feat!: [downloader] c-apiとadditional-librariesのディレクトリを掘る

.github/workflows/test.yml

@@ -180,6 +180,7 @@ jobs:
           cargo xtask update-c-header --verify
           git diff
 
+  # TODO: "build-and-test-…"にする
   build-unix-cpp-example:
     strategy:
       fail-fast: false
@@ -200,10 +201,10 @@ jobs:
         run: cargo build -p voicevox_core_c_api --features load-onnxruntime -v
       - name: 必要なfileをunix用exampleのディレクトリに移動させる
         run: |
-          mkdir -p example/cpp/unix/voicevox_core/
+          mkdir -p example/cpp/unix/voicevox_core/c_api
           sed 's:^//\(#define VOICEVOX_LOAD_ONNXRUNTIME\)$:\1:' \
             crates/voicevox_core_c_api/include/voicevox_core.h \
-            > example/cpp/unix/voicevox_core/voicevox_core.h
+            > example/cpp/unix/voicevox_core/c_api/voicevox_core.h
           cp -v target/debug/libvoicevox_core.{so,dylib} example/cpp/unix/voicevox_core/ || true
           cp -v target/debug/libonnxruntime.so.* example/cpp/unix/voicevox_core/ || true
           cp -v target/debug/libonnxruntime.*.dylib example/cpp/unix/voicevox_core/ || true
@@ -221,6 +222,7 @@ jobs:
           cmake -S . -B build
           cmake --build build
 
+  # TODO: "build-and-test-…"にする
   build-windows-cpp-example:
     strategy:
       fail-fast: false

crates/downloader/src/main.rs

@@ -371,15 +371,15 @@ async fn main() -> anyhow::Result<()> {
         tasks.spawn(download_and_extract_from_gh(
             c_api,
             Stripping::FirstDir,
-            &output,
+            output.join("c_api"),
             &progresses,
         )?);
     }
     if let Some(onnxruntime) = onnxruntime {
         tasks.spawn(download_and_extract_from_gh(
             onnxruntime,
             Stripping::FirstDir,
-            &output.join("onnxruntime"),
+            output.join("onnxruntime"),
             &progresses,
         )?);
     }
@@ -388,23 +388,19 @@ async fn main() -> anyhow::Result<()> {
             tasks.spawn(download_and_extract_from_gh(
                 additional_libraries,
                 Stripping::FirstDir,
-                &output,
+                output.join("additional_libraries"),
                 &progresses,
             )?);
         }
     }
     if let Some(models) = models {
-        tasks.spawn(download_models(
-            models,
-            &output.join("models"),
-            &progresses,
-        )?);
+        tasks.spawn(download_models(models, output.join("models"), &progresses)?);
     }
     if targets.contains(&DownloadTarget::Dict) {
         tasks.spawn(download_and_extract_from_url(
             &OPEN_JTALK_DIC_URL,
             Stripping::None,
-            &output,
+            output.join("dict"),
             &progresses,
         )?);
     }
@@ -704,10 +700,9 @@ fn download_and_extract_from_gh(
         ..
     }: GhAsset,
     stripping: Stripping,
-    output: &Path,
+    output: PathBuf,
     progresses: &MultiProgress,
 ) -> anyhow::Result<impl Future<Output = anyhow::Result<()>>> {
-    let output = output.to_owned();
     let archive_kind = ArchiveKind::from_filename(&name)?;
     let pb = add_progress_bar(progresses, size as _, name);
 
@@ -734,10 +729,9 @@ fn download_and_extract_from_gh(
 fn download_and_extract_from_url(
     url: &'static Url,
     stripping: Stripping,
-    output: &Path,
+    output: PathBuf,
     progresses: &MultiProgress,
 ) -> anyhow::Result<impl Future<Output = anyhow::Result<()>>> {
-    let output = output.to_owned();
     let name = url
         .path_segments()
         .and_then(|s| s.last())
@@ -769,10 +763,9 @@ fn download_models(
         models,
         ..
     }: ModelsWithTerms,
-    output: &Path,
+    output: PathBuf,
     progresses: &MultiProgress,
 ) -> anyhow::Result<impl Future<Output = anyhow::Result<()>>> {
-    let output = output.to_owned();
     let reqwest = reqwest::Client::builder().build()?;
 
     let models = models

docs/guide/user/usage.md

@@ -65,11 +65,11 @@ from pprint import pprint
 from voicevox_core.blocking import Onnxruntime, OpenJtalk, Synthesizer, VoiceModelFile
 
