WIP: integrate vcs to new gui 2

client/demo/src/components/demo/904-3_FileUploader.tsx

@@ -1,6 +1,6 @@
 import React, { useEffect, useMemo, useState } from "react";
 import { useAppState } from "../../001_provider/001_AppStateProvider";
-import { FileUploadSetting, InitialFileUploadSetting, ModelFileKind, ModelUploadSetting, VoiceChangerType, fileSelector } from "@dannadori/voice-changer-client-js";
+import { ModelFileKind, ModelUploadSetting, VoiceChangerType, fileSelector } from "@dannadori/voice-changer-client-js";
 import { useMessageBuilder } from "../../hooks/useMessageBuilder";
 import { ModelSlotManagerDialogScreen } from "./904_ModelSlotManagerDialog";
 import { checkExtention, trimfileName } from "../../utils/utils";
@@ -54,15 +54,31 @@ export const FileUploaderScreen = (props: FileUploaderScreenProps) => {
 
         const checkModelSetting = (setting: ModelUploadSetting) => {
             if (setting.voiceChangerType == "RVC") {
-                // const enough = !!setting.files.find(x => { return x.kind == "rvcModel" }) &&
-                //     !!setting.files.find(x => { return x.kind == "rvcIndex" })
-                // return enough
                 const enough = !!setting.files.find(x => { return x.kind == "rvcModel" })
                 return enough
+            } else if (setting.voiceChangerType == "MMVCv13") {
+                const enough = !!setting.files.find(x => { return x.kind == "mmvcv13Model" }) &&
+                    !!setting.files.find(x => { return x.kind == "mmvcv13Config" })
+                return enough
+            } else if (setting.voiceChangerType == "MMVCv15") {
+                const enough = !!setting.files.find(x => { return x.kind == "mmvcv15Model" }) &&
+                    !!setting.files.find(x => { return x.kind == "mmvcv15Config" })
+                return enough
+            } else if (setting.voiceChangerType == "so-vits-svc-40") {
+                const enough = !!setting.files.find(x => { return x.kind == "soVitsSvc40Config" }) &&
+                    !!setting.files.find(x => { return x.kind == "soVitsSvc40Model" })
+                return enough
+            } else if (setting.voiceChangerType == "DDSP-SVC") {
+                const enough = !!setting.files.find(x => { return x.kind == "ddspSvcModel" }) &&
+                    !!setting.files.find(x => { return x.kind == "ddspSvcModelConfig" }) &&
+                    !!setting.files.find(x => { return x.kind == "ddspSvcDiffusion" }) &&
+                    !!setting.files.find(x => { return x.kind == "ddspSvcDiffusionConfig" })
+                return enough
             }
+            return false
         }
 
-        const generateFileRow = (setting: ModelUploadSetting, title: string, kind: ModelFileKind, ext: string[]) => {
+        const generateFileRow = (setting: ModelUploadSetting, title: string, kind: ModelFileKind, ext: string[], dir: string = "") => {
             const selectedFile = setting.files.find(x => { return x.kind == kind })
             const selectedFilename = selectedFile?.file.name || ""
             return (
@@ -81,7 +97,7 @@ export const FileUploaderScreen = (props: FileUploaderScreenProps) => {
                         if (selectedFile) {
                             selectedFile.file = file
                         } else {
-                            setting.files.push({ kind: kind, file: file })
+                            setting.files.push({ kind: kind, file: file, dir: dir })
                         }
                         setUploadSetting({ ...setting })
                     }}>
@@ -96,6 +112,21 @@ export const FileUploaderScreen = (props: FileUploaderScreenProps) => {
             if (vcType == "RVC") {
                 rows.push(generateFileRow(uploadSetting!, "Model", "rvcModel", ["pth", "onnx"]))
                 rows.push(generateFileRow(uploadSetting!, "Index", "rvcIndex", ["index", "bin"]))
+            } else if (vcType == "MMVCv13") {
+                rows.push(generateFileRow(uploadSetting!, "Config", "mmvcv13Config", ["json"]))
+                rows.push(generateFileRow(uploadSetting!, "Model", "mmvcv13Model", ["pth", "onnx"]))
+            } else if (vcType == "MMVCv15") {
+                rows.push(generateFileRow(uploadSetting!, "Config", "mmvcv15Config", ["json"]))
+                rows.push(generateFileRow(uploadSetting!, "Model", "mmvcv15Model", ["pth", "onnx"]))
+            } else if (vcType == "so-vits-svc-40") {
+                rows.push(generateFileRow(uploadSetting!, "Config", "soVitsSvc40Config", ["json"]))
+                rows.push(generateFileRow(uploadSetting!, "Model", "soVitsSvc40Model", ["pth"]))
+                rows.push(generateFileRow(uploadSetting!, "Cluster", "soVitsSvc40Cluster", ["pth", "pt"]))
+            } else if (vcType == "DDSP-SVC") {
+                rows.push(generateFileRow(uploadSetting!, "Config", "ddspSvcModelConfig", ["yaml"], "model/"))
+                rows.push(generateFileRow(uploadSetting!, "Model", "ddspSvcModel", ["pth", "pt"], "model/"))
+                rows.push(generateFileRow(uploadSetting!, "Config(diff)", "ddspSvcDiffusionConfig", ["yaml"], "diff/"))
+                rows.push(generateFileRow(uploadSetting!, "Model(diff)", "ddspSvcDiffusion", ["pth", "pt"], "diff/"))
             }
             return rows
         }

