WIP:diffusion svc config

client/demo/src/components/demo/components2/101-7_diffusion-svcSettingArea.tsx

@@ -0,0 +1,81 @@
+import React, { useMemo } from "react";
+import { useAppState } from "../../../001_provider/001_AppStateProvider";
+import { DiffusionSVCModelSlot } from "@dannadori/voice-changer-client-js";
+
+export type DiffusionSVCSettingAreaProps = {};
+
+export const DiffusionSVCSettingArea = (_props: DiffusionSVCSettingAreaProps) => {
+    const { serverSetting } = useAppState();
+
+    const selected = useMemo(() => {
+        if (serverSetting.serverSetting.modelSlotIndex == undefined) {
+            return;
+        }
+        return serverSetting.serverSetting.modelSlots[serverSetting.serverSetting.modelSlotIndex];
+    }, [serverSetting.serverSetting.modelSlotIndex, serverSetting.serverSetting.modelSlots]);
+
+    const settingArea = useMemo(() => {
+        if (!selected) {
+            return <></>;
+        }
+
+        if (selected.voiceChangerType != "Diffusion-SVC") {
+            return <></>;
+        }
+
+        const kStepRow = (
+            <div className="character-area-control">
+                <div className="character-area-control-title">k-step:</div>
+                <div className="character-area-control-field">
+                    <div className="character-area-slider-control">
+                        <span className="character-area-slider-control-kind"></span>
+                        <span className="character-area-slider-control-slider">
+                            <input
+                                type="range"
+                                min="0"
+                                max={(selected as DiffusionSVCModelSlot).kStepMax}
+                                step="1"
+                                value={serverSetting.serverSetting.kStep}
+                                onChange={(e) => {
+                                    serverSetting.updateServerSettings({ ...serverSetting.serverSetting, kStep: Number(e.target.value) });
+                                }}
+                            ></input>
+                        </span>
+                        <span className="character-area-slider-control-val">{serverSetting.serverSetting.kStep}</span>
+                    </div>
+                </div>
+            </div>
+        );
+        const speedUpRow = (
+            <div className="character-area-control">
+                <div className="character-area-control-title">speedup</div>
+                <div className="character-area-control-field">
+                    <div className="character-area-slider-control">
+                        <span className="character-area-slider-control-kind"></span>
+                        <span className="character-area-slider-control-slider">
+                            <input
+                                type="range"
+                                min="0"
+                                max={serverSetting.serverSetting.kStep}
+                                step="1"
+                                value={serverSetting.serverSetting.speedUp}
+                                onChange={(e) => {
+                                    serverSetting.updateServerSettings({ ...serverSetting.serverSetting, speedUp: Number(e.target.value) });
+                                }}
+                            ></input>
+                        </span>
+                        <span className="character-area-slider-control-val">{serverSetting.serverSetting.speedUp}</span>
+                    </div>
+                </div>
+            </div>
+        );
+        return (
+            <>
+                {kStepRow}
+                {speedUpRow}
+            </>
+        );
+    }, [serverSetting.serverSetting, serverSetting.updateServerSettings, selected]);
+
+    return settingArea;
+};