 # 1. Synthesizerの初期化
-open_jtalk_dict_dir = "open_jtalk_dic_utf_8-1.11"
+open_jtalk_dict_dir = "dict/open_jtalk_dic_utf_8-1.11"
 synthesizer = Synthesizer(Onnxruntime.load_once(), OpenJtalk(open_jtalk_dict_dir))
 
 # 2. 音声モデルの読み込み
-with VoiceModelFile.open("model/0.vvm") as model:
+with VoiceModelFile.open("models/vvms/0.vvm") as model:
     synthesizer.load_voice_model(model)
 
 # 3. テキスト音声合成
@@ -89,7 +89,7 @@ AIエンジンの`Onnxruntime`のインスタンスと、辞書などを取り
 VVM ファイルから`VoiceModelFile`インスタンスを作成し、`Synthesizer`に読み込ませます。その VVM ファイルにどの声が含まれているかは`VoiceModelFile`の`.metas`や[音声モデルと声の対応表](https://github.com/VOICEVOX/voicevox_fat_resource/blob/main/core/model/README.md#%E9%9F%B3%E5%A3%B0%E3%83%A2%E3%83%87%E3%83%ABvvm%E3%83%95%E3%82%A1%E3%82%A4%E3%83%AB%E3%81%A8%E5%A3%B0%E3%82%AD%E3%83%A3%E3%83%A9%E3%82%AF%E3%82%BF%E3%83%BC%E3%82%B9%E3%82%BF%E3%82%A4%E3%83%AB%E5%90%8D%E3%81%A8%E3%82%B9%E3%82%BF%E3%82%A4%E3%83%AB-id-%E3%81%AE%E5%AF%BE%E5%BF%9C%E8%A1%A8)で確認できます。
 
 ```python
-with VoiceModelFile.open("model/0.vvm") as model:
+with VoiceModelFile.open("models/vvms/0.vvm") as model:
     pprint(model.metas)
 ```
 

example/cpp/unix/CMakeLists.txt

@@ -5,9 +5,9 @@ project(SimpleTTS)
 add_executable(simple_tts simple_tts.cpp)
 set_property(TARGET simple_tts PROPERTY CXX_STANDARD 17)
 
-file(GLOB ONNXRUNTIME_SHARED_LIB ./libonnxruntime.so.* ./libonnxruntime.*.dylib)
-target_link_directories(simple_tts PRIVATE ./voicevox_core)
+file(GLOB ONNXRUNTIME_SHARED_LIB ./libonnxruntime.so.* ./libonnxruntime.*.dylib) # TODO: なにこれ
+target_link_directories(simple_tts PRIVATE ./voicevox_core/c_api/)
 
 
-file(GLOB CORE_LIB ./voicevox_core/libvoicevox_core.so.* ./voicevox_core/libvoicevox_core.*.dylib)
+file(GLOB CORE_LIB ./voicevox_core/c_api/libvoicevox_core.so.* ./voicevox_core/c_api/libvoicevox_core.*.dylib) # TODO: なにこれ
 target_link_libraries(simple_tts voicevox_core)

example/cpp/unix/simple_tts.cpp

@@ -3,7 +3,7 @@
 #include <iostream>
 #include <string>
 
-#include "voicevox_core/voicevox_core.h"
+#include "voicevox_core/c_api/voicevox_core.h"
 
 #define STYLE_ID 0
 #define OUTPUT_WAV_NAME "audio.wav"
@@ -14,14 +14,18 @@ int main(int argc, char *argv[]) {
     return 0;
   }
 