client/lib/src/ServerConfigurator.ts

@@ -57,13 +57,13 @@ export class ServerConfigurator {
         return info
     }
 
-    uploadFile2 = async (file: File, onprogress: (progress: number, end: boolean) => void) => {
+    uploadFile2 = async (dir: string, file: File, onprogress: (progress: number, end: boolean) => void) => {
         const url = this.serverUrl + "/upload_file"
         onprogress(0, false)
         const size = 1024 * 1024;
         let index = 0; // index値
         const fileLength = file.size
-        const filename = file.name
+        const filename = dir + file.name
         const fileChunkNum = Math.ceil(fileLength / size)
 
         while (true) {

client/lib/src/VoiceChangerClient.ts

@@ -290,8 +290,8 @@ export class VoiceChangerClient {
     uploadFile = (buf: ArrayBuffer, filename: string, onprogress: (progress: number, end: boolean) => void) => {
         return this.configurator.uploadFile(buf, filename, onprogress)
     }
-    uploadFile2 = (file: File, onprogress: (progress: number, end: boolean) => void) => {
+    uploadFile2 = (dir: string, file: File, onprogress: (progress: number, end: boolean) => void) => {
-        return this.configurator.uploadFile2(file, onprogress)
+        return this.configurator.uploadFile2(dir, file, onprogress)
     }
     concatUploadedFile = (filename: string, chunkNum: number) => {
         return this.configurator.concatUploadedFile(filename, chunkNum)

client/lib/src/hooks/useServerSetting.ts

@@ -41,6 +41,7 @@ export type ModelFileKind = typeof ModelFileKind[keyof typeof ModelFileKind]
 export type ModelFile = {
     file: File,
     kind: ModelFileKind
+    dir: string
 }
 
 export type ModelUploadSetting = {
@@ -296,7 +297,7 @@ export const useServerSetting = (props: UseServerSettingProps): ServerSettingSta
             if (!props.voiceChangerClient) return
             console.log("uploading..1.", file)
             console.log("uploading..2.", file.name)
-            const num = await props.voiceChangerClient.uploadFile2(file, onprogress)
+            const num = await props.voiceChangerClient.uploadFile2(dir, file, onprogress)
             const res = await props.voiceChangerClient.concatUploadedFile(dir + file.name, num)
             console.log("uploaded", num, res)
         }
@@ -319,11 +320,11 @@ export const useServerSetting = (props: UseServerSettingProps): ServerSettingSta
                     const progOffset = 100 * i * progRate
                     await _uploadFile2(setting.files[i].file, (progress: number, _end: boolean) => {
                         setUploadProgress(progress * progRate + progOffset)
-                    })
+                    }, setting.files[i].dir)
                 }
             }
             const params: ModelUploadSettingForServer = {
-                ...setting, files: setting.files.map((f) => { return { name: f.file.name, kind: f.kind } })
+                ...setting, files: setting.files.map((f) => { return { name: f.file.name, kind: f.kind, dir: f.dir } })
             }
 
             const loadPromise = props.voiceChangerClient.loadModel(

server/data/ModelSlot.py

@@ -38,7 +38,72 @@ class RVCModelSlot(ModelSlot):
     iconFile: str = ""
 
