fix(vocoder): rank-stable anchor for macOS 14 Espresso (FluidAudio#836)

Replace reduce_mean(keep_dims=false) -> rank-0 -> expand_dims with a single reduce_mean(axes=[1,2]) producing [1] directly. macOS 14's Espresso runtime mangles the rank-0 intermediate into rank 2, failing every prediction under all compute-unit presets. anchor is discarded by all hosts; x_pre verified bit-exact vs previous artifacts. Applies to ANE, ANE-zh, ANE-ja (ja vocoder is byte-identical to en).

ANE-ja/KokoroVocoder.mlmodelc/analytics/coremldata.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8c7c1a25a46ad46b1068905ece8f841c4bd23df23306551bad175d0da28ae74b
+oid sha256:1cc4f1e6436597c6d458a845570d9d0ace458813ff90f225d08e0e713496b463
 size 243

ANE-ja/KokoroVocoder.mlmodelc/coremldata.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e73aaf146c7543c0f75f544ac52e3287fa4eaa7e9a04bf3ac6a94d0023f16c00
+oid sha256:71bac80d2f2f077fefc0d66cdc39c0faaa96540b3a6149093ef29de01b1f8764
 size 626

ANE-ja/KokoroVocoder.mlmodelc/metadata.json

@@ -1,147 +1,143 @@
 [
   {
-    "metadataOutputVersion" : "3.0",
-    "storagePrecision" : "Mixed (Float16, Palettized (8 bits))",
-    "outputSchema" : [
+    "metadataOutputVersion": "3.0",
+    "storagePrecision": "Mixed (Float16, Palettized (8 bits))",
+    "outputSchema": [
       {
-        "hasShapeFlexibility" : "0",
-        "isOptional" : "0",
-        "dataType" : "Float16",
-        "formattedType" : "MultiArray (Float16 1)",
-        "shortDescription" : "",
-        "shape" : "[1]",
-        "name" : "anchor",
-        "type" : "MultiArray"
+        "hasShapeFlexibility": "0",
+        "isOptional": "0",
+        "dataType": "Float16",
+        "formattedType": "MultiArray (Float16 1)",
+        "shortDescription": "",
+        "shape": "[1]",
+        "name": "anchor",
+        "type": "MultiArray"
       },
       {
-        "hasShapeFlexibility" : "0",
-        "isOptional" : "0",
-        "dataType" : "Float16",
-        "formattedType" : "MultiArray (Float16)",
-        "shortDescription" : "",
-        "shape" : "[]",
-        "name" : "x_pre",
-        "type" : "MultiArray"
+        "hasShapeFlexibility": "0",
+        "isOptional": "0",
+        "dataType": "Float16",
+        "formattedType": "MultiArray (Float16)",
+        "shortDescription": "",
+        "shape": "[]",
+        "name": "x_pre",
+        "type": "MultiArray"
       }
     ],
-    "modelParameters" : [
-
-    ],
-    "specificationVersion" : 8,
-    "mlProgramOperationTypeHistogram" : {
-      "Ios17.mul" : 161,
-      "Ios17.linear" : 46,
-      "Ios17.cos" : 36,
-      "Ios16.constexprLutToDense" : 101,
-      "Ios17.conv" : 54,
-      "Ios17.leakyRelu" : 13,
-      "Ios17.concat" : 6,
-      "Ios17.add" : 193,
-      "Ios17.convTranspose" : 3,
-      "Ios17.sliceByIndex" : 2,
-      "UpsampleNearestNeighbor" : 1,
-      "Ios16.reduceMean" : 1,
-      "Ios17.expandDims" : 4,
-      "Ios17.instanceNorm" : 42,
-      "Ios17.reshape" : 46,
-      "Split" : 46,
-      "Ios17.squeeze" : 1
+    "modelParameters": [],
+    "specificationVersion": 8,
+    "mlProgramOperationTypeHistogram": {
+      "Ios17.mul": 161,
+      "Ios17.linear": 46,
+      "Ios17.cos": 36,
+      "Ios16.constexprLutToDense": 101,
+      "Ios17.conv": 54,
+      "Ios17.leakyRelu": 13,
+      "Ios17.concat": 6,
+      "Ios17.add": 193,
+      "Ios17.convTranspose": 3,
+      "Ios17.sliceByIndex": 2,
+      "UpsampleNearestNeighbor": 1,
+      "Ios16.reduceMean": 1,
+      "Ios17.expandDims": 3,
+      "Ios17.instanceNorm": 42,
+      "Ios17.reshape": 46,
+      "Split": 46,
+      "Ios17.squeeze": 1
     },
-    "computePrecision" : "Mixed (Float16, Float32, Int32)",
-    "isUpdatable" : "0",
-    "stateSchema" : [
-
-    ],
-    "availability" : {
-      "macOS" : "14.0",
-      "tvOS" : "17.0",
-      "visionOS" : "1.0",
-      "watchOS" : "10.0",
-      "iOS" : "17.0",
-      "macCatalyst" : "17.0"
+    "computePrecision": "Mixed (Float16, Float32, Int32)",
+    "isUpdatable": "0",
+    "stateSchema": [],
+    "availability": {
+      "macOS": "14.0",
+      "tvOS": "17.0",
+      "visionOS": "1.0",
+      "watchOS": "10.0",
+      "iOS": "17.0",
+      "macCatalyst": "17.0"
     },
-    "modelType" : {
-      "name" : "MLModelType_mlProgram"
+    "modelType": {
+      "name": "MLModelType_mlProgram"
     },
-    "userDefinedMetadata" : {
-      "com.github.apple.coremltools.conversion_date" : "2026-04-27",
-      "com.github.apple.coremltools.source" : "torch==2.11.0",
-      "com.github.apple.coremltools.version" : "9.0",
-      "com.github.apple.coremltools.source_dialect" : "TorchScript"
+    "userDefinedMetadata": {
+      "com.github.apple.coremltools.conversion_date": "2026-04-27",
+      "com.github.apple.coremltools.source": "torch==2.11.0",
+      "com.github.apple.coremltools.version": "9.0",
+      "com.github.apple.coremltools.source_dialect": "TorchScript"
