cafe waiter npc and mic detect speech more

3d-generation-pipeline/README.md

@@ -1,6 +1,7 @@
 ### TODO
 * klaviatuuriga võiks saada teleportida mängijat interactionite algustesse
 * quest marker peaks jätma nähtavad noolekesed enda liikumise teele
+* korralik teleportation minigame'idesse, kasutades TeleportProviderit, nagu button
 * igal pool, kus mängija peab rääkima, peaks olema:
 	* helilained räägitava heli amplituudi põhjal
 	* "loading" dünaamiline ikoon kui mängija ootab protsessimise järel

Assets/_PROJECT/Prefabs/XR/Bow/TargetUFO.prefab

@@ -8,212 +8,8 @@ GameObject:
   m_PrefabAsset: {fileID: 0}
   serializedVersion: 6
   m_Component:
-  - component: {fileID: 8461444523124913307}
-  - component: {fileID: 3572526038880022669}
-  - component: {fileID: 3766976745949113692}
-  - component: {fileID: 1047001759896168042}
-  - component: {fileID: 4197516043068701935}
-  - component: {fileID: 8992144926730888769}
-  - component: {fileID: 2113851326702507156}
-  m_Layer: 0
-  m_Name: TargetUFO
-  m_TagString: Untagged
-  m_Icon: {fileID: 0}
-  m_NavMeshLayer: 0
-  m_StaticEditorFlags: 0
-  m_IsActive: 1
---- !u!4 &8461444523124913307
-Transform:
-  m_ObjectHideFlags: 0
-  m_CorrespondingSourceObject: {fileID: 0}
-  m_PrefabInstance: {fileID: 0}
-  m_PrefabAsset: {fileID: 0}
-  m_GameObject: {fileID: 93219228833127587}
-  m_LocalRotation: {x: 0, y: 0, z: 0, w: 1}
-  m_LocalPosition: {x: 1.398, y: 1.1, z: 8.292}
-  m_LocalScale: {x: 1, y: 1, z: 1}
-  m_ConstrainProportionsScale: 0
-  m_Children:
-  - {fileID: 7933779445518253981}
-  m_Father: {fileID: 0}
-  m_RootOrder: -1
-  m_LocalEulerAnglesHint: {x: 0, y: 0, z: 0}
---- !u!114 &3572526038880022669
-MonoBehaviour:
-  m_ObjectHideFlags: 0
-  m_CorrespondingSourceObject: {fileID: 0}
-  m_PrefabInstance: {fileID: 0}
-  m_PrefabAsset: {fileID: 0}
-  m_GameObject: {fileID: 93219228833127587}
-  m_Enabled: 1
-  m_EditorHideFlags: 0
-  m_Script: {fileID: 11500000, guid: 3b311379c72a5ae4b8936e3b7283dd7a, type: 3}
-  m_Name: 
-  m_EditorClassIdentifier: 
-  _componentIndexCache: 0
-  _addedNetworkObject: {fileID: 1047001759896168042}
-  _networkObjectCache: {fileID: 1047001759896168042}
-  pointsText: {fileID: 1237163760934993282, guid: 105635d7165dacd47956f38546d4a2ea,
-    type: 3}
-  endPosition: {x: 0, y: 0, z: 0}
-  forwardSpeed: 2
---- !u!54 &3766976745949113692
-Rigidbody:
-  m_ObjectHideFlags: 0
-  m_CorrespondingSourceObject: {fileID: 0}
-  m_PrefabInstance: {fileID: 0}
-  m_PrefabAsset: {fileID: 0}
-  m_GameObject: {fileID: 93219228833127587}
-  serializedVersion: 4
-  m_Mass: 1
-  m_Drag: 0
-  m_AngularDrag: 0
-  m_CenterOfMass: {x: 0, y: 0, z: 0}
-  m_InertiaTensor: {x: 1, y: 1, z: 1}
-  m_InertiaRotation: {x: 0, y: 0, z: 0, w: 1}
-  m_IncludeLayers:
-    serializedVersion: 2
-    m_Bits: 0
-  m_ExcludeLayers:
-    serializedVersion: 2
-    m_Bits: 0
-  m_ImplicitCom: 1
-  m_ImplicitTensor: 1
-  m_UseGravity: 0
-  m_IsKinematic: 1
-  m_Interpolate: 0
-  m_Constraints: 126
-  m_CollisionDetection: 0
---- !u!114 &1047001759896168042
-MonoBehaviour:
-  m_ObjectHideFlags: 0
-  m_CorrespondingSourceObject: {fileID: 0}
-  m_PrefabInstance: {fileID: 0}
-  m_PrefabAsset: {fileID: 0}
-  m_GameObject: {fileID: 93219228833127587}
-  m_Enabled: 1
-  m_EditorHideFlags: 0
-  m_Script: {fileID: 11500000, guid: 26b716c41e9b56b4baafaf13a523ba2e, type: 3}
-  m_Name: 