client/demo/src/components/demo/components2/101_CharacterArea.tsx

@@ -1,88 +1,95 @@
-import React, { useEffect, useMemo, useState } from "react"
-import { useAppState } from "../../../001_provider/001_AppStateProvider"
-import { useGuiState } from "../001_GuiStateProvider"
-import { OnnxExporterInfo } from "@dannadori/voice-changer-client-js"
-import { useMessageBuilder } from "../../../hooks/useMessageBuilder"
-import { TuningArea } from "./101-1_TuningArea"
-import { IndexArea } from "./101-2_IndexArea"
-import { SpeakerArea } from "./101-3_SpeakerArea"
-import { F0FactorArea } from "./101-4_F0FactorArea"
-import { SoVitsSVC40SettingArea } from "./101-5_so-vits-svc40SettingArea"
-import { DDSPSVC30SettingArea } from "./101-6_ddsp-svc30SettingArea"
-
-export type CharacterAreaProps = {
-}
+import React, { useEffect, useMemo, useState } from "react";
+import { useAppState } from "../../../001_provider/001_AppStateProvider";
+import { useGuiState } from "../001_GuiStateProvider";
+import { OnnxExporterInfo } from "@dannadori/voice-changer-client-js";
+import { useMessageBuilder } from "../../../hooks/useMessageBuilder";
+import { TuningArea } from "./101-1_TuningArea";
+import { IndexArea } from "./101-2_IndexArea";
+import { SpeakerArea } from "./101-3_SpeakerArea";
+import { F0FactorArea } from "./101-4_F0FactorArea";
+import { SoVitsSVC40SettingArea } from "./101-5_so-vits-svc40SettingArea";
+import { DDSPSVC30SettingArea } from "./101-6_ddsp-svc30SettingArea";
+import { DiffusionSVCSettingArea } from "./101-7_diffusion-svcSettingArea";
 
+export type CharacterAreaProps = {};
 
 export const CharacterArea = (_props: CharacterAreaProps) => {
-    const { serverSetting, initializedRef, volume, bufferingTime, performance, setting, setVoiceChangerClientSetting, start, stop } = useAppState()
-    const guiState = useGuiState()
-    const messageBuilderState = useMessageBuilder()
+    const { serverSetting, initializedRef, volume, bufferingTime, performance, setting, setVoiceChangerClientSetting, start, stop } = useAppState();
+    const guiState = useGuiState();
+    const messageBuilderState = useMessageBuilder();
 
     useMemo(() => {
-        messageBuilderState.setMessage(__filename, "terms_of_use", { "ja": "利用規約", "en": "terms of use" })
-        messageBuilderState.setMessage(__filename, "export_to_onnx", { "ja": "onnx出力", "en": "export to onnx" })
-        messageBuilderState.setMessage(__filename, "save_default", { "ja": "設定保存", "en": "save setting" })
-        messageBuilderState.setMessage(__filename, "alert_onnx", { "ja": "ボイチェン中はonnx出力できません", "en": "cannot export onnx when voice conversion is enabled" })
-    }, [])
+        messageBuilderState.setMessage(__filename, "terms_of_use", { ja: "利用規約", en: "terms of use" });
+        messageBuilderState.setMessage(__filename, "export_to_onnx", { ja: "onnx出力", en: "export to onnx" });
+        messageBuilderState.setMessage(__filename, "save_default", { ja: "設定保存", en: "save setting" });
+        messageBuilderState.setMessage(__filename, "alert_onnx", { ja: "ボイチェン中はonnx出力できません", en: "cannot export onnx when voice conversion is enabled" });
+    }, []);
 
     const selected = useMemo(() => {
         if (serverSetting.serverSetting.modelSlotIndex == undefined) {
-            return
+            return;
         }
-        return serverSetting.serverSetting.modelSlots[serverSetting.serverSetting.modelSlotIndex]
-    }, [serverSetting.serverSetting.modelSlotIndex, serverSetting.serverSetting.modelSlots])
-
+        return serverSetting.serverSetting.modelSlots[serverSetting.serverSetting.modelSlotIndex];
+    }, [serverSetting.serverSetting.modelSlotIndex, serverSetting.serverSetting.modelSlots]);
 
     useEffect(() => {
-        const vol = document.getElementById("status-vol") as HTMLSpanElement
-        const buf = document.getElementById("status-buf") as HTMLSpanElement
-        const res = document.getElementById("status-res") as HTMLSpanElement
+        const vol = document.getElementById("status-vol") as HTMLSpanElement;
+        const buf = document.getElementById("status-buf") as HTMLSpanElement;
+        const res = document.getElementById("status-res") as HTMLSpanElement;
         if (!vol || !buf || !res) {
-            return
+            return;
         }
-        vol.innerText = volume.toFixed(4)
-        buf.innerText = bufferingTime.toString()
-        res.innerText = performance.responseTime.toString()
-
-    }, [volume, bufferingTime, performance])
+        vol.innerText = volume.toFixed(4);
+        buf.innerText = bufferingTime.toString();
+        res.innerText = performance.responseTime.toString();
+    }, [volume, bufferingTime, performance]);
 
     const portrait = useMemo(() => {
         if (!selected) {
-            return <></>
+            return <></>;
         }
 
