mirror of
https://github.com/w-okada/voice-changer.git
synced 2025-01-23 13:35:12 +03:00
add sample diffusion svc
This commit is contained in:
w-okada
parent
a0e890a438
commit
d3e49efd69
2
client/demo/dist/index.js
vendored
2
client/demo/dist/index.js
vendored
File diff suppressed because one or more lines are too long
26
client/demo/package-lock.json
generated
26
client/demo/package-lock.json
generated
@ -9,7 +9,7 @@
|
||||
"version": "1.0.0",
|
||||
"license": "ISC",
|
||||
"dependencies": {
|
||||
"@dannadori/voice-changer-client-js": "^1.0.161",
|
||||
"@dannadori/voice-changer-client-js": "^1.0.163",
|
||||
"@fortawesome/fontawesome-svg-core": "^6.4.0",
|
||||
"@fortawesome/free-brands-svg-icons": "^6.4.0",
|
||||
"@fortawesome/free-regular-svg-icons": "^6.4.0",
|
||||
@ -35,7 +35,7 @@
|
||||
"eslint": "^8.45.0",
|
||||
"eslint-config-prettier": "^8.8.0",
|
||||
"eslint-plugin-prettier": "^5.0.0",
|
||||
"eslint-plugin-react": "^7.32.2",
|
||||
"eslint-plugin-react": "^7.33.0",
|
||||
"eslint-webpack-plugin": "^4.0.1",
|
||||
"html-loader": "^4.2.0",
|
||||
"html-webpack-plugin": "^5.5.3",
|
||||
@ -48,7 +48,7 @@
|
||||
"ts-loader": "^9.4.4",
|
||||
"tsconfig-paths": "^4.2.0",
|
||||
"typescript": "^5.1.6",
|
||||
"webpack": "^5.88.1",
|
||||
"webpack": "^5.88.2",
|
||||
"webpack-cli": "^5.1.4",
|
||||
"webpack-dev-server": "^4.15.1"
|
||||
}
|
||||
@ -2708,9 +2708,9 @@
|
||||
}
|
||||
},
|
||||
"node_modules/@dannadori/voice-changer-client-js": {
|
||||
"version": "1.0.161",
|
||||
"resolved": "https://registry.npmjs.org/@dannadori/voice-changer-client-js/-/voice-changer-client-js-1.0.161.tgz",
|
||||
"integrity": "sha512-09TncUiM+NTnU3F8FnPuVqI/OQcQ6KEvUW0ErSuZtXLWIn6M7Ol+F+tvdlIa+5/fahsZle40Ujaz+5oq1g6uXQ==",
|
||||
"version": "1.0.163",
|
||||
"resolved": "https://registry.npmjs.org/@dannadori/voice-changer-client-js/-/voice-changer-client-js-1.0.163.tgz",
|
||||
"integrity": "sha512-ZSY+ppO96pEe6ouwBI9vCkOY7PjnWcxpGmW5ZyuHZTyhIZK4ArBrKyRs+hfo4oJIqVfkgvrv3RblFDV3BNnNow==",
|
||||
"dependencies": {
|
||||
"@types/readable-stream": "^2.3.15",
|
||||
"amazon-chime-sdk-js": "^3.15.0",
|
||||
@ -6232,9 +6232,9 @@
|
||||
}
|
||||
},
|
||||
"node_modules/eslint-plugin-react": {
|
||||
"version": "7.32.2",
|
||||
"resolved": "https://registry.npmjs.org/eslint-plugin-react/-/eslint-plugin-react-7.32.2.tgz",
|
||||
"integrity": "sha512-t2fBMa+XzonrrNkyVirzKlvn5RXzzPwRHtMvLAtVZrt8oxgnTQaYbU6SXTOO1mwQgp1y5+toMSKInnzGr0Knqg==",
|
||||
"version": "7.33.0",
|
||||
"resolved": "https://registry.npmjs.org/eslint-plugin-react/-/eslint-plugin-react-7.33.0.tgz",
|
||||
"integrity": "sha512-qewL/8P34WkY8jAqdQxsiL82pDUeT7nhs8IsuXgfgnsEloKCT4miAV9N9kGtx7/KM9NH/NCGUE7Edt9iGxLXFw==",
|
||||
"dev": true,
|
||||
"dependencies": {
|
||||
"array-includes": "^3.1.6",
|
||||
@ -6250,7 +6250,7 @@
|
||||
"object.values": "^1.1.6",
|
||||
"prop-types": "^15.8.1",
|
||||
"resolve": "^2.0.0-next.4",
|
||||
"semver": "^6.3.0",
|
||||
"semver": "^6.3.1",
|
||||
"string.prototype.matchall": "^4.0.8"
|
||||
},
|
||||
"engines": {
|
||||
@ -11748,9 +11748,9 @@
|
||||
}
|
||||
},
|
||||
"node_modules/webpack": {
|
||||
"version": "5.88.1",
|
||||
"resolved": "https://registry.npmjs.org/webpack/-/webpack-5.88.1.tgz",
|
||||
"integrity": "sha512-FROX3TxQnC/ox4N+3xQoWZzvGXSuscxR32rbzjpXgEzWudJFEJBpdlkkob2ylrv5yzzufD1zph1OoFsLtm6stQ==",
|
||||
"version": "5.88.2",
|
||||
"resolved": "https://registry.npmjs.org/webpack/-/webpack-5.88.2.tgz",
|
||||
"integrity": "sha512-JmcgNZ1iKj+aiR0OvTYtWQqJwq37Pf683dY9bVORwVbUrDhLhdn/PlO2sHsFHPkj7sHNQF3JwaAkp49V+Sq1tQ==",
|
||||
"dev": true,
|
||||
"dependencies": {
|
||||
"@types/eslint-scope": "^3.7.3",
|
||||
|
||||