-  m_EditorClassIdentifier: 
-  <IsNested>k__BackingField: 0
-  <ComponentIndex>k__BackingField: 0
-  <PredictedSpawn>k__BackingField: {fileID: 0}
-  _networkBehaviours:
-  - {fileID: 3572526038880022669}
-  - {fileID: 4197516043068701935}
-  - {fileID: 7294680161384397297}
-  - {fileID: 3316891016740450456}
-  - {fileID: 8062271296867124751}
-  - {fileID: 8618473685913268443}
-  <ParentNetworkObject>k__BackingField: {fileID: 0}
-  <ChildNetworkObjects>k__BackingField: []
-  _isNetworked: 1
-  _isGlobal: 0
-  _initializeOrder: 0
-  _defaultDespawnType: 1
-  NetworkObserver: {fileID: 0}
-  <PrefabId>k__BackingField: 9
-  <SpawnableCollectionId>k__BackingField: 0
-  _scenePathHash: 0
-  <SceneId>k__BackingField: 0
-  <AssetPathHash>k__BackingField: 12713957617246967544
-  _sceneNetworkObjects: []
---- !u!114 &4197516043068701935
-MonoBehaviour:
-  m_ObjectHideFlags: 0
-  m_CorrespondingSourceObject: {fileID: 0}
-  m_PrefabInstance: {fileID: 0}
-  m_PrefabAsset: {fileID: 0}
-  m_GameObject: {fileID: 93219228833127587}
-  m_Enabled: 1
-  m_EditorHideFlags: 0
-  m_Script: {fileID: 11500000, guid: a2836e36774ca1c4bbbee976e17b649c, type: 3}
-  m_Name: 
-  m_EditorClassIdentifier: 
-  _componentIndexCache: 1
-  _addedNetworkObject: {fileID: 1047001759896168042}
-  _networkObjectCache: {fileID: 1047001759896168042}
-  _synchronizeParent: 0
-  _packing:
-    Position: 1
-    Rotation: 1
-    Scale: 0
-  _interpolation: 2
-  _extrapolation: 2
-  _enableTeleport: 0
-  _teleportThreshold: 1
-  _clientAuthoritative: 1
-  _sendToOwner: 1
-  _synchronizePosition: 1
-  _positionSnapping:
-    X: 0
-    Y: 0
-    Z: 0
-  _synchronizeRotation: 1
-  _rotationSnapping:
-    X: 0
-    Y: 0
-    Z: 0
-  _synchronizeScale: 1
-  _scaleSnapping:
-    X: 0
-    Y: 0
-    Z: 0
---- !u!136 &8992144926730888769
-CapsuleCollider:
-  m_ObjectHideFlags: 0
-  m_CorrespondingSourceObject: {fileID: 0}
-  m_PrefabInstance: {fileID: 0}
-  m_PrefabAsset: {fileID: 0}
-  m_GameObject: {fileID: 93219228833127587}
-  m_Material: {fileID: 0}
-  m_IncludeLayers:
-    serializedVersion: 2
-    m_Bits: 0
-  m_ExcludeLayers:
-    serializedVersion: 2
-    m_Bits: 0
-  m_LayerOverridePriority: 0
-  m_IsTrigger: 0
-  m_ProvidesContacts: 0
-  m_Enabled: 1
-  serializedVersion: 2
-  m_Radius: 1.07
-  m_Height: 0.97
-  m_Direction: 1
-  m_Center: {x: 0.16, y: -0.51, z: 0}
---- !u!65 &2113851326702507156
-BoxCollider:
-  m_ObjectHideFlags: 0
-  m_CorrespondingSourceObject: {fileID: 0}
-  m_PrefabInstance: {fileID: 0}
-  m_PrefabAsset: {fileID: 0}
-  m_GameObject: {fileID: 93219228833127587}
-  m_Material: {fileID: 0}
-  m_IncludeLayers:
-    serializedVersion: 2
-    m_Bits: 0
-  m_ExcludeLayers:
-    serializedVersion: 2
-    m_Bits: 0
-  m_LayerOverridePriority: 0
-  m_IsTrigger: 0
-  m_ProvidesContacts: 0
-  m_Enabled: 1
-  serializedVersion: 3
-  m_Size: {x: 4.5, y: 0.5, z: 3.77}
-  m_Center: {x: 0.12, y: -0.24, z: 0.06}
---- !u!1 &1503281482029840645
-GameObject:
-  m_ObjectHideFlags: 0
-  m_CorrespondingSourceObject: {fileID: 0}
-  m_PrefabInstance: {fileID: 0}
-  m_PrefabAsset: {fileID: 0}
-  serializedVersion: 6
-  m_Component:
-  - component: {fileID: 7933779445518253981}
-  - component: {fileID: 7294680161384397297}
+  - component: {fileID: 3337587390139541606}
+  - component: {fileID: 627565954663756710}
   m_Layer: 0
   m_Name: SportModel
   m_TagString: Untagged

