Updated Sentis Version from 2.1.2 and added DLL for MacOS (Apple Chips)

Assets/Piper/Samples/PiperSample.cs

@@ -14,7 +14,7 @@ public class PiperSample : MonoBehaviour
 
         private AudioSource _source;
 
-        private void Awake()
+        private void Awake() 
         {
             _source = GetComponent<AudioSource>();
             input.onSubmit.AddListener(OnInputSubmit);

Assets/Piper/Samples/PiperSample.unity

@@ -13,7 +13,7 @@ OcclusionCullingSettings:
 --- !u!104 &2
 RenderSettings:
   m_ObjectHideFlags: 0
-  serializedVersion: 9
+  serializedVersion: 10
   m_Fog: 0
   m_FogColor: {r: 0.5, g: 0.5, b: 0.5, a: 1}
   m_FogMode: 3
@@ -38,13 +38,12 @@ RenderSettings:
   m_ReflectionIntensity: 1
   m_CustomReflection: {fileID: 0}
   m_Sun: {fileID: 705507994}
-  m_IndirectSpecularColor: {r: 0.44657898, g: 0.4964133, b: 0.5748178, a: 1}
+  m_IndirectSpecularColor: {r: 0, g: 0, b: 0, a: 1}
   m_UseRadianceAmbientProbe: 0
 --- !u!157 &3
 LightmapSettings:
   m_ObjectHideFlags: 0
   serializedVersion: 12
-  m_GIWorkflowMode: 1
   m_GISettings:
     serializedVersion: 2
     m_BounceScale: 1
@@ -67,9 +66,6 @@ LightmapSettings:
     m_LightmapParameters: {fileID: 0}
     m_LightmapsBakeMode: 1
     m_TextureCompression: 1
-    m_FinalGather: 0
-    m_FinalGatherFiltering: 1
-    m_FinalGatherRayCount: 256
     m_ReflectionCompression: 2
     m_MixedBakeMode: 2
     m_BakeBackend: 1
@@ -367,9 +363,8 @@ Light:
   m_PrefabAsset: {fileID: 0}
   m_GameObject: {fileID: 705507993}
   m_Enabled: 1
-  serializedVersion: 10
+  serializedVersion: 11
   m_Type: 1
-  m_Shape: 0
   m_Color: {r: 1, g: 0.95686275, b: 0.8392157, a: 1}
   m_Intensity: 1
   m_Range: 10
@@ -419,8 +414,12 @@ Light:
   m_BoundingSphereOverride: {x: 0, y: 0, z: 0, w: 0}
   m_UseBoundingSphereOverride: 0
   m_UseViewFrustumForShadowCasterCull: 1
+  m_ForceVisible: 0
   m_ShadowRadius: 0
   m_ShadowAngle: 0
+  m_LightUnit: 1
+  m_LuxAtDistance: 1
+  m_EnableSpotReflector: 1
 --- !u!4 &705507995
 Transform:
   m_ObjectHideFlags: 0
@@ -639,6 +638,9 @@ MeshRenderer:
   m_ReflectionProbeUsage: 1
   m_RayTracingMode: 2
   m_RayTraceProcedural: 0
+  m_RayTracingAccelStructBuildFlagsOverride: 0
+  m_RayTracingAccelStructBuildFlags: 1
+  m_SmallMeshCulling: 1
   m_RenderingLayerMask: 1
   m_RendererPriority: 0
   m_Materials:
@@ -840,6 +842,7 @@ AudioSource:
   serializedVersion: 4
   OutputAudioMixerGroup: {fileID: 0}
   m_audioClip: {fileID: 0}
+  m_Resource: {fileID: 0}
   m_PlayOnAwake: 0
   m_Volume: 1
   m_Pitch: 1
@@ -1159,9 +1162,9 @@ MonoBehaviour:
   m_Name: 
   m_EditorClassIdentifier: 
   backend: 512
-  model: {fileID: 5022602860645237092, guid: 808a451b47032994c9dc7906b4c27ec7, type: 3}
+  model: {fileID: 5022602860645237092, guid: 4e2afd77f290c4dd68adb6d86976ace9, type: 3}
   voice: en-us
-  sampleRate: 22050
+  sampleRate: 16000
 --- !u!1 &1795061063
 GameObject:
   m_ObjectHideFlags: 0

