diff --git a/modules/features/teamsbot/service.py b/modules/features/teamsbot/service.py
index c92756fc..503cf347 100644
--- a/modules/features/teamsbot/service.py
+++ b/modules/features/teamsbot/service.py
@@ -87,6 +87,7 @@ class TeamsbotService:
         self._lastTranscriptSpeaker: Optional[str] = None
         self._lastTranscriptText: Optional[str] = None
         self._lastTranscriptId: Optional[str] = None
+        self._lastSttTime: float = 0.0
         self._lastBotResponseText: Optional[str] = None
         self._lastBotResponseTs: float = 0.0
 
@@ -623,12 +624,14 @@ class TeamsbotService:
 
         # Differential transcript writing:
         # audioCapture from same speaker → append text (merge STT chunks into one block)
-        # other sources → always create a new record
+        # Start a new block after a pause (>5s gap between STT results)
+        sttPauseThreshold = 5.0
         isMerge = (
             source == "audioCapture"
             and self._lastTranscriptSpeaker == speaker
             and self._lastTranscriptText is not None
             and self._lastTranscriptId is not None
+            and (time.time() - self._lastSttTime) < sttPauseThreshold
         )
 
         if isMerge:
@@ -680,6 +683,9 @@ class TeamsbotService:
                 count = session.get("transcriptSegmentCount", 0) + 1
                 interface.updateSession(sessionId, {"transcriptSegmentCount": count})
 
+        if source == "audioCapture":
+            self._lastSttTime = time.time()
+
         displayText = self._lastTranscriptText if isMerge else text
         await _emitSessionEvent(sessionId, "transcript", {
             "id": createdTranscript.get("id"),