diff --git a/voice_recognition.py b/voice_recognition.py
index d6b9db5..00eb40b 100644
--- a/voice_recognition.py
+++ b/voice_recognition.py
@@ -32,8 +32,11 @@ SAMPLE_WIDTH = OpusDecoder.SAMPLE_SIZE // OpusDecoder.CHANNELS
 SAMPLING_RATE = OpusDecoder.SAMPLING_RATE
 
 
+#rotate file after there is 0.5s between last received pcm for user.
+#delete messsages after user disconnect
+
 logger = logging.getLogger("discord")
-location = "/home/pi/Conjurer/"
+location = "/home/pi/Conjurer/transcripts/"
 
 
 class WaveWriter:
@@ -43,7 +46,7 @@ class WaveWriter:
         self.present_file_id = 0
         self.file_name_past = None
 
-        self.file_name_present = location + self.username + str(self.present_file_id)
+        self.file_name_present = location + self.username + "_" + str(self.present_file_id)
         self.transcript_file_present : wave.Wave_write = wave.open(
             self.file_name_present, "wb"
         )
@@ -51,7 +54,7 @@ class WaveWriter:
         self.transcript_file_present.setsampwidth(SAMPLE_WIDTH)
         self.transcript_file_present.setframerate(SAMPLING_RATE)
 
-        self.file_name_future = location + self.username + str(self.present_file_id + 1)
+        self.file_name_future = location + self.username + "_" + str(self.present_file_id + 1)
         self.transcript_file_future : wave.Wave_write = wave.open(
             self.file_name_future, "wb"
         )
@@ -66,9 +69,7 @@ class WaveWriter:
         self.present_file_id += 1
         self.file_name_present = self.file_name_future
         self.transcript_file_present = self.transcript_file_future
-        self.transcript_file_future = (
-            location + self.username + str(self.present_file_id + 1)
-        )
+        self.file_name_future = location + self.username + "_" + str(self.present_file_id + 1)
         self.transcript_file_future : wave.Wave_write = wave.open(
             self.file_name_future, "wb"
         )
@@ -106,12 +107,15 @@ class SRBuffer(voice_recv.AudioSink):
 
     def on_user_disconnect(self,username):
         self.wavewriter[username].cleanup()
+        self.wavewriter.pop(username)
+
+    def rotate_user(self,username):
+        self.wavewriter[username].rotate()
 
     def wants_opus(self) -> bool:
         return False
 
     def write(self, user, data) -> None:
-        logger.info("user: %s", user)
         self.wavewriter[str(user.id)].writeframes(data.pcm)
 
     def cleanup(self) -> None:
@@ -160,10 +164,14 @@ class Transcriber(commands.Cog):
     async def on_voice_state_update(self, user, before, after):
         if before.channel is None:
             logger.info("User %s connected to channel %s", user, after.channel.name)
+            self.wsink.on_user_connect(user.id)
         elif after.channel is  None:
             logger.info("User %s disconnected from channel %s", user, before.channel.name)
+            self.wsink.on_user_disconnect(user.id)
+        elif after.self_mute or after.mute:
+            self.wsink.rotate_user(user.id)
         else:
-            logger.info("User VC status changed %s", user)
+            logger.info("User VC status changed %s", user.id)
             logger.info("Before %s", before)
             logger.info("After %s", after)