diff --git a/record_and_predict.py b/record_and_predict.py
index f99e676..d5ba55f 100644
--- a/record_and_predict.py
+++ b/record_and_predict.py
@@ -104,6 +104,7 @@ def record_and_predict():
     # Segment assembly state
     segment = []             # list of float32 chunks (includes pre, speech, trailing silence)
     speech_active = False
+    turn_started = False
     trailing_silence = 0
     since_trigger_chunks = 0
 
@@ -130,15 +131,23 @@ def record_and_predict():
             is_speech = vad.prob(f32) > VAD_THRESHOLD
 
             if not speech_active:
-                # Warmup pre-speech buffer until trigger
-                pre_buffer.append(f32)
+                # It could mean two cases:
+                # 1. There are already some VAD spans, but it haven't reach a turn end
+                # 2. No VAD span has been identified
+                if not turn_started and not is_speech:
+                    # Warmup pre-speech buffer until trigger
+                    pre_buffer.append(f32)
                 if is_speech:
-                    # Trigger: start a new segment with pre-speech
-                    segment = list(pre_buffer)
+                    if not turn_started:
+                        # Trigger: start a new segment with pre-speech
+                        segment = list(pre_buffer)
+                        since_trigger_chunks = 0
                     segment.append(f32)
+                    since_trigger_chunks += 1
                     speech_active = True
                     trailing_silence = 0
-                    since_trigger_chunks = 1
+                    turn_started = True
+                    print(">> VAD span started")
             else:
                 # Already in a segment
                 segment.append(f32)
@@ -152,13 +161,16 @@ def record_and_predict():
                 if trailing_silence >= stop_chunks or since_trigger_chunks >= max_chunks:
                     # Pause capture while we process
                     stream.stop_stream()
-                    _process_segment(np.concatenate(segment, dtype=np.float32))
-                    # Reset for next segment
-                    segment.clear()
+                    print(">> VAD span ended")
+                    if _process_segment(np.concatenate(segment, dtype=np.float32)):
+                        # Reset for next segment
+                        segment.clear()
+                        trailing_silence = 0
+                        since_trigger_chunks = 0
+                        pre_buffer.clear()
+                        turn_started = False
+                        print("[WARN] Turn ended")
                     speech_active = False
-                    trailing_silence = 0
-                    since_trigger_chunks = 0
-                    pre_buffer.clear()
                     stream.start_stream()
                     print("Listening for speech...")
 
@@ -173,7 +185,7 @@ def record_and_predict():
 def _process_segment(segment_audio_f32: np.ndarray):
     if segment_audio_f32.size == 0:
         print("Captured empty audio segment, skipping prediction.")
-        return
+        return 0
 
     if DEBUG_SAVE_WAV:
         wavfile.write(TEMP_OUTPUT_WAV, RATE, (segment_audio_f32 * 32767.0).astype(np.int16))
@@ -192,6 +204,7 @@ def _process_segment(segment_audio_f32: np.ndarray):
     print(f"Prediction: {'Complete' if pred == 1 else 'Incomplete'}")
     print(f"Probability of complete: {prob:.4f}")
     print(f"Inference time: {dt_ms:.2f} ms")
+    return pred
 
 
 if __name__ == "__main__":