最適な話者ピッチ範囲を送れるように

run.py

@@ -54,6 +54,7 @@
     Speaker,
     SpeakerInfo,
     StyleIdNotFoundError,
+    StylePitchRange,
     SupportedDevicesInfo,
     UserDictWord,
     VvlibManifest,
@@ -537,6 +538,21 @@ def multi_synthesis(
             background=BackgroundTask(delete_file, f.name),
         )
 
+    @app.post(
+        "/optimal_pitch",
+        response_model=StylePitchRange,
+        tags=["その他"],
+        summary="指定したスタイルに対して最適なピッチ範囲を得る",
+    )
+    def optimal_pitch(style_id: StyleId) -> StylePitchRange:
+        try:
+            low, high = metas_store.get_style_optimal_pitch_range(style_id)
+            return StylePitchRange(low=low, high=high)
+        except StyleIdNotFoundError as e:
+            raise HTTPException(
+                status_code=404, detail=f"該当するスタイル(style_id={e.style_id})が見つかりません"
+            )
+
     @app.post(
         "/morphable_targets",
         response_model=list[dict[str, MorphableTargetInfo]],

speaker_info/35b2c544-660e-401e-b503-0e14c635303a/metas.json

@@ -1,3 +1,10 @@
 {
-  "supported_features": { "permitted_synthesis_morphing": "NOTHING" }
-}
+    "supported_features": {},
+    "range": [
+        {
+            "style_id": 8,
+            "low": 5.07,
+            "high": 6.5
+        }
+    ]
+}

speaker_info/388f246b-8c41-4ac1-8e2d-5d79f3ff56d9/metas.json

@@ -1,3 +1,17 @@
 {
-  "supported_features": { "permitted_synthesis_morphing": "SELF_ONLY" }
-}
+    "supported_features": {
+        "permitted_synthesis_morphing": "SELF_ONLY"
+    },
+    "range": [
+        {
+            "style_id": 1,
+            "low": 5.38,
+            "high": 6.44
+        },
+        {
+            "style_id": 3,
+            "low": 5.11,
+            "high": 6.5
+        }
+    ]
+}

speaker_info/7ffcb7ce-00ec-4bdc-82cd-45a8889e43ff/metas.json

@@ -1 +1,17 @@
-{}
+{
+    "supported_features": {
+        "permitted_synthesis_morphing": "SELF_ONLY"
+    },
+    "range": [
+        {
+            "style_id": 2,
+            "low": 5.16,
+            "high": 6.2
+        },
+        {
+            "style_id": 0,
+            "low": 5.21,
+            "high": 6.13
+        }
+    ]
+}

speaker_info/b1a81618-b27b-40d2-b0ea-27a9ad408c4b/metas.json

@@ -1,3 +1,4 @@
 {
-  "supported_features": { "permitted_synthesis_morphing": "ALL" }
-}
+    "supported_features": {},
+    "range": []
+}

voicevox_engine/metas/Metas.py

@@ -6,6 +6,7 @@
 # NOTE: 循環importを防ぐためにとりあえずここに書いている
 # FIXME: 他のmodelに依存せず、全modelから参照できる場所に配置する
 StyleId = NewType("StyleId", int)
+PitchRange = NewType("PitchRange", float)
 StyleType = Literal["talk", "singing_teacher", "frame_decode", "sing"]
 
 
@@ -28,6 +29,12 @@ class SpeakerStyle(BaseModel):
     )
 
 
+class SpeakerOptimalPitchRangeItem(BaseModel):
+    style_id: StyleId
+    high: PitchRange
+    low: PitchRange
+
+
 class SpeakerSupportPermittedSynthesisMorphing(str, Enum):
     ALL = "ALL"  # 全て許可
     SELF_ONLY = "SELF_ONLY"  # 同じ話者内でのみ許可
@@ -67,6 +74,9 @@ class EngineSpeaker(BaseModel):
     supported_features: SpeakerSupportedFeatures = Field(
         title="話者の対応機能", default_factory=SpeakerSupportedFeatures
     )
+    range: List[SpeakerOptimalPitchRangeItem] = Field(
+        title="話者の最適ピッチ範囲", default_factory=list[SpeakerOptimalPitchRangeItem]
+    )
 
 
 class Speaker(CoreSpeaker, EngineSpeaker):

voicevox_engine/metas/MetasStore.py

@@ -11,6 +11,7 @@
     StyleId,
     StyleType,
 )
+from voicevox_engine.model import StyleIdNotFoundError
 
 if TYPE_CHECKING:
     from voicevox_engine.core.core_adapter import CoreAdapter
@@ -61,6 +62,14 @@ def load_combined_metas(self, core: "CoreAdapter") -> List[Speaker]:
             for speaker_meta in core_metas
         ]
 
+    def get_style_optimal_pitch_range(self, style_id: StyleId) -> tuple[float, float]:
+        for meta in self._loaded_metas.values():
+            for range in meta.range:
+                if range.style_id == style_id:
+                    return (range.low, range.high)
+        else:
+            raise StyleIdNotFoundError(style_id=style_id)
+
 
 def construct_lookup(
     speakers: List[Speaker],

voicevox_engine/model.py

@@ -155,6 +155,11 @@ class MorphableTargetInfo(BaseModel):
     # reason: Optional[str] = Field(title="is_morphableがfalseである場合、その理由")
 
 
+class StylePitchRange(BaseModel):
+    low: float
+    high: float
+
+
 class StyleIdNotFoundError(LookupError):
     def __init__(self, style_id: int, *args: object, **kywrds: object) -> None:
         self.style_id = style_id