追加: 文内無音倍率 (VOICEVOX#2352)

Co-authored-by: Hiroshiba <[email protected]> Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Hiroshiba · Nov 15, 2024 · 258c791 · 258c791
1 parent 7043359
commit 258c791
Show file tree

Hide file tree

Showing 12 changed files with 149 additions and 19 deletions.
diff --git a/src/backend/common/ConfigManager.ts b/src/backend/common/ConfigManager.ts
@@ -265,6 +265,17 @@ const migrations: [string, (store: Record<string, unknown>) => unknown][] = [
       return config;
     },
   ],
+  [
+    ">=0.22",
+    (config) => {
+      // プリセットに文内無音倍率を追加
+      const presets = config.presets as ConfigType["presets"];
+      for (const preset of Object.values(presets.items)) {
+        if (preset == undefined) throw new Error("preset == undefined");
+        preset.pauseLengthScale = 1;
+      }
+    },
+  ],
 ];
 
 export type Metadata = {

diff --git a/src/components/Talk/AudioInfo.vue b/src/components/Talk/AudioInfo.vue
@@ -434,6 +434,24 @@ const parameterConfigs = computed<ParameterConfig[]>(() => [
       }),
     key: "volumeScale",
   },
+  {
+    label: "文内無音倍率",
+    sliderProps: {
+      modelValue: () => query.value?.pauseLengthScale ?? null,
+      disable: () => uiLocked.value,
+      max: SLIDER_PARAMETERS.PAUSE_LENGTH_SCALE.max,
+      min: SLIDER_PARAMETERS.PAUSE_LENGTH_SCALE.min,
+      step: SLIDER_PARAMETERS.PAUSE_LENGTH_SCALE.step,
+      scrollStep: SLIDER_PARAMETERS.PAUSE_LENGTH_SCALE.scrollStep,
+      scrollMinStep: SLIDER_PARAMETERS.PAUSE_LENGTH_SCALE.scrollMinStep,
+    },
+    onChange: (pauseLengthScale: number) =>
+      store.actions.COMMAND_MULTI_SET_AUDIO_PAUSE_LENGTH_SCALE({
+        audioKeys: selectedAudioKeys.value,
+        pauseLengthScale,
+      }),
+    key: "pauseLengthScale",
+  },
   {
     label: "開始無音",
     sliderProps: {

diff --git a/src/domain/project/index.ts b/src/domain/project/index.ts
@@ -125,6 +125,7 @@ export const migrateProjectFileObject = async (
     for (const audioItemsKey in projectData.audioItems) {
       if (projectData.audioItems[audioItemsKey].query != null) {
         projectData.audioItems[audioItemsKey].query.volumeScale = 1;
+        projectData.audioItems[audioItemsKey].query.pauseLengthScale = 1;
         projectData.audioItems[audioItemsKey].query.prePhonemeLength = 0.1;
         projectData.audioItems[audioItemsKey].query.postPhonemeLength = 0.1;
         projectData.audioItems[audioItemsKey].query.outputSamplingRate =
@@ -302,6 +303,13 @@ export const migrateProjectFileObject = async (
     projectData.song.trackOrder = Object.keys(newTracks);
   }
 
+  if (semver.satisfies(projectAppVersion, "<0.22.0", semverSatisfiesOptions)) {
+    // 文内無音倍率の追加
+    for (const audioItemsKey in projectData.talk.audioItems) {
+      projectData.talk.audioItems[audioItemsKey].query.pauseLengthScale = 1;
+    }
+  }
+
   // Validation check
   // トークはvalidateTalkProjectで検証する
   // ソングはSET_SCOREの中の`isValidScore`関数で検証される

diff --git a/src/domain/project/schema.ts b/src/domain/project/schema.ts
@@ -33,6 +33,7 @@ const audioQuerySchema = z.object({
   pitchScale: z.number(),
   intonationScale: z.number(),
   volumeScale: z.number(),
+  pauseLengthScale: z.number(),
   prePhonemeLength: z.number(),
   postPhonemeLength: z.number(),
   outputSamplingRate: z.union([z.number(), z.literal("engineDefault")]),