     },
-    "inputSchema" : [
+    "inputSchema": [
       {
-        "dataType" : "Float16",
-        "hasShapeFlexibility" : "1",
-        "isOptional" : "0",
-        "shapeFlexibility" : "1 × 512 × 1...2000",
-        "shapeRange" : "[[1, 1], [512, 512], [1, 2000]]",
-        "formattedType" : "MultiArray (Float16 1 × 512 × 120)",
-        "type" : "MultiArray",
-        "shape" : "[1, 512, 120]",
-        "name" : "asr",
-        "shortDescription" : ""
+        "dataType": "Float16",
+        "hasShapeFlexibility": "1",
+        "isOptional": "0",
+        "shapeFlexibility": "1 \u00d7 512 \u00d7 1...2000",
+        "shapeRange": "[[1, 1], [512, 512], [1, 2000]]",
+        "formattedType": "MultiArray (Float16 1 \u00d7 512 \u00d7 120)",
+        "type": "MultiArray",
+        "shape": "[1, 512, 120]",
+        "name": "asr",
+        "shortDescription": ""
       },
       {
-        "dataType" : "Float16",
-        "hasShapeFlexibility" : "1",
-        "isOptional" : "0",
-        "shapeFlexibility" : "1 × 2...4000",
-        "shapeRange" : "[[1, 1], [2, 4000]]",
-        "formattedType" : "MultiArray (Float16 1 × 240)",
-        "type" : "MultiArray",
-        "shape" : "[1, 240]",
-        "name" : "F0_curve",
-        "shortDescription" : ""
+        "dataType": "Float16",
+        "hasShapeFlexibility": "1",
+        "isOptional": "0",
+        "shapeFlexibility": "1 \u00d7 2...4000",
+        "shapeRange": "[[1, 1], [2, 4000]]",
+        "formattedType": "MultiArray (Float16 1 \u00d7 240)",
+        "type": "MultiArray",
+        "shape": "[1, 240]",
+        "name": "F0_curve",
+        "shortDescription": ""
       },
       {
-        "dataType" : "Float16",
-        "hasShapeFlexibility" : "1",
-        "isOptional" : "0",
-        "shapeFlexibility" : "1 × 2...4000",
-        "shapeRange" : "[[1, 1], [2, 4000]]",
-        "formattedType" : "MultiArray (Float16 1 × 240)",
-        "type" : "MultiArray",
-        "shape" : "[1, 240]",
-        "name" : "N_pred",
-        "shortDescription" : ""
+        "dataType": "Float16",
+        "hasShapeFlexibility": "1",
+        "isOptional": "0",
+        "shapeFlexibility": "1 \u00d7 2...4000",
+        "shapeRange": "[[1, 1], [2, 4000]]",
+        "formattedType": "MultiArray (Float16 1 \u00d7 240)",
+        "type": "MultiArray",
+        "shape": "[1, 240]",
+        "name": "N_pred",
+        "shortDescription": ""
       },
       {
-        "dataType" : "Float16",
-        "hasShapeFlexibility" : "1",
-        "isOptional" : "0",
-        "shapeFlexibility" : "1 × 256 × 1...40000",
-        "shapeRange" : "[[1, 1], [256, 256], [1, 40000]]",
-        "formattedType" : "MultiArray (Float16 1 × 256 × 2400)",
-        "type" : "MultiArray",
-        "shape" : "[1, 256, 2400]",
-        "name" : "x_source_0",
-        "shortDescription" : ""
+        "dataType": "Float16",
+        "hasShapeFlexibility": "1",
+        "isOptional": "0",
+        "shapeFlexibility": "1 \u00d7 256 \u00d7 1...40000",
+        "shapeRange": "[[1, 1], [256, 256], [1, 40000]]",
+        "formattedType": "MultiArray (Float16 1 \u00d7 256 \u00d7 2400)",
+        "type": "MultiArray",
+        "shape": "[1, 256, 2400]",
+        "name": "x_source_0",
+        "shortDescription": ""
       },
       {
-        "dataType" : "Float16",
-        "hasShapeFlexibility" : "1",
-        "isOptional" : "0",
-        "shapeFlexibility" : "1 × 128 × 1...240001",
-        "shapeRange" : "[[1, 1], [128, 128], [1, 240001]]",
-        "formattedType" : "MultiArray (Float16 1 × 128 × 14401)",
-        "type" : "MultiArray",
-        "shape" : "[1, 128, 14401]",
-        "name" : "x_source_1",
-        "shortDescription" : ""
+        "dataType": "Float16",
+        "hasShapeFlexibility": "1",
+        "isOptional": "0",
+        "shapeFlexibility": "1 \u00d7 128 \u00d7 1...240001",
+        "shapeRange": "[[1, 1], [128, 128], [1, 240001]]",
+        "formattedType": "MultiArray (Float16 1 \u00d7 128 \u00d7 14401)",
+        "type": "MultiArray",
+        "shape": "[1, 128, 14401]",
+        "name": "x_source_1",
+        "shortDescription": ""
       },
       {
-        "hasShapeFlexibility" : "0",
-        "isOptional" : "0",
-        "dataType" : "Float16",
-        "formattedType" : "MultiArray (Float16 1 × 128)",
-        "shortDescription" : "",
-        "shape" : "[1, 128]",
-        "name" : "style_timbre",
-        "type" : "MultiArray"
+        "hasShapeFlexibility": "0",
+        "isOptional": "0",
+        "dataType": "Float16",
+        "formattedType": "MultiArray (Float16 1 \u00d7 128)",
+        "shortDescription": "",
+        "shape": "[1, 128]",
+        "name": "style_timbre",
+        "type": "MultiArray"
       }
     ],
-    "generatedClassName" : "KokoroVocoder",
-    "method" : "predict"
+    "generatedClassName": "KokoroVocoder",
+    "method": "predict"
   }
 ]

