モーフィングの許可に関する設定およびAPIを追加

run.py

@@ -40,11 +40,12 @@
     ParseKanaError,
     Speaker,
     SpeakerInfo,
+    SpeakerNotFoundError,
     SupportedDevicesInfo,
     UserDictWord,
     WordTypes,
 )
-from voicevox_engine.morphing import synthesis_morphing
+from voicevox_engine.morphing import is_synthesis_morphing_permitted, synthesis_morphing
 from voicevox_engine.morphing import (
     synthesis_morphing_parameter as _synthesis_morphing_parameter,
 )
@@ -483,6 +484,34 @@ def multi_synthesis(
             background=BackgroundTask(delete_file, f.name),
         )
 
+    @app.get(
+        "/is_morphable",
+        response_model=bool,
+        tags=["音声合成"],
+        summary="2人の話者でモーフィングが可能かどうか返す",
+    )
+    def is_morphable(
+        base_speaker: int,
+        target_speaker: int,
+        core_version: Optional[str] = None,
+    ):
+        """
+        指定された2人の話者でモーフィング機能を利用可能か返します。
+        モーフィングの許可/禁止は`/speakers`の`speaker.supportedFeatures.synthesisMorphing`に記載されています。
+        プロパティが存在しない場合は、モーフィングが許可されているとみなします。
+        """
+        engine = get_engine(core_version)
+
+        try:
+            is_permitted = is_synthesis_morphing_permitted(
+                engine, root_dir / "speaker_info", base_speaker, target_speaker
+            )
+            return is_permitted
+        except SpeakerNotFoundError as e:
+            raise HTTPException(
+                status_code=404, detail=f"該当する話者(speaker={e.speaker})が見つかりません"
+            )
+
     @app.post(
         "/synthesis_morphing",
         response_class=FileResponse,
@@ -509,6 +538,20 @@ def _synthesis_morphing(
         """
         engine = get_engine(core_version)
 
+        try:
+            is_permitted = is_synthesis_morphing_permitted(
+                engine, root_dir / "speaker_info", base_speaker, target_speaker
+            )
+            if not is_permitted:
+                raise HTTPException(
+                    status_code=400,
+                    detail="指定された話者ペアでのモーフィングはできません",
+                )
+        except SpeakerNotFoundError as e:
+            raise HTTPException(
+                status_code=404, detail=f"該当する話者(speaker={e.speaker})が見つかりません"
+            )
+
         # 生成したパラメータはキャッシュされる
         morph_param = synthesis_morphing_parameter(
             engine=engine,

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

@@ -1,6 +1,7 @@
 {
     "speakerName": "dummy3",
     "speakerUuid": "dummy3uuid",
+    "supportedFeatures": {"permitedSynthesisMorphing": "NOTHING"},
     "styles": [
         {
             "styleId": 8

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

@@ -1,6 +1,7 @@
 {
     "speakerName": "dummy2",
     "speakerUuid": "dummy2uuid",
+    "supportedFeatures": {"permitedSynthesisMorphing": "SELF_ONLY"},
     "styles": [
         {
             "styleId": 3

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

@@ -1,6 +1,7 @@
 {
     "speakerName": "dummy4",
     "speakerUuid": "dummy4uuid",
+    "supportedFeatures": {"permitedSynthesisMorphing": "ALL"},
     "styles": [
         {
             "styleId": 9

voicevox_engine/model.py

@@ -113,13 +113,36 @@ class SpeakerStyle(BaseModel):
     id: int = Field(title="スタイルID")
 
 
+class SpeakerSupporPermitedSynthesisMorphing(str, Enum):
+    ALL = "ALL"  # 全て許可
+    SELF_ONLY = "SELF_ONLY"  # 同じ話者内でのみ許可
+    NOTHING = "NOTHING"  # 全て禁止
+
+    @classmethod
+    def _missing_(cls, value: object) -> "SpeakerSupporPermitedSynthesisMorphing":
+        return SpeakerSupporPermitedSynthesisMorphing.ALL
+
+
+class SpeakerSupportedFeatures(BaseModel):
+    """
+    話者の対応機能の情報
+    """
+
+    permited_synthesis_morphing: Optional[
+        SpeakerSupporPermitedSynthesisMorphing
+    ] = Field(
+        title="モーフィング機能への対応", default=SpeakerSupporPermitedSynthesisMorphing(None)
+    )
+
+
 class Speaker(BaseModel):
     """
     スピーカー情報
     """
 
     name: str = Field(title="名前")
     speaker_uuid: str = Field(title="スピーカーのUUID")
+    supported_features: Optional[SpeakerSupportedFeatures] = Field(title="スピーカーの対応機能")
     styles: List[SpeakerStyle] = Field(title="スピーカースタイルの一覧")
     version: str = Field("スピーカーのバージョン")
 
