Change routine to deliver back PCM rather than creating a wav and directly playing

autious · autious · commit c297e15df945 · 2020-01-31T23:45:35.000+01:00
diff --git a/Assets/Libraries/eSpeakWrapper/Client.cs b/Assets/Libraries/eSpeakWrapper/Client.cs
@@ -59,20 +59,46 @@ enum SpeechFlags
 
         static bool Initialized = false;
 
+        public static int sampleRate;
+
         public static void Initialize(string path)
         {
-            var result = espeak_Initialize(AudioOutput.SynchronousPlayback, 0, path, 0);
-            //var result = espeak_Initialize(AudioOutput.Retrieval, 0, path, 0); // TODO allow receiving audio data..
+            //var result = espeak_Initialize(AudioOutput.SynchronousPlayback, 0, path, 0);
+            var result = espeak_Initialize(AudioOutput.Retrieval, 0, path, 0); // TODO allow receiving audio data..
             if (result == (int)Error.EE_INTERNAL_ERROR)
             {
                 throw new Exception(string.Format("Could not initialize ESpeak. Maybe there is no espeak data at {0}?", path));
+            } else {
+                sampleRate = result;
             }
 
             espeak_SetSynthCallback(EventHandler.Handle);
 
             Initialized = true;
         }
 
+        public static bool VoiceFinished() {
+            bool ret = false;
+            EventHandler.audio_files_mutex.WaitOne();
+            if(EventHandler.audio_files.Count > 0) {
+                ret = true; 
+            }
+            EventHandler.audio_files_mutex.ReleaseMutex();
+            return ret;
+        }
+
+        public static byte[] PopVoice() {
+            byte[] ret = null;
+            EventHandler.audio_files_mutex.WaitOne();
+            if(EventHandler.audio_files.Count > 0) {
+                ret = EventHandler.audio_files[0];
+                EventHandler.audio_files.RemoveAt(0);
+            }
+            EventHandler.audio_files_mutex.ReleaseMutex();
+            return ret;
+            
+        }
+
         public static bool SetRate(int rate)
         {
             if (rate < 80 && rate > 450)
diff --git a/Assets/Libraries/eSpeakWrapper/EventHandler.cs b/Assets/Libraries/eSpeakWrapper/EventHandler.cs
@@ -4,15 +4,20 @@
 using System.Runtime.InteropServices;
 using System.Collections.Generic;
 using System.Text;
+using System.Threading;
 
 namespace ESpeakWrapper
 {
     class EventHandler
     {
-
         public delegate int SynthCallback(IntPtr wavePtr, int bufferLength, IntPtr eventsPtr);
+        public delegate void OnVoiceFinished();
+
+        public static OnVoiceFinished ovf = null;
 
         static MemoryStream Stream;
+        public static Mutex audio_files_mutex = new Mutex();
+        public static List<byte[]> audio_files = new List<byte[]>();
 
         public static int Handle(IntPtr wavePtr, int bufferLength, IntPtr eventsPtr)
         {
@@ -22,14 +27,20 @@ public static int Handle(IntPtr wavePtr, int bufferLength, IntPtr eventsPtr)
             // crash, but console log gets printed to output.log
             //Debug.Log("Buffer length is " + bufferLength);
 
+            //Assume that synthesiz is final if buffer length is zero?
             if (bufferLength == 0)
             {
                 //var file = new FileStream("alarm01.wav", FileMode.Open);
                 //Stream.Seek(0, SeekOrigin.Begin);
                 //file.CopyTo(Stream);
 
-                PlayAudio();
-                Console.Write(ConvertHeadersToString(Stream.GetBuffer()));
+                //PlayAudio();
+                //Console.Write(ConvertHeadersToString(Stream.GetBuffer()));
+
+                Stream.Flush();
+                audio_files_mutex.WaitOne();
+                audio_files.Add(Stream.ToArray());
+                audio_files_mutex.ReleaseMutex();
                 Stream.Dispose();
                 return 0;
             }
@@ -76,7 +87,7 @@ static int WriteAudioToStream(IntPtr wavePtr, int bufferLength)
             if (Stream == null)
             {
                 Stream = new MemoryStream();
-                InitializeStream();
+                //InitializeStream();
             }
 
             byte[] audio = new byte[bufferLength * 2];
@@ -86,6 +97,7 @@ static int WriteAudioToStream(IntPtr wavePtr, int bufferLength)
             return 0;
         }
 