Assets/_PROJECT/Scripts/ModeGeneration/ArcheryRange/MicrophoneStand.cs

@@ -61,6 +61,7 @@ public class MicrophoneStand : MonoBehaviour
     private void OnPlayerSpeechUpdated(string text)
     {
         outputText.text = text;
+        OnPlayerFinishedSpeaking?.Invoke();
     }
 
     private void OnPlayerSpeechFinished(string playerText)

Assets/_PROJECT/Scripts/ModeGeneration/FMODWhisperBridge.cs

@@ -5,18 +5,19 @@ using UnityEngine;
 using FMOD;
 using FMODUnity;
 using Whisper;          // WhisperManager, WhisperStream, WhisperResult
-using Whisper.Utils;    // AudioChunk
+using Whisper.Utils;
+using Debug = UnityEngine.Debug;    // AudioChunk
 
 /// <summary>
-/// Capture microphone with FMOD and feed chunks to WhisperStream (no Unity Microphone).
-/// Also (optionally) plays the recorded sound back via FMOD loopback.
+/// FMOD mic is initialized once (Start) and runs continuously in a ring buffer.
+/// Whisper stream is created when ActivateRecording() is called and disposed on DeactivateRecording().
+/// Optional loopback can be paused/resumed instead of starting/stopping to avoid stalls.
 /// </summary>
-
 public class FMODWhisperBridge : MonoBehaviour
 {
     [Header("Whisper")]
     [SerializeField] private WhisperManager whisper;        // assign in Inspector
-    [SerializeField] private bool useVadInStream = false;   // let WhisperStream do VAD or not
+    [SerializeField] private bool useVadInStream = true;   // let WhisperStream do VAD or not
 
     [Header("FMOD capture")]
     [Tooltip("Recording device index (0 = default)")]
@@ -29,6 +30,8 @@ public class FMODWhisperBridge : MonoBehaviour
 
     [Header("Loopback (monitor your voice)")]
     public bool playLoopback = true;
+    [Tooltip("If true, loopback plays only while active; otherwise it’s always on.")]
+    public bool loopbackOnlyWhenActive = true;
     [Range(0f, 2f)] public float loopbackVolume = 1.0f;
 
     public delegate void OnWhisperSegmentUpdatedDelegate(string result);
@@ -55,8 +58,11 @@ public class FMODWhisperBridge : MonoBehaviour
 
     // temp conversion buffer
     private float[] _floatTmp = new float[0];
+    private short[] _shortOverlay;
 