 
-ModelSlots: TypeAlias = Union[ModelSlot, RVCModelSlot]
+@dataclass
+class MMVCv13ModelSlot(ModelSlot):
+    voiceChangerType: VoiceChangerType = "MMVCv13"
+    modelFile: str = ""
+    configFile: str = ""
+    srcId: int = 107
+    dstId: int = 100
+    isONNX: bool = False
+    samplingRate: int = 24000
+
+    name: str = ""
+    description: str = ""
+    iconFile: str = ""
+
+
+@dataclass
+class MMVCv15ModelSlot(ModelSlot):
+    voiceChangerType: VoiceChangerType = "MMVCv15"
+    modelFile: str = ""
+    configFile: str = ""
+    srcId: int = 0
+    dstId: int = 101
+    isONNX: bool = False
+    samplingRate: int = 24000
+
+    name: str = ""
+    description: str = ""
+    iconFile: str = ""
+
+
+@dataclass
+class SoVitsSvc40ModelSlot(ModelSlot):
+    voiceChangerType: VoiceChangerType = "so-vits-svc-40"
+    modelFile: str = ""
+    configFile: str = ""
+    clusterFile: str = ""
+    dstId: int = 0
+    isONNX: bool = False
+
+    name: str = ""
+    description: str = ""
+    credit: str = ""
+    termsOfUseUrl: str = ""
+    sampleId: str = ""
+    iconFile: str = ""
+
+
+@dataclass
+class DDSPSVCModelSlot(ModelSlot):
+    voiceChangerType: VoiceChangerType = "DDSP-SVC"
+    modelFile: str = ""
+    configFile: str = ""
+    diffModelFile: str = ""
+    diffConfigFile: str = ""
+    dstId: int = 0
+    isONNX: bool = False
+
+    name: str = ""
+    description: str = ""
+    credit: str = ""
+    termsOfUseUrl: str = ""
+    sampleId: str = ""
+    iconFile: str = ""
+
+
+ModelSlots: TypeAlias = Union[ModelSlot, RVCModelSlot, MMVCv13ModelSlot, MMVCv15ModelSlot, SoVitsSvc40ModelSlot, DDSPSVCModelSlot]
 
 
 def loadSlotInfo(model_dir: str, slotIndex: int) -> ModelSlots:
@@ -50,6 +115,14 @@ def loadSlotInfo(model_dir: str, slotIndex: int) -> ModelSlots:
     slotInfo = ModelSlot(**{k: v for k, v in jsonDict.items() if k in ModelSlot.__annotations__})
     if slotInfo.voiceChangerType == "RVC":
         return RVCModelSlot(**jsonDict)
+    elif slotInfo.voiceChangerType == "MMVCv13":
+        return MMVCv13ModelSlot(**jsonDict)
+    elif slotInfo.voiceChangerType == "MMVCv15":
+        return MMVCv15ModelSlot(**jsonDict)
+    elif slotInfo.voiceChangerType == "so-vits-svc-40":
+        return SoVitsSvc40ModelSlot(**jsonDict)
+    elif slotInfo.voiceChangerType == "DDSP-SVC":
+        return DDSPSVCModelSlot(**jsonDict)
     else:
         return ModelSlot()
 

server/voice_changer/DDSP_SVC/DDSP_SVC.py

@@ -3,6 +3,7 @@ import os
 from dataclasses import asdict
 import numpy as np
 import torch
+from data.ModelSlot import DDSPSVCModelSlot
 from voice_changer.DDSP_SVC.ModelSlot import ModelSlot
 
 from voice_changer.DDSP_SVC.deviceManager.DeviceManager import DeviceManager
@@ -21,7 +22,7 @@ from diffusion.infer_gt_mel import DiffGtMel  # type: ignore
 
 from voice_changer.utils.VoiceChangerModel import AudioInOut
 from voice_changer.utils.VoiceChangerParams import VoiceChangerParams
-from voice_changer.utils.LoadModelParams import LoadModelParams
+from voice_changer.utils.LoadModelParams import LoadModelParams, LoadModelParams2
 from voice_changer.DDSP_SVC.DDSP_SVCSetting import DDSP_SVCSettings
 from voice_changer.RVC.embedder.EmbedderManager import EmbedderManager
 
@@ -44,11 +45,7 @@ def phase_vocoder(a, b, fade_out, fade_in):
     deltaphase = deltaphase - 2 * np.pi * torch.floor(deltaphase / 2 / np.pi + 0.5)
     w = 2 * np.pi * torch.arange(n // 2 + 1).to(a) + deltaphase
     t = torch.arange(n).unsqueeze(-1).to(a) / n
-    result = (
+    result = a * (fade_out**2) + b * (fade_in**2) + torch.sum(absab * torch.cos(w * t + phia), -1) * fade_out * fade_in / n
-        a * (fade_out**2)
-        + b * (fade_in**2)
-        + torch.sum(absab * torch.cos(w * t + phia), -1) * fade_out * fade_in / n
-    )
     return result
 
 
@@ -102,9 +99,7 @@ class DDSP_SVC:
     def reloadModel(self):
         self.device = self.deviceManager.getDevice(self.settings.gpu)
         modelFile = self.settings.modelSlots[self.settings.modelSlotIndex].modelFile
-        diffusionFile = self.settings.modelSlots[
+        diffusionFile = self.settings.modelSlots[self.settings.modelSlotIndex].diffusionFile
-            self.settings.modelSlotIndex
-        ].diffusionFile
 
         self.svc_model = SvcDDSP()
         self.svc_model.setVCParams(self.params)
@@ -144,15 +139,11 @@ class DDSP_SVC:
         # newData = newData.astype(np.float32)
 
         if self.audio_buffer is not None:
-            self.audio_buffer = np.concatenate(
+            self.audio_buffer = np.concatenate([self.audio_buffer, newData], 0)  # 過去のデータに連結
-                [self.audio_buffer, newData], 0
-            )  # 過去のデータに連結
         else:
             self.audio_buffer = newData
 
