Maximum number of engine caches</string>
+    <string name="thread_num">Number of threads</string>
+    <string name="thread_num_summary">Maximum number of threads per engine</string>
 </resources>
\ No newline at end of file