voicevox: init at 0.20.0 (#319403)

TomaSajt · web-flow · commit 3add6c84dba5 · 2024-12-21T12:10:03.000+01:00
diff --git a/pkgs/by-name/vo/voicevox-core/package.nix b/pkgs/by-name/vo/voicevox-core/package.nix
@@ -0,0 +1,74 @@
+{
+  lib,
+  stdenv,
+  fetchzip,
+  autoPatchelfHook,
+}:
+
+let
+  platformInfoTable = {
+    "x86_64-linux" = {
+      id = "linux-x64";
+      hash = "sha256-/PD5e0bWgnIsIrvyOypoJw30VkgbOFWV1NJpPS2G0WM=";
+    };
+    "aarch64-linux" = {
+      id = "linux-arm64";
+      hash = "sha256-zfiorXZyIISZPXPwmcdYeHceDmQXkUhsvTkNZScg648=";
+    };
+    "x86_64-darwin" = {
+      id = "osx-x64";
+      hash = "sha256-cdNdV1fVPkz6B7vtKZiPsLQGqnIiDtYa9KTcwSkjdJg=";
+    };
+    "aarch64-darwin" = {
+      id = "osx-arm64";
+      hash = "sha256-Z1dq2t/HBQulbPF23ZCihOrcZHMpTXEQ6yXKORZaFPk=";
+    };
+  };
+
+  platformInfo =
+    platformInfoTable.${stdenv.hostPlatform.system}
+      or (throw "Unsupported system: ${stdenv.hostPlatform.system}");
+in
+stdenv.mkDerivation (finalAttrs: {
+  pname = "voicevox-core";
+  version = "0.15.4";
+
+  # Note: Only the prebuilt binaries are able to decrypt the encrypted voice models
+  src = fetchzip {
+    url = "https://github.com/VOICEVOX/voicevox_core/releases/download/${finalAttrs.version}/voicevox_core-${platformInfo.id}-cpu-${finalAttrs.version}.zip";
+    inherit (platformInfo) hash;
+  };
+
+  nativeBuildInputs = lib.optionals stdenv.hostPlatform.isLinux [ autoPatchelfHook ];
+
+  buildInputs = [ stdenv.cc.cc.lib ];
+
+  installPhase = ''
+    runHook preInstall
+
+    install -Dm755 libonnxruntime.* libvoicevox_core.* -t $out/lib
+    install -Dm644 model/* -t $out/lib/model
+    install -Dm644 *.h -t $out/include
+    install -Dm644 README.txt -t $out/share/doc/voicevox-core
+
+    runHook postInstall
+  '';
+
+  meta = {
+    changelog = "https://github.com/VOICEVOX/voicevox_core/releases/tag/${finalAttrs.version}";
+    description = "Core library for the VOICEVOX speech synthesis software";
+    homepage = "https://github.com/VOICEVOX/voicevox_core";
+    license = with lib.licenses; [
+      mit
+      ({
+        name = "VOICEVOX Core Library Terms of Use";
+        url = "https://github.com/VOICEVOX/voicevox_resource/blob/main/core/README.md";
+        free = false;
+        redistributable = true;
+      })
+    ];
+    maintainers = with lib.maintainers; [ tomasajt ];
+    platforms = lib.attrNames platformInfoTable;
+    sourceProvenance = with lib.sourceTypes; [ binaryNativeCode ];
+  };
+})
diff --git a/pkgs/by-name/vo/voicevox-engine/make-installable.patch b/pkgs/by-name/vo/voicevox-engine/make-installable.patch
@@ -0,0 +1,21 @@
+diff --git a/pyproject.toml b/pyproject.toml
+index fa23446..6a7705c 100644
+--- a/pyproject.toml
++++ b/pyproject.toml
+@@ -40,7 +40,15 @@ use_parentheses = true
+ datas = "datas" # PyInstaller's argument
+ 
+ [tool.poetry]
+-package-mode = false
++name = "voicevox-engine"
++version = "@version@"
++authors = []
++description = ""
++packages = [ { include = "voicevox_engine" } ]
++include = [ "resources/**/*", "run.py", "engine_manifest.json", "presets.yaml"]
++
++[tool.poetry.scripts]
++voicevox-engine = "run:main"
+ 
+ [tool.poetry.dependencies]
+ python = "~3.11"
