PublicMirrors/voice-changer
1
0
Fork 0
mirror of https://github.com/w-okada/voice-changer.git synced 2025-02-02 16:23:58 +03:00
Code Issues Actions Packages Projects Releases Wiki Activity

hubert/content vec uploader

Browse Source
This commit is contained in:
wataru 2023-03-13 21:07:35 +09:00
parent dd30c70f87
commit d5cc5c9e28
12 changed files with 1001 additions and 82 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
client/demo_so-vits-svc_40v2/dist/index.html vendored
View File

@ -1 +1,10 @@
<!doctype html><html style="width:100%;height:100%;overflow:hidden"><head><meta charset="utf-8"/><title>Voice Changer Client Demo</title><script defer="defer" src="index.js"></script></head><body style="width:100%;height:100%;margin:0"><div id="app" style="width:100%;height:100%"></div></body></html>
<!DOCTYPE html>
<html style="width: 100%; height: 100%; overflow: hidden">
<head>
<meta charset="utf-8" />
<title>Voice Changer Client Demo</title>
<script defer src="index.js"></script></head>
<body style="width: 100%; height: 100%; margin: 0px">
<div id="app" style="width: 100%; height: 100%"></div>
</body>
</html>

892
client/demo_so-vits-svc_40v2/dist/index.js vendored
View File

File diff suppressed because one or more lines are too long

31
client/demo_so-vits-svc_40v2/dist/index.js.LICENSE.txt vendored
View File

@ -1,31 +0,0 @@
/*! regenerator-runtime -- Copyright (c) 2014-present, Facebook, Inc. -- license (MIT): https://github.com/facebook/regenerator/blob/main/LICENSE */
/**
* @license React
* react-dom.production.min.js
*
* Copyright (c) Facebook, Inc. and its affiliates.
*
* This source code is licensed under the MIT license found in the
* LICENSE file in the root directory of this source tree.
*/
/**
* @license React
* react.production.min.js
*
* Copyright (c) Facebook, Inc. and its affiliates.
*
* This source code is licensed under the MIT license found in the
* LICENSE file in the root directory of this source tree.
*/
/**
* @license React
* scheduler.production.min.js
*
* Copyright (c) Facebook, Inc. and its affiliates.
*
* This source code is licensed under the MIT license found in the
* LICENSE file in the root directory of this source tree.
*/

48
client/demo_so-vits-svc_40v2/src/102_model_setting.tsx
View File

@ -64,26 +64,48 @@ export const useModelSettingArea = (): ServerSettingState => {
configFile: null
})
}
const onOnnxFileLoadClicked = async () => {
const onHubertFileLoadClicked = async () => {
const file = await fileSelector("")
if (file.name.endsWith(".onnx") == false) {
alert("モデルファイルの拡張子はonnxである必要があります。")
if (file.name.endsWith(".pth") == false) {
alert("モデルファイルの拡張子はpthである必要があります。")
return
}
appState.serverSetting.setFileUploadSetting({
...appState.serverSetting.fileUploadSetting,
onnxModel: {
hubertTorchModel: {
file: file
}
})
}
const onOnnxFileClearClicked = () => {
const onHubertFileClearClicked = () => {
appState.serverSetting.setFileUploadSetting({
...appState.serverSetting.fileUploadSetting,
onnxModel: null
hubertTorchModel: null
})
}
// const onOnnxFileLoadClicked = async () => {
// const file = await fileSelector("")
// if (file.name.endsWith(".onnx") == false) {
// alert("モデルファイルの拡張子はonnxである必要があります。")
// return
// }
// appState.serverSetting.setFileUploadSetting({
// ...appState.serverSetting.fileUploadSetting,
// onnxModel: {
// file: file
// }
// })
// }
// const onOnnxFileClearClicked = () => {
// appState.serverSetting.setFileUploadSetting({
// ...appState.serverSetting.fileUploadSetting,
// onnxModel: null
// })
// }
const onModelUploadClicked = async () => {
appState.serverSetting.loadModel()
}
@ -93,7 +115,9 @@ export const useModelSettingArea = (): ServerSettingState => {
const uploadButtonLabel = appState.serverSetting.isUploading ? "wait..." : "upload"
const configFilenameText = appState.serverSetting.fileUploadSetting.configFile?.filename || appState.serverSetting.fileUploadSetting.configFile?.file?.name || ""
const onnxModelFilenameText = appState.serverSetting.fileUploadSetting.onnxModel?.filename || appState.serverSetting.fileUploadSetting.onnxModel?.file?.name || ""
const hubertModelFilenameText = appState.serverSetting.fileUploadSetting.hubertTorchModel?.filename || appState.serverSetting.fileUploadSetting.hubertTorchModel?.file?.name || ""
// const onnxModelFilenameText = appState.serverSetting.fileUploadSetting.onnxModel?.filename || appState.serverSetting.fileUploadSetting.onnxModel?.file?.name || ""
const pyTorchFilenameText = appState.serverSetting.fileUploadSetting.pyTorchModel?.filename || appState.serverSetting.fileUploadSetting.pyTorchModel?.file?.name || ""
const uploadingStatus = appState.serverSetting.isUploading ?
@ -125,6 +149,16 @@ export const useModelSettingArea = (): ServerSettingState => {
<div className="body-button left-margin-1" onClick={onConfigFileClearClicked}>clear</div>
</div>
</div>
<div className="body-row split-3-3-4 left-padding-1 guided">
<div className="body-item-title left-padding-2">hubert(.pth)</div>
<div className="body-item-text">
<div>{hubertModelFilenameText}</div>
</div>
<div className="body-button-container">
<div className="body-button" onClick={onHubertFileLoadClicked}>select</div>
<div className="body-button left-margin-1" onClick={onHubertFileClearClicked}>clear</div>
</div>
</div>
{/* <div className="body-row split-3-3-4 left-padding-1 guided">
<div className="body-item-title left-padding-2">Onnx(.onnx)</div>
<div className="body-item-text">