@ -36,7 +36,7 @@
|
||||
"eslint": "^8.45.0",
|
||||
"eslint-config-prettier": "^8.8.0",
|
||||
"eslint-plugin-prettier": "^5.0.0",
|
||||
"eslint-plugin-react": "^7.32.2",
|
||||
"eslint-plugin-react": "^7.33.0",
|
||||
"eslint-webpack-plugin": "^4.0.1",
|
||||
"html-loader": "^4.2.0",
|
||||
"html-webpack-plugin": "^5.5.3",
|
||||
@ -54,7 +54,7 @@
|
||||
"webpack-dev-server": "^4.15.1"
|
||||
},
|
||||
"dependencies": {
|
||||
"@dannadori/voice-changer-client-js": "^1.0.162",
|
||||
"@dannadori/voice-changer-client-js": "^1.0.163",
|
||||
"@fortawesome/fontawesome-svg-core": "^6.4.0",
|
||||
"@fortawesome/free-brands-svg-icons": "^6.4.0",
|
||||
"@fortawesome/free-regular-svg-icons": "^6.4.0",
|
||||
|
||||
@ -223,7 +223,7 @@ export const MainScreen = (props: MainScreenProps) => {
|
||||
iconArea = generateIconArea(index, slotInfo.iconFile, true);
|
||||
nameRow = generateNameRow(index, slotInfo.name, slotInfo.termsOfUseUrl);
|
||||
fileRows.push(generateFileRow("model", slotInfo.modelFile));
|
||||
infoRow = generateInfoRow(`tune:${slotInfo.defaultTune},ks(max):${slotInfo.kStepMax},ks:${slotInfo.defaultKstep}, sp:${slotInfo.defaultSpeedup}, l:${slotInfo.nLayers},${slotInfo.nnLayers},`);
|
||||
infoRow = generateInfoRow(`tune:${slotInfo.defaultTune},mks:${slotInfo.kStepMax},ks:${slotInfo.defaultKstep}, sp:${slotInfo.defaultSpeedup}, l:${slotInfo.nLayers},${slotInfo.nnLayers},`);
|
||||
} else {
|
||||
iconArea = generateIconArea(index, "/assets/icons/blank.png", false);
|
||||
nameRow = generateNameRow(index, "", "");
|
||||
|
||||
@ -1,78 +1,94 @@
|
||||
import React, { useMemo, useState } from "react";
|
||||
import { useAppState } from "../../001_provider/001_AppStateProvider";
|
||||
import { ModelUploadSetting } from "@dannadori/voice-changer-client-js";
|
||||
import { DiffusionSVCSampleModel, ModelUploadSetting, RVCSampleModel, VoiceChangerType } from "@dannadori/voice-changer-client-js";
|
||||
import { useMessageBuilder } from "../../hooks/useMessageBuilder";
|
||||
import { ModelSlotManagerDialogScreen } from "./904_ModelSlotManagerDialog";
|
||||
|
||||
export type SampleDownloaderScreenProps = {
|
||||
screen: ModelSlotManagerDialogScreen
|
||||
targetIndex: number
|
||||
close: () => void
|
||||
backToSlotManager: () => void
|
||||
|
||||
}
|
||||
screen: ModelSlotManagerDialogScreen;
|
||||
targetIndex: number;
|
||||
close: () => void;
|
||||
backToSlotManager: () => void;
|
||||
};
|
||||
|
||||
export const SampleDownloaderScreen = (props: SampleDownloaderScreenProps) => {
|
||||
const { serverSetting } = useAppState()
|
||||
const [lang, setLang] = useState<string>("All")
|
||||
const messageBuilderState = useMessageBuilder()
|
||||
const { serverSetting } = useAppState();
|
||||
const [lang, setLang] = useState<string>("All");
|
||||
const messageBuilderState = useMessageBuilder();
|
||||
useMemo(() => {
|
||||
messageBuilderState.setMessage(__filename, "header_message", { "ja": "サンプルをダウンロードしてください. 対象：", "en": "Download Sample for" })
|
||||
messageBuilderState.setMessage(__filename, "lang", { "ja": "言語", "en": "Lang" })
|
||||
messageBuilderState.setMessage(__filename, "back", { "ja": "戻る", "en": "back" })
|
||||
messageBuilderState.setMessage(__filename, "terms_of_use", { "ja": "利用規約", "en": "terms of use" })
|
||||
messageBuilderState.setMessage(__filename, "download", { "ja": "ダウンロード", "en": "download" })
|
||||
}, [])
|
||||
|
||||
|
||||
|
||||
messageBuilderState.setMessage(__filename, "header_message", { ja: "サンプルをダウンロードしてください. 対象：", en: "Download Sample for" });
|
||||
messageBuilderState.setMessage(__filename, "lang", { ja: "言語", en: "Lang" });
|
||||
messageBuilderState.setMessage(__filename, "back", { ja: "戻る", en: "back" });