diff --git a/src/store/audio.ts b/src/store/audio.ts
@@ -11,6 +11,7 @@ import {
   AudioCommandStoreTypes,
   transformCommandStore,
   FetchAudioResult,
+  EditorAudioQuery,
 } from "./type";
 import {
   buildAudioFileNameFromRawData,
@@ -34,6 +35,7 @@ import {
   isMorphable,
 } from "./audioGenerate";
 import { ContinuousPlayer } from "./audioContinuousPlayer";
+import { convertAudioQueryFromEngineToEditor } from "./proxy";
 import {
   convertHiraToKana,
   convertLongVowel,
@@ -739,6 +741,8 @@ export const audioStore = createPartialStore<AudioStoreTypes>({
           baseAudioItem.query.prePhonemeLength;
         newAudioItem.query.postPhonemeLength =
           baseAudioItem.query.postPhonemeLength;
+        newAudioItem.query.pauseLengthScale =
+          baseAudioItem.query.pauseLengthScale;
         newAudioItem.query.outputSamplingRate =
           baseAudioItem.query.outputSamplingRate;
         newAudioItem.query.outputStereo = baseAudioItem.query.outputStereo;
@@ -891,6 +895,23 @@ export const audioStore = createPartialStore<AudioStoreTypes>({
     },
   },
 
+  SET_AUDIO_PAUSE_LENGTH_SCALE: {
+    mutation(
+      state,
+      {
+        audioKey,
+        pauseLengthScale,
+      }: {
+        audioKey: AudioKey;
+        pauseLengthScale: number;
+      },
+    ) {
+      const query = state.audioItems[audioKey].query;
+      if (query == undefined) throw new Error("query == undefined");
+      query.pauseLengthScale = pauseLengthScale;
+    },
+  },
+
   SET_AUDIO_PRE_PHONEME_LENGTH: {
     mutation(
       state,
@@ -949,13 +970,16 @@ export const audioStore = createPartialStore<AudioStoreTypes>({
   SET_AUDIO_QUERY: {
     mutation(
       state,
-      { audioKey, audioQuery }: { audioKey: AudioKey; audioQuery: AudioQuery },
+      {
+        audioKey,
+        audioQuery,
+      }: { audioKey: AudioKey; audioQuery: EditorAudioQuery },
     ) {
       state.audioItems[audioKey].query = audioQuery;
     },
     action(
       { mutations },
-      payload: { audioKey: AudioKey; audioQuery: AudioQuery },
+      payload: { audioKey: AudioKey; audioQuery: EditorAudioQuery },
     ) {
       mutations.SET_AUDIO_QUERY(payload);
     },
@@ -974,11 +998,13 @@ export const audioStore = createPartialStore<AudioStoreTypes>({
         .INSTANTIATE_ENGINE_CONNECTOR({
           engineId,
         })
-        .then((instance) =>
-          instance.invoke("audioQueryAudioQueryPost")({
-            text,
-            speaker: styleId,
-          }),
+        .then(async (instance) =>
+          convertAudioQueryFromEngineToEditor(
+            await instance.invoke("audioQueryAudioQueryPost")({
+              text,
+              speaker: styleId,
+            }),
+          ),
         )
         .catch((error) => {
           window.backend.logError(
@@ -1271,7 +1297,9 @@ export const audioStore = createPartialStore<AudioStoreTypes>({
           length += m.consonantLength != undefined ? m.consonantLength : 0;
           length += m.vowelLength;
         });
-        length += phrase.pauseMora ? phrase.pauseMora.vowelLength : 0;
+        length += phrase.pauseMora
+          ? phrase.pauseMora.vowelLength * query.pauseLengthScale
+          : 0;
         // post phoneme lengthは最後のアクセント句の一部として扱う
         if (i === accentPhrases.length - 1) {
           length += query.postPhonemeLength;
@@ -1919,7 +1947,7 @@ export const audioCommandStore = transformCommandStore(
         payload: { audioKey: AudioKey; text: string } & (
           | { update: "Text" }
           | { update: "AccentPhrases"; accentPhrases: AccentPhrase[] }
-          | { update: "AudioQuery"; query: AudioQuery }
+          | { update: "AudioQuery"; query: EditorAudioQuery }
         ),
       ) {
         audioStore.mutations.SET_AUDIO_TEXT(draft, {
@@ -2025,7 +2053,7 @@ export const audioCommandStore = transformCommandStore(
               }
             | {
                 update: "AudioQuery";
-                query: AudioQuery;
+                query: EditorAudioQuery;
               }
             | {
                 update: "OnlyVoice";
@@ -2089,7 +2117,7 @@ export const audioCommandStore = transformCommandStore(
             }
           | {
               update: "AudioQuery";
-              query: AudioQuery;
+              query: EditorAudioQuery;
             }
           | {
               update: "OnlyVoice";
@@ -2100,7 +2128,7 @@ export const audioCommandStore = transformCommandStore(
           try {
             const audioItem = state.audioItems[audioKey];
             if (audioItem.query == undefined) {
-              const query: AudioQuery = await actions.FETCH_AUDIO_QUERY({
+              const query = await actions.FETCH_AUDIO_QUERY({
                 text: audioItem.text,
                 engineId: voice.engineId,
                 styleId: voice.styleId,
@@ -2711,6 +2739,29 @@ export const audioCommandStore = transformCommandStore(
       },
     },
 