4
client/lib/src/ServerConfigurator.ts
View File

@ -109,13 +109,15 @@ export class ServerConfigurator {
})
}
loadModel = async (configFilename: string, pyTorchModelFilename: string | null, onnxModelFilename: string | null) => {
loadModel = async (configFilename: string, pyTorchModelFilename: string | null, onnxModelFilename: string | null, hubertTorchModelFilename: string | null) => {
const url = this.serverUrl + "/load_model"
const info = new Promise<ServerInfo>(async (resolve) => {
const formData = new FormData();
formData.append("pyTorchModelFilename", pyTorchModelFilename || "-");
formData.append("onnxModelFilename", onnxModelFilename || "-");
formData.append("configFilename", configFilename);
formData.append("hubertTorchModelFilename", hubertTorchModelFilename || "-");
const request = new Request(url, {
method: 'POST',
body: formData,

4
client/lib/src/VoiceChangerClient.ts
View File

@ -243,8 +243,8 @@ export class VoiceChangerClient {
concatUploadedFile = (filename: string, chunkNum: number) => {
return this.configurator.concatUploadedFile(filename, chunkNum)
}
loadModel = (configFilename: string, pyTorchModelFilename: string | null, onnxModelFilename: string | null) => {
return this.configurator.loadModel(configFilename, pyTorchModelFilename, onnxModelFilename)
loadModel = (configFilename: string, pyTorchModelFilename: string | null, onnxModelFilename: string | null, hubertTorchModelFilename: string | null) => {
return this.configurator.loadModel(configFilename, pyTorchModelFilename, onnxModelFilename, hubertTorchModelFilename)
}
//## Worklet ##//