diff --git a/pkgs/by-name/vo/voicevox-engine/mecab-remove-deprecated.patch b/pkgs/by-name/vo/voicevox-engine/mecab-remove-deprecated.patch
@@ -0,0 +1,92 @@
+diff --git a/lib/open_jtalk/src/mecab/src/char_property.h b/lib/open_jtalk/src/mecab/src/char_property.h
+index 35f4b05..9c904ba 100644
+--- a/lib/open_jtalk/src/mecab/src/char_property.h
++++ b/lib/open_jtalk/src/mecab/src/char_property.h
+@@ -37,7 +37,7 @@ class CharProperty {
+   inline const char *seekToOtherType(const char *begin, const char *end,
+                                      CharInfo c, CharInfo *fail,
+                                      size_t *mblen, size_t *clen) const {
+-    register const char *p =  begin;
++    const char *p =  begin;
+     *clen = 0;
+     while (p != end && c.isKindOf(*fail = getCharInfo(p, end, mblen))) {
+       p += *mblen;
+diff --git a/lib/open_jtalk/src/mecab/src/darts.h b/lib/open_jtalk/src/mecab/src/darts.h
+index 91b2eae..d6736cf 100644
+--- a/lib/open_jtalk/src/mecab/src/darts.h
++++ b/lib/open_jtalk/src/mecab/src/darts.h
+@@ -404,10 +404,10 @@ class DoubleArrayImpl {
+     T result;
+     set_result(result, -1, 0);
+ 
+-    register array_type_  b = array_[node_pos].base;
+-    register array_u_type_ p;
++    array_type_  b = array_[node_pos].base;
++    array_u_type_ p;
+ 
+-    for (register size_t i = 0; i < len; ++i) {
++    for (size_t i = 0; i < len; ++i) {
+       p = b +(node_u_type_)(key[i]) + 1;
+       if (static_cast<array_u_type_>(b) == array_[p].check)
+         b = array_[p].base;
+@@ -431,12 +431,12 @@ class DoubleArrayImpl {
+                             size_t node_pos = 0) const {
+     if (!len) len = length_func_()(key);
+ 
+-    register array_type_  b   = array_[node_pos].base;
+-    register size_t     num = 0;
+-    register array_type_  n;
+-    register array_u_type_ p;
++    array_type_  b   = array_[node_pos].base;
++    size_t     num = 0;
++    array_type_  n;
++    array_u_type_ p;
+ 
+-    for (register size_t i = 0; i < len; ++i) {
++    for (size_t i = 0; i < len; ++i) {
+       p = b;  // + 0;
+       n = array_[p].base;
+       if ((array_u_type_) b == array_[p].check && n < 0) {
+@@ -469,8 +469,8 @@ class DoubleArrayImpl {
+                       size_t len = 0) const {
+     if (!len) len = length_func_()(key);
+ 
+-    register array_type_  b = array_[node_pos].base;
+-    register array_u_type_ p;
++    array_type_  b = array_[node_pos].base;
++    array_u_type_ p;
+ 
+     for (; key_pos < len; ++key_pos) {
+       p = b +(node_u_type_)(key[key_pos]) + 1;
+diff --git a/lib/open_jtalk/src/mecab/src/dictionary.cpp b/lib/open_jtalk/src/mecab/src/dictionary.cpp
+index 5717d4d..3ab6e1f 100644
+--- a/lib/open_jtalk/src/mecab/src/dictionary.cpp
++++ b/lib/open_jtalk/src/mecab/src/dictionary.cpp
+@@ -66,7 +66,7 @@ int progress_bar_darts(size_t current, size_t total) {
+ }
+ 
+ template <typename T1, typename T2>
+-struct pair_1st_cmp: public std::binary_function<bool, T1, T2> {
++struct pair_1st_cmp {
+   bool operator()(const std::pair<T1, T2> &x1,
+                   const std::pair<T1, T2> &x2)  {
+     return x1.first < x2.first;
+diff --git a/lib/open_jtalk/src/mecab/src/viterbi.cpp b/lib/open_jtalk/src/mecab/src/viterbi.cpp
+index 6277fe9..5ccefb7 100644