|
||||
messageBuilderState.setMessage(__filename, "terms_of_use", { ja: "利用規約", en: "terms of use" });
|
||||
messageBuilderState.setMessage(__filename, "download", { ja: "ダウンロード", en: "download" });
|
||||
}, []);
|
||||
|
||||
/////////////////////////////////////////
|
||||
// Sample Downloader
|
||||
/////////////////////////////////////////
|
||||
const screen = useMemo(() => {
|
||||
if (props.screen != "SampleDownloader") {
|
||||
return <></>
|
||||
return <></>;
|
||||
}
|
||||
if (!serverSetting.serverSetting.modelSlots) {
|
||||
return <></>
|
||||
return <></>;
|
||||
}
|
||||
|
||||
const langs = serverSetting.serverSetting.sampleModels.reduce((prev, cur) => {
|
||||
const langs = serverSetting.serverSetting.sampleModels.reduce(
|
||||
(prev, cur) => {
|
||||
if (prev.includes(cur.lang) == false) {
|
||||
prev.push(cur.lang)
|
||||
prev.push(cur.lang);
|
||||
}
|
||||
return prev
|
||||
}, ["All"] as string[])
|
||||
const langOptions = (
|
||||
langs.map(x => {
|
||||
return <option key={x} value={x}>{x}</option>
|
||||
})
|
||||
)
|
||||
|
||||
const onDownloadSampleClicked = async (id: string) => {
|
||||
return prev;
|
||||
},
|
||||
["All"] as string[],
|
||||
);
|
||||
const langOptions = langs.map((x) => {
|
||||
return (
|
||||
<option key={x} value={x}>
|
||||
{x}
|
||||
</option>
|
||||
);
|
||||
});
|
||||
const onDownloadSampleClicked = async (id: string, voiceChangerType: VoiceChangerType, params: any) => {
|
||||
const uploadParams: ModelUploadSetting = {
|
||||
voiceChangerType: "RVC",
|
||||
voiceChangerType: voiceChangerType,
|
||||
slot: props.targetIndex,
|
||||
isSampleMode: true,
|
||||
sampleId: id,
|
||||
files: [],
|
||||
params: {
|
||||
rvcIndexDownload: true
|
||||
}
|
||||
}
|
||||
params: params,
|
||||
};
|
||||
try {
|
||||
await serverSetting.uploadModel(uploadParams)
|
||||
await serverSetting.uploadModel(uploadParams);
|
||||
} catch (e) {
|
||||
alert(e)
|
||||
alert(e);
|
||||
}
|
||||
props.backToSlotManager()
|
||||
props.backToSlotManager();
|
||||
// setMode("localFile")
|
||||
}
|
||||
const options = (
|
||||
serverSetting.serverSetting.sampleModels.filter(x => { return lang == "All" ? true : x.lang == lang }).map((x, index) => {
|
||||
const termOfUseUrlLink = x.termsOfUseUrl && x.termsOfUseUrl.length > 0 ? <a href={x.termsOfUseUrl} target="_blank" rel="noopener noreferrer" className="body-item-text-small">[{messageBuilderState.getMessage(__filename, "terms_of_use")}]</a> : <></>
|
||||
};
|
||||
const options = serverSetting.serverSetting.sampleModels
|
||||
.filter((x) => {
|
||||
return lang == "All" ? true : x.lang == lang;
|
||||
})
|
||||
.map((x, index) => {
|
||||
const termOfUseUrlLink =
|
||||
x.termsOfUseUrl && x.termsOfUseUrl.length > 0 ? (
|
||||
<a href={x.termsOfUseUrl} target="_blank" rel="noopener noreferrer" className="body-item-text-small">
|
||||
[{messageBuilderState.getMessage(__filename, "terms_of_use")}]
|
||||
</a>
|
||||
) : (
|
||||
<></>
|
||||
);
|
||||
|
||||
let info = ``;
|
||||
if (x.voiceChangerType == "RVC") {
|
||||
const sampleInfo = x as RVCSampleModel;
|
||||
info = `type:${sampleInfo.modelType}, f0:${sampleInfo.f0 ? "f0" : "nof0"}, sr:${sampleInfo.sampleRate}`;
|
||||
} else if (x.voiceChangerType == "Diffusion-SVC") {
|
||||
const sampleInfo = x as DiffusionSVCSampleModel;
|
||||
info = `native_l:${sampleInfo.numOfNativeLayers}, diff_l:${sampleInfo.numOfDiffLayers}, max_kstep:${sampleInfo.maxKStep}`;
|
||||
}
|
||||
return (
|
||||
<div key={index} className="model-slot">
|
||||
<img src={x.icon} className="model-slot-icon"></img>
|
||||
@ -82,21 +98,36 @@ export const SampleDownloaderScreen = (props: SampleDownloaderScreenProps) => {