22
client/lib/src/hooks/useServerSetting.ts
View File

@ -14,12 +14,14 @@ export type FileUploadSetting = {
pyTorchModel: ModelData | null
onnxModel: ModelData | null
configFile: ModelData | null
hubertTorchModel: ModelData | null
}
const InitialFileUploadSetting: FileUploadSetting = {
pyTorchModel: null,
configFile: null,
onnxModel: null,
hubertTorchModel: null
}
export type UseServerSettingProps = {
@ -144,6 +146,10 @@ export const useServerSetting = (props: UseServerSettingProps): ServerSettingSta
alert("Configファイルを指定する必要があります。")
return
}
if (props.clientType == "so_vits_svc_40v2c" && !fileUploadSetting.hubertTorchModel) {
alert("content vecのファイルを指定する必要があります。")
return
}
if (!props.voiceChangerClient) return
setUploadProgress(0)
@ -162,9 +168,13 @@ export const useServerSetting = (props: UseServerSettingProps): ServerSettingSta
fileUploadSetting.configFile.data = await fileUploadSetting.configFile.file!.arrayBuffer()
fileUploadSetting.configFile.filename = await fileUploadSetting.configFile.file!.name
}
if (props.clientType == "so_vits_svc_40v2c" && !fileUploadSetting.hubertTorchModel!.data) {
fileUploadSetting.hubertTorchModel!.data = await fileUploadSetting.hubertTorchModel!.file!.arrayBuffer()
fileUploadSetting.hubertTorchModel!.filename = await fileUploadSetting.hubertTorchModel!.file!.name
}
// ファイルをサーバにアップロード
const models = [fileUploadSetting.onnxModel, fileUploadSetting.pyTorchModel].filter(x => { return x != null }) as ModelData[]
const models = [fileUploadSetting.onnxModel, fileUploadSetting.pyTorchModel, fileUploadSetting.hubertTorchModel].filter(x => { return x != null }) as ModelData[]
for (let i = 0; i < models.length; i++) {
const progRate = 1 / models.length
const progOffset = 100 * i * progRate
@ -178,13 +188,17 @@ export const useServerSetting = (props: UseServerSettingProps): ServerSettingSta
console.log(progress, end)
})
const loadPromise = props.voiceChangerClient.loadModel(fileUploadSetting.configFile.filename!, fileUploadSetting.pyTorchModel?.filename || null, fileUploadSetting.onnxModel?.filename || null)
const loadPromise = props.voiceChangerClient.loadModel(fileUploadSetting.configFile.filename!, fileUploadSetting.pyTorchModel?.filename || null, fileUploadSetting.onnxModel?.filename || null, fileUploadSetting.hubertTorchModel?.filename || null)
// サーバでロード中にキャッシュにセーブ
const saveData: FileUploadSetting = {
pyTorchModel: fileUploadSetting.pyTorchModel ? { data: fileUploadSetting.pyTorchModel.data, filename: fileUploadSetting.pyTorchModel.filename } : null,
onnxModel: fileUploadSetting.onnxModel ? { data: fileUploadSetting.onnxModel.data, filename: fileUploadSetting.onnxModel.filename } : null,
configFile: { data: fileUploadSetting.configFile.data, filename: fileUploadSetting.configFile.filename }
configFile: { data: fileUploadSetting.configFile.data, filename: fileUploadSetting.configFile.filename },
hubertTorchModel: fileUploadSetting.hubertTorchModel ? {
data: fileUploadSetting.hubertTorchModel.data, filename: fileUploadSetting.hubertTorchModel.filename
} : null
}
setItem(INDEXEDDB_KEY_MODEL_DATA, saveData)
@ -193,7 +207,7 @@ export const useServerSetting = (props: UseServerSettingProps): ServerSettingSta
setIsUploading(false)
reloadServerInfo()
}
}, [fileUploadSetting, props.voiceChangerClient])
}, [fileUploadSetting, props.voiceChangerClient, props.clientType])
const reloadServerInfo = useMemo(() => {
return async () => {

8
server/MMVCServerSIO.py
View File

@ -39,6 +39,7 @@ def setupArgParser():
default=False, help="run on colab")
parser.add_argument("--modelType", type=str,
default="MMVCv15", help="model type: MMVCv13, MMVCv15, so-vits-svc-40v2")
parser.add_argument("--hubert", type=str, help="path to hubert model")
return parser
@ -79,6 +80,7 @@ PORT = args.p
CONFIG = args.c
MODEL = args.m if args.m != None else None
ONNX_MODEL = args.o if args.o != None else None
HUBERT_MODEL = args.hubert if args.hubert != None else None
MODEL_TYPE = os.environ.get('MODEL_TYPE', None)
if MODEL_TYPE == None:
MODEL_TYPE = args.modelType
@ -100,7 +102,11 @@ if args.colab == True:
voiceChangerManager = VoiceChangerManager.get_instance()
if CONFIG and (MODEL or ONNX_MODEL):
voiceChangerManager.loadModel(CONFIG, MODEL, ONNX_MODEL)
if MODEL_TYPE == "MMVCv15" or MODEL_TYPE == "MMVCv13":
voiceChangerManager.loadModel(CONFIG, MODEL, ONNX_MODEL, None)
else:
voiceChangerManager.loadModel(CONFIG, MODEL, ONNX_MODEL, HUBERT_MODEL)
app_fastapi = MMVC_Rest.get_instance(voiceChangerManager)
app_socketio = MMVC_SocketIOApp.get_instance(app_fastapi, voiceChangerManager)