ANE-ja/KokoroVocoder.mlmodelc/model.mil

@@ -1,5 +1,5 @@
 program(1.0)
-[buildInfo = dict<tensor<string, []>, tensor<string, []>>({{"coremlc-component-MIL", "3520.4.1"}, {"coremlc-version", "3520.5.1"}})]
+[buildInfo = dict<tensor<string, []>, tensor<string, []>>({{"coremlc-component-MIL", "3520.4.1"}, {"coremlc-version", "3520.5.1.2"}})]
 {
     func main<ios17>(tensor<fp16, [1, ?]> F0_curve, tensor<fp16, [1, ?]> N_pred, tensor<fp16, [1, 512, ?]> asr, tensor<fp16, [1, 128]> style_timbre, tensor<fp16, [1, 256, ?]> x_source_0, tensor<fp16, [1, 128, ?]> x_source_1) [FlexibleShapeInformation = tuple<tuple<tensor<string, []>, dict<tensor<string, []>, tensor<int32, [?]>>>, tuple<tensor<string, []>, dict<tensor<string, []>, list<tensor<int32, [2]>, ?>>>>((("DefaultShapes", {{"F0_curve", [1, 240]}, {"N_pred", [1, 240]}, {"asr", [1, 512, 120]}, {"x_source_0", [1, 256, 2400]}, {"x_source_1", [1, 128, 14401]}}), ("RangeDims", {{"F0_curve", [[1, 1], [2, 4000]]}, {"N_pred", [[1, 1], [2, 4000]]}, {"asr", [[1, 1], [512, 512], [1, 2000]]}, {"x_source_0", [[1, 1], [256, 256], [1, 40000]]}, {"x_source_1", [[1, 1], [128, 128], [1, 240001]]}})))] {
             tensor<int32, [1]> input_1_axes_0 = const()[name = tensor<string, []>("input_1_axes_0"), val = tensor<int32, [1]>([1])];
@@ -1533,8 +1533,7 @@ program(1.0)
             tensor<fp32, []> var_2519 = const()[name = tensor<string, []>("op_2519"), val = tensor<fp32, []>(0x1.47ae14p-7)];
             tensor<fp16, [1, 128, ?]> x_pre = leaky_relu(alpha = var_2519, x = _inversed_input_cast_fp16)[name = tensor<string, []>("x_pre_cast_fp16")];
             tensor<bool, []> var_2522_keep_dims_0 = const()[name = tensor<string, []>("op_2522_keep_dims_0"), val = tensor<bool, []>(false)];
-            tensor<fp16, []> var_2522_cast_fp16 = reduce_mean(keep_dims = var_2522_keep_dims_0, x = x_pre)[name = tensor<string, []>("op_2522_cast_fp16")];
-            tensor<int32, [1]> var_2524_axes_0 = const()[name = tensor<string, []>("op_2524_axes_0"), val = tensor<int32, [1]>([0])];
-            tensor<fp16, [1]> anchor = expand_dims(axes = var_2524_axes_0, x = var_2522_cast_fp16)[name = tensor<string, []>("op_2524_cast_fp16")];
+            tensor<int32, [2]> var_2524_axes_0 = const()[name = tensor<string, []>("op_2524_axes_0"), val = tensor<int32, [2]>([1, 2])];
+            tensor<fp16, [1]> anchor = reduce_mean(axes = var_2524_axes_0, keep_dims = var_2522_keep_dims_0, x = x_pre)[name = tensor<string, []>("op_2522_cast_fp16")];
         } -> (anchor, x_pre);
 }

ANE-zh/KokoroVocoder.mlmodelc/analytics/coremldata.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:12bf39f5117a2fe2645b0158d75d71a99b328b237f1add55a1511d0e2ee3b456
+oid sha256:80953c7c1509d2f2d16e822a1a64e173addf98810e34df868c340bd184e10d56
 size 243

ANE-zh/KokoroVocoder.mlmodelc/coremldata.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:24ad83fed32d94eff2b3d9d70159140057ebad500b2a7f8fb54b0be0204b1cd2
+oid sha256:64c49e62863d689ca4482681373e6466cc6cdb02edadd30c5a33ab19dc9e7ab3
 size 627

