diff --git a/android/SherpaOnnxTts/app/src/main/java/com/k2fsa/sherpa/onnx/Tts.kt b/android/SherpaOnnxTts/app/src/main/java/com/k2fsa/sherpa/onnx/Tts.kt
index 132e8b06a..d08e803a5 100644
--- a/android/SherpaOnnxTts/app/src/main/java/com/k2fsa/sherpa/onnx/Tts.kt
+++ b/android/SherpaOnnxTts/app/src/main/java/com/k2fsa/sherpa/onnx/Tts.kt
@@ -23,7 +23,7 @@ data class OfflineTtsModelConfig(
 data class OfflineTtsConfig(
     var model: OfflineTtsModelConfig,
     var ruleFsts: String = "",
-    var maxNumSentences: Int = 2,
+    var maxNumSentences: Int = 1,
 )
 
 class GeneratedAudio(
diff --git a/scripts/apk/generate-tts-apk-script.py b/scripts/apk/generate-tts-apk-script.py
index 958f55475..fd24be887 100755
--- a/scripts/apk/generate-tts-apk-script.py
+++ b/scripts/apk/generate-tts-apk-script.py
@@ -33,6 +33,23 @@ class TtsModel:
     data_dir: Optional[str] = None
 
 
+def get_coqui_models() -> List[TtsModel]:
+    # English (coqui-ai/TTS)
+    models = [
+        TtsModel(model_dir="vits-coqui-en-ljspeech"),
+        TtsModel(model_dir="vits-coqui-en-ljspeech-neon"),
+        TtsModel(model_dir="vits-coqui-en-vctk"),
+        #  TtsModel(model_dir="vits-coqui-en-jenny"),
+    ]
+
+    for m in models:
+        m.data_dir = m.model_dir + "/" + "espeak-ng-data"
+        m.model_name = "model.onnx"
+        m.lang = "en"
+
+    return models
+
+
 def get_piper_models() -> List[TtsModel]:
     models = [
         TtsModel(model_dir="vits-piper-ar_JO-kareem-low"),
@@ -137,6 +154,7 @@ def get_piper_models() -> List[TtsModel]:
         TtsModel(model_dir="vits-piper-vi_VN-vivos-x_low"),
         TtsModel(model_dir="vits-piper-zh_CN-huayan-medium"),
     ]
+
     for m in models:
         m.data_dir = m.model_dir + "/" + "espeak-ng-data"
         m.model_name = m.model_dir[len("vits-piper-") :] + ".onnx"
@@ -145,7 +163,7 @@ def get_piper_models() -> List[TtsModel]:
     return models
 
 
-def get_all_models() -> List[TtsModel]:
+def get_vits_models() -> List[TtsModel]:
     return [
         # Chinese
         TtsModel(
@@ -202,12 +220,6 @@ def get_all_models() -> List[TtsModel]:
             lang="zh",
             rule_fsts="vits-zh-hf-theresa/rule.fst",
         ),
-        # English (coqui-ai/TTS)
-        # fmt: off
-        TtsModel(model_dir="vits-coqui-en-ljspeech", model_name="model.onnx", lang="en"),
-        TtsModel(model_dir="vits-coqui-en-ljspeech-neon", model_name="model.onnx", lang="en"),
-        TtsModel(model_dir="vits-coqui-en-vctk", model_name="model.onnx", lang="en"),
-        #  TtsModel(model_dir="vits-coqui-en-jenny", model_name="model.onnx", lang="en"),
         # English (US)
         TtsModel(model_dir="vits-vctk", model_name="vits-vctk.onnx", lang="en"),
         TtsModel(model_dir="vits-ljs", model_name="vits-ljs.onnx", lang="en"),
@@ -225,8 +237,11 @@ def main():
         s = f.read()
     template = environment.from_string(s)
     d = dict()
-    #  all_model_list = get_all_models()
-    all_model_list = get_piper_models()
+
+    all_model_list = get_vits_models()
+    all_model_list += get_piper_models()
+    all_model_list += get_coqui_models()
+
     num_models = len(all_model_list)
 
     num_per_runner = num_models // total
diff --git a/sherpa-onnx/csrc/piper-phonemize-lexicon.cc b/sherpa-onnx/csrc/piper-phonemize-lexicon.cc
index c27d92a11..20512e6cb 100644
--- a/sherpa-onnx/csrc/piper-phonemize-lexicon.cc
+++ b/sherpa-onnx/csrc/piper-phonemize-lexicon.cc
@@ -122,6 +122,8 @@ static std::vector<int64_t> CoquiPhonemesToIds(
   int32_t eos_id = meta_data.eos_id;
   int32_t blank_id = meta_data.blank_id;
   int32_t add_blank = meta_data.add_blank;
+  int32_t comma_id = token2id.at(',');
+  SHERPA_ONNX_LOGE("comma id: %d", comma_id);
 
   std::vector<int64_t> ans;
   if (add_blank) {
@@ -158,6 +160,9 @@ static std::vector<int64_t> CoquiPhonemesToIds(
     }
   }
 
+  // add a comma at the end of a sentence so that we can have a longer pause.
+  ans.push_back(comma_id);
+
   if (use_eos_bos) {
     ans.push_back(eos_id);
   }