|
||||
<div className="model-slot-detail-row-value">{x.name}</div>
|
||||
<div className="">{termOfUseUrlLink}</div>
|
||||
</div>
|
||||
<div className="model-slot-detail-row">
|
||||
<div className="model-slot-detail-row-label">VC Type: </div>
|
||||
<div className="model-slot-detail-row-value">{x.voiceChangerType}</div>
|
||||
<div className=""></div>
|
||||
</div>
|
||||
<div className="model-slot-detail-row">
|
||||
<div className="model-slot-detail-row-label">info: </div>
|
||||
<div className="model-slot-detail-row-value">{x.modelType},{x.f0 ? "f0" : "nof0"},{x.sampleRate}</div>
|
||||
<div className="model-slot-detail-row-value">{info}</div>
|
||||
<div className=""></div>
|
||||
</div>
|
||||
</div>
|
||||
<div className="model-slot-buttons">
|
||||
<div className="model-slot-button" onClick={() => { onDownloadSampleClicked(x.id) }}>
|
||||
<div
|
||||
className="model-slot-button"
|
||||
onClick={() => {
|
||||
if (x.voiceChangerType == "RVC") {
|
||||
onDownloadSampleClicked(x.id, x.voiceChangerType, {
|
||||
rvcIndexDownload: true,
|
||||
});
|
||||
} else if (x.voiceChangerType == "Diffusion-SVC") {
|
||||
onDownloadSampleClicked(x.id, x.voiceChangerType, {});
|
||||
}
|
||||
}}
|
||||
>
|
||||
{messageBuilderState.getMessage(__filename, "download")}
|
||||
</div>
|
||||
</div>
|
||||
</div>
|
||||
)
|
||||
})
|
||||
)
|
||||
);
|
||||
});
|
||||
|
||||
return (
|
||||
<div className="dialog-frame">
|
||||
@ -104,32 +135,32 @@ export const SampleDownloaderScreen = (props: SampleDownloaderScreenProps) => {
|
||||
<div className="dialog-fixed-size-content">
|
||||
<div className="model-slot-header">
|
||||
{messageBuilderState.getMessage(__filename, "header_message")} Slot[{props.targetIndex}]
|
||||
<span onClick={() => { props.backToSlotManager() }} className="model-slot-header-button">
|
||||
<span
|
||||
onClick={() => {
|
||||
props.backToSlotManager();
|
||||
}}
|
||||
className="model-slot-header-button"
|
||||
>
|
||||
<<{messageBuilderState.getMessage(__filename, "back")}
|
||||
</span>
|
||||
</div>
|
||||
<div>{messageBuilderState.getMessage(__filename, "lang")}:
|
||||
<select value={lang} onChange={(e) => { setLang(e.target.value) }}>
|
||||
<div>
|
||||
{messageBuilderState.getMessage(__filename, "lang")}:
|
||||
<select
|
||||
value={lang}
|
||||
onChange={(e) => {
|
||||
setLang(e.target.value);
|
||||
}}
|
||||
>
|
||||
{langOptions}
|
||||
</select>
|
||||
</div>
|
||||
|
||||
<div className="model-slot-container">
|
||||
{options}
|
||||
<div className="model-slot-container">{options}</div>
|
||||
</div>
|
||||
</div>
|
||||
</div>
|
||||
)
|
||||
|
||||
}, [
|
||||
props.screen,
|
||||
props.targetIndex,
|
||||
lang,
|
||||
serverSetting.serverSetting
|
||||
])
|
||||
|
||||
|
||||
);
|
||||
}, [props.screen, props.targetIndex, lang, serverSetting.serverSetting]);
|
||||
|
||||
return screen;
|
||||
|
||||
};
|
||||
|
||||
4
client/lib/package-lock.json
generated
4
client/lib/package-lock.json
generated
@ -1,12 +1,12 @@
|
||||
{
|
||||
"name": "@dannadori/voice-changer-client-js",
|
||||
"version": "1.0.162",
|
||||
"version": "1.0.163",
|
||||
"lockfileVersion": 3,
|
||||
"requires": true,
|
||||
"packages": {
|
||||
"": {
|
||||
"name": "@dannadori/voice-changer-client-js",
|
||||
"version": "1.0.162",
|
||||
"version": "1.0.163",
|
||||
"license": "ISC",
|
||||
"dependencies": {
|
||||
"@types/readable-stream": "^2.3.15",
|
||||
|
||||
@ -1,6 +1,6 @@
|
||||
{
|
||||
"name": "@dannadori/voice-changer-client-js",
|
||||
"version": "1.0.162",
|
||||
"version": "1.0.163",
|
||||
"description": "",
|
||||
"main": "dist/index.js",
|
||||
"directories": {
|
||||
|
||||