ANE-zh/KokoroVocoder.mlmodelc/metadata.json

@@ -1,147 +1,143 @@
 [
   {
-    "metadataOutputVersion" : "3.0",
-    "storagePrecision" : "Mixed (Float16, Palettized (8 bits))",
-    "outputSchema" : [
+    "metadataOutputVersion": "3.0",
+    "storagePrecision": "Mixed (Float16, Palettized (8 bits))",
+    "outputSchema": [
       {
-        "hasShapeFlexibility" : "0",
-        "isOptional" : "0",
-        "dataType" : "Float16",
-        "formattedType" : "MultiArray (Float16 1)",
-        "shortDescription" : "",
-        "shape" : "[1]",
-        "name" : "anchor",
-        "type" : "MultiArray"
+        "hasShapeFlexibility": "0",
+        "isOptional": "0",
+        "dataType": "Float16",
+        "formattedType": "MultiArray (Float16 1)",
+        "shortDescription": "",
+        "shape": "[1]",
+        "name": "anchor",
+        "type": "MultiArray"
       },
       {
-        "hasShapeFlexibility" : "0",
-        "isOptional" : "0",
-        "dataType" : "Float16",
-        "formattedType" : "MultiArray (Float16)",
-        "shortDescription" : "",
-        "shape" : "[]",
-        "name" : "x_pre",
-        "type" : "MultiArray"
+        "hasShapeFlexibility": "0",
+        "isOptional": "0",
+        "dataType": "Float16",
+        "formattedType": "MultiArray (Float16)",
+        "shortDescription": "",
+        "shape": "[]",
+        "name": "x_pre",
+        "type": "MultiArray"
       }
     ],
-    "modelParameters" : [
-
-    ],
-    "specificationVersion" : 8,
-    "mlProgramOperationTypeHistogram" : {
-      "Ios17.mul" : 161,
-      "Ios17.linear" : 46,
-      "Ios17.cos" : 36,
-      "Ios16.constexprLutToDense" : 101,
-      "Ios17.conv" : 54,
-      "Ios17.leakyRelu" : 13,
-      "Ios17.concat" : 6,
-      "Ios17.add" : 193,
-      "Ios17.convTranspose" : 3,
-      "Ios17.sliceByIndex" : 2,
-      "UpsampleNearestNeighbor" : 1,
-      "Ios16.reduceMean" : 1,
-      "Ios17.expandDims" : 4,
-      "Ios17.instanceNorm" : 42,
-      "Ios17.reshape" : 46,
-      "Split" : 46,
-      "Ios17.squeeze" : 1
+    "modelParameters": [],
+    "specificationVersion": 8,
+    "mlProgramOperationTypeHistogram": {
+      "Ios17.mul": 161,
+      "Ios17.linear": 46,
+      "Ios17.cos": 36,
+      "Ios16.constexprLutToDense": 101,
+      "Ios17.conv": 54,
+      "Ios17.leakyRelu": 13,
+      "Ios17.concat": 6,
+      "Ios17.add": 193,
+      "Ios17.convTranspose": 3,
+      "Ios17.sliceByIndex": 2,
+      "UpsampleNearestNeighbor": 1,
+      "Ios16.reduceMean": 1,
+      "Ios17.expandDims": 3,
+      "Ios17.instanceNorm": 42,
+      "Ios17.reshape": 46,
+      "Split": 46,
+      "Ios17.squeeze": 1
     },
-    "computePrecision" : "Mixed (Float16, Float32, Int32)",
-    "isUpdatable" : "0",
-    "stateSchema" : [
-
-    ],
-    "availability" : {
-      "macOS" : "14.0",
-      "tvOS" : "17.0",
-      "visionOS" : "1.0",
-      "watchOS" : "10.0",
-      "iOS" : "17.0",
-      "macCatalyst" : "17.0"
+    "computePrecision": "Mixed (Float16, Float32, Int32)",
+    "isUpdatable": "0",
+    "stateSchema": [],
+    "availability": {
+      "macOS": "14.0",
+      "tvOS": "17.0",
+      "visionOS": "1.0",
+      "watchOS": "10.0",
+      "iOS": "17.0",
+      "macCatalyst": "17.0"
     },
-    "modelType" : {
-      "name" : "MLModelType_mlProgram"
+    "modelType": {
+      "name": "MLModelType_mlProgram"
     },
-    "userDefinedMetadata" : {
-      "com.github.apple.coremltools.conversion_date" : "2026-05-03",
-      "com.github.apple.coremltools.source" : "torch==2.11.0",
-      "com.github.apple.coremltools.version" : "9.0",
-      "com.github.apple.coremltools.source_dialect" : "TorchScript"
+    "userDefinedMetadata": {
+      "com.github.apple.coremltools.conversion_date": "2026-05-03",
+      "com.github.apple.coremltools.source": "torch==2.11.0",
+      "com.github.apple.coremltools.version": "9.0",
+      "com.github.apple.coremltools.source_dialect": "TorchScript"
     },
-    "inputSchema" : [
+    "inputSchema": [
       {
-        "dataType" : "Float16",
-        "hasShapeFlexibility" : "1",
-        "isOptional" : "0",
-        "shapeFlexibility" : "1 × 512 × 1...2000",
-        "shapeRange" : "[[1, 1], [512, 512], [1, 2000]]",
-        "formattedType" : "MultiArray (Float16 1 × 512 × 133)",
-        "type" : "MultiArray",
-        "shape" : "[1, 512, 133]",
-        "name" : "asr",
-        "shortDescription" : ""
+        "dataType": "Float16",
+        "hasShapeFlexibility": "1",
+        "isOptional": "0",
+        "shapeFlexibility": "1 \u00d7 512 \u00d7 1...2000",
+        "shapeRange": "[[1, 1], [512, 512], [1, 2000]]",
+        "formattedType": "MultiArray (Float16 1 \u00d7 512 \u00d7 133)",
+        "type": "MultiArray",
+        "shape": "[1, 512, 133]",
+        "name": "asr",
+        "shortDescription": ""
       },
       {
-        "dataType" : "Float16",
-        "hasShapeFlexibility" : "1",
-        "isOptional" : "0",
-        "shapeFlexibility" : "1 × 2...4000",
-        "shapeRange" : "[[1, 1], [2, 4000]]",
-        "formattedType" : "MultiArray (Float16 1 × 266)",
-        "type" : "MultiArray",
-        "shape" : "[1, 266]",
-        "name" : "F0_curve",
-        "shortDescription" : ""
+        "dataType": "Float16",
+        "hasShapeFlexibility": "1",
+        "isOptional": "0",
+        "shapeFlexibility": "1 \u00d7 2...4000",
+        "shapeRange": "[[1, 1], [2, 4000]]",
+        "formattedType": "MultiArray (Float16 1 \u00d7 266)",
+        "type": "MultiArray",
+        "shape": "[1, 266]",
+        "name": "F0_curve",
+        "shortDescription": ""
       },
       {
-        "dataType" : "Float16",
-        "hasShapeFlexibility" : "1",
-        "isOptional" : "0",
-        "shapeFlexibility" : "1 × 2...4000",
-        "shapeRange" : "[[1, 1], [2, 4000]]",
-        "formattedType" : "MultiArray (Float16 1 × 266)",
-        "type" : "MultiArray",
-        "shape" : "[1, 266]",
-        "name" : "N_pred",
-        "shortDescription" : ""
+        "dataType": "Float16",
+        "hasShapeFlexibility": "1",
+        "isOptional": "0",
+        "shapeFlexibility": "1 \u00d7 2...4000",
+        "shapeRange": "[[1, 1], [2, 4000]]",
+        "formattedType": "MultiArray (Float16 1 \u00d7 266)",
+        "type": "MultiArray",
+        "shape": "[1, 266]",
+        "name": "N_pred",
+        "shortDescription": ""
       },
       {
-        "dataType" : "Float16",
-        "hasShapeFlexibility" : "1",
-        "isOptional" : "0",
-        "shapeFlexibility" : "1 × 256 × 1...40000",
-        "shapeRange" : "[[1, 1], [256, 256], [1, 40000]]",
-        "formattedType" : "MultiArray (Float16 1 × 256 × 2660)",
-        "type" : "MultiArray",
-        "shape" : "[1, 256, 2660]",
-        "name" : "x_source_0",
-        "shortDescription" : ""
+        "dataType": "Float16",
+        "hasShapeFlexibility": "1",
+        "isOptional": "0",
+        "shapeFlexibility": "1 \u00d7 256 \u00d7 1...40000",
+        "shapeRange": "[[1, 1], [256, 256], [1, 40000]]",
+        "formattedType": "MultiArray (Float16 1 \u00d7 256 \u00d7 2660)",
+        "type": "MultiArray",
+        "shape": "[1, 256, 2660]",
+        "name": "x_source_0",
+        "shortDescription": ""
       },
       {
-        "dataType" : "Float16",
-        "hasShapeFlexibility" : "1",
-        "isOptional" : "0",
-        "shapeFlexibility" : "1 × 128 × 1...240001",
-        "shapeRange" : "[[1, 1], [128, 128], [1, 240001]]",
-        "formattedType" : "MultiArray (Float16 1 × 128 × 15961)",
-        "type" : "MultiArray",
-        "shape" : "[1, 128, 15961]",
-        "name" : "x_source_1",
-        "shortDescription" : ""
+        "dataType": "Float16",
+        "hasShapeFlexibility": "1",
+        "isOptional": "0",
+        "shapeFlexibility": "1 \u00d7 128 \u00d7 1...240001",
+        "shapeRange": "[[1, 1], [128, 128], [1, 240001]]",
+        "formattedType": "MultiArray (Float16 1 \u00d7 128 \u00d7 15961)",
+        "type": "MultiArray",
+        "shape": "[1, 128, 15961]",
+        "name": "x_source_1",
+        "shortDescription": ""
       },
       {
-        "hasShapeFlexibility" : "0",
-        "isOptional" : "0",
-        "dataType" : "Float16",
-        "formattedType" : "MultiArray (Float16 1 × 128)",
-        "shortDescription" : "",
-        "shape" : "[1, 128]",
-        "name" : "style_timbre",
-        "type" : "MultiArray"
+        "hasShapeFlexibility": "0",
+        "isOptional": "0",
+        "dataType": "Float16",
+        "formattedType": "MultiArray (Float16 1 \u00d7 128)",
+        "shortDescription": "",
+        "shape": "[1, 128]",
+        "name": "style_timbre",
+        "type": "MultiArray"
       }
     ],
-    "generatedClassName" : "KokoroVocoder",
-    "method" : "predict"
+    "generatedClassName": "KokoroVocoder",
+    "method": "predict"
   }
 ]