+        /*
         static void InitializeStream()
         {
             var ascii = Encoding.ASCII;
@@ -115,7 +127,9 @@ static void InitializeStream()
             // audio size: will fill this in the PlayAudio() method
             Stream.Write(BitConverter.GetBytes(0), 0, 4);
         }
+        */
 
+        /*
         static void PlayAudio()
         {
             Stream.Seek(4, SeekOrigin.Begin);
@@ -125,23 +139,28 @@ static void PlayAudio()
             Stream.Write(BitConverter.GetBytes(Stream.Length - 44), 0, 4);
 
             Stream.Seek(0, SeekOrigin.Begin); // have to do this, otherwise the player will give a bogus error
-            var player = new SoundPlayer(Stream);
+            //var player = new SoundPlayer(Stream);
 
-            try
-            {
-                player.Play();
-            }
-            catch (Exception e)
-            {
-                Console.WriteLine(e);
-            }
+            //try
+            //{
+            //    player.Play();
+            //}
+            //catch (Exception e)
+            //{
+            //    Console.WriteLine(e);
+            //}
 
             // NOTE this gets always generated.. we dont need?
-            using (var file = new FileStream("test.wav", FileMode.Create))
-            {
-                Stream.WriteTo(file);
+            //using (var file = new FileStream("test.wav", FileMode.Create))
+            //{
+            //    Stream.WriteTo(file);
+            //}
+
+            if(ovf != null) {
+                ovf();
             }
         }
+        */
 