+    // activation flag
     private bool isRecordingActivated = false;
+    private bool _skipOneFeedFrame = false;
 
     private void Awake()
     {
@@ -66,85 +72,195 @@ public class FMODWhisperBridge : MonoBehaviour
 
     private async void Start()
     {
-        // Query device info to get native rate/channels.
-        // (FMOD: getRecordDriverInfo gives you system rate & speaker mode)
+        // -------------- FMOD initialize ONCE --------------
+        // Query device info
         string name;
         Guid guid;
         SPEAKERMODE sm;
         int smChannels;
         DRIVER_STATE driverState;
-        // signature: getRecordDriverInfo(id, out name, nameLen, out guid, out systemrate, out speakermode, out speakermodechannels, out driverState)
-        _core.getRecordDriverInfo(recordDriverId, out name, 256, out guid, out _nativeRate, out sm, out smChannels, out driverState);
-        _nativeChannels = channels > 0 ? channels : smChannels;
-        UnityEngine.Debug.Log($"[FMOD→Whisper] Using input device #{recordDriverId}: \"{name}\"  rate={_nativeRate} ch={_nativeChannels}");
 
-        // Build a user sound buffer that FMOD will fill (OPENUSER | LOOP_NORMAL).
+        var res = _core.getRecordDriverInfo(
+            recordDriverId,
+            out name, 256,
+            out guid,
+            out _nativeRate,
+            out sm,
+            out smChannels,
+            out driverState
+        );
+
+        if (res != RESULT.OK)
+        {
+            Debug.LogError($"[FMOD→Whisper] getRecordDriverInfo failed: {res}");
+            return;
+        }
+
+        _nativeChannels = channels > 0 ? channels : smChannels;
+        int rate = (_nativeRate > 0) ? _nativeRate : desiredSampleRate;
+        Debug.Log($"[FMOD→Whisper] Using input device #{recordDriverId}: \"{name}\"  rate={rate} ch={_nativeChannels}");
+
+        // Build user sound (ring buffer) — multiple seconds
         CREATESOUNDEXINFO ex = new CREATESOUNDEXINFO
         {
             cbsize = Marshal.SizeOf(typeof(CREATESOUNDEXINFO)),
             numchannels = _nativeChannels,
-            defaultfrequency = (_nativeRate > 0) ? _nativeRate : desiredSampleRate,
+            defaultfrequency = rate,
             format = SOUND_FORMAT.PCM16,
-            length = (uint)(((_nativeRate > 0 ? _nativeRate : desiredSampleRate) * _nativeChannels) * sizeof(short)) // seconds=1 (we loop)
+            length = (uint)(rate * _nativeChannels * sizeof(short) * bufferLengthSec)
         };
 
-        _core.createSound("", MODE.OPENUSER | MODE.LOOP_NORMAL | MODE.CREATESAMPLE, ref ex, out _recSound);
+        res = _core.createSound("", MODE.OPENUSER | MODE.LOOP_NORMAL | MODE.CREATESAMPLE, ref ex, out _recSound);
+        if (res != RESULT.OK)
+        {
+            Debug.LogError($"[FMOD→Whisper] createSound failed: {res}");
+            return;
+        }
+
         _recSound.getLength(out _soundPcmLength, TIMEUNIT.PCM);
 
-        // Start FMOD recording into that sound (looping ring buffer).
-        _core.recordStart(recordDriverId, _recSound, true);
-        UnityEngine.Debug.Log("[FMOD→Whisper] Recording started.");
+        // Start recording (looping)
+        res = _core.recordStart(recordDriverId, _recSound, true);
+        if (res != RESULT.OK)
+        {
+            Debug.LogError($"[FMOD→Whisper] recordStart failed: {res}");
+            _recSound.release();
+            _recSound.clearHandle();
+            return;
+        }
 
-        // Optional loopback playback using FMOD (plays same sound ring buffer).
+        // Initialize record position to avoid a huge first delta
+        _core.getRecordPosition(recordDriverId, out _lastRecordPos);
+        Debug.Log("[FMOD→Whisper] Recording started.");
+
+        // Loopback channel (optional). Start once; pause when inactive if desired.
         _core.getMasterChannelGroup(out _masterGroup);
         if (playLoopback)
         {
-            _core.playSound(_recSound, _masterGroup, false, out _playChannel);
-            _playChannel.setMode(MODE._2D);
-            _playChannel.setVolume(loopbackVolume);
-            UnityEngine.Debug.Log("[FMOD→Whisper] Loopback playback started.");
+            res = _core.playSound(_recSound, _masterGroup, false, out _playChannel);
+            if (res == RESULT.OK && _playChannel.hasHandle())
+            {
+                _playChannel.setMode(MODE._2D);
+                _playChannel.setVolume(loopbackVolume);
+                if (loopbackOnlyWhenActive) _playChannel.setPaused(true); // keep muted until Activate
+                Debug.Log("[FMOD→Whisper] Loopback playback ready.");
+            }
+            else
+            {
+                Debug.LogWarning($"[FMOD→Whisper] playSound failed or channel invalid: {res}");
+            }
         }
 