ANE-zh/KokoroVocoder.mlmodelc/model.mil

@@ -1,5 +1,5 @@
 program(1.0)
-[buildInfo = dict<tensor<string, []>, tensor<string, []>>({{"coremlc-component-MIL", "3520.4.1"}, {"coremlc-version", "3520.5.1"}})]
+[buildInfo = dict<tensor<string, []>, tensor<string, []>>({{"coremlc-component-MIL", "3520.4.1"}, {"coremlc-version", "3520.5.1.2"}})]
 {
     func main<ios17>(tensor<fp16, [1, ?]> F0_curve, tensor<fp16, [1, ?]> N_pred, tensor<fp16, [1, 512, ?]> asr, tensor<fp16, [1, 128]> style_timbre, tensor<fp16, [1, 256, ?]> x_source_0, tensor<fp16, [1, 128, ?]> x_source_1) [FlexibleShapeInformation = tuple<tuple<tensor<string, []>, dict<tensor<string, []>, tensor<int32, [?]>>>, tuple<tensor<string, []>, dict<tensor<string, []>, list<tensor<int32, [2]>, ?>>>>((("DefaultShapes", {{"F0_curve", [1, 266]}, {"N_pred", [1, 266]}, {"asr", [1, 512, 133]}, {"x_source_0", [1, 256, 2660]}, {"x_source_1", [1, 128, 15961]}}), ("RangeDims", {{"F0_curve", [[1, 1], [2, 4000]]}, {"N_pred", [[1, 1], [2, 4000]]}, {"asr", [[1, 1], [512, 512], [1, 2000]]}, {"x_source_0", [[1, 1], [256, 256], [1, 40000]]}, {"x_source_1", [[1, 1], [128, 128], [1, 240001]]}})))] {
             tensor<int32, [1]> input_1_axes_0 = const()[name = tensor<string, []>("input_1_axes_0"), val = tensor<int32, [1]>([1])];
@@ -1533,8 +1533,7 @@ program(1.0)
             tensor<fp32, []> var_2519 = const()[name = tensor<string, []>("op_2519"), val = tensor<fp32, []>(0x1.47ae14p-7)];
             tensor<fp16, [1, 128, ?]> x_pre = leaky_relu(alpha = var_2519, x = _inversed_input_cast_fp16)[name = tensor<string, []>("x_pre_cast_fp16")];
             tensor<bool, []> var_2522_keep_dims_0 = const()[name = tensor<string, []>("op_2522_keep_dims_0"), val = tensor<bool, []>(false)];
-            tensor<fp16, []> var_2522_cast_fp16 = reduce_mean(keep_dims = var_2522_keep_dims_0, x = x_pre)[name = tensor<string, []>("op_2522_cast_fp16")];
-            tensor<int32, [1]> var_2524_axes_0 = const()[name = tensor<string, []>("op_2524_axes_0"), val = tensor<int32, [1]>([0])];
-            tensor<fp16, [1]> anchor = expand_dims(axes = var_2524_axes_0, x = var_2522_cast_fp16)[name = tensor<string, []>("op_2524_cast_fp16")];
+            tensor<int32, [2]> var_2524_axes_0 = const()[name = tensor<string, []>("op_2524_axes_0"), val = tensor<int32, [2]>([1, 2])];
+            tensor<fp16, [1]> anchor = reduce_mean(axes = var_2524_axes_0, keep_dims = var_2522_keep_dims_0, x = x_pre)[name = tensor<string, []>("op_2522_cast_fp16")];
         } -> (anchor, x_pre);
 }