-        convertSize = (
+        convertSize = inputSize + crossfadeSize + solaSearchFrame + self.settings.extraConvertSize
-            inputSize + crossfadeSize + solaSearchFrame + self.settings.extraConvertSize
-        )
 
         # if convertSize % self.hop_size != 0:  # モデルの出力のホップサイズで切り捨てが発生するので補う。
         #     convertSize = convertSize + (self.hop_size - (convertSize % self.hop_size))
@@ -187,8 +178,7 @@ class DDSP_SVC:
             f0_min=50,
             f0_max=1100,
             # safe_prefix_pad_length=0,  # TBD なにこれ？
-            safe_prefix_pad_length=self.settings.extraConvertSize
+            safe_prefix_pad_length=self.settings.extraConvertSize / self.svc_model.args.data.sampling_rate,
-            / self.svc_model.args.data.sampling_rate,
             diff_model=self.diff_model,
             diff_acc=self.settings.diffAcc,  # TBD なにこれ？
             diff_spk_id=self.settings.diffSpkId,
@@ -196,9 +186,7 @@ class DDSP_SVC:
             # diff_use_dpm=True if self.settings.useDiffDpm == 1 else False,  # TBD なにこれ？
             method=self.settings.diffMethod,
             k_step=self.settings.kStep,  # TBD なにこれ？
-            diff_silence=True
+            diff_silence=True if self.settings.useDiffSilence == 1 else False,  # TBD なにこれ？
-            if self.settings.useDiffSilence == 1
-            else False,  # TBD なにこれ？
         )
 
         return _audio.cpu().numpy() * 32768.0
@@ -210,9 +198,21 @@ class DDSP_SVC:
             audio = self._pyTorch_inference(data)
         return audio
 
-    # def destroy(self):
+    @classmethod
-    #     del self.net_g
+    def loadModel2(cls, props: LoadModelParams2):
-    #     del self.onnx_session
+        slotInfo: DDSPSVCModelSlot = DDSPSVCModelSlot()
+        for file in props.files:
+            if file.kind == "ddspSvcModelConfig":
+                slotInfo.configFile = file.name
+            elif file.kind == "ddspSvcModel":
+                slotInfo.modelFile = file.name
+            elif file.kind == "ddspSvcDiffusionConfig":
+                slotInfo.diffConfigFile = file.name
+            elif file.kind == "ddspSvcDiffusion":
+                slotInfo.diffModelFile = file.name
+        slotInfo.isONNX = slotInfo.modelFile.endswith(".onnx")
+        slotInfo.name = os.path.splitext(os.path.basename(slotInfo.modelFile))[0]
+        return slotInfo
 
     def __del__(self):
         del self.net_g

server/voice_changer/MMVCv13/MMVCv13.py

@@ -1,7 +1,8 @@
 import sys
 import os
+from data.ModelSlot import MMVCv13ModelSlot
 
-from voice_changer.utils.LoadModelParams import LoadModelParams
+from voice_changer.utils.LoadModelParams import LoadModelParams, LoadModelParams2
 from voice_changer.utils.VoiceChangerModel import AudioInOut
 
 if sys.platform.startswith("darwin"):
@@ -77,13 +78,7 @@ class MMVCv13:
 
         # PyTorchモデル生成
         if self.settings.pyTorchModelFile is not None:
-            self.net_g = SynthesizerTrn(
+            self.net_g = SynthesizerTrn(len(symbols), self.hps.data.filter_length // 2 + 1, self.hps.train.segment_size // self.hps.data.hop_length, n_speakers=self.hps.data.n_speakers, **self.hps.model)
-                len(symbols),
-                self.hps.data.filter_length // 2 + 1,
-                self.hps.train.segment_size // self.hps.data.hop_length,
-                n_speakers=self.hps.data.n_speakers,
-                **self.hps.model
-            )
             self.net_g.eval()
             load_checkpoint(self.settings.pyTorchModelFile, self.net_g, None)
 
@@ -154,9 +149,7 @@ class MMVCv13:
     def get_info(self):
         data = asdict(self.settings)
 
-        data["onnxExecutionProviders"] = (
+        data["onnxExecutionProviders"] = self.onnx_session.get_providers() if self.onnx_session is not None else []
-            self.onnx_session.get_providers() if self.onnx_session is not None else []
-        )
         files = ["configFile", "pyTorchModelFile", "onnxModelFile"]
         for f in files:
             if data[f] is not None and os.path.exists(data[f]):
@@ -193,9 +186,7 @@ class MMVCv13:
         newData = newData.astype(np.float32) / self.hps.data.max_wav_value
 
         if self.audio_buffer is not None:
-            self.audio_buffer = np.concatenate(
+            self.audio_buffer = np.concatenate([self.audio_buffer, newData], 0)  # 過去のデータに連結
-                [self.audio_buffer, newData], 0
-            )  # 過去のデータに連結
         else:
             self.audio_buffer = newData
 