-        // Create Whisper stream WITHOUT MicrophoneRecord, just from (freq, channels).
-        // We'll push AudioChunk manually.
-        // NOTE: WhisperStream’s sliding window is governed by manager’s stepSec/keepSec/lengthSec.
-        _stream = await whisper.CreateStream(ex.defaultfrequency, _nativeChannels);
-        _stream.OnResultUpdated += (txt) =>
+        // No Whisper stream here. It will be created on ActivateRecording().
+        await System.Threading.Tasks.Task.Yield();
+    }
+
+    /// <summary>
+    /// Creates a fresh Whisper stream and starts feeding audio (FMOD already recording).
+    /// </summary>
+    public async void ActivateRecording()
+    {
+        if (isRecordingActivated)
         {
-            //OnWhisperResultProcessed?.Invoke(txt);
-        };
+            Debug.Log("[FMOD→Whisper] ActivateRecording ignored (already active).");
+            return;
+        }
+
+        if (!_recSound.hasHandle())
+        {
+            Debug.LogError("[FMOD→Whisper] FMOD not initialized or recording not running.");
+            return;
+        }
+
+        int rate = (_nativeRate > 0) ? _nativeRate : desiredSampleRate;
+
+        try
+        {
+            _stream = await whisper.CreateStream(rate, _nativeChannels);
+        }
+        catch (Exception e)
+        {
+            Debug.LogError($"[FMOD→Whisper] CreateStream exception: {e}");
+            _stream = null;
+            _streamStarted = false;
+            return;
+        }
+
+        // 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);
         };
 
-        // If you want Whisper to respect VAD, enable in manager or set useVad (manager controls stream params).
         whisper.useVad = useVadInStream;
 
         _stream.StartStream();
         _streamStarted = true;
 
-        // prepare temp arrays roughly 100ms of audio
-        EnsureTmpCapacity((ex.defaultfrequency / 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 (buffer cleared; reading from current head).");
+
+    }
+
+    /// <summary>
+    /// Stops and disposes the Whisper stream only. FMOD keeps recording.
+    /// </summary>
+    public void DeactivateRecording()
+    {
+        if (!isRecordingActivated && !_streamStarted)
+            return;
+
+        isRecordingActivated = false;
+
+        // Pause loopback if it should only be active during recording
+        if (playLoopback && loopbackOnlyWhenActive && _playChannel.hasHandle())
+            _playChannel.setPaused(true);
+
+        // Tear down Whisper stream
+        if (_streamStarted)
+        {
+            try { _stream.StopStream(); } catch { /* ignore */ }
+            _streamStarted = false;
+        }
+        _stream = null;
+
+        Debug.Log("[FMOD→Whisper] Stream deactivated (Whisper stopped; FMOD still recording).");
     }
 
     private void Update()
     {
-        if (!isRecordingActivated) return;
+        // Always tick FMOD
         if (_core.handle != IntPtr.Zero) _core.update();
-        if (!_streamStarted || !_recSound.hasHandle()) return;
+        if (!_recSound.hasHandle()) return;
 
-        // How many samples recorded since last frame?
+        // Compute how many samples recorded since last frame.
         uint recPos;
         _core.getRecordPosition(recordDriverId, out recPos);
 
@@ -152,45 +268,54 @@ public class FMODWhisperBridge : MonoBehaviour
             ? (recPos - _lastRecordPos)
             : (recPos + _soundPcmLength - _lastRecordPos);
 
-        if (deltaSamples == 0) return;
+        if (deltaSamples == 0)
+        {
+            // Even if 0, keep last pos
+            _lastRecordPos = recPos;
+            return;
+        }
 