@ -296,7 +296,7 @@ export type ServerInfo = VoiceChangerServerSetting & {
|
||||
modelSlots: ModelSlotUnion[]
|
||||
serverAudioInputDevices: ServerAudioDevice[]
|
||||
serverAudioOutputDevices: ServerAudioDevice[]
|
||||
sampleModels: RVCSampleModel[]
|
||||
sampleModels: (RVCSampleModel|DiffusionSVCSampleModel)[]
|
||||
gpus: {
|
||||
id: number,
|
||||
name: string,
|
||||
@ -306,24 +306,37 @@ export type ServerInfo = VoiceChangerServerSetting & {
|
||||
|
||||
}
|
||||
|
||||
|
||||
export type RVCSampleModel = {
|
||||
export type SampleModel = {
|
||||
id: string
|
||||
name: string
|
||||
modelUrl: string
|
||||
indexUrl: string
|
||||
featureUrl: string
|
||||
termsOfUseUrl: string
|
||||
credit: string
|
||||
description: string
|
||||
voiceChangerType: VoiceChangerType
|
||||
lang: string
|
||||
tag: string[]
|
||||
name: string
|
||||
modelUrl: string
|
||||
termsOfUseUrl: string
|
||||
icon: string
|
||||
f0: boolean
|
||||
credit: string
|
||||
description: string
|
||||
sampleRate: number
|
||||
modelType: string
|
||||
f0: boolean
|
||||
}
|
||||
|
||||
export type RVCSampleModel =SampleModel & {
|
||||
indexUrl: string
|
||||
featureUrl: string
|
||||
}
|
||||
|
||||
|
||||
export type DiffusionSVCSampleModel =SampleModel & {
|
||||
numOfDiffLayers: number
|
||||
numOfNativeLayers: number
|
||||
maxKStep: number
|
||||
}
|
||||
|
||||
|
||||
|
||||
|
||||
export const DefaultServerSetting: ServerInfo = {
|
||||
// VC Common
|
||||
inputSampleRate: 48000,
|
||||
|
||||
@ -101,11 +101,14 @@ def getSampleJsonAndModelIds(mode: RVCSampleMode):
|
||||
# "https://huggingface.co/wok000/vcclient_model/raw/main/samples_0002.json",
|
||||
"https://huggingface.co/wok000/vcclient_model/raw/main/samples_0003_t2.json",
|
||||
"https://huggingface.co/wok000/vcclient_model/raw/main/samples_0003_o2.json",
|
||||
"https://huggingface.co/wok000/vcclient_model/raw/main/samples_0003_d2.json",
|
||||
], [
|
||||
("Tsukuyomi-chan_o", {"useIndex": False}),
|
||||
("Amitaro_o", {"useIndex": False}),
|
||||
("KikotoMahiro_o", {"useIndex": False}),
|
||||
("TokinaShigure_o", {"useIndex": True}),
|
||||
("diffusion_combo_spk5_nl3_dl20_k50_500ep", {}),
|
||||
("diffusion_combo_spk5_nl3_dl1_k50_500ep", {}),
|
||||
]
|
||||
elif mode == "testOfficial":
|
||||
return [
|
||||
|
||||
@ -11,6 +11,7 @@ class ModelSample:
|
||||
|
||||
@dataclass
|
||||
class RVCModelSample(ModelSample):
|
||||
id: str = ""
|
||||
voiceChangerType: VoiceChangerType = "RVC"
|
||||
lang: str = ""
|
||||
tag: list[str] = field(default_factory=lambda: [])
|
||||
@ -27,11 +28,34 @@ class RVCModelSample(ModelSample):
|
||||
f0: bool = True
|
||||
|
||||
|
||||
ModelSamples: TypeAlias = Union[ModelSample, RVCModelSample]
|
||||
@dataclass
|
||||
class DiffusionSVCModelSample(ModelSample):
|
||||
id: str = ""
|
||||
voiceChangerType: VoiceChangerType = "Diffusion-SVC"
|
||||
lang: str = ""
|
||||
tag: list[str] = field(default_factory=lambda: [])
|
||||
name: str = ""
|
||||
modelUrl: str = ""
|
||||
termsOfUseUrl: str = ""
|
||||
icon: str = ""
|
||||
credit: str = ""
|
||||
description: str = ""
|
||||
|
||||
sampleRate: int = 48000
|
||||
modelType: str = ""
|
||||
f0: bool = True
|
||||
numOfDiffLayers: int = 20
|
||||
numOfNativeLayers: int = 3
|
||||
maxKStep: int = 50
|
||||
|
||||
|
||||
ModelSamples: TypeAlias = Union[ModelSample, RVCModelSample, DiffusionSVCModelSample]
|
||||
|
||||
|
||||
def generateModelSample(params: Any) -> ModelSamples:
|
||||
if params["voiceChangerType"] == "RVC":
|
||||
return RVCModelSample(**params)
|
||||
return RVCModelSample(**{k: v for k, v in params.items() if k in RVCModelSample.__annotations__})
|
||||