@@ -204,9 +195,7 @@ class MMVCv13:
         # if convertSize < 8192:
         #     convertSize = 8192
         if convertSize % self.hps.data.hop_length != 0:  # モデルの出力のホップサイズで切り捨てが発生するので補う。
-            convertSize = convertSize + (
+            convertSize = convertSize + (self.hps.data.hop_length - (convertSize % self.hps.data.hop_length))
-                self.hps.data.hop_length - (convertSize % self.hps.data.hop_length)
-            )
 
         convertOffset = -1 * convertSize
         self.audio_buffer = self.audio_buffer[convertOffset:]  # 変換対象の部分だけ抽出
@@ -238,7 +227,9 @@ class MMVCv13:
                     "sid_src": sid_src.numpy(),
                     "sid_tgt": sid_tgt1.numpy(),
                 },
-            )[0][0, 0]
+            )[
+                0
+            ][0, 0]
             * self.hps.data.max_wav_value
         )
         return audio1
@@ -254,19 +245,10 @@ class MMVCv13:
             dev = torch.device("cuda", index=self.settings.gpu)
 
         with torch.no_grad():
-            x, x_lengths, spec, spec_lengths, y, y_lengths, sid_src = [
+            x, x_lengths, spec, spec_lengths, y, y_lengths, sid_src = [x.to(dev) for x in data]
-                x.to(dev) for x in data
-            ]
             sid_target = torch.LongTensor([self.settings.dstId]).to(dev)
 
-            audio1 = (
+            audio1 = self.net_g.to(dev).voice_conversion(spec, spec_lengths, sid_src=sid_src, sid_tgt=sid_target)[0, 0].data * self.hps.data.max_wav_value
-                self.net_g.to(dev)
-                .voice_conversion(
-                    spec, spec_lengths, sid_src=sid_src, sid_tgt=sid_target
-                )[0, 0]
-                .data
-                * self.hps.data.max_wav_value
-            )
             result = audio1.float().cpu().numpy()
 
         return result
@@ -278,6 +260,18 @@ class MMVCv13:
             audio = self._pyTorch_inference(data)
         return audio
 
+    @classmethod
+    def loadModel2(cls, props: LoadModelParams2):
+        slotInfo: MMVCv13ModelSlot = MMVCv13ModelSlot()
+        for file in props.files:
+            if file.kind == "mmvcv13Model":
+                slotInfo.modelFile = file.name
+            elif file.kind == "mmvcv13Config":
+                slotInfo.configFile = file.name
+        slotInfo.isONNX = slotInfo.modelFile.endswith(".onnx")
+        slotInfo.name = os.path.splitext(os.path.basename(slotInfo.modelFile))[0]
+        return slotInfo
+
     def __del__(self):
         del self.net_g
         del self.onnx_session

server/voice_changer/MMVCv15/MMVCv15.py

@@ -1,7 +1,8 @@
 import sys
 import os
+from data.ModelSlot import MMVCv15ModelSlot
 
-from voice_changer.utils.LoadModelParams import LoadModelParams
+from voice_changer.utils.LoadModelParams import LoadModelParams, LoadModelParams2
 from voice_changer.utils.VoiceChangerModel import AudioInOut
 
 if sys.platform.startswith("darwin"):
@@ -172,12 +173,7 @@ class MMVCv15:
     def get_info(self):
         data = asdict(self.settings)
 
-        data["onnxExecutionProviders"] = (
+        data["onnxExecutionProviders"] = self.onnx_session.get_providers() if self.settings.onnxModelFile != "" and self.settings.onnxModelFile is not None else []
-            self.onnx_session.get_providers()
-            if self.settings.onnxModelFile != ""
-            and self.settings.onnxModelFile is not None
-            else []
-        )
         files = ["configFile", "pyTorchModelFile", "onnxModelFile"]
         for f in files:
             if data[f] is not None and os.path.exists(data[f]):
@@ -195,9 +191,7 @@ class MMVCv15:
     def _get_f0(self, detector: str, newData: AudioInOut):
         audio_norm_np = newData.astype(np.float64)
         if detector == "dio":
-            _f0, _time = pw.dio(
+            _f0, _time = pw.dio(audio_norm_np, self.hps.data.sampling_rate, frame_period=5.5)
-                audio_norm_np, self.hps.data.sampling_rate, frame_period=5.5
-            )
             f0 = pw.stonemask(audio_norm_np, _f0, _time, self.hps.data.sampling_rate)
         else:
             f0, t = pw.harvest(
@@ -207,9 +201,7 @@ class MMVCv15:
                 f0_floor=71.0,
                 f0_ceil=1000.0,
             )
-        f0 = convert_continuos_f0(
+        f0 = convert_continuos_f0(f0, int(audio_norm_np.shape[0] / self.hps.data.hop_length))
-            f0, int(audio_norm_np.shape[0] / self.hps.data.hop_length)
-        )
         f0 = torch.from_numpy(f0.astype(np.float32))
         return f0
 