-        // We’ll read that region (16-bit) and convert to float[] [-1..1].
-        // Calculate byte range to lock in sound buffer
-        uint bytesToRead = deltaSamples * (uint)_nativeChannels * 2; // 16-bit = 2 bytes
+        // If not active, we *still* advance the ring (so we don't backlog data),
+        // but we *don't* push chunks to Whisper.
+        bool shouldFeed = isRecordingActivated && _streamStarted && _stream != null;
+
+        // Calculate byte range to lock (16-bit)
+        uint bytesToRead = deltaSamples * (uint)_nativeChannels * 2; // 2 bytes per sample
         uint startBytes = _lastRecordPos * (uint)_nativeChannels * 2;
 
         IntPtr p1, p2;
         uint len1, len2;
-        // Lock can wrap — FMOD splits into p1/p2.
-        _recSound.@lock(startBytes, bytesToRead, out p1, out p2, out len1, out len2);
+
+        var r = _recSound.@lock(startBytes, bytesToRead, out p1, out p2, out len1, out len2);
+        if (r != RESULT.OK)
+        {
+            _lastRecordPos = recPos;
+            return;
+        }
 
         try
         {
-            // Convert both parts to float and push to Whisper
-            if (len1 > 0) CopyPcm16ToFloatAndFeed(p1, len1);
-            if (len2 > 0) CopyPcm16ToFloatAndFeed(p2, len2);
+            if (shouldFeed && !_skipOneFeedFrame)
+            {
+                if (len1 > 0) CopyPcm16ToFloatAndFeed(p1, len1);
+                if (len2 > 0) CopyPcm16ToFloatAndFeed(p2, len2);
+            }
+            // 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;
+
     }
 
-    public void ActivateRecording()
+    private string PostProcessInput(string input)
     {
-        isRecordingActivated = true;
-    }
-
-    public void DeactivateRecording()
-    {
-        isRecordingActivated = false;
-    }
-
-    private bool IsSpeechMeaningful(string userText)
-    {
-        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)
@@ -198,26 +323,20 @@ public class FMODWhisperBridge : MonoBehaviour
         int samples = (int)(byteLen / 2); // 2 bytes per sample
         EnsureTmpCapacity(samples);
 
-        // Marshal the 16-bit PCM into managed space
-        // We pin a short[] overlay to avoid copying twice
-        int shorts = samples;
-        int byteCount = (int)byteLen;
+        EnsureShortOverlay(samples, out short[] sBuf);
+        Marshal.Copy(src, sBuf, 0, samples);
 
-        // Use Marshal.Copy into a short[] then convert to float[-1..1]
-        // (You can also unsafe copy for speed if needed.)
-        EnsureShortOverlay(shorts, out short[] sBuf);
-        Marshal.Copy(src, sBuf, 0, shorts);
-
-        for (int i = 0; i < shorts; i++)
+        // Convert to float [-1..1] (no downmix change from your original)
+        for (int i = 0; i < samples; i++)
         {
-            // 32768f avoids clipping at -32768
             _floatTmp[i] = Mathf.Clamp(sBuf[i] / 32768f, -1f, 1f);
         }
 
-        // Build a chunk for WhisperStream; with VAD off, IsVoiceDetected=true is fine.
+        // TODO (optional): downmix to mono and/or run a light gate before feeding.
+        // For now we keep your original behavior:
         var chunk = new AudioChunk
         {
-            Data = _floatTmp.AsSpan(0, shorts).ToArray(),
+            Data = _floatTmp.AsSpan(0, samples).ToArray(),
             Frequency = (_nativeRate > 0) ? _nativeRate : desiredSampleRate,
             Channels = _nativeChannels,
             IsVoiceDetected = true
@@ -226,7 +345,6 @@ public class FMODWhisperBridge : MonoBehaviour
         _stream.AddToStream(chunk);
     }
 
-    private short[] _shortOverlay;
     private void EnsureShortOverlay(int samples, out short[] buf)
     {
         if (_shortOverlay == null || _shortOverlay.Length < samples)
@@ -242,13 +360,71 @@ public class FMODWhisperBridge : MonoBehaviour
 
     private void OnDisable()
     {
-        if (_streamStarted)
+        // Stop Whisper (if active)
+        DeactivateRecording();
+
+        // Stop/purge FMOD resources (since object is going away)
+        if (_playChannel.hasHandle())
         {
-            _stream.StopStream();
-            _streamStarted = false;
+            try { _playChannel.stop(); } catch { /* ignore */ }
+            _playChannel.clearHandle();
+        }
+        if (_recSound.hasHandle())
+        {
+            try { _core.recordStop(recordDriverId); } catch { /* ignore */ }
+            try { _recSound.release(); } catch { /* ignore */ }
+            _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;
         }
 
