cafe waiter npc and mic detect speech more

Assets/_PROJECT/Scripts/ModeGeneration/FMODWhisperBridge.cs

@@ -62,6 +62,7 @@ public class FMODWhisperBridge : MonoBehaviour
 
     // activation flag
     private bool isRecordingActivated = false;
+    private bool _skipOneFeedFrame = false;
 
     private void Awake()
     {
@@ -188,13 +189,15 @@ public class FMODWhisperBridge : MonoBehaviour
         // Wire events
         _stream.OnSegmentUpdated += (seg) =>
         {
-            if (IsSpeechMeaningful(seg.Result))
-                OnWhisperSegmentUpdated?.Invoke(seg.Result);
+            string cleanedText = PostProcessInput(seg.Result);
+            if (!string.IsNullOrEmpty(cleanedText))
+                OnWhisperSegmentUpdated?.Invoke(cleanedText);
         };
         _stream.OnSegmentFinished += (seg) =>
         {
-            if (IsSpeechMeaningful(seg.Result))
-                OnWhisperSegmentFinished?.Invoke(seg.Result);
+            string cleanedText = PostProcessInput(seg.Result);
+            if (!string.IsNullOrEmpty(cleanedText))
+                OnWhisperSegmentFinished?.Invoke(cleanedText);
         };
 
         whisper.useVad = useVadInStream;
@@ -202,15 +205,28 @@ public class FMODWhisperBridge : MonoBehaviour
         _stream.StartStream();
         _streamStarted = true;
 
-        // Unpause loopback if it's meant to play only while active
-        if (playLoopback && loopbackOnlyWhenActive && _playChannel.hasHandle())
-            _playChannel.setPaused(false);
 
-        // Prepare temp arrays roughly 100ms of audio
-        EnsureTmpCapacity((rate / 10) * _nativeChannels);
+        // --- NEW: Clear the ring buffer and reset read pointer ---
+        // Pause loopback while we clear (optional, but avoids clicks)
+        if (playLoopback && _playChannel.hasHandle())
+            _playChannel.setPaused(true);
+
+        // Clear buffer bytes
+        ClearRecordRingBuffer();
+
+        // Reset our read pointer to the current write head
+        _core.getRecordPosition(recordDriverId, out _lastRecordPos);
+
+        // We’ll skip feeding for one frame to guarantee a clean start
+        _skipOneFeedFrame = true;
+
+        // Unpause loopback if we want it active during recording
+        if (playLoopback && _playChannel.hasHandle() && (!loopbackOnlyWhenActive || isRecordingActivated))
+            _playChannel.setPaused(loopbackOnlyWhenActive ? false : _playChannel.getPaused(out var paused) == FMOD.RESULT.OK && paused ? false : false);
 
         isRecordingActivated = true;
-        Debug.Log("[FMOD→Whisper] Stream activated (Whisper started; FMOD was already recording).");
+        Debug.Log("[FMOD→Whisper] Stream activated (buffer cleared; reading from current head).");
+
     }
 
     /// <summary>
@@ -269,36 +285,37 @@ public class FMODWhisperBridge : MonoBehaviour
 
         IntPtr p1, p2;
         uint len1, len2;
+
         var r = _recSound.@lock(startBytes, bytesToRead, out p1, out p2, out len1, out len2);
         if (r != RESULT.OK)
         {
-            // If lock fails, still advance last position to avoid spin
             _lastRecordPos = recPos;
             return;
         }
 
         try
         {
-            if (shouldFeed)
+            if (shouldFeed && !_skipOneFeedFrame)
             {
                 if (len1 > 0) CopyPcm16ToFloatAndFeed(p1, len1);
                 if (len2 > 0) CopyPcm16ToFloatAndFeed(p2, len2);
             }
-            // else: just discard; we’re only keeping the ring fresh.
+            // If skipping, we just discard this frame to ensure no stale data leaks.
         }
         finally
         {
             _recSound.unlock(p1, p2, len1, len2);
         }
 
+        if (_skipOneFeedFrame) _skipOneFeedFrame = false;
+
         _lastRecordPos = recPos;
+
     }
 
-    private bool IsSpeechMeaningful(string userText)
+    private string PostProcessInput(string input)
     {
-        return !string.IsNullOrEmpty(userText)
-               && !userText.Contains("BLANK_AUDIO")
-               && !userText.Trim().Equals("[ Silence ]");
+        return input.Replace("[silence]", "").Replace("[ Silence ]", "").Replace("BLANK_AUDIO", "").Replace("[", "").Replace("]", "").Trim();
     }
 
     private void CopyPcm16ToFloatAndFeed(IntPtr src, uint byteLen)
@@ -359,4 +376,55 @@ public class FMODWhisperBridge : MonoBehaviour
             _recSound.clearHandle();
         }
     }
+
+
+    private void ClearRecordRingBuffer()
+    {
+        if (!_recSound.hasHandle() || _soundPcmLength == 0) return;
+
+        uint totalBytes = _soundPcmLength * (uint)_nativeChannels * 2; // PCM16
+        IntPtr p1, p2;
+        uint len1, len2;
+
+        // Lock the whole buffer (start=0, length=totalBytes)
+        var r = _recSound.@lock(0, totalBytes, out p1, out p2, out len1, out len2);
+        if (r != FMOD.RESULT.OK)
+        {
+            Debug.LogWarning($"[FMOD→Whisper] Could not lock ring buffer to clear: {r}");
+            return;
+        }
+
+        try
+        {
+            if (len1 > 0)
+            {
+                // zero p1
+                // We’ll reuse a static zero array to avoid allocating huge buffers repeatedly
+                ZeroMem(p1, (int)len1);
+            }
+            if (len2 > 0)
+            {
+                ZeroMem(p2, (int)len2);
+            }
+        }
+        finally
+        {
+            _recSound.unlock(p1, p2, len1, len2);
+        }
+    }
+
+    // cheap zeroing helper (avoids allocating len-sized arrays each time)
+    private static readonly byte[] _zeroChunk = new byte[16 * 1024]; // 16 KB
+    private static void ZeroMem(IntPtr dst, int byteLen)
+    {
+        int offset = 0;
+        while (byteLen > 0)
+        {
+            int n = Math.Min(_zeroChunk.Length, byteLen);
+            Marshal.Copy(_zeroChunk, 0, dst + offset, n);
+            offset += n;
+            byteLen -= n;
+        }
+    }
+
 }