@@ -145,6 +168,12 @@ class SpeakerInfo(BaseModel):
     style_infos: List[StyleInfo] = Field(title="スタイルの追加情報")
 
 
+class SpeakerNotFoundError(LookupError):
+    def __init__(self, speaker: int, *args: object, **kywrds: object) -> None:
+        self.speaker = speaker
+        super().__init__(f"speaker {speaker} is not found.", *args, **kywrds)
+
+
 class DownloadableLibrary(BaseModel):
     """
     ダウンロード可能な音声ライブラリの情報（最新情報をwebで取得することを考慮して、ローカルの情報はない）

voicevox_engine/morphing.py

@@ -1,10 +1,18 @@
+import json
 from copy import deepcopy
 from dataclasses import dataclass
+from pathlib import Path
 
 import numpy as np
 import pyworld as pw
 
-from .model import AudioQuery
+from voicevox_engine.synthesis_engine.synthesis_engine_base import SynthesisEngineBase
+
+from .model import (
+    AudioQuery,
+    SpeakerNotFoundError,
+    SpeakerSupporPermitedSynthesisMorphing,
+)
 from .synthesis_engine import SynthesisEngine
 
 
@@ -43,6 +51,80 @@ def create_morphing_parameter(
     )
 
 
+def is_synthesis_morphing_permitted(
+    engine: SynthesisEngineBase,
+    speaker_info_folder: Path,
+    base_speaker: int,
+    target_speaker: int,
+) -> bool:
+    """
+    指定されたspeakerがモーフィング可能かどうか返す
+    speakerが見つからない場合はSpeakerNotFoundErrorを送出する
+    """
+
+    core_speakers = json.loads(engine.speakers)
+    base_speaker_core_info, target_speaker_core_info = None, None
+    for speaker in core_speakers:
+        style_id_arr = tuple(style["id"] for style in speaker["styles"])
+        if base_speaker_core_info is None and base_speaker in style_id_arr:
+            base_speaker_core_info = speaker
+        if target_speaker_core_info is None and target_speaker in style_id_arr:
+            target_speaker_core_info = speaker
+
+    if base_speaker_core_info is None or target_speaker_core_info is None:
+        raise SpeakerNotFoundError(
+            base_speaker if base_speaker_core_info is None else target_speaker
+        )
+
+    base_speaker_uuid = base_speaker_core_info["speaker_uuid"]
+    target_speaker_uuid = target_speaker_core_info["speaker_uuid"]
+
+    # FIXME: engineのmetasロード処理をPresetLoaderのように纏める
+    base_speaker_engine_info = json.loads(
+        (speaker_info_folder / f"{base_speaker_uuid}" / "metas.json").read_text(
+            encoding="utf-8"
+        )
+    )
+    target_speaker_engine_info = json.loads(
+        (speaker_info_folder / f"{target_speaker_uuid}" / "metas.json").read_text(
+            encoding="utf-8"
+        )
+    )
+
+    # FIXME: 他にsupported_featuresができたら共通化する
+    base_speaker_morphing_info: SpeakerSupporPermitedSynthesisMorphing = (
+        base_speaker_engine_info.get("supportedFeatures", dict()).get(
+            "permitedSynthesisMorphing", SpeakerSupporPermitedSynthesisMorphing(None)
+        )
+    )
+
+    target_speaker_morphing_info: SpeakerSupporPermitedSynthesisMorphing = (
+        target_speaker_engine_info.get("supportedFeatures", dict()).get(
+            "permitedSynthesisMorphing", SpeakerSupporPermitedSynthesisMorphing(None)
+        )
+    )
+
+    # 禁止されている場合はFalse
+    if (
+        base_speaker_morphing_info == SpeakerSupporPermitedSynthesisMorphing.NOTHING
+        or target_speaker_morphing_info
+        == SpeakerSupporPermitedSynthesisMorphing.NOTHING
+    ):
+        return False
+    # 同一話者のみの場合は同一話者判定
+    if (
+        base_speaker_morphing_info == SpeakerSupporPermitedSynthesisMorphing.SELF_ONLY
+        or target_speaker_morphing_info
+        == SpeakerSupporPermitedSynthesisMorphing.SELF_ONLY
+    ):
+        return base_speaker_uuid == target_speaker_uuid
+    # 念のため許可されているかチェック
+    return (
+        base_speaker_morphing_info == SpeakerSupporPermitedSynthesisMorphing.ALL
+        and target_speaker_morphing_info == SpeakerSupporPermitedSynthesisMorphing.ALL
+    )
+
+
 def synthesis_morphing_parameter(
     engine: SynthesisEngine,
     query: AudioQuery,