@@ -237,9 +229,7 @@ class MMVCv15:
         newData = newData.astype(np.float32) / self.hps.data.max_wav_value
 
         if self.audio_buffer is not None:
-            self.audio_buffer = np.concatenate(
+            self.audio_buffer = np.concatenate([self.audio_buffer, newData], 0)  # 過去のデータに連結
-                [self.audio_buffer, newData], 0
-            )  # 過去のデータに連結
         else:
             self.audio_buffer = newData
 
@@ -248,9 +238,7 @@ class MMVCv15:
         # if convertSize < 8192:
         #     convertSize = 8192
         if convertSize % self.hps.data.hop_length != 0:  # モデルの出力のホップサイズで切り捨てが発生するので補う。
-            convertSize = convertSize + (
+            convertSize = convertSize + (self.hps.data.hop_length - (convertSize % self.hps.data.hop_length))
-                self.hps.data.hop_length - (convertSize % self.hps.data.hop_length)
-            )
 
         # ONNX は固定長
         if self.settings.framework == "ONNX":
@@ -290,16 +278,15 @@ class MMVCv15:
                     "sid_src": sid_src.numpy(),
                     "sid_tgt": sid_tgt1.numpy(),
                 },
-            )[0][0, 0]
+            )[
+                0
+            ][0, 0]
             * self.hps.data.max_wav_value
         )
         return audio1
 
     def _pyTorch_inference(self, data):
-        if (
+        if self.settings.pyTorchModelFile == "" or self.settings.pyTorchModelFile is None:
-            self.settings.pyTorchModelFile == ""
-            or self.settings.pyTorchModelFile is None
-        ):
             print("[Voice Changer] No pyTorch session.")
             raise NoModeLoadedException("pytorch")
 
@@ -316,12 +303,7 @@ class MMVCv15:
             sid_src = sid_src.to(dev)
             sid_target = torch.LongTensor([self.settings.dstId]).to(dev)
 
-            audio1 = (
+            audio1 = self.net_g.to(dev).voice_conversion(spec, spec_lengths, f0, sid_src, sid_target)[0, 0].data * self.hps.data.max_wav_value
-                self.net_g.to(dev)
-                .voice_conversion(spec, spec_lengths, f0, sid_src, sid_target)[0, 0]
-                .data
-                * self.hps.data.max_wav_value
-            )
             result = audio1.float().cpu().numpy()
         return result
 
@@ -336,6 +318,18 @@ class MMVCv15:
             print(_e)
             raise ONNXInputArgumentException()
 
+    @classmethod
+    def loadModel2(cls, props: LoadModelParams2):
+        slotInfo: MMVCv15ModelSlot = MMVCv15ModelSlot()
+        for file in props.files:
+            if file.kind == "mmvcv15Model":
+                slotInfo.modelFile = file.name
+            elif file.kind == "mmvcv15Config":
+                slotInfo.configFile = file.name
+        slotInfo.isONNX = slotInfo.modelFile.endswith(".onnx")
+        slotInfo.name = os.path.splitext(os.path.basename(slotInfo.modelFile))[0]
+        return slotInfo
+
     def __del__(self):
         del self.net_g
         del self.onnx_session

server/voice_changer/SoVitsSvc40/SoVitsSvc40.py

@@ -1,7 +1,8 @@
 import sys
 import os
+from data.ModelSlot import SoVitsSvc40ModelSlot
 
-from voice_changer.utils.LoadModelParams import LoadModelParams
+from voice_changer.utils.LoadModelParams import LoadModelParams, LoadModelParams2
 from voice_changer.utils.VoiceChangerModel import AudioInOut
 from voice_changer.utils.VoiceChangerParams import VoiceChangerParams
 
@@ -97,11 +98,7 @@ class SoVitsSvc40:
             self.settings.pyTorchModelFile = modelFile
             self.settings.onnxModelFile = None
 
-        clusterTorchModel = (
+        clusterTorchModel = params["files"]["soVitsSvc40Cluster"] if "soVitsSvc40Cluster" in params["files"] else None
-            params["files"]["soVitsSvc40Cluster"]
-            if "soVitsSvc40Cluster" in params["files"]
-            else None
-        )
 
         content_vec_path = self.params.content_vec_500
         content_vec_onnx_path = self.params.content_vec_500_onnx
@@ -212,9 +209,7 @@ class SoVitsSvc40:
     def get_info(self):
         data = asdict(self.settings)
 