         static string PrintBytes(byte[] byteArray)
         {
diff --git a/Assets/Scenes/espeak.unity b/Assets/Scenes/espeak.unity
@@ -54,14 +54,15 @@ LightmapSettings:
     m_EnableBakedLightmaps: 0
     m_EnableRealtimeLightmaps: 0
   m_LightmapEditorSettings:
-    serializedVersion: 10
+    serializedVersion: 12
     m_Resolution: 2
     m_BakeResolution: 10
     m_AtlasSize: 512
     m_AO: 0
     m_AOMaxDistance: 1
     m_CompAOExponent: 1
     m_CompAOExponentDirect: 0
+    m_ExtractAmbientOcclusion: 0
     m_Padding: 2
     m_LightmapParameters: {fileID: 0}
     m_LightmapsBakeMode: 1
@@ -76,10 +77,16 @@ LightmapSettings:
     m_PVRDirectSampleCount: 32
     m_PVRSampleCount: 256
     m_PVRBounces: 2
+    m_PVREnvironmentSampleCount: 256
+    m_PVREnvironmentReferencePointCount: 2048
+    m_PVRFilteringMode: 2
+    m_PVRDenoiserTypeDirect: 0
+    m_PVRDenoiserTypeIndirect: 0
+    m_PVRDenoiserTypeAO: 0
     m_PVRFilterTypeDirect: 0
     m_PVRFilterTypeIndirect: 0
     m_PVRFilterTypeAO: 0
-    m_PVRFilteringMode: 1
+    m_PVREnvironmentMIS: 0
     m_PVRCulling: 0
     m_PVRFilteringGaussRadiusDirect: 1
     m_PVRFilteringGaussRadiusIndirect: 5
@@ -88,6 +95,7 @@ LightmapSettings:
     m_PVRFilteringAtrousPositionSigmaIndirect: 2
     m_PVRFilteringAtrousPositionSigmaAO: 1
     m_ShowResolutionOverlay: 1
+    m_ExportTrainingData: 0
   m_LightingDataAsset: {fileID: 0}
   m_UseShadowmask: 1
 --- !u!196 &4
@@ -150,9 +158,10 @@ Camera:
   m_ClearFlags: 2
   m_BackGroundColor: {r: 0.19215687, g: 0.3019608, b: 0.4745098, a: 0}
   m_projectionMatrixMode: 1
+  m_GateFitMode: 2
+  m_FOVAxisMode: 0
   m_SensorSize: {x: 36, y: 24}
   m_LensShift: {x: 0, y: 0}
-  m_GateFitMode: 2
   m_FocalLength: 50
   m_NormalizedViewPortRect:
     serializedVersion: 2
@@ -223,7 +232,7 @@ MonoBehaviour:
   m_Script: {fileID: 11500000, guid: 0896d1324347bd349886948e2acc0cbe, type: 3}
   m_Name: 
   m_EditorClassIdentifier: 
-  sayAtStart: Welcome!
+  sayAtStart: Testing
 --- !u!4 &548749181
 Transform:
   m_ObjectHideFlags: 0
@@ -238,6 +247,133 @@ Transform:
   m_Father: {fileID: 0}
   m_RootOrder: 2
   m_LocalEulerAnglesHint: {x: 0, y: 0, z: 0}
+--- !u!1 &1440771956
+GameObject:
+  m_ObjectHideFlags: 0
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
+  serializedVersion: 6
+  m_Component:
+  - component: {fileID: 1440771958}
+  - component: {fileID: 1440771957}
+  m_Layer: 0
+  m_Name: Audio Source
+  m_TagString: Untagged
+  m_Icon: {fileID: 0}
+  m_NavMeshLayer: 0
+  m_StaticEditorFlags: 0
+  m_IsActive: 1
+--- !u!82 &1440771957
+AudioSource:
+  m_ObjectHideFlags: 0
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
+  m_GameObject: {fileID: 1440771956}
+  m_Enabled: 1
+  serializedVersion: 4
+  OutputAudioMixerGroup: {fileID: 0}
+  m_audioClip: {fileID: 0}
+  m_PlayOnAwake: 1
+  m_Volume: 1
+  m_Pitch: 1
+  Loop: 0
+  Mute: 0
+  Spatialize: 0
+  SpatializePostEffects: 0
+  Priority: 128
+  DopplerLevel: 1
+  MinDistance: 1
+  MaxDistance: 500
+  Pan2D: 0
+  rolloffMode: 0
+  BypassEffects: 0
+  BypassListenerEffects: 0
+  BypassReverbZones: 0
+  rolloffCustomCurve:
+    serializedVersion: 2
+    m_Curve:
+    - serializedVersion: 3
+      time: 0
+      value: 1
+      inSlope: 0
+      outSlope: 0
+      tangentMode: 0
+      weightedMode: 0
+      inWeight: 0.33333334
+      outWeight: 0.33333334
+    - serializedVersion: 3
+      time: 1
+      value: 0
+      inSlope: 0
+      outSlope: 0
+      tangentMode: 0
+      weightedMode: 0
+      inWeight: 0.33333334
+      outWeight: 0.33333334
+    m_PreInfinity: 2
+    m_PostInfinity: 2
+    m_RotationOrder: 4
+  panLevelCustomCurve:
+    serializedVersion: 2
+    m_Curve:
+    - serializedVersion: 3
+      time: 0
+      value: 0
+      inSlope: 0
+      outSlope: 0
+      tangentMode: 0
+      weightedMode: 0
+      inWeight: 0.33333334
+      outWeight: 0.33333334
+    m_PreInfinity: 2
+    m_PostInfinity: 2
+    m_RotationOrder: 4
+  spreadCustomCurve:
+    serializedVersion: 2
+    m_Curve:
+    - serializedVersion: 3
+      time: 0
+      value: 0
+      inSlope: 0
+      outSlope: 0
+      tangentMode: 0
+      weightedMode: 0
+      inWeight: 0.33333334
+      outWeight: 0.33333334
+    m_PreInfinity: 2
+    m_PostInfinity: 2
+    m_RotationOrder: 4
+  reverbZoneMixCustomCurve:
+    serializedVersion: 2
+    m_Curve:
+    - serializedVersion: 3
+      time: 0
+      value: 1
+      inSlope: 0
+      outSlope: 0
+      tangentMode: 0
+      weightedMode: 0
+      inWeight: 0.33333334
+      outWeight: 0.33333334
+    m_PreInfinity: 2
+    m_PostInfinity: 2
+    m_RotationOrder: 4
+--- !u!4 &1440771958
+Transform:
+  m_ObjectHideFlags: 0
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
+  m_GameObject: {fileID: 1440771956}
+  m_LocalRotation: {x: 0, y: 0, z: 0, w: 1}
+  m_LocalPosition: {x: -2.093978, y: 2.580687, z: 1.2901945}
+  m_LocalScale: {x: 1, y: 1, z: 1}
+  m_Children: []
+  m_Father: {fileID: 0}
+  m_RootOrder: 3
+  m_LocalEulerAnglesHint: {x: 0, y: 0, z: 0}
 --- !u!1 &1665498520
 GameObject:
   m_ObjectHideFlags: 0
@@ -292,6 +428,7 @@ MonoBehaviour:
   m_Name: 
   m_EditorClassIdentifier: 
   voiceID: Tweaky
+  audio_source: {fileID: 1440771957}
 --- !u!4 &1665498523
 Transform:
   m_ObjectHideFlags: 0
diff --git a/Assets/Scripts/Speech.cs b/Assets/Scripts/Speech.cs