+    COMMAND_MULTI_SET_AUDIO_PAUSE_LENGTH_SCALE: {
+      mutation(
+        draft,
+        payload: {
+          audioKeys: AudioKey[];
+          pauseLengthScale: number;
+        },
+      ) {
+        for (const audioKey of payload.audioKeys) {
+          audioStore.mutations.SET_AUDIO_PAUSE_LENGTH_SCALE(draft, {
+            audioKey,
+            pauseLengthScale: payload.pauseLengthScale,
+          });
+        }
+      },
+      action(
+        { mutations },
+        payload: { audioKeys: AudioKey[]; pauseLengthScale: number },
+      ) {
+        mutations.COMMAND_MULTI_SET_AUDIO_PAUSE_LENGTH_SCALE(payload);
+      },
+    },
+
     COMMAND_MULTI_SET_AUDIO_PRE_PHONEME_LENGTH: {
       mutation(
         draft,

diff --git a/src/store/audioGenerate.ts b/src/store/audioGenerate.ts
@@ -100,7 +100,11 @@ export async function generateLabFromAudioQuery(
     });
     if (accentPhrase.pauseMora != undefined) {
       labString += timestamp.toFixed() + " ";
-      timestamp += (accentPhrase.pauseMora.vowelLength * 10000000) / speedScale;
+      timestamp +=
+        (accentPhrase.pauseMora.vowelLength *
+          audioQuery.pauseLengthScale *
+          10000000) /
+        speedScale;
       labString += timestamp.toFixed() + " ";
       labString += accentPhrase.pauseMora.vowel + "\n";
     }

diff --git a/src/store/preset.ts b/src/store/preset.ts
@@ -222,6 +222,7 @@ export const presetStore = createPartialStore<PresetStoreTypes>({
           pitchScale: 0.0,
           intonationScale: 1.0,
           volumeScale: 1.0,
+          pauseLengthScale: 1,
           prePhonemeLength: 0.1,
           postPhonemeLength: 0.1,
         };

diff --git a/src/store/proxy.ts b/src/store/proxy.ts
@@ -43,6 +43,7 @@ const proxyStoreCreator = (_engineFactory: IEngineConnectorFactory) => {
   return proxyStore;
 };
 
+/** AudioQueryをエンジン用に変換する */
 export const convertAudioQueryFromEditorToEngine = (
   editorAudioQuery: EditorAudioQuery,
   defaultOutputSamplingRate: number,
@@ -56,4 +57,14 @@ export const convertAudioQueryFromEditorToEngine = (
   };
 };
 
+/** AudioQueryをエディタ用に変換する */
+export const convertAudioQueryFromEngineToEditor = (
+  engineAudioQuery: AudioQuery,
+): EditorAudioQuery => {
+  return {
+    ...engineAudioQuery,
+    pauseLengthScale: engineAudioQuery.pauseLengthScale ?? 1,
+  };
+};
+
 export const proxyStore = proxyStoreCreator(OpenAPIEngineConnectorFactory);
diff --git a/src/store/type.ts b/src/store/type.ts
@@ -76,8 +76,12 @@ import {
 /**
  * エディタ用のAudioQuery
  */
-export type EditorAudioQuery = Omit<AudioQuery, "outputSamplingRate"> & {
+export type EditorAudioQuery = Omit<
+  AudioQuery,
+  "outputSamplingRate" | "pauseLengthScale"
+> & {
   outputSamplingRate: number | "engineDefault";
+  pauseLengthScale: number; // エンジンと違って必須
 };
 
 export type AudioItem = {
@@ -290,6 +294,10 @@ export type AudioStoreTypes = {
     mutation: { audioKey: AudioKey; volumeScale: number };
   };
 