elif params["voiceChangerType"] == "Diffusion-SVC":
|
||||
return DiffusionSVCModelSample(**{k: v for k, v in params.items() if k in DiffusionSVCModelSample.__annotations__})
|
||||
else:
|
||||
return ModelSample(**{k: v for k, v in params.items() if k in ModelSample.__annotations__})
|
||||
|
||||
@ -5,21 +5,29 @@ from typing import Any, Tuple
|
||||
|
||||
from const import RVCSampleMode, getSampleJsonAndModelIds
|
||||
from data.ModelSample import ModelSamples, generateModelSample
|
||||
from data.ModelSlot import RVCModelSlot
|
||||
from data.ModelSlot import DiffusionSVCModelSlot, ModelSlot, RVCModelSlot
|
||||
from voice_changer.DiffusionSVC.DiffusionSVCModelSlotGenerator import DiffusionSVCModelSlotGenerator
|
||||
from voice_changer.ModelSlotManager import ModelSlotManager
|
||||
from voice_changer.RVC.RVCModelSlotGenerator import RVCModelSlotGenerator
|
||||
from downloader.Downloader import download, download_no_tqdm
|
||||
|
||||
|
||||
def downloadInitialSamples(mode: RVCSampleMode, model_dir: str):
|
||||
print("1111111111111")
|
||||
sampleJsonUrls, sampleModels = getSampleJsonAndModelIds(mode)
|
||||
print("11111111111112")
|
||||
sampleJsons = _downloadSampleJsons(sampleJsonUrls)
|
||||
print("11111111111113")
|
||||
if os.path.exists(model_dir):
|
||||
print("[Voice Changer] model_dir is already exists. skip download samples.")
|
||||
return
|
||||
print("11111111111114")
|
||||
samples = _generateSampleList(sampleJsons)
|
||||
print("11111111111115")
|
||||
slotIndex = list(range(len(sampleModels)))
|
||||
print("11111111111116")
|
||||
_downloadSamples(samples, sampleModels, model_dir, slotIndex)
|
||||
print("11111111111117")
|
||||
|
||||
pass
|
||||
|
||||
@ -90,6 +98,7 @@ def _downloadSamples(samples: list[ModelSamples], sampleModelIds: list[Tuple[str
|
||||
|
||||
# 検出されたら、、、
|
||||
slotDir = os.path.join(model_dir, str(targetSlotIndex))
|
||||
slotInfo: ModelSlot = ModelSlot()
|
||||
if sample.voiceChangerType == "RVC":
|
||||
slotInfo: RVCModelSlot = RVCModelSlot()
|
||||
|
||||
@ -148,6 +157,50 @@ def _downloadSamples(samples: list[ModelSamples], sampleModelIds: list[Tuple[str
|
||||
slotInfo.defaultProtect = 0.5
|
||||
slotInfo.isONNX = slotInfo.modelFile.endswith(".onnx")
|
||||
modelSlotManager.save_model_slot(targetSlotIndex, slotInfo)
|
||||
elif sample.voiceChangerType == "Diffusion-SVC":
|
||||
slotInfo: DiffusionSVCModelSlot = DiffusionSVCModelSlot()
|
||||
|
||||
os.makedirs(slotDir, exist_ok=True)
|
||||
modelFilePath = os.path.join(
|
||||
slotDir,
|
||||
os.path.basename(sample.modelUrl),
|
||||
)
|
||||
downloadParams.append(
|
||||
{
|
||||
"url": sample.modelUrl,
|
||||
"saveTo": modelFilePath,
|
||||
"position": line_num,
|
||||
}
|
||||
)
|
||||
slotInfo.modelFile = modelFilePath
|
||||
line_num += 1
|
||||
|
||||
if hasattr(sample, "icon") and sample.icon != "":
|
||||
iconPath = os.path.join(
|
||||
slotDir,
|
||||
os.path.basename(sample.icon),
|
||||
)
|
||||
downloadParams.append(
|
||||
{
|
||||
"url": sample.icon,
|
||||
"saveTo": iconPath,
|
||||
"position": line_num,
|
||||
}
|
||||
)
|
||||
slotInfo.iconFile = iconPath
|
||||
line_num += 1
|
||||
|
||||
slotInfo.sampleId = sample.id
|
||||
slotInfo.credit = sample.credit
|
||||
slotInfo.description = sample.description
|
||||
slotInfo.name = sample.name
|
||||
slotInfo.termsOfUseUrl = sample.termsOfUseUrl
|
||||
slotInfo.defaultTune = 0
|
||||
slotInfo.defaultKstep = 0
|
||||
slotInfo.defaultSpeedup = 0
|
||||
slotInfo.kStepMax = 0
|
||||
slotInfo.isONNX = slotInfo.modelFile.endswith(".onnx")
|
||||
modelSlotManager.save_model_slot(targetSlotIndex, slotInfo)
|
||||
else:
|
||||
print(f"[Voice Changer] {sample.voiceChangerType} is not supported.")