-        data["onnxExecutionProviders"] = (
+        data["onnxExecutionProviders"] = self.onnx_session.get_providers() if self.onnx_session is not None else []
-            self.onnx_session.get_providers() if self.onnx_session is not None else []
-        )
         files = ["configFile", "pyTorchModelFile", "onnxModelFile"]
         for f in files:
             if data[f] is not None and os.path.exists(data[f]):
@@ -246,9 +241,7 @@ class SoVitsSvc40:
             )
 
         if wav_44k.shape[0] % self.hps.data.hop_length != 0:
-            print(
+            print(f" !!! !!! !!! wav size not multiple of hopsize: {wav_44k.shape[0] / self.hps.data.hop_length}")
-                f" !!! !!! !!! wav size not multiple of hopsize: {wav_44k.shape[0] / self.hps.data.hop_length}"
-            )
 
         f0, uv = utils.interpolate_f0(f0)
         f0 = torch.FloatTensor(f0)
@@ -257,14 +250,10 @@ class SoVitsSvc40:
         f0 = f0.unsqueeze(0)
         uv = uv.unsqueeze(0)
 
-        wav16k_numpy = librosa.resample(
+        wav16k_numpy = librosa.resample(audio_buffer, orig_sr=self.hps.data.sampling_rate, target_sr=16000)
-            audio_buffer, orig_sr=self.hps.data.sampling_rate, target_sr=16000
-        )
         wav16k_tensor = torch.from_numpy(wav16k_numpy)
 
-        if (
+        if (self.settings.gpu < 0 or self.gpu_num == 0) or self.settings.framework == "ONNX":
-            self.settings.gpu < 0 or self.gpu_num == 0
-        ) or self.settings.framework == "ONNX":
             dev = torch.device("cpu")
         else:
             dev = torch.device("cuda", index=self.settings.gpu)
@@ -282,44 +271,27 @@ class SoVitsSvc40:
             if self.hps.model.ssl_dim == 768:
                 self.hubert_model = self.hubert_model.to(dev)
                 wav16k_tensor = wav16k_tensor.to(dev)
-                c = get_hubert_content_layer9(
+                c = get_hubert_content_layer9(self.hubert_model, wav_16k_tensor=wav16k_tensor)
-                    self.hubert_model, wav_16k_tensor=wav16k_tensor
-                )
             else:
                 self.hubert_model = self.hubert_model.to(dev)
                 wav16k_tensor = wav16k_tensor.to(dev)
-                c = utils.get_hubert_content(
+                c = utils.get_hubert_content(self.hubert_model, wav_16k_tensor=wav16k_tensor)
-                    self.hubert_model, wav_16k_tensor=wav16k_tensor
-                )
 
         uv = uv.to(dev)
         f0 = f0.to(dev)
 
         c = utils.repeat_expand_2d(c.squeeze(0), f0.shape[1])
 
-        if (
+        if self.settings.clusterInferRatio != 0 and hasattr(self, "cluster_model") and self.cluster_model is not None:
-            self.settings.clusterInferRatio != 0
+            speaker = [key for key, value in self.settings.speakers.items() if value == self.settings.dstId]
-            and hasattr(self, "cluster_model")
-            and self.cluster_model is not None
-        ):
-            speaker = [
-                key
-                for key, value in self.settings.speakers.items()
-                if value == self.settings.dstId
-            ]
             if len(speaker) != 1:
                 pass
                 # print("not only one speaker found.", speaker)
             else:
-                cluster_c = cluster.get_cluster_center_result(
+                cluster_c = cluster.get_cluster_center_result(self.cluster_model, c.cpu().numpy().T, speaker[0]).T
-                    self.cluster_model, c.cpu().numpy().T, speaker[0]
-                ).T
                 cluster_c = torch.FloatTensor(cluster_c).to(dev)
                 c = c.to(dev)
-                c = (
+                c = self.settings.clusterInferRatio * cluster_c + (1 - self.settings.clusterInferRatio) * c
-                    self.settings.clusterInferRatio * cluster_c
-                    + (1 - self.settings.clusterInferRatio) * c
-                )
 
         c = c.unsqueeze(0)
         return c, f0, uv
@@ -334,20 +306,14 @@ class SoVitsSvc40:
         newData = newData.astype(np.float32) / self.hps.data.max_wav_value
 
         if self.audio_buffer is not None:
-            self.audio_buffer = np.concatenate(
+            self.audio_buffer = np.concatenate([self.audio_buffer, newData], 0)  # 過去のデータに連結
-                [self.audio_buffer, newData], 0
-            )  # 過去のデータに連結
         else:
             self.audio_buffer = newData
 
