Update app.py

app.py

@@ -28,16 +28,19 @@ for name in list_repo_files(repo_id="balacoon/tts"):
 
 def text_to_speech(text):
     with tempfile.NamedTemporaryFile(delete=False, suffix='.wav') as temp_file:
+        if len(text) > 1024:
+            # truncate the text
+            text_str = text[:1024]
+        else:
+            text_str = text
         with locker:
-            tts = TTS(os.path.join(os.getcwd(), tts_model_str))
-            if len(text) > 1024:
-                # truncate the text
-                text_str = text[:1024]
-            else:
-                text_str = text
-            audio_data = tts.synthesize(text_str, "92")
-        temp_file.write(np.ascontiguousarray(audio_data))
-    return temp_file.name
+            samples = tts.synthesize(text_str, "92")
+        output_file = temp_file.name
+        with wave.open(f"{output_file}", "w") as fp:
+            fp.setparams((1, 2, tts.get_sampling_rate(), len(samples), "NONE", "NONE"))
+            samples = np.ascontiguousarray(samples)
+            fp.writeframes(samples)
+    return output_file
 
 def combine_audio_files(audio_files):
     combined = AudioSegment.empty()