|
||||
|
||||
@ -171,3 +224,9 @@ def _downloadSamples(samples: list[ModelSamples], sampleModelIds: list[Tuple[str
|
||||
slotInfo = RVCModelSlotGenerator._setInfoByPytorch(slotInfo)
|
||||
|
||||
modelSlotManager.save_model_slot(targetSlotIndex, slotInfo)
|
||||
elif slotInfo.voiceChangerType == "Diffusion-SVC":
|
||||
if slotInfo.isONNX:
|
||||
pass
|
||||
else:
|
||||
slotInfo = DiffusionSVCModelSlotGenerator._setInfoByPytorch(slotInfo)
|
||||
modelSlotManager.save_model_slot(targetSlotIndex, slotInfo)
|
||||
|
||||
76
server/samples_0003_d2.json
Normal file
76
server/samples_0003_d2.json
Normal file
@ -0,0 +1,76 @@
|
||||
{
|
||||
"Diffusion-SVC": [
|
||||
{
|
||||
"id": "diffusion_combo_spk5_nl3_dl20_k50_500ep",
|
||||
"voiceChangerType": "Diffusion-SVC",
|
||||
"lang": "ja-JP",
|
||||
"tag": ["diffsuion-svc", "torch"],
|
||||
"name": "初期メン(3,20)",
|
||||
"modelUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/diffusion-svc/combo_spk5_nl3_dl20_k50_500ep.ptc.bin",
|
||||
"termsOfUseUrl": "https://huggingface.co/wok000/vcclient_model/raw/main/diffusion-svc/term_of_use.txt",
|
||||
"icon": "https://huggingface.co/wok000/vcclient_model/resolve/main/diffusion-svc/diffusion-svc.png",
|
||||
"credit": "あみたろ,つくよみちゃん,黄琴海月,黄琴まひろ,刻鳴時雨",
|
||||
"description": "",
|
||||
"sampleRate": 44100,
|
||||
"modelType": "combo",
|
||||
"f0": true,
|
||||
"numOfDiffLayers":20,
|
||||
"numOfNativeLayers":3,
|
||||
"maxKStep":50
|
||||
},
|
||||
{
|
||||
"id": "diffusion_combo_spk5_nl3_dl10_k50_500ep",
|
||||
"voiceChangerType": "Diffusion-SVC",
|
||||
"lang": "ja-JP",
|
||||
"tag": ["diffsuion-svc", "torch"],
|
||||
"name": "初期メン(3,10)",
|
||||
"modelUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/diffusion-svc/combo_spk5_nl3_dl10_k50_500ep.ptc.bin",
|
||||
"termsOfUseUrl": "https://huggingface.co/wok000/vcclient_model/raw/main/diffusion-svc/term_of_use.txt",
|
||||
"icon": "https://huggingface.co/wok000/vcclient_model/resolve/main/diffusion-svc/diffusion-svc.png",
|
||||
"credit": "あみたろ,つくよみちゃん,黄琴海月,黄琴まひろ,刻鳴時雨",
|
||||
"description": "",
|
||||
"sampleRate": 44100,
|
||||
"modelType": "combo",
|
||||
"f0": true,
|
||||
"numOfDiffLayers":10,
|
||||
"numOfNativeLayers":3,
|
||||
"maxKStep":50
|
||||
},
|
||||
{
|
||||
"id": "diffusion_combo_spk5_nl3_dl1_k50_500ep",
|
||||
"voiceChangerType": "Diffusion-SVC",
|
||||
"lang": "ja-JP",
|
||||
"tag": ["diffsuion-svc", "torch"],
|
||||
"name": "初期メン(3,1)",
|
||||
"modelUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/diffusion-svc/combo_spk5_nl3_dl1_k50_500ep.ptc.bin",
|
||||
"termsOfUseUrl": "https://huggingface.co/wok000/vcclient_model/raw/main/diffusion-svc/term_of_use.txt",
|
||||
"icon": "https://huggingface.co/wok000/vcclient_model/resolve/main/diffusion-svc/diffusion-svc.png",
|
||||
"credit": "あみたろ,つくよみちゃん,黄琴海月,黄琴まひろ,刻鳴時雨",
|
||||
"description": "",
|
||||
"sampleRate": 44100,
|
||||
"modelType": "combo",
|
||||
"f0": true,
|
||||
"numOfDiffLayers":1,
|
||||
"numOfNativeLayers":3,
|
||||
"maxKStep":50
|
||||
},
|
||||
{
|
||||
"id": "diffusion_combo_spk5_nl1_dl1_k50_500ep",
|
||||
"voiceChangerType": "Diffusion-SVC",
|
||||
"lang": "ja-JP",
|
||||
"tag": ["diffsuion-svc", "torch"],