-        convertSize = (
+        convertSize = inputSize + crossfadeSize + solaSearchFrame + self.settings.extraConvertSize
-            inputSize + crossfadeSize + solaSearchFrame + self.settings.extraConvertSize
-        )
 
         if convertSize % self.hps.data.hop_length != 0:  # モデルの出力のホップサイズで切り捨てが発生するので補う。
-            convertSize = convertSize + (
+            convertSize = convertSize + (self.hps.data.hop_length - (convertSize % self.hps.data.hop_length))
-                self.hps.data.hop_length - (convertSize % self.hps.data.hop_length)
-            )
 
         convertOffset = -1 * convertSize
         self.audio_buffer = self.audio_buffer[convertOffset:]  # 変換対象の部分だけ抽出
@@ -389,9 +355,7 @@ class SoVitsSvc40:
                     "f0": f0.astype(np.float32),
                     "uv": uv.astype(np.float32),
                     "g": sid_target.astype(np.int64),
-                    "noise_scale": np.array([self.settings.noiseScale]).astype(
+                    "noise_scale": np.array([self.settings.noiseScale]).astype(np.float32),
-                        np.float32
-                    ),
                     # "predict_f0": np.array([self.settings.dstId]).astype(np.int64),
                 },
             )[0][0, 0]
@@ -457,6 +421,20 @@ class SoVitsSvc40:
 
         return audio
 
+    @classmethod
+    def loadModel2(cls, props: LoadModelParams2):
+        slotInfo: SoVitsSvc40ModelSlot = SoVitsSvc40ModelSlot()
+        for file in props.files:
+            if file.kind == "soVitsSvc40Config":
+                slotInfo.configFile = file.name
+            elif file.kind == "soVitsSvc40Model":
+                slotInfo.modelFile = file.name
+            elif file.kind == "soVitsSvc40Cluster":
+                slotInfo.clusterFile = file.name
+        slotInfo.isONNX = slotInfo.modelFile.endswith(".onnx")
+        slotInfo.name = os.path.splitext(os.path.basename(slotInfo.modelFile))[0]
+        return slotInfo
+
     def __del__(self):
         del self.net_g
         del self.onnx_session

server/voice_changer/VoiceChangerManager.py

@@ -95,22 +95,48 @@ class VoiceChangerManager(ServerDeviceCallbacks):
             # Dataを展開
             params = LoadModelParams2(**paramDict)
             params.files = [LoadModelParamFile(**x) for x in paramDict["files"]]
+
             # ファイルをslotにコピー
             for file in params.files:
                 print("FILE", file)
-                srcPath = os.path.join(UPLOAD_DIR, file.name)
+                srcPath = os.path.join(UPLOAD_DIR, file.dir, file.name)
-                dstDir = os.path.join(self.params.model_dir, str(params.slot))
+                dstDir = os.path.join(
+                    self.params.model_dir,
+                    str(params.slot),
+                    file.dir,
+                )
                 dstPath = os.path.join(dstDir, file.name)
                 os.makedirs(dstDir, exist_ok=True)
                 print(f"move to {srcPath} -> {dstPath}")
                 shutil.move(srcPath, dstPath)
                 file.name = dstPath
+
             # メタデータ作成(各VCで定義)
             if params.voiceChangerType == "RVC":
                 from voice_changer.RVC.RVC import RVC  # 起動時にインポートするとパラメータが取れない。
 
                 slotInfo = RVC.loadModel2(params)
                 self.modelSlotManager.save_model_slot(params.slot, slotInfo)
+            elif params.voiceChangerType == "MMVCv13":
+                from voice_changer.MMVCv13.MMVCv13 import MMVCv13
+
+                slotInfo = MMVCv13.loadModel2(params)
+                self.modelSlotManager.save_model_slot(params.slot, slotInfo)
+            elif params.voiceChangerType == "MMVCv15":
+                from voice_changer.MMVCv15.MMVCv15 import MMVCv15
+
+                slotInfo = MMVCv15.loadModel2(params)
+                self.modelSlotManager.save_model_slot(params.slot, slotInfo)
+            elif params.voiceChangerType == "so-vits-svc-40":
+                from voice_changer.SoVitsSvc40.SoVitsSvc40 import SoVitsSvc40
+
+                slotInfo = SoVitsSvc40.loadModel2(params)
+                self.modelSlotManager.save_model_slot(params.slot, slotInfo)
+            elif params.voiceChangerType == "DDSP-SVC":
+                from voice_changer.DDSP_SVC.DDSP_SVC import DDSP_SVC
+
+                slotInfo = DDSP_SVC.loadModel2(params)
+                self.modelSlotManager.save_model_slot(params.slot, slotInfo)
             print("params", params)
 
         else:

server/voice_changer/utils/LoadModelParams.py

@@ -43,6 +43,7 @@ LoadModelParamFileKind: TypeAlias = Literal[
 class LoadModelParamFile:
     name: str
     kind: LoadModelParamFileKind
+    dir: str
 
 
 @dataclass