-  std::string open_jtalk_dict_path("voicevox_core/open_jtalk_dic_utf_8-1.11");
+  std::string open_jtalk_dict_path(
+      "voicevox_core/dict/open_jtalk_dic_utf_8-1.11");
   std::string text(argv[1]);
 
   std::cout << "coreの初期化中..." << std::endl;
 
   auto initialize_options = voicevox_make_default_initialize_options();
   const VoicevoxOnnxruntime* onnxruntime;
   auto load_ort_options = voicevox_make_default_load_onnxruntime_options();
+  std::string ort_filename = "./voicevox_core/onnxruntime/lib/";
+  ort_filename += voicevox_get_onnxruntime_lib_versioned_filename();
+  load_ort_options.filename = ort_filename.c_str();
   auto result = voicevox_onnxruntime_load_once(load_ort_options, &onnxruntime);
   if (result != VOICEVOX_RESULT_OK){
     std::cerr << voicevox_error_result_to_message(result) << std::endl;
@@ -42,7 +46,7 @@ int main(int argc, char *argv[]) {
   voicevox_open_jtalk_rc_delete(open_jtalk);
 
   for (auto const& entry :
-       std::filesystem::directory_iterator{"./voicevox_core/model"}) {
+       std::filesystem::directory_iterator{"./voicevox_core/models/vvms"}) {
     const auto path = entry.path();
     if (path.extension() != ".vvm") {
       continue;

example/cpp/windows/README.md

@@ -22,9 +22,9 @@ Visual Studio Installerを使用しインストールしてください。
 
 - simple_tts\bin\x64\Debug に配置
   - voicevox_core.dll
-  - onnxruntime.dll
-  - onnxruntime_providers_shared.dll
-  - modelフォルダ
+  - onnxruntime.dll (TODO: 要らないはず)
+  - onnxruntime_providers_shared.dll (TODO: 要らないはず)
+  - modelフォルダ (TODO: 要らないはず)
 
 - simple_tts\lib\x64 に配置
   - voicevox_core.lib

example/python/README.md

@@ -18,8 +18,6 @@ https://github.com/VOICEVOX/voicevox_core/releases/latest
 
 2. ダウンローダーを使って環境構築します。
 
-FIXME: 今は`--exclude core`がある
-
 linux/mac の場合
 
 download-linux-x64 のところはアーキテクチャや OS によって適宜読み替えてください。
@@ -29,18 +27,14 @@ https://github.com/VOICEVOX/voicevox_core/releases/latest#%E3%83%80%E3%82%A6%E3%
 ❯ binary=download-linux-x64
 ❯ curl -sSfL https://github.com/VOICEVOX/voicevox_core/releases/latest/download/${binary} -o download
 ❯ chmod +x download
-❯ ./download -o ./example/python
-❯ # いくつかのファイルは不要なので消すことができます
-❯ #rm -r ./example/python/{model,VERSION,*voicevox_core*}
+❯ ./download -o ./example/python --exclude c-api
 ```
 
 windows の場合
 
 ```console
 ❯ Invoke-WebRequest https://github.com/VOICEVOX/voicevox_core/releases/latest/download/download-windows-x64.exe -OutFile ./download.exe
-❯ ./download -o ./example/python
-❯ # いくつかのファイルは不要なので消すことができます
-❯ #Remove-Item -Recurse ./example/python/model,./example/python/VERSION,./example/python/*voicevox_core*
+❯ ./download -o ./example/python --exclude c-api
 ```
 
 TODO:
@@ -71,16 +65,23 @@ optional arguments:
 ## 実行例
 