|
||||
"name": "初期メン(1,1)",
|
||||
"modelUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/diffusion-svc/combo_spk5_nl1_dl1_k50_500ep.ptc.bin",
|
||||
"termsOfUseUrl": "https://huggingface.co/wok000/vcclient_model/raw/main/diffusion-svc/term_of_use.txt",
|
||||
"icon": "https://huggingface.co/wok000/vcclient_model/resolve/main/diffusion-svc/diffusion-svc.png",
|
||||
"credit": "あみたろ,つくよみちゃん,黄琴海月,黄琴まひろ,刻鳴時雨",
|
||||
"description": "",
|
||||
"sampleRate": 44100,
|
||||
"modelType": "combo",
|
||||
"f0": true,
|
||||
"numOfDiffLayers":1,
|
||||
"numOfNativeLayers":1,
|
||||
"maxKStep":50
|
||||
}
|
||||
]
|
||||
}
|
||||
@ -45,6 +45,7 @@ class DiffusionSVC(VoiceChangerModel):
|
||||
self.settings.tran = self.slotInfo.defaultTune
|
||||
self.settings.dstId = self.slotInfo.dstId
|
||||
self.settings.kStep = self.slotInfo.defaultKstep
|
||||
self.settings.speedUp = self.slotInfo.defaultSpeedup
|
||||
|
||||
print("[Voice Changer] [DiffusionSVC] Initializing... done")
|
||||
|
||||
|
||||
@ -9,6 +9,15 @@ from voice_changer.DiffusionSVC.inferencer.diffusion_svc_model.diffusion.unit2me
|
||||
from voice_changer.utils.LoadModelParams import LoadModelParams
|
||||
from voice_changer.utils.ModelSlotGenerator import ModelSlotGenerator
|
||||
|
||||
def get_divisors(n):
|
||||
divisors = []
|
||||
for i in range(1, int(n**0.5)+1):
|
||||
if n % i == 0:
|
||||
divisors.append(i)
|
||||
if i != n // i:
|
||||
divisors.append(n //i)
|
||||
return sorted(divisors)
|
||||
|
||||
|
||||
class DiffusionSVCModelSlotGenerator(ModelSlotGenerator):
|
||||
@classmethod
|
||||
@ -20,7 +29,7 @@ class DiffusionSVCModelSlotGenerator(ModelSlotGenerator):
|
||||
slotInfo.defaultTune = 0
|
||||
slotInfo.isONNX = slotInfo.modelFile.endswith(".onnx")
|
||||
slotInfo.name = os.path.splitext(os.path.basename(slotInfo.modelFile))[0]
|
||||
slotInfo.iconFile = "/assets/icons/noimage.png"
|
||||
# slotInfo.iconFile = "/assets/icons/noimage.png"
|
||||
slotInfo.embChannels = 768
|
||||
|
||||
if slotInfo.isONNX:
|
||||
@ -35,7 +44,9 @@ class DiffusionSVCModelSlotGenerator(ModelSlotGenerator):
|
||||
slot.kStepMax = diff_args.model.k_step_max
|
||||
slot.nLayers = diff_args.model.n_layers
|
||||
slot.nnLayers = naive_args.model.n_layers
|
||||
diff_args.model.n_spk
|
||||
slot.defaultKstep = slot.kStepMax
|
||||
divs = get_divisors(slot.defaultKstep)
|
||||
slot.defaultSpeedup = divs[-2]
|
||||
slot.speakers = {(x+1): f"user{x+1}" for x in range(diff_args.model.n_spk)}
|
||||
return slot
|
||||
|
||||
|
||||
@ -119,7 +119,8 @@ class VoiceChangerManager(ServerDeviceCallbacks):
|
||||
def loadModel(self, params: LoadModelParams):
|
||||
if params.isSampleMode:
|
||||
# サンプルダウンロード
|
||||
downloadSample(self.params.sample_mode, params.sampleId, self.params.model_dir, params.slot, {"useIndex": params.params["rvcIndexDownload"]})
|
||||
print("[Voice Changer] sample download....", params)
|
||||
downloadSample(self.params.sample_mode, params.sampleId, self.params.model_dir, params.slot, params.params)
|
||||
self.modelSlotManager.getAllSlotInfo(reload=True)
|
||||
info = {"status": "OK"}
|
||||
return info
|
||||
|
||||
Loading…
Reference in New Issue
Block a user