-        if (_playChannel.hasHandle()) { _playChannel.stop(); _playChannel.clearHandle(); }
-        if (_recSound.hasHandle()) { _core.recordStop(recordDriverId); _recSound.release(); _recSound.clearHandle(); }
+        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;
+        }
+    }
+
 }

Assets/_PROJECT/Scripts/ModeGeneration/ModelGenerationUtils.cs

@@ -3,11 +3,15 @@ using System;
 using System.Diagnostics;
 using System.IO;
 using System.Threading.Tasks;
+using Unity.XR.CoreUtils;
 using UnityEngine;
+using UnityEngine.InputSystem;
+using UnityEngine.SceneManagement;
 
 public class ModelGenerationUtils : MonoBehaviour
 {
     public static ModelGenerationUtils Instance { get; private set; }
+    public Transform[] minigamePoints;
 
     private void Awake()
     {
@@ -23,9 +27,38 @@ public class ModelGenerationUtils : MonoBehaviour
     // Update is called once per frame
     void Update()
     {
-        
+        // For teleporting player to different minigames, used during expos
+
+        // No keyboard present (e.g., Quest standalone without a paired keyboard)
+        if (Keyboard.current == null) return;
+
+        if (Keyboard.current.digit1Key.wasPressedThisFrame)
+        {
+            XROrigin xrOrigin = FindObjectOfType<XROrigin>(includeInactive: true);
+            xrOrigin.transform.position = minigamePoints[0].position;
+        }
+            
+
+        if (Keyboard.current.digit2Key.wasPressedThisFrame)
+        {
+            XROrigin xrOrigin = FindObjectOfType<XROrigin>(includeInactive: true);
+            xrOrigin.transform.position = minigamePoints[1].position;
+        }
+
+        if (Keyboard.current.digit3Key.wasPressedThisFrame)
+        {
+            XROrigin xrOrigin = FindObjectOfType<XROrigin>(includeInactive: true);
+            xrOrigin.transform.position = minigamePoints[2].position;
+        }
+
+        if (Keyboard.current.digit0Key.wasPressedThisFrame)
+        {
+            string currentSceneName = SceneManager.GetActiveScene().name;
+            SceneManager.LoadScene(currentSceneName);
+        }
     }
 
+
     /**
      * Generate model by starting a new Python subprocess
      * NOT USED IN LATEST VERSION

Assets/_PROJECT/Scripts/ModeGeneration/NPCs/CafeWaiterNPC.cs

@@ -50,9 +50,9 @@ public class CafeWaiterNPC : NPCController
         {
             SpeakVoiceLine(0);
 
+            fmodWhisperBridge.ActivateRecording();
             fmodWhisperBridge.OnWhisperSegmentUpdated += OnPlayerSpeechUpdated;
             fmodWhisperBridge.OnWhisperSegmentFinished += OnPlayerSpeechFinished;
-            fmodWhisperBridge.ActivateRecording();
 
             notepad.SetActive(true);
             state = 1;
@@ -76,7 +76,7 @@ public class CafeWaiterNPC : NPCController
 