24
server/restapi/MMVC_Rest_Fileuploader.py
View File

@ -1,4 +1,5 @@
import os,shutil
import os
import shutil
from typing import Union
from fastapi import APIRouter
from fastapi.encoders import jsonable_encoder
@ -8,23 +9,24 @@ from fastapi import HTTPException, FastAPI, UploadFile, File, Form
from restapi.mods.FileUploader import upload_file, concat_file_chunks
from voice_changer.VoiceChangerManager import VoiceChangerManager
from const import MODEL_DIR ,UPLOAD_DIR
from const import MODEL_DIR, UPLOAD_DIR
os.makedirs(UPLOAD_DIR, exist_ok=True)
os.makedirs(MODEL_DIR, exist_ok=True)
class MMVC_Rest_Fileuploader:
def __init__(self, voiceChangerManager:VoiceChangerManager):
def __init__(self, voiceChangerManager: VoiceChangerManager):
self.voiceChangerManager = voiceChangerManager
self.router = APIRouter()
self.router.add_api_route("/info", self.get_info, methods=["GET"])
self.router.add_api_route("/upload_file", self.post_upload_file, methods=["POST"])
self.router.add_api_route("/concat_uploaded_file", self.post_concat_uploaded_file, methods=["POST"])
self.router.add_api_route("/update_setteings",self.post_update_setteings, methods=["POST"])
self.router.add_api_route("/update_setteings", self.post_update_setteings, methods=["POST"])
self.router.add_api_route("/load_model", self.post_load_model, methods=["POST"])
self.router.add_api_route("/load_model_for_train", self.post_load_model_for_train, methods=["POST"])
self.router.add_api_route("/extract_voices", self.post_extract_voices, methods=["POST"])
self.onnx_provider=""
self.onnx_provider = ""
def post_upload_file(self, file: UploadFile = File(...), filename: str = Form(...)):
res = upload_file(UPLOAD_DIR, file, filename)
@ -41,7 +43,7 @@ class MMVC_Rest_Fileuploader:
json_compatible_item_data = jsonable_encoder(info)
return JSONResponse(content=json_compatible_item_data)
def post_update_setteings(self, key:str=Form(...), val:Union[int, str, float]=Form(...)):
def post_update_setteings(self, key: str = Form(...), val: Union[int, str, float] = Form(...)):
print("post_update_setteings", key, val)
info = self.voiceChangerManager.update_setteings(key, val)
json_compatible_item_data = jsonable_encoder(info)
@ -51,19 +53,20 @@ class MMVC_Rest_Fileuploader:
self,
pyTorchModelFilename: str = Form(...),
onnxModelFilename: str = Form(...),
configFilename: str = Form(...)
configFilename: str = Form(...),
hubertTorchModelFilename: str = Form(...)
):
print("Hubert:", hubertTorchModelFilename)
pyTorchModelFilePath = os.path.join(UPLOAD_DIR, pyTorchModelFilename) if pyTorchModelFilename != "-" else None
onnxModelFilePath = os.path.join(UPLOAD_DIR, onnxModelFilename) if onnxModelFilename != "-" else None
configFilePath = os.path.join(UPLOAD_DIR, configFilename)
hubertTorchModelFilePath = os.path.join(UPLOAD_DIR, hubertTorchModelFilename) if hubertTorchModelFilename != "-" else None
info = self.voiceChangerManager.loadModel(configFilePath, pyTorchModelFilePath, onnxModelFilePath)
info = self.voiceChangerManager.loadModel(configFilePath, pyTorchModelFilePath, onnxModelFilePath, hubertTorchModelFilePath)
json_compatible_item_data = jsonable_encoder(info)
return JSONResponse(content=json_compatible_item_data)
# return {"load": f"{configFilePath}, {pyTorchModelFilePath}, {onnxModelFilePath}"}
def post_load_model_for_train(
self,
modelGFilename: str = Form(...),
@ -86,4 +89,3 @@ class MMVC_Rest_Fileuploader:
UPLOAD_DIR, zipFilename, zipFileChunkNum, UPLOAD_DIR)
shutil.unpack_archive(zipFilePath, "MMVC_Trainer/dataset/textful/")
return {"Zip file unpacked": f"{zipFilePath}"}