Assets/Piper/Scripts/PiperManager.cs

@@ -1,6 +1,5 @@
 using System.Collections.Generic;
 using System.IO;
-using System.Linq;
 using System.Threading.Tasks;
 using Unity.Sentis;
 using UnityEngine;
@@ -16,15 +15,16 @@ public class PiperManager : MonoBehaviour
         public int sampleRate = 22050;
 
         private Model _runtimeModel;
-        private IWorker _worker;
+        private Worker _worker;
 
         private void Awake()
         {
             var espeakPath = Path.Combine(Application.streamingAssetsPath, "espeak-ng-data");
             PiperWrapper.InitPiper(espeakPath);
 
             _runtimeModel = ModelLoader.Load(model);
-            _worker = WorkerFactory.CreateWorker(backend, _runtimeModel);
+            _worker = new Worker(_runtimeModel, backend);
+
         }
 
         public async Task<AudioClip> TextToSpeech(string text)
@@ -41,30 +41,28 @@ public async Task<AudioClip> TextToSpeech(string text)
 
             var inputLengthsShape = new TensorShape(1);
             var scalesShape = new TensorShape(3);
-            using var scalesTensor = new TensorFloat(scalesShape, new float[] { 0.667f, 1f, 0.8f });
+            using var scalesTensor = new Tensor<float>(scalesShape, new [] { 1f, 1f, 1f });
 
             var audioBuffer = new List<float>();
-            for (int i = 0; i < phonemes.Sentences.Length; i++) 
+            foreach (var sentence in phonemes.Sentences)
             {
-                var sentence = phonemes.Sentences[i];
-
                 var inputPhonemes = sentence.PhonemesIds;
                 var inputShape = new TensorShape(1, inputPhonemes.Length);
-                using var inputTensor = new TensorInt(inputShape, inputPhonemes);
-                using var inputLengthsTensor = new TensorInt(inputLengthsShape, new int[] { inputPhonemes.Length });
-
-                var input = new Dictionary<string, Tensor>();
-                input.Add("input", inputTensor);
-                input.Add("input_lengths", inputLengthsTensor);
-                input.Add("scales", scalesTensor);
+                using var inputTensor = new Tensor<int>(inputShape, inputPhonemes);
+                using var inputLengthsTensor = new Tensor<int>(inputLengthsShape, new [] { inputPhonemes.Length });
 
-                _worker.Execute(input);
+                _worker.Schedule(inputTensor, inputLengthsTensor, scalesTensor);
 
-                using var outputTensor = _worker.PeekOutput() as TensorFloat;
-                await outputTensor.MakeReadableAsync();
+                using var outputTensor = _worker.PeekOutput() as Tensor<float>;
+                if (outputTensor != null)
+                {
+                    await outputTensor.ReadbackAndCloneAsync();
+
+                    var output = outputTensor.DownloadToArray();
+                    Debug.Log(output.Length);
 
-                var output = outputTensor.ToReadOnlyArray();
-                audioBuffer.AddRange(output);
+                    audioBuffer.AddRange(output);
+                }
             }
 
             Debug.Log($"Finished piper inference: {sw.ElapsedMilliseconds} ms");

Assets/Piper/Scripts/PiperNative.cs

@@ -1,5 +1,6 @@
 using System;
 using System.Runtime.InteropServices;
+using UnityEngine;
 
 namespace Piper.Native
 {
@@ -20,8 +21,15 @@ public unsafe struct PiperProcessedTextNative
 
     public static unsafe class PiperNative
     {
-        private const string LibraryName = "piper_phonemize";
-
+        private const string LibraryName =
+
+#if UNITY_STANDALONE_OSX
+            "libpiper_phonemize";
+
+#elif UNITY_STANDALONE_WIN
+            "piper_phonemize";
+#endif
+
         [DllImport(LibraryName)]
         public static extern int init_piper(string dataPath);
 
@@ -33,6 +41,10 @@ public static unsafe class PiperNative
 
         [DllImport(LibraryName)]
         public static extern void free_piper();
+        public static void PrintLibraryName()
+        {
+            Debug.Log(LibraryName);
+        }
 
     }
 }

Assets/Piper/Scripts/PiperWrapper.cs

@@ -46,7 +46,7 @@ public static bool InitPiper(string datapath)
                 Debug.LogError($"Provided espeak data path \"{datapath}\" doesn't exist!");
                 return false;
             }
-
+            PiperNative.PrintLibraryName();
             var code = PiperNative.init_piper(datapath);
             if (code < 0)
             {