+  SET_AUDIO_PAUSE_LENGTH_SCALE: {
+    mutation: { audioKey: AudioKey; pauseLengthScale: number };
+  };
+
   SET_AUDIO_PRE_PHONEME_LENGTH: {
     mutation: { audioKey: AudioKey; prePhonemeLength: number };
   };
@@ -329,16 +337,16 @@ export type AudioStoreTypes = {
   };
 
   SET_AUDIO_QUERY: {
-    mutation: { audioKey: AudioKey; audioQuery: AudioQuery };
-    action(payload: { audioKey: AudioKey; audioQuery: AudioQuery }): void;
+    mutation: { audioKey: AudioKey; audioQuery: EditorAudioQuery };
+    action(payload: { audioKey: AudioKey; audioQuery: EditorAudioQuery }): void;
   };
 
   FETCH_AUDIO_QUERY: {
     action(payload: {
       text: string;
       engineId: EngineId;
       styleId: StyleId;
-    }): Promise<AudioQuery>;
+    }): Promise<EditorAudioQuery>;
   };
 
   SET_AUDIO_VOICE: {
@@ -506,7 +514,7 @@ export type AudioCommandStoreTypes = {
     mutation: { audioKey: AudioKey; text: string } & (
       | { update: "Text" }
       | { update: "AccentPhrases"; accentPhrases: AccentPhrase[] }
-      | { update: "AudioQuery"; query: AudioQuery }
+      | { update: "AudioQuery"; query: EditorAudioQuery }
     );
     action(payload: { audioKey: AudioKey; text: string }): void;
   };
@@ -522,7 +530,7 @@ export type AudioCommandStoreTypes = {
           }
         | {
             update: "AudioQuery";
-            query: AudioQuery;
+            query: EditorAudioQuery;
           }
         | {
             update: "OnlyVoice";
@@ -627,6 +635,11 @@ export type AudioCommandStoreTypes = {
     action(payload: { audioKeys: AudioKey[]; volumeScale: number }): void;
   };
 
+  COMMAND_MULTI_SET_AUDIO_PAUSE_LENGTH_SCALE: {
+    mutation: { audioKeys: AudioKey[]; pauseLengthScale: number };
+    action(payload: { audioKeys: AudioKey[]; pauseLengthScale: number }): void;
+  };
+
   COMMAND_MULTI_SET_AUDIO_PRE_PHONEME_LENGTH: {
     mutation: { audioKeys: AudioKey[]; prePhonemeLength: number };
     action(payload: { audioKeys: AudioKey[]; prePhonemeLength: number }): void;

diff --git a/src/store/utility.ts b/src/store/utility.ts
@@ -105,6 +105,16 @@ export const SLIDER_PARAMETERS = {
     scrollStep: () => 0.1,
     scrollMinStep: () => 0.01,
   },
+  /**
+   *  文内無音(倍率)パラメータの定義
+   */
+  PAUSE_LENGTH_SCALE: {
+    max: () => 2,
+    min: () => 0,
+    step: () => 0.01,
+    scrollStep: () => 0.1,
+    scrollMinStep: () => 0.01,
+  },
   /**
    *  モーフィングレートパラメータの定義
    */

diff --git a/src/type/preload.ts b/src/type/preload.ts
@@ -414,6 +414,7 @@ export type Preset = {
   pitchScale: number;
   intonationScale: number;
   volumeScale: number;
+  pauseLengthScale: number;
   prePhonemeLength: number;
   postPhonemeLength: number;
   morphingInfo?: MorphingInfo;
@@ -643,6 +644,7 @@ export const configSchema = z
               pitchScale: z.number(),
               intonationScale: z.number(),
               volumeScale: z.number(),
+              pauseLengthScale: z.number(),
               prePhonemeLength: z.number(),
               postPhonemeLength: z.number(),
               morphingInfo: z

diff --git a/tests/e2e/browser/スクリーンショット.spec.ts-snapshots/トーク画面-browser-win32.png b/tests/e2e/browser/スクリーンショット.spec.ts-snapshots/トーク画面-browser-win32.png