-        const icon = selected.iconFile.length > 0 ? selected.iconFile : "./assets/icons/human.png"
-        const selectedTermOfUseUrlLink = selected.termsOfUseUrl ? <a href={selected.termsOfUseUrl} target="_blank" rel="noopener noreferrer" className="portrait-area-terms-of-use-link">[{messageBuilderState.getMessage(__filename, "terms_of_use")}]</a> : <></>
-
+        const icon = selected.iconFile.length > 0 ? selected.iconFile : "./assets/icons/human.png";
+        const selectedTermOfUseUrlLink = selected.termsOfUseUrl ? (
+            <a href={selected.termsOfUseUrl} target="_blank" rel="noopener noreferrer" className="portrait-area-terms-of-use-link">
+                [{messageBuilderState.getMessage(__filename, "terms_of_use")}]
+            </a>
+        ) : (
+            <></>
+        );
 
         return (
             <div className="portrait-area">
                 <div className="portrait-container">
                     <img className="portrait" src={icon} alt={selected.name} />
                     <div className="portrait-area-status">
-                        <p><span className="portrait-area-status-vctype">{selected.voiceChangerType}</span></p>
-                        <p>vol: <span id="status-vol">0</span></p>
-                        <p>buf: <span id="status-buf">0</span> ms</p>
-                        <p>res: <span id="status-res">0</span> ms</p>
-                    </div>
-                    <div className="portrait-area-terms-of-use">
-                        {selectedTermOfUseUrlLink}
+                        <p>
+                            <span className="portrait-area-status-vctype">{selected.voiceChangerType}</span>
+                        </p>
+                        <p>
+                            vol: <span id="status-vol">0</span>
+                        </p>
+                        <p>
+                            buf: <span id="status-buf">0</span> ms
+                        </p>
+                        <p>
+                            res: <span id="status-res">0</span> ms
+                        </p>
                     </div>
+                    <div className="portrait-area-terms-of-use">{selectedTermOfUseUrlLink}</div>
                 </div>
             </div>
-        )
-    }, [selected])
+        );
+    }, [selected]);
 
-
-    const [startWithAudioContextCreate, setStartWithAudioContextCreate] = useState<boolean>(false)
+    const [startWithAudioContextCreate, setStartWithAudioContextCreate] = useState<boolean>(false);
     useEffect(() => {
         if (!startWithAudioContextCreate) {
-            return
+            return;
         }
-        guiState.setIsConverting(true)
-        start()
-    }, [startWithAudioContextCreate])
+        guiState.setIsConverting(true);
+        start();
+    }, [startWithAudioContextCreate]);
 