+--- a/lib/open_jtalk/src/mecab/src/viterbi.cpp
++++ b/lib/open_jtalk/src/mecab/src/viterbi.cpp
+@@ -318,11 +318,11 @@ template <bool IsAllPath> bool connect(size_t pos, Node *rnode,
+                                        const Connector *connector,
+                                        Allocator<Node, Path> *allocator) {
+   for (;rnode; rnode = rnode->bnext) {
+-    register long best_cost = 2147483647;
++    long best_cost = 2147483647;
+     Node* best_node = 0;
+     for (Node *lnode = end_node_list[pos]; lnode; lnode = lnode->enext) {
+-      register int lcost = connector->cost(lnode, rnode);  // local cost
+-      register long cost = lnode->cost + lcost;
++      int lcost = connector->cost(lnode, rnode);  // local cost
++      long cost = lnode->cost + lcost;
+ 
+       if (cost < best_cost) {
+         best_node  = lnode;
diff --git a/pkgs/by-name/vo/voicevox-engine/package.nix b/pkgs/by-name/vo/voicevox-engine/package.nix
@@ -0,0 +1,126 @@
+{
+  lib,
+  fetchFromGitHub,
+  python3Packages,
+  replaceVars,
+  voicevox-core,
+}:
+
+python3Packages.buildPythonApplication rec {
+  pname = "voicevox-engine";
+  version = "0.20.0";
+  pyproject = true;
+
+  src = fetchFromGitHub {
+    owner = "VOICEVOX";
+    repo = "voicevox_engine";
+    rev = "refs/tags/${version}";
+    hash = "sha256-Gib5R7oleg+XXyu2V65EqrflQ1oiAR7a09a0MFhSITc=";
+  };
+
+  patches = [
+    # the upstream package only uses poetry for dependency management, not for package definition
+    # this patch makes the package installable via poetry-core
+    (replaceVars ./make-installable.patch {
+      inherit version;
+    })
+  ];
+
+  build-system = with python3Packages; [
+    poetry-core
+  ];
+
+  dependencies =
+    [
+      passthru.pyopenjtalk
+    ]
+    ++ (with python3Packages; [
+      numpy
+      fastapi
+      jinja2
+      python-multipart
+      uvicorn
+      soundfile
+      pyyaml
+      pyworld
+      semver
+      platformdirs
+      soxr
+      pydantic
+      starlette
+    ]);
+
+  pythonRemoveDeps = [
+    # upstream wants fastapi-slim, but we provide fastapi instead
+    "fastapi-slim"
+  ];
+
+  pythonRelaxDeps = true;
+
+  preConfigure = ''
+    # copy demo metadata to temporary directory
+    mv resources/character_info test_character_info
+
+    # populate the `character_info` directory with the actual model metadata instead of the demo metadata
+    cp -r --no-preserve=all ${passthru.resources}/character_info resources/character_info
+
+    # the `character_info` directory copied from `resources` doesn't exactly have the expected format,
+    # so we transform them to be acceptable by `voicevox-engine`
+    pushd resources/character_info
+    for dir in *; do
+        # remove unused large files
+        rm $dir/*/*.png_large
+
+        # rename directory from "$name_$uuid" to "$uuid"
+        mv $dir ''${dir#*"_"}
+    done
+    popd
+  '';
+
+  makeWrapperArgs = [
+    ''--add-flags "--voicelib_dir=${voicevox-core}/lib"''
+  ];
+
+  preCheck = ''
+    # some tests assume $HOME actually exists
+    export HOME=$(mktemp -d)
+
+    # since the actual metadata files have been installed to `$out` by this point,
+    # we can move the demo metadata back to its place for the tests to succeed