 ```console
-❯ python ./run.py ../../crates/test_util/data/model/sample.vvm
-[DEBUG] __main__: voicevox_core.supported_devices()=SupportedDevices(cpu=True, cuda=False, dml=False)
-[INFO] __main__: Initializing (acceleration_mode=<AccelerationMode.AUTO: 'AUTO'>, open_jtalk_dict_dir=PosixPath('open_jtalk_dic_utf_8-1.11'))
-[DEBUG] __main__: synthesizer.metas=[]
-[DEBUG] __main__: synthesizer.is_gpu_mode=False
-[INFO] __main__: Loading `../../crates/test_util/data/model/sample.vvm`
+❯ python ./run.py ./models/vvms/0.vvm
+[INFO] __main__: Loading ONNX Runtime (args.onnxruntime='./onnxruntime/lib/libvoicevox_onnxruntime.so.1.17.3')
+[DEBUG] __main__: onnxruntime.supported_devices()=SupportedDevices(cpu=True, cuda=True, dml=False)
+[INFO] __main__: Initializing (args.mode=<AccelerationMode.AUTO: 'AUTO'>, args.dict_dir=PosixPath('dict/open_jtalk_dic_utf_8-1.11'))
+[INFO] voicevox_core.synthesizer: GPUをテストします:
+[INFO] voicevox_core.synthesizer:   * CUDA (device_id=0): OK
+[INFO] voicevox_core.synthesizer:   * DirectML (device_id=0): 現在ロードされているONNX Runtimeでは利用できません
+[INFO] voicevox_core.synthesizer: CUDA (device_id=0)を利用します
+[DEBUG] __main__: synthesizer.metas()=[]
+[DEBUG] __main__: synthesizer.is_gpu_mode=True
+[INFO] __main__: Loading `models/vvms/0.vvm`
+[WARNING] ort.environment: Some nodes were not assigned to the preferred execution providers which may or may not have an negative impact on performance. e.g. ORT explicitly assigns shape related ops to CPU to improve perf.
+[DEBUG] voicevox_core_python_api: Closing a VoiceModelFile
 [INFO] __main__: Creating an AudioQuery from 'この音声は、ボイスボックスを使用して、出力されています。'
-[INFO] __main__: Synthesizing with {"accent_phrases": [{"moras": [{"text": "コ", "consonant": "k", "consonant_length": 0.0556899, "vowel": "o", "vowel_length": 0.075180575, "pitch": 5.542309}, {"text": "ノ", "consonant": "n", "consonant_length": 0.06551014, "vowel": "o", "vowel_length": 0.09984577, "pitch": 5.6173983}], "accent": 2, "pause_mora": null, "is_interrogative": false}, {"moras": [{"text": "オ", "consonant": null, "consonant_length": null, "vowel": "o", "vowel_length": 0.116150305, "pitch": 5.7063766}, {"text": "ン", "consonant": null, "consonant_length": null, "vowel": "N", "vowel_length": 0.044380233, "pitch": 5.785717}, {"text": "セ", "consonant": "s", "consonant_length": 0.07719758, "vowel": "e", "vowel_length": 0.08653869, "pitch": 5.662092}, {"text": "エ", "consonant": null, "consonant_length": null, "vowel": "e", "vowel_length": 0.08311573, "pitch": 5.532917}, {"text": "ワ", "consonant": "w", "consonant_length": 0.06373148, "vowel": "a", "vowel_length": 0.16219379, "pitch": 5.293258}], "accent": 1, "pause_mora": {"text": "、", "consonant": null, "consonant_length": null, "vowel": "pau", "vowel_length": 0.35826492, "pitch": 0.0}, "is_interrogative": false}, {"moras": [{"text": "ボ", "consonant": "b", "consonant_length": 0.047082342, "vowel": "o", "vowel_length": 0.12611786, "pitch": 5.583892}, {"text": "イ", "consonant": null, "consonant_length": null, "vowel": "i", "vowel_length": 0.059451744, "pitch": 5.7947493}, {"text": "ス", "consonant": "s", "consonant_length": 0.089278996, "vowel": "u", "vowel_length": 0.11847979, "pitch": 5.818695}, {"text": "ボ", "consonant": "b", "consonant_length": 0.06535433, "vowel": "o", "vowel_length": 0.120458946, "pitch": 5.7965107}, {"text": "ッ", "consonant": null, "consonant_length": null, "vowel": "cl", "vowel_length": 0.06940381, "pitch": 0.0}, {"text": "ク", "consonant": "k", "consonant_length": 0.053739145, "vowel": "U", "vowel_length": 0.05395376, "pitch": 0.0}, {"text": "ス", "consonant": "s", "consonant_length": 0.10222931, "vowel": "u", "vowel_length": 0.071811065, "pitch": 5.8024883}, {"text": "オ", "consonant": null, "consonant_length": null, "vowel": "o", "vowel_length": 0.11092262, "pitch": 5.5036163}], "accent": 