     const startControl = useMemo(() => {
         const onStartClicked = async () => {
@@ -90,82 +97,86 @@ export const CharacterArea = (_props: CharacterAreaProps) => {
                 if (!initializedRef.current) {
                     while (true) {
                         await new Promise<void>((resolve) => {
-                            setTimeout(resolve, 500)
-                        })
+                            setTimeout(resolve, 500);
+                        });
                         if (initializedRef.current) {
-                            break
+                            break;
                         }
                     }
-                    setStartWithAudioContextCreate(true)
+                    setStartWithAudioContextCreate(true);
                 } else {
-                    guiState.setIsConverting(true)
-                    await start()
+                    guiState.setIsConverting(true);
+                    await start();
                 }
             } else {
-                serverSetting.updateServerSettings({ ...serverSetting.serverSetting, serverAudioStated: 1 })
-                guiState.setIsConverting(true)
+                serverSetting.updateServerSettings({ ...serverSetting.serverSetting, serverAudioStated: 1 });
+                guiState.setIsConverting(true);
             }
-        }
+        };
         const onStopClicked = async () => {
             if (serverSetting.serverSetting.enableServerAudio == 0) {
-                guiState.setIsConverting(false)
-                await stop()
+                guiState.setIsConverting(false);
+                await stop();
             } else {
-                guiState.setIsConverting(false)
-                serverSetting.updateServerSettings({ ...serverSetting.serverSetting, serverAudioStated: 0 })
+                guiState.setIsConverting(false);
+                serverSetting.updateServerSettings({ ...serverSetting.serverSetting, serverAudioStated: 0 });
             }
-        }
-        const startClassName = guiState.isConverting ? "character-area-control-button-active" : "character-area-control-button-stanby"
-        const stopClassName = guiState.isConverting ? "character-area-control-button-stanby" : "character-area-control-button-active"
+        };
+        const startClassName = guiState.isConverting ? "character-area-control-button-active" : "character-area-control-button-stanby";
+        const stopClassName = guiState.isConverting ? "character-area-control-button-stanby" : "character-area-control-button-active";
 
         return (
             <div className="character-area-control">
                 <div className="character-area-control-buttons">
-                    <div onClick={onStartClicked} className={startClassName}>start</div>
-                    <div onClick={onStopClicked} className={stopClassName}>stop</div>
+                    <div onClick={onStartClicked} className={startClassName}>
+                        start
+                    </div>
+                    <div onClick={onStopClicked} className={stopClassName}>
+                        stop
+                    </div>
                 </div>
             </div>
-        )
-    }, [
-        guiState.isConverting,
-        start,
-        stop,
-        serverSetting.serverSetting,
-        serverSetting.updateServerSettings
-    ])
+        );
+    }, [guiState.isConverting, start, stop, serverSetting.serverSetting, serverSetting.updateServerSettings]);
 
     const gainControl = useMemo(() => {
-        const currentInputGain = serverSetting.serverSetting.enableServerAudio == 0 ? setting.voiceChangerClientSetting.inputGain : serverSetting.serverSetting.serverInputAudioGain
-        const inputValueUpdatedAction = serverSetting.serverSetting.enableServerAudio == 0 ?
-            async (val: number) => {
-                await setVoiceChangerClientSetting({ ...setting.voiceChangerClientSetting, inputGain: val })
-            } :
-            async (val: number) => {
-                await serverSetting.updateServerSettings({ ...serverSetting.serverSetting, serverInputAudioGain: val })
-            }
-
-        const currentOutputGain = serverSetting.serverSetting.enableServerAudio == 0 ? setting.voiceChangerClientSetting.outputGain : serverSetting.serverSetting.serverOutputAudioGain
-        const outputValueUpdatedAction = serverSetting.serverSetting.enableServerAudio == 0 ?
-            async (val: number) => {
-                await setVoiceChangerClientSetting({ ...setting.voiceChangerClientSetting, outputGain: val })
-            } :
-            async (val: number) => {
-                await serverSetting.updateServerSettings({ ...serverSetting.serverSetting, serverOutputAudioGain: val })
-            }
+        const currentInputGain = serverSetting.serverSetting.enableServerAudio == 0 ? setting.voiceChangerClientSetting.inputGain : serverSetting.serverSetting.serverInputAudioGain;
+        const inputValueUpdatedAction =
+            serverSetting.serverSetting.enableServerAudio == 0
+                ? async (val: number) => {
+                      await setVoiceChangerClientSetting({ ...setting.voiceChangerClientSetting, inputGain: val });
+                  }
+                : async (val: number) => {
+                      await serverSetting.updateServerSettings({ ...serverSetting.serverSetting, serverInputAudioGain: val });
+                  };
 