ANE-zh/KokoroVocoder.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fb36344f47c7c0da6e39b928edd8e2d0393246d84cf807e5d74855eef37dfebd
-size 255855
+oid sha256:2f75fb1b4053a8232ca8b9ff66f03b57e07e3e434ea1379c8a7d3f1a2ed3ffe2
+size 255742

ANE-zh/KokoroVocoder.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "43E7FECC-0F8F-4FFD-8BEA-805B89C6839F": {
+        "4133C074-3053-45F2-AC5B-F5F7E3B18557": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Weights",
             "name": "weights",
             "path": "com.apple.CoreML/weights"
         },
-        "D8E47693-058B-43E2-BD65-9D1DDE555FD0": {
+        "53B59843-E250-456F-93B0-21CEC56666F1": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Specification",
             "name": "model.mlmodel",
             "path": "com.apple.CoreML/model.mlmodel"
         }
     },
-    "rootModelIdentifier": "D8E47693-058B-43E2-BD65-9D1DDE555FD0"
+    "rootModelIdentifier": "53B59843-E250-456F-93B0-21CEC56666F1"
 }

ANE/KokoroVocoder.mlmodelc/analytics/coremldata.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8c7c1a25a46ad46b1068905ece8f841c4bd23df23306551bad175d0da28ae74b
+oid sha256:1cc4f1e6436597c6d458a845570d9d0ace458813ff90f225d08e0e713496b463
 size 243

ANE/KokoroVocoder.mlmodelc/coremldata.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e73aaf146c7543c0f75f544ac52e3287fa4eaa7e9a04bf3ac6a94d0023f16c00
+oid sha256:71bac80d2f2f077fefc0d66cdc39c0faaa96540b3a6149093ef29de01b1f8764
 size 626