     private void OnPlayerSpeechFinished(string playerText)
     {
-        if (Time.time < lastPlayerVoiceUpdateTime + 0.5f)
+        if (Time.time < lastPlayerVoiceUpdateTime + 1.0f)
         {
             return;
         }
@@ -85,23 +85,34 @@ public class CafeWaiterNPC : NPCController
         if (state == 1)
         {
             // Show transcription and ask whether it is correct
+            fmodWhisperBridge.DeactivateRecording();
             notepadText.text = playerText;
-            notepad.transform.DOLocalRotate(notepadFlippedRotation, 0.5f);
+            notepad.transform.DOLocalRotate(notepadFlippedRotation, 0.5f).OnComplete(() =>
+            {
+                fmodWhisperBridge.ActivateRecording();
+                fmodWhisperBridge.OnWhisperSegmentUpdated += OnPlayerSpeechUpdated;
+                fmodWhisperBridge.OnWhisperSegmentFinished += OnPlayerSpeechFinished;
+            });
             SpeakVoiceLine(1);
             state = 2;
         } else if (state == 2)
         {
+            fmodWhisperBridge.DeactivateRecording();
             // Flip notepad back
             notepad.transform.DOLocalRotate(notepadOriginalRotation, 0.5f);
             // if player answered positively, bring food, otherwise ask again
-            if (playerText.ToLower().Contains("yes"))
+            if (playerText.ToLower().Contains("ye"))
             {
+                
                 SpeakVoiceLine(2);
                 Invoke("BringFood", 1f);
                 state = 3;
             } else
             {
                 SpeakVoiceLine(3);
+                fmodWhisperBridge.ActivateRecording();
+                fmodWhisperBridge.OnWhisperSegmentUpdated += OnPlayerSpeechUpdated;
+                fmodWhisperBridge.OnWhisperSegmentFinished += OnPlayerSpeechFinished;
                 state = 1;
             }
         }
@@ -113,6 +124,14 @@ public class CafeWaiterNPC : NPCController
         if (state == 1)
         {
             notepadText.text = playerText;
+
+            // For now, when something is transcribed, treat it as player finished speaking
+            OnPlayerSpeechFinished(playerText);
+        }
+        // faster reaction to player answering yes/no
+        else if (state == 2)
+        {
+            OnPlayerSpeechFinished(playerText);
         }
     }
 

Assets/_PROJECT/Scripts/ModeGeneration/ReleasableButton.cs

@@ -5,6 +5,8 @@ public class ReleasableButton : MonoBehaviour
 {
     public delegate void OnButtonPressedDelegate();
     public event OnButtonPressedDelegate OnButtonPressed;
+    public delegate void OnButtonReleasedDelegate();
+    public event OnButtonReleasedDelegate OnButtonReleased;
 
     public Transform movableParts;
     public float moveDuration = 0.25f;
@@ -37,7 +39,6 @@ public class ReleasableButton : MonoBehaviour
     {
         if (!isButtonDown && collider.gameObject.tag.EndsWith("Hand"))
         {
-            Debug.Log("collided with: " + collider.gameObject.name);
             Activate();
             OnButtonPressed?.Invoke();
         }
@@ -47,8 +48,8 @@ public class ReleasableButton : MonoBehaviour
     {
         if (isButtonDown && collider.gameObject.tag.EndsWith("Hand"))
         {
-            Debug.Log("collider exited: " + collider.gameObject.name);
             Deactivate();
+            OnButtonReleased?.Invoke();
         }
     }
 

Assets/_PROJECT/Scripts/ModeGeneration/ShapeDetection/RadioTransmitter.cs

@@ -18,6 +18,7 @@ public class RadioTransmitter : XRGrabInteractable
     void Start()
     {
         radioButton.OnButtonPressed += OnRadioButtonPressed;
+        radioButton.OnButtonReleased += OnRadioButtonReleased;
     }
 
     // Update is called once per frame
@@ -38,12 +39,20 @@ public class RadioTransmitter : XRGrabInteractable
         fmodWhisperBridge.OnWhisperSegmentFinished += OnPlayerSpeechFinished;
         AudioManager.Instance.PlayAttachedInstance(FMODEvents.Instance.RadioButton, gameObject);
         fmodWhisperBridge.ActivateRecording();
-        // TODO: deactivate when button is released
+    }
+
+    private void OnRadioButtonReleased()
+    {
+        fmodWhisperBridge.OnWhisperSegmentUpdated -= OnPlayerSpeechUpdated;
+        fmodWhisperBridge.OnWhisperSegmentFinished -= OnPlayerSpeechFinished;
+        AudioManager.Instance.PlayAttachedInstance(FMODEvents.Instance.RadioButton, gameObject);
+        fmodWhisperBridge.DeactivateRecording();
     }
 
     private void OnPlayerSpeechUpdated(string text)
     {
         computerScreen.text = text;
+        OnPlayerFinishedSpeaking?.Invoke();
     }
 
     private void OnPlayerSpeechFinished(string playerText)