+        const currentOutputGain = serverSetting.serverSetting.enableServerAudio == 0 ? setting.voiceChangerClientSetting.outputGain : serverSetting.serverSetting.serverOutputAudioGain;
+        const outputValueUpdatedAction =
+            serverSetting.serverSetting.enableServerAudio == 0
+                ? async (val: number) => {
+                      await setVoiceChangerClientSetting({ ...setting.voiceChangerClientSetting, outputGain: val });
+                  }
+                : async (val: number) => {
+                      await serverSetting.updateServerSettings({ ...serverSetting.serverSetting, serverOutputAudioGain: val });
+                  };
 
         return (
             <div className="character-area-control">
-                <div className="character-area-control-title">
-                    GAIN:
-                </div>
+                <div className="character-area-control-title">GAIN:</div>
                 <div className="character-area-control-field">
                     <div className="character-area-slider-control">
                         <span className="character-area-slider-control-kind">in</span>
                         <span className="character-area-slider-control-slider">
-                            <input type="range" min="0.1" max="10.0" step="0.1" value={currentInputGain} onChange={(e) => {
-                                inputValueUpdatedAction(Number(e.target.value))
-                            }}></input>
+                            <input
+                                type="range"
+                                min="0.1"
+                                max="10.0"
+                                step="0.1"
+                                value={currentInputGain}
+                                onChange={(e) => {
+                                    inputValueUpdatedAction(Number(e.target.value));
+                                }}
+                            ></input>
                         </span>
                         <span className="character-area-slider-control-val">{currentInputGain}</span>
                     </div>
@@ -173,64 +184,72 @@ export const CharacterArea = (_props: CharacterAreaProps) => {
                     <div className="character-area-slider-control">
                         <span className="character-area-slider-control-kind">out</span>
                         <span className="character-area-slider-control-slider">
-                            <input type="range" min="0.1" max="10.0" step="0.1" value={currentOutputGain} onChange={(e) => {
-                                outputValueUpdatedAction(Number(e.target.value))
-                            }}></input>
+                            <input
+                                type="range"
+                                min="0.1"
+                                max="10.0"
+                                step="0.1"
+                                value={currentOutputGain}
+                                onChange={(e) => {
+                                    outputValueUpdatedAction(Number(e.target.value));
+                                }}
+                            ></input>
                         </span>
                         <span className="character-area-slider-control-val">{currentOutputGain}</span>
                     </div>
-
                 </div>
             </div>
-        )
-    }, [serverSetting.serverSetting, setting, setVoiceChangerClientSetting, serverSetting.updateServerSettings])
-
-
+        );
+    }, [serverSetting.serverSetting, setting, setVoiceChangerClientSetting, serverSetting.updateServerSettings]);
 
     const modelSlotControl = useMemo(() => {
         if (!selected) {
-            return <></>
+            return <></>;
         }
         const onUpdateDefaultClicked = async () => {
-            await serverSetting.updateModelDefault()
-        }
+            await serverSetting.updateModelDefault();
+        };
 
         const onnxExportButtonAction = async () => {
             if (guiState.isConverting) {
-                alert(messageBuilderState.getMessage(__filename, "alert_onnx"))
-                return
+                alert(messageBuilderState.getMessage(__filename, "alert_onnx"));
+                return;
             }
 
-            document.getElementById("dialog")?.classList.add("dialog-container-show")
-            guiState.stateControls.showWaitingCheckbox.updateState(true)
-            const res = await serverSetting.getOnnx() as OnnxExporterInfo
-            const a = document.createElement("a")
-            a.href = res.path
+            document.getElementById("dialog")?.classList.add("dialog-container-show");
+            guiState.stateControls.showWaitingCheckbox.updateState(true);
+            const res = (await serverSetting.getOnnx()) as OnnxExporterInfo;
+            const a = document.createElement("a");
+            a.href = res.path;
             a.download = res.filename;
             document.body.appendChild(a);
             a.click();
             document.body.removeChild(a);
-            guiState.stateControls.showWaitingCheckbox.updateState(false)
+            guiState.stateControls.showWaitingCheckbox.updateState(false);
+        };
 