5
server/voice_changer/SoVitsSvc40v2/SoVitsSvc40v2.py
View File

@ -58,12 +58,13 @@ class SoVitsSvc40v2:
self.gpu_num = torch.cuda.device_count()
self.prevVol = 0
def loadModel(self, config: str, pyTorch_model_file: str = None, onnx_model_file: str = None):
def loadModel(self, config: str, pyTorch_model_file: str = None, onnx_model_file: str = None, hubertTorchModel: str = None):
self.settings.configFile = config
self.hps = utils.get_hparams_from_file(config)
# hubert model
vec_path = "hubert/checkpoint_best_legacy_500.pt"
# vec_path = "hubert/checkpoint_best_legacy_500.pt"
vec_path = hubertTorchModel
models, saved_cfg, task = checkpoint_utils.load_model_ensemble_and_task(
[vec_path],
suffix="",

24
server/voice_changer/VoiceChanger.py
View File

@ -48,15 +48,15 @@ class VoiceChanger():
self.currentCrossFadeOverlapSize = 0 # setting
self.crossfadeSize = 0 # calculated
modelType = getModelType()
print("[VoiceChanger] activate model type:", modelType)
if modelType == "MMVCv15":
self.modelType = getModelType()
print("[VoiceChanger] activate model type:", self.modelType)
if self.modelType == "MMVCv15":
from voice_changer.MMVCv15.MMVCv15 import MMVCv15
self.voiceChanger = MMVCv15()
elif modelType == "MMVCv13":
elif self.modelType == "MMVCv13":
from voice_changer.MMVCv13.MMVCv13 import MMVCv13
self.voiceChanger = MMVCv13()
elif modelType == "so-vits-svc-40v2":
elif self.modelType == "so-vits-svc-40v2":
from voice_changer.SoVitsSvc40v2.SoVitsSvc40v2 import SoVitsSvc40v2
self.voiceChanger = SoVitsSvc40v2()
@ -70,8 +70,11 @@ class VoiceChanger():
print(f"VoiceChanger Initialized (GPU_NUM:{self.gpu_num}, mps_enabled:{self.mps_enabled})")
def loadModel(self, config: str, pyTorch_model_file: str = None, onnx_model_file: str = None):
def loadModel(self, config: str, pyTorch_model_file: str = None, onnx_model_file: str = None, hubertTorchModel: str = None):
if self.modelType == "MMVCv15" or self.modelType == "MMVCv13":
return self.voiceChanger.loadModel(config, pyTorch_model_file, onnx_model_file)
else: # so-vits-svc-40v2
return self.voiceChanger.loadModel(config, pyTorch_model_file, onnx_model_file, hubertTorchModel)
def get_info(self):
data = asdict(self.settings)
@ -165,15 +168,6 @@ class VoiceChanger():
print_convert_processing(
f" Crossfade data size: crossfade:{crossfadeSize}, crossfade setting:{self.settings.crossFadeOverlapSize}, input size:{inputSize}")
# if convertSize < 8192:
# convertSize = 8192
# if convertSize % processing_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 + (processing_hop_length - (convertSize % processing_hop_length))
# overlapSize = min(self.settings.crossFadeOverlapSize, inputSize)
# cropRange = (-1 * (inputSize + overlapSize), -1 * overlapSize)
print_convert_processing(f" Convert data size of {inputSize + crossfadeSize} (+ extra size)")
print_convert_processing(f" will be cropped:{-1 * (inputSize + crossfadeSize)}, {-1 * (crossfadeSize)}")

4
server/voice_changer/VoiceChangerManager.py
View File

@ -10,8 +10,8 @@ class VoiceChangerManager():
cls._instance.voiceChanger = VoiceChanger()
return cls._instance
def loadModel(self, config, model, onnx_model):
info = self.voiceChanger.loadModel(config, model, onnx_model)
def loadModel(self, config, model, onnx_model, hubertTorchModel):
info = self.voiceChanger.loadModel(config, model, onnx_model, hubertTorchModel)
info["status"] = "OK"
return info