+    rm -r resources/character_info
+    mv test_character_info resources/character_info
+  '';
+
+  disabledTests = [
+    # this test checks the behaviour of openapi
+    # one of the functions returns a slightly different output due to openapi version differences
+    "test_OpenAPIの形が変わっていないことを確認"
+  ];
+
+  nativeCheckInputs = with python3Packages; [
+    pytestCheckHook
+    syrupy
+    httpx
+  ];
+
+  passthru = {
+    resources = fetchFromGitHub {
+      owner = "VOICEVOX";
+      repo = "voicevox_resource";
+      rev = "refs/tags/${version}";
+      hash = "sha256-m888DF9qgGbK30RSwNnAoT9D0tRJk6cD5QY72FRkatM=";
+    };
+
+    pyopenjtalk = python3Packages.callPackage ./pyopenjtalk.nix { };
+  };
+
+  meta = {
+    changelog = "https://github.com/VOICEVOX/voicevox_engine/releases/tag/${version}";
+    description = "Engine for the VOICEVOX speech synthesis software";
+    homepage = "https://github.com/VOICEVOX/voicevox_engine";
+    license = lib.licenses.lgpl3Only;
+    mainProgram = "voicevox-engine";
+    maintainers = with lib.maintainers; [ tomasajt ];
+    platforms = lib.platforms.linux ++ lib.platforms.darwin;
+  };
+}
diff --git a/pkgs/by-name/vo/voicevox-engine/pyopenjtalk.nix b/pkgs/by-name/vo/voicevox-engine/pyopenjtalk.nix
@@ -0,0 +1,75 @@
+{
+  lib,
+  python,
+  buildPythonPackage,
+  fetchFromGitHub,
+  fetchzip,
+  setuptools,
+  cython_0,
+  cmake,
+  numpy,
+  oldest-supported-numpy,
+  six,
+  tqdm,
+}:
+
+let
+  dic-dirname = "open_jtalk_dic_utf_8-1.11";
+  dic-src = fetchzip {
+    url = "https://github.com/r9y9/open_jtalk/releases/download/v1.11.1/${dic-dirname}.tar.gz";
+    hash = "sha256-+6cHKujNEzmJbpN9Uan6kZKsPdwxRRzT3ZazDnCNi3s=";
+  };
+in
+buildPythonPackage {
+  pname = "pyopenjtalk";
+  version = "0-unstable-2023-09-08";
+  pyproject = true;
+
+  src = fetchFromGitHub {
+    owner = "VOICEVOX";
+    repo = "pyopenjtalk";
+    rev = "b35fc89fe42948a28e33aed886ea145a51113f88";
+    hash = "sha256-DbZkCMdirI6wSRUQSJrkojyjGmViqGeQPO0kSKiw2gE=";
+    fetchSubmodules = true;
+  };
+
+  patches = [
+    # this patch fixes the darwin build
+    # open_jtalk uses mecab, which uses the register keyword and std::binary_function, which are not allowed in c++17
+    # this patch removes them
+    ./mecab-remove-deprecated.patch
+  ];
+
+  postPatch = ''
+    substituteInPlace pyproject.toml \
+        --replace-fail 'setuptools<v60.0' 'setuptools'
+  '';
+
+  build-system = [
+    setuptools
+    cython_0
+    cmake
+    numpy
+    oldest-supported-numpy
+  ];
+
+  dontUseCmakeConfigure = true;
+
+  dependencies = [
+    setuptools # imports pkg_resources at runtime
+    numpy
+    six
+    tqdm
+  ];
+
+  postInstall = ''
+    ln -s ${dic-src} $out/${python.sitePackages}/pyopenjtalk/${dic-dirname}
+  '';
+
+  meta = {
+    description = "VOICEVOX's fork of the pyopenjtalk text-to-speech library";
+    homepage = "https://github.com/VOICEVOX/pyopenjtalk";
+    license = lib.licenses.mit;
+    maintainers = with lib.maintainers; [ tomasajt ];
+  };
+}
diff --git a/pkgs/by-name/vo/voicevox/hardcode-paths.patch b/pkgs/by-name/vo/voicevox/hardcode-paths.patch
diff --git a/pkgs/by-name/vo/voicevox/package.nix b/pkgs/by-name/vo/voicevox/package.nix