-        }
-
-        const exportOnnx = selected.voiceChangerType == "RVC" && selected.modelFile.endsWith("pth") ? (
-            <div className="character-area-button" onClick={onnxExportButtonAction}>{messageBuilderState.getMessage(__filename, "export_to_onnx")}</div>
-        ) : <></>
+        const exportOnnx =
+            selected.voiceChangerType == "RVC" && selected.modelFile.endsWith("pth") ? (
+                <div className="character-area-button" onClick={onnxExportButtonAction}>
+                    {messageBuilderState.getMessage(__filename, "export_to_onnx")}
+                </div>
+            ) : (
+                <></>
+            );
         return (
             <div className="character-area-control">
-                <div className="character-area-control-title">
-
-                </div>
+                <div className="character-area-control-title"></div>
                 <div className="character-area-control-field">
                     <div className="character-area-buttons">
-                        <div className="character-area-button" onClick={onUpdateDefaultClicked}>{messageBuilderState.getMessage(__filename, "save_default")}</div>
+                        <div className="character-area-button" onClick={onUpdateDefaultClicked}>
+                            {messageBuilderState.getMessage(__filename, "save_default")}
+                        </div>
                         {exportOnnx}
                     </div>
                 </div>
             </div>
-        )
-    }, [selected, serverSetting.getOnnx, serverSetting.updateModelDefault])
+        );
+    }, [selected, serverSetting.getOnnx, serverSetting.updateModelDefault]);
 
     const characterArea = useMemo(() => {
         return (
@@ -245,11 +264,12 @@ export const CharacterArea = (_props: CharacterAreaProps) => {
                     <F0FactorArea />
                     <SoVitsSVC40SettingArea />
                     <DDSPSVC30SettingArea />
+                    <DiffusionSVCSettingArea />
                     {modelSlotControl}
                 </div>
             </div>
-        )
-    }, [portrait, startControl, gainControl, modelSlotControl])
+        );
+    }, [portrait, startControl, gainControl, modelSlotControl]);
 
-    return characterArea
-}
+    return characterArea;
+};

client/lib/src/const.ts

@@ -122,6 +122,8 @@ export const ServerSettingKey = {
     "kStep": "kStep",
     "threshold": "threshold",
 
+    "speedUp": "speedUp",
+
     "inputSampleRate": "inputSampleRate",
     "enableDirectML": "enableDirectML",
 } as const
@@ -181,6 +183,9 @@ export type VoiceChangerServerSetting = {
     kStep: number// DDSP-SVC
     threshold: number// DDSP-SVC
 
+    speedUp: number // Diffusion-SVC
+
+
     inputSampleRate: InputSampleRate
     enableDirectML: number
 }
@@ -260,7 +265,21 @@ export type DDSPSVCModelSlot = ModelSlot & {
     speakers: { [key: number]: string }
 }
 
-export type ModelSlotUnion = RVCModelSlot | MMVCv13ModelSlot | MMVCv15ModelSlot | SoVitsSvc40ModelSlot | DDSPSVCModelSlot
+
+export type DiffusionSVCModelSlot = ModelSlot & {
+    modelFile: string
+    dstId: number
+
+    samplingRate: number
+
+    defaultTune: number
+    defaultKstep : number
+    defaultSpeedup: number
+    kStepMax: number
+    speakers: { [key: number]: string }
+}
+
+export type ModelSlotUnion = RVCModelSlot | MMVCv13ModelSlot | MMVCv15ModelSlot | SoVitsSvc40ModelSlot | DDSPSVCModelSlot | DiffusionSVCModelSlot
 
 type ServerAudioDevice = {
     kind: "audioinput" | "audiooutput",
@@ -362,6 +381,8 @@ export const DefaultServerSetting: ServerInfo = {
     kStep: 120,
     threshold: -45,
 