4, "pause_mora": null, "is_interrogative": false}, {"moras": [{"text": "シ", "consonant": "sh", "consonant_length": 0.09327768, "vowel": "i", "vowel_length": 0.09126951, "pitch": 5.369444}, {"text": "ヨ", "consonant": "y", "consonant_length": 0.06251812, "vowel": "o", "vowel_length": 0.07805054, "pitch": 5.5021667}, {"text": "オ", "consonant": null, "consonant_length": null, "vowel": "o", "vowel_length": 0.09904325, "pitch": 5.5219536}], "accent": 3, "pause_mora": null, "is_interrogative": false}, {"moras": [{"text": "シ", "consonant": "sh", "consonant_length": 0.04879771, "vowel": "I", "vowel_length": 0.06514315, "pitch": 0.0}, {"text": "テ", "consonant": "t", "consonant_length": 0.0840496, "vowel": "e", "vowel_length": 0.19438823, "pitch": 5.4875555}], "accent": 2, "pause_mora": {"text": "、", "consonant": null, "consonant_length": null, "vowel": "pau", "vowel_length": 0.35208154, "pitch": 0.0}, "is_interrogative": false}, {"moras": [{"text": "シュ", "consonant": "sh", "consonant_length": 0.05436731, "vowel": "U", "vowel_length": 0.06044446, "pitch": 0.0}, {"text": "ツ", "consonant": "ts", "consonant_length": 0.102865085, "vowel": "u", "vowel_length": 0.057028636, "pitch": 5.6402535}, {"text": "リョ", "consonant": "ry", "consonant_length": 0.058293864, "vowel": "o", "vowel_length": 0.080050275, "pitch": 5.6997967}, {"text": "ク", "consonant": "k", "consonant_length": 0.054767884, "vowel": "U", "vowel_length": 0.042932786, "pitch": 0.0}], "accent": 2, "pause_mora": null, "is_interrogative": false}, {"moras": [{"text": "サ", "consonant": "s", "consonant_length": 0.08067487, "vowel": "a", "vowel_length": 0.07377973, "pitch": 5.652378}, {"text": "レ", "consonant": "r", "consonant_length": 0.040600352, "vowel": "e", "vowel_length": 0.079322875, "pitch": 5.6290326}, {"text": "テ", "consonant": "t", "consonant_length": 0.06773268, "vowel": "e", "vowel_length": 0.08347456, "pitch": 5.6427326}], "accent": 3, "pause_mora": null, "is_interrogative": false}, {"moras": [{"text": "イ", "consonant": null, "consonant_length": null, "vowel": "i", "vowel_length": 0.07542324, "pitch": 5.641289}, {"text": "マ", "consonant": "m", "consonant_length": 0.066299975, "vowel": "a", "vowel_length": 0.107257664, "pitch": 5.6201453}, {"text": "ス", "consonant": "s", "consonant_length": 0.07186453, "vowel": "U", "vowel_length": 0.1163103, "pitch": 0.0}], "accent": 2, "pause_mora": null, "is_interrogative": false}], "speed_scale": 1.0, "pitch_scale": 0.0, "intonation_scale": 1.0, "volume_scale": 1.0, "pre_phoneme_length": 0.1, "post_phoneme_length": 0.1, "output_sampling_rate": 24000, "output_stereo": false, "kana": "コノ'/オ'ンセエワ、ボイスボ'ッ_クスオ/シヨオ'/_シテ'、_シュツ' リョ_ク/サレテ'/イマ'_ス"}
+[INFO] __main__: Synthesizing with {"accent_phrases": […], "speed_scale": 1.0, "pitch_scale": 0.0, "intonation_scale": 1.0, "volume_scale": 1.0, "pre_phoneme_length": 0.1, "post_phoneme_length": 0.1, "output_sampling_rate": 24000, "output_stereo": false, "pause_length": null, "pause_length_scale": 1.0, "kana": "コノ'/オ'ンセエワ、ボイスボ'ッ_クスオ/シヨオ'/_シテ'、_シュツ'リョ_ク/サレテ'/イマ'_ス"}
 [INFO] __main__: Wrote `output.wav`
-[DEBUG] voicevox_core_python_api: Destructing a VoicevoxCore
+[WARNING] voicevox_core_python_api: デストラクタにより`Synthesizer`のクローズを行います。通常は、可能な限り`__exit__`でクローズするようにして下さい
 ```
 
 正常に実行されれば音声合成の結果である wav ファイルが生成されます。

example/python/run.py

@@ -36,12 +36,12 @@ def parse_args() -> "Args":
         )
         argparser.add_argument(
             "--onnxruntime",
-            default=Onnxruntime.LIB_VERSIONED_FILENAME,
+            default=f"./onnxruntime/lib/{Onnxruntime.LIB_VERSIONED_FILENAME}",
             help="ONNX Runtimeのライブラリのfilename",
         )
         argparser.add_argument(
             "--dict-dir",
-            default="./open_jtalk_dic_utf_8-1.11",
+            default="./dict/open_jtalk_dic_utf_8-1.11",
             type=Path,
             help="Open JTalkの辞書ディレクトリ",
         )