話者ごとのスタイルの中のボイスサンプルやアイコンを必須じゃなくする

engine_manifest_assets/downloadable_libraries.json

@@ -24,6 +24,7 @@
         },
         "speaker_info": {
           "policy": "",
+          "icon": "",
           "portrait": "",
           "style_infos": [
             {

run.py

@@ -34,7 +34,7 @@
 from voicevox_engine.engine_manifest.EngineManifest import EngineManifest
 from voicevox_engine.engine_manifest.EngineManifestLoader import EngineManifestLoader
 from voicevox_engine.library_manager import LibraryManager
-from voicevox_engine.metas.Metas import StyleId
+from voicevox_engine.metas.Metas import StyleId, StyleInfo
 from voicevox_engine.metas.MetasStore import (
     MetasStore,
     construct_lookup,
@@ -885,6 +885,7 @@ def _speaker_info(
         #   speaker_info/
         #       {speaker_uuid_0}/
         #           policy.md
+        #           icon.png
         #           portrait.png
         #           icons/
         #               {id_0}.png
@@ -921,6 +922,9 @@ def _speaker_info(
             # speaker policy
             policy_path = speaker_path / "policy.md"
             policy = policy_path.read_text("utf-8")
+            # speaker icon
+            icon_path = speaker_path / "icon.png"
+            icon = b64encode_str(icon_path.read_bytes())
             # speaker portrait
             portrait_path = speaker_path / "portrait.png"
             portrait = b64encode_str(portrait_path.read_bytes())
@@ -930,29 +934,31 @@ def _speaker_info(
                 id = style.id
                 # style icon
                 style_icon_path = speaker_path / "icons" / f"{id}.png"
-                icon = b64encode_str(style_icon_path.read_bytes())
+                style_icon = None
+                if style_icon_path.exists():
+                    style_icon = b64encode_str(style_icon_path.read_bytes())
                 # style portrait
                 style_portrait_path = speaker_path / "portraits" / f"{id}.png"
                 style_portrait = None
                 if style_portrait_path.exists():
                     style_portrait = b64encode_str(style_portrait_path.read_bytes())
                 # voice samples
-                voice_samples = [
-                    b64encode_str(
-                        (
-                            speaker_path
-                            / "voice_samples/{}_{}.wav".format(id, str(j + 1).zfill(3))
-                        ).read_bytes()
-                    )
+                voice_sample_paths = [
+                    speaker_path / "voice_samples" / f"{id}_{str(j + 1).zfill(3)}.wav"
                     for j in range(3)
                 ]
+                voice_samples = None
+                if all([p.exists() for p in voice_sample_paths]):
+                    voice_samples = [
+                        b64encode_str(p.read_bytes()) for p in voice_sample_paths
+                    ]
                 style_infos.append(
-                    {
-                        "id": id,
-                        "icon": icon,
-                        "portrait": style_portrait,
-                        "voice_samples": voice_samples,
-                    }
+                    StyleInfo(
+                        id=id,
+                        icon=style_icon,
+                        portrait=style_portrait,
+                        voice_samples=voice_samples,
+                    )
                 )
         except FileNotFoundError:
             import traceback
@@ -964,6 +970,7 @@ def _speaker_info(
 
         ret_data = SpeakerInfo(
             policy=policy,
+            icon=icon,
             portrait=portrait,
             style_infos=style_infos,
         )

voicevox_engine/dev/core/mock.py

@@ -22,14 +22,15 @@ def __init__(
     def metas(self) -> str:
         return json.dumps(
             [
-                # トーク２つ・ハミング２つ
+                # トーク２つ・ハミング２つ・ソングティーチャー１つ
                 {
                     "name": "dummy1",
                     "styles": [
                         {"name": "style0", "id": 0},
                         {"name": "style1", "id": 2},
                         {"name": "style2", "id": 4, "type": "frame_decode"},
                         {"name": "style3", "id": 6, "type": "frame_decode"},
+                        {"name": "style4", "id": 6000, "type": "singing_teacher"},
                     ],
                     "speaker_uuid": "7ffcb7ce-00ec-4bdc-82cd-45a8889e43ff",
                     "version": "mock",

voicevox_engine/metas/Metas.py

@@ -84,11 +84,13 @@ class StyleInfo(BaseModel):
     """
 
     id: StyleId = Field(title="スタイルID")
-    icon: str = Field(title="当該スタイルのアイコンをbase64エンコードしたもの")
+    icon: Optional[str] = Field(
+        title="当該スタイルのアイコンをbase64エンコードしたもの"
+    )
     portrait: Optional[str] = Field(
         title="当該スタイルのportrait.pngをbase64エンコードしたもの"
     )
-    voice_samples: List[str] = Field(
+    voice_samples: Optional[List[str]] = Field(
         title="voice_sampleのwavファイルをbase64エンコードしたもの"
     )
 
@@ -99,5 +101,6 @@ class SpeakerInfo(BaseModel):
     """
 
     policy: str = Field(title="policy.md")
+    icon: str = Field(title="アイコンをbase64エンコードしたもの")
     portrait: str = Field(title="portrait.pngをbase64エンコードしたもの")
     style_infos: List[StyleInfo] = Field(title="スタイルの追加情報")