ANE/KokoroVocoder.mlmodelc/metadata.json

@@ -1,147 +1,143 @@
 [
   {
-    "metadataOutputVersion" : "3.0",
-    "storagePrecision" : "Mixed (Float16, Palettized (8 bits))",
-    "outputSchema" : [
+    "metadataOutputVersion": "3.0",
+    "storagePrecision": "Mixed (Float16, Palettized (8 bits))",
+    "outputSchema": [
       {
-        "hasShapeFlexibility" : "0",
-        "isOptional" : "0",
-        "dataType" : "Float16",
-        "formattedType" : "MultiArray (Float16 1)",
-        "shortDescription" : "",
-        "shape" : "[1]",
-        "name" : "anchor",
-        "type" : "MultiArray"
+        "hasShapeFlexibility": "0",
+        "isOptional": "0",
+        "dataType": "Float16",
+        "formattedType": "MultiArray (Float16 1)",
+        "shortDescription": "",
+        "shape": "[1]",
+        "name": "anchor",
+        "type": "MultiArray"
       },
       {
-        "hasShapeFlexibility" : "0",
-        "isOptional" : "0",
-        "dataType" : "Float16",
-        "formattedType" : "MultiArray (Float16)",
-        "shortDescription" : "",
-        "shape" : "[]",
-        "name" : "x_pre",
-        "type" : "MultiArray"
+        "hasShapeFlexibility": "0",
+        "isOptional": "0",
+        "dataType": "Float16",
+        "formattedType": "MultiArray (Float16)",
+        "shortDescription": "",
+        "shape": "[]",
+        "name": "x_pre",
+        "type": "MultiArray"
       }
     ],
-    "modelParameters" : [
-
-    ],
-    "specificationVersion" : 8,
-    "mlProgramOperationTypeHistogram" : {
-      "Ios17.mul" : 161,
-      "Ios17.linear" : 46,
-      "Ios17.cos" : 36,
-      "Ios16.constexprLutToDense" : 101,
-      "Ios17.conv" : 54,
-      "Ios17.leakyRelu" : 13,
-      "Ios17.concat" : 6,
-      "Ios17.add" : 193,
-      "Ios17.convTranspose" : 3,
-      "Ios17.sliceByIndex" : 2,
-      "UpsampleNearestNeighbor" : 1,
-      "Ios16.reduceMean" : 1,
-      "Ios17.expandDims" : 4,
-      "Ios17.instanceNorm" : 42,
-      "Ios17.reshape" : 46,
-      "Split" : 46,
-      "Ios17.squeeze" : 1
+    "modelParameters": [],
+    "specificationVersion": 8,
+    "mlProgramOperationTypeHistogram": {
+      "Ios17.mul": 161,
+      "Ios17.linear": 46,
+      "Ios17.cos": 36,
+      "Ios16.constexprLutToDense": 101,
+      "Ios17.conv": 54,
+      "Ios17.leakyRelu": 13,
+      "Ios17.concat": 6,
+      "Ios17.add": 193,
+      "Ios17.convTranspose": 3,
+      "Ios17.sliceByIndex": 2,
+      "UpsampleNearestNeighbor": 1,
+      "Ios16.reduceMean": 1,
+      "Ios17.expandDims": 3,
+      "Ios17.instanceNorm": 42,
+      "Ios17.reshape": 46,
+      "Split": 46,
+      "Ios17.squeeze": 1
     },
-    "computePrecision" : "Mixed (Float16, Float32, Int32)",
-    "isUpdatable" : "0",
-    "stateSchema" : [
-
-    ],
-    "availability" : {
-      "macOS" : "14.0",
-      "tvOS" : "17.0",
-      "visionOS" : "1.0",
-      "watchOS" : "10.0",
-      "iOS" : "17.0",
-      "macCatalyst" : "17.0"
+    "computePrecision": "Mixed (Float16, Float32, Int32)",
+    "isUpdatable": "0",
+    "stateSchema": [],
+    "availability": {
+      "macOS": "14.0",
+      "tvOS": "17.0",
+      "visionOS": "1.0",
+      "watchOS": "10.0",
+      "iOS": "17.0",
+      "macCatalyst": "17.0"
     },
-    "modelType" : {
-      "name" : "MLModelType_mlProgram"
+    "modelType": {
+      "name": "MLModelType_mlProgram"
     },
-    "userDefinedMetadata" : {
-      "com.github.apple.coremltools.conversion_date" : "2026-04-27",
-      "com.github.apple.coremltools.source" : "torch==2.11.0",
-      "com.github.apple.coremltools.version" : "9.0",
-      "com.github.apple.coremltools.source_dialect" : "TorchScript"
+    "userDefinedMetadata": {
+      "com.github.apple.coremltools.conversion_date": "2026-04-27",
+      "com.github.apple.coremltools.source": "torch==2.11.0",
+      "com.github.apple.coremltools.version": "9.0",
+      "com.github.apple.coremltools.source_dialect": "TorchScript"
     },
-    "inputSchema" : [
+    "inputSchema": [
       {
-        "dataType" : "Float16",
-        "hasShapeFlexibility" : "1",
-        "isOptional" : "0",
-        "shapeFlexibility" : "1 × 512 × 1...2000",
-        "shapeRange" : "[[1, 1], [512, 512], [1, 2000]]",
-        "formattedType" : "MultiArray (Float16 1 × 512 × 120)",
-        "type" : "MultiArray",
-        "shape" : "[1, 512, 120]",
-        "name" : "asr",
-        "shortDescription" : ""
+        "dataType": "Float16",
+        "hasShapeFlexibility": "1",
+        "isOptional": "0",
+        "shapeFlexibility": "1 \u00d7 512 \u00d7 1...2000",
+        "shapeRange": "[[1, 1], [512, 512], [1, 2000]]",
+        "formattedType": "MultiArray (Float16 1 \u00d7 512 \u00d7 120)",
+        "type": "MultiArray",
+        "shape": "[1, 512, 120]",
+        "name": "asr",
+        "shortDescription": ""
       },
       {
-        "dataType" : "Float16",
-        "hasShapeFlexibility" : "1",
-        "isOptional" : "0",
-        "shapeFlexibility" : "1 × 2...4000",
-        "shapeRange" : "[[1, 1], [2, 4000]]",
-        "formattedType" : "MultiArray (Float16 1 × 240)",
-        "type" : "MultiArray",
-        "shape" : "[1, 240]",
-        "name" : "F0_curve",
-        "shortDescription" : ""
+        "dataType": "Float16",
+        "hasShapeFlexibility": "1",
+        "isOptional": "0",
+        "shapeFlexibility": "1 \u00d7 2...4000",
+        "shapeRange": "[[1, 1], [2, 4000]]",
+        "formattedType": "MultiArray (Float16 1 \u00d7 240)",
+        "type": "MultiArray",
+        "shape": "[1, 240]",
+        "name": "F0_curve",
+        "shortDescription": ""
       },
       {
-        "dataType" : "Float16",
-        "hasShapeFlexibility" : "1",
-        "isOptional" : "0",
-        "shapeFlexibility" : "1 × 2...4000",
-        "shapeRange" : "[[1, 1], [2, 4000]]",
-        "formattedType" : "MultiArray (Float16 1 × 240)",
-        "type" : "MultiArray",
-        "shape" : "[1, 240]",
-        "name" : "N_pred",
-        "shortDescription" : ""
+        "dataType": "Float16",
+        "hasShapeFlexibility": "1",
+        "isOptional": "0",
+        "shapeFlexibility": "1 \u00d7 2...4000",
+        "shapeRange": "[[1, 1], [2, 4000]]",
+        "formattedType": "MultiArray (Float16 1 \u00d7 240)",
+        "type": "MultiArray",
+        "shape": "[1, 240]",
+        "name": "N_pred",
+        "shortDescription": ""
       },
       {
-        "dataType" : "Float16",
-        "hasShapeFlexibility" : "1",
-        "isOptional" : "0",
-        "shapeFlexibility" : "1 × 256 × 1...40000",
-        "shapeRange" : "[[1, 1], [256, 256], [1, 40000]]",
-        "formattedType" : "MultiArray (Float16 1 × 256 × 2400)",
-        "type" : "MultiArray",
-        "shape" : "[1, 256, 2400]",
-        "name" : "x_source_0",
-        "shortDescription" : ""
+        "dataType": "Float16",
+        "hasShapeFlexibility": "1",
+        "isOptional": "0",
+        "shapeFlexibility": "1 \u00d7 256 \u00d7 1...40000",
+        "shapeRange": "[[1, 1], [256, 256], [1, 40000]]",
+        "formattedType": "MultiArray (Float16 1 \u00d7 256 \u00d7 2400)",
+        "type": "MultiArray",
+        "shape": "[1, 256, 2400]",
+        "name": "x_source_0",
+        "shortDescription": ""
       },
       {
-        "dataType" : "Float16",
-        "hasShapeFlexibility" : "1",
-        "isOptional" : "0",
-        "shapeFlexibility" : "1 × 128 × 1...240001",
-        "shapeRange" : "[[1, 1], [128, 128], [1, 240001]]",
-        "formattedType" : "MultiArray (Float16 1 × 128 × 14401)",
-        "type" : "MultiArray",
-        "shape" : "[1, 128, 14401]",
-        "name" : "x_source_1",
-        "shortDescription" : ""
+        "dataType": "Float16",
+        "hasShapeFlexibility": "1",
+        "isOptional": "0",
+        "shapeFlexibility": "1 \u00d7 128 \u00d7 1...240001",
+        "shapeRange": "[[1, 1], [128, 128], [1, 240001]]",
+        "formattedType": "MultiArray (Float16 1 \u00d7 128 \u00d7 14401)",
+        "type": "MultiArray",
+        "shape": "[1, 128, 14401]",
+        "name": "x_source_1",
+        "shortDescription": ""
       },
       {
-        "hasShapeFlexibility" : "0",
-        "isOptional" : "0",
-        "dataType" : "Float16",
-        "formattedType" : "MultiArray (Float16 1 × 128)",
-        "shortDescription" : "",
-        "shape" : "[1, 128]",
-        "name" : "style_timbre",
-        "type" : "MultiArray"
+        "hasShapeFlexibility": "0",
+        "isOptional": "0",
+        "dataType": "Float16",
+        "formattedType": "MultiArray (Float16 1 \u00d7 128)",
+        "shortDescription": "",
+        "shape": "[1, 128]",
+        "name": "style_timbre",
+        "type": "MultiArray"
       }
     ],
-    "generatedClassName" : "KokoroVocoder",
-    "method" : "predict"
+    "generatedClassName": "KokoroVocoder",
+    "method": "predict"
   }
 ]