+    speedUp: 10,
+
     enableDirectML: 0,
     // 
     status: "ok",

server/voice_changer/DiffusionSVC/DiffusionSVC.py

@@ -32,7 +32,7 @@ class DiffusionSVC(VoiceChangerModel):
         self.feature_buffer: FeatureInOut | None = None
         self.prevVol = 0.0
         self.slotInfo = slotInfo
-        
+
     def initialize(self):
         print("[Voice Changer] [DiffusionSVC] Initializing... ")
 
@@ -138,7 +138,7 @@ class DiffusionSVC(VoiceChangerModel):
 
         if self.pipeline is None:
             return np.zeros(convertSize).astype(np.int16) * np.sqrt(vol)
-        
+
         # device = self.pipeline.device
         # audio = torch.from_numpy(audio).to(device=device, dtype=torch.float32)
         # audio = self.resampler16K(audio)
@@ -146,6 +146,8 @@ class DiffusionSVC(VoiceChangerModel):
         f0_up_key = self.settings.tran
         protect = 0
 
+        kStep = self.settings.kStep
+        speedUp = self.settings.speedUp
         embOutputLayer = 12
         useFinalProj = False
         silenceFrontSec = self.settings.extraConvertSize / self.slotInfo.samplingRate if self.settings.silenceFront else 0.  # extaraConvertSize(既にモデルのサンプリングレートにリサンプリング済み)の秒数。モデルのサンプリングレートで処理(★１)。
@@ -157,6 +159,8 @@ class DiffusionSVC(VoiceChangerModel):
                 pitchf,
                 feature,
                 f0_up_key,
+                kStep,
+                speedUp,
                 silenceFrontSec,
                 embOutputLayer,
                 useFinalProj,
@@ -172,7 +176,7 @@ class DiffusionSVC(VoiceChangerModel):
             # raise e
 
         return
-    
+
     def __del__(self):
         del self.pipeline
 

server/voice_changer/DiffusionSVC/DiffusionSVCSettings.py

@@ -11,8 +11,8 @@ class DiffusionSVCSettings:
     silentThreshold: float = 0.00001
     extraConvertSize: int = 1024 * 4
 
-    kstep: int = 20
-    speedup: int = 10
+    kStep: int = 20
+    speedUp: int = 10
 
     silenceFront: int = 1  # 0:off, 1:on
     modelSamplingRate: int = 44100
@@ -26,7 +26,8 @@ class DiffusionSVCSettings:
         "dstId",
         "tran",
         "extraConvertSize",
-        "kstep",
+        "kStep",
+        "speedUp",
         "silenceFront",
     ]
     floatData = ["silentThreshold"]

server/voice_changer/DiffusionSVC/inferencer/DiffusionSVCInferencer.py

@@ -105,8 +105,8 @@ class DiffusionSVCInferencer(Inferencer):
         volume: torch.Tensor,
         mask: torch.Tensor,
         sid: torch.Tensor,
-        infer_speedup: int,
         k_step: int,
+        infer_speedup: int,
         silence_front: float,
     ) -> torch.Tensor:
         gt_spec = self.naive_model_call(feats, pitch, volume, spk_id=sid, spk_mix_dict=None, aug_shift=0, spk_emb=None)

server/voice_changer/DiffusionSVC/pipeline/Pipeline.py

@@ -105,6 +105,8 @@ class Pipeline(object):
         pitchf,  # np.array [m]
         feature,  # np.array [m, feat]
         f0_up_key,
+        k_step,
+        infer_speedup,
         silence_front,
         embOutputLayer,
         useFinalProj,
@@ -164,8 +166,8 @@ class Pipeline(object):
                                 volume,
                                 mask,
                                 sid,
-                                infer_speedup=10,
-                                k_step=20,
+                                k_step,
+                                infer_speedup,
                                 silence_front=silence_front
                                 ).to(dtype=torch.float32),
                             -1.0,