ANE/KokoroVocoder.mlmodelc/model.mil

@@ -1,5 +1,5 @@
 program(1.0)
-[buildInfo = dict<tensor<string, []>, tensor<string, []>>({{"coremlc-component-MIL", "3520.4.1"}, {"coremlc-version", "3520.5.1"}})]
+[buildInfo = dict<tensor<string, []>, tensor<string, []>>({{"coremlc-component-MIL", "3520.4.1"}, {"coremlc-version", "3520.5.1.2"}})]
 {
     func main<ios17>(tensor<fp16, [1, ?]> F0_curve, tensor<fp16, [1, ?]> N_pred, tensor<fp16, [1, 512, ?]> asr, tensor<fp16, [1, 128]> style_timbre, tensor<fp16, [1, 256, ?]> x_source_0, tensor<fp16, [1, 128, ?]> x_source_1) [FlexibleShapeInformation = tuple<tuple<tensor<string, []>, dict<tensor<string, []>, tensor<int32, [?]>>>, tuple<tensor<string, []>, dict<tensor<string, []>, list<tensor<int32, [2]>, ?>>>>((("DefaultShapes", {{"F0_curve", [1, 240]}, {"N_pred", [1, 240]}, {"asr", [1, 512, 120]}, {"x_source_0", [1, 256, 2400]}, {"x_source_1", [1, 128, 14401]}}), ("RangeDims", {{"F0_curve", [[1, 1], [2, 4000]]}, {"N_pred", [[1, 1], [2, 4000]]}, {"asr", [[1, 1], [512, 512], [1, 2000]]}, {"x_source_0", [[1, 1], [256, 256], [1, 40000]]}, {"x_source_1", [[1, 1], [128, 128], [1, 240001]]}})))] {
             tensor<int32, [1]> input_1_axes_0 = const()[name = tensor<string, []>("input_1_axes_0"), val = tensor<int32, [1]>([1])];
@@ -1533,8 +1533,7 @@ program(1.0)
             tensor<fp32, []> var_2519 = const()[name = tensor<string, []>("op_2519"), val = tensor<fp32, []>(0x1.47ae14p-7)];
             tensor<fp16, [1, 128, ?]> x_pre = leaky_relu(alpha = var_2519, x = _inversed_input_cast_fp16)[name = tensor<string, []>("x_pre_cast_fp16")];
             tensor<bool, []> var_2522_keep_dims_0 = const()[name = tensor<string, []>("op_2522_keep_dims_0"), val = tensor<bool, []>(false)];
-            tensor<fp16, []> var_2522_cast_fp16 = reduce_mean(keep_dims = var_2522_keep_dims_0, x = x_pre)[name = tensor<string, []>("op_2522_cast_fp16")];
-            tensor<int32, [1]> var_2524_axes_0 = const()[name = tensor<string, []>("op_2524_axes_0"), val = tensor<int32, [1]>([0])];
-            tensor<fp16, [1]> anchor = expand_dims(axes = var_2524_axes_0, x = var_2522_cast_fp16)[name = tensor<string, []>("op_2524_cast_fp16")];
+            tensor<int32, [2]> var_2524_axes_0 = const()[name = tensor<string, []>("op_2524_axes_0"), val = tensor<int32, [2]>([1, 2])];
+            tensor<fp16, [1]> anchor = reduce_mean(axes = var_2524_axes_0, keep_dims = var_2522_keep_dims_0, x = x_pre)[name = tensor<string, []>("op_2522_cast_fp16")];
         } -> (anchor, x_pre);
 }

ANE/KokoroVocoder.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5435ae53f7d14e998ceca6574fefc934954b2c0460e00a32dbc8179a2c1a50f7
-size 255854
+oid sha256:18481f9531abd0d01cf6fe6b8837b98263e8a63d87db56324d49ed57cc5c75fa
+size 255741

ANE/KokoroVocoder.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "8B8CAD76-5234-4AC3-9F94-3C90B26C1548": {
+        "511D8B74-593C-4045-BFEF-041A9B672CD4": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Weights",
             "name": "weights",
             "path": "com.apple.CoreML/weights"
         },
-        "A680BF92-9E9C-4D0B-AD35-4EDFFEDCBA56": {
+        "7E37220E-3B3B-46AB-9567-C022BED16A19": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Specification",
             "name": "model.mlmodel",
             "path": "com.apple.CoreML/model.mlmodel"
         }
     },
-    "rootModelIdentifier": "A680BF92-9E9C-4D0B-AD35-4EDFFEDCBA56"
+    "rootModelIdentifier": "7E37220E-3B3B-46AB-9567-C022BED16A19"
 }