Merge pull request #246856 from mweinelt/piper-tts-1.2.0

piper-tts: 0.0.2 -> 1.2.0

8 files changed, 163 insertions, 42 deletions

diff --git a/pkgs/development/libraries/piper-phonemize/default.nix b/pkgs/development/libraries/piper-phonemize/default.nix
new file mode 100644
index 0000000000000..fd1c1ae34b4c5
--- /dev/null
+++ b/pkgs/development/libraries/piper-phonemize/default.nix
@@ -0,0 +1,60 @@
+{ lib
+, stdenv
+, fetchFromGitHub
+
+# build
+, cmake
+, pkg-config
+
+# runtime
+, espeak-ng
+, onnxruntime
+}:
+
+let
+  espeak-ng' = espeak-ng.overrideAttrs (oldAttrs: {
+    version = "1.52-dev";
+    src = fetchFromGitHub {
+      owner = "rhasspy";
+      repo = "espeak-ng";
+      rev = "61504f6b76bf9ebbb39b07d21cff2a02b87c99ff";
+      hash = "sha256-RBHL11L5uazAFsPFwul2QIyJREXk9Uz8HTZx9JqmyIQ=";
+    };
+
+    patches = [
+      ./espeak-mbrola.patch
+    ];
+  });
+in
+stdenv.mkDerivation rec {
+  pname = "piper-phonemize";
+  version = "1.1.0";
+
+  src = fetchFromGitHub {
+    owner = "rhasspy";
+    repo = "piper-phonemize";
+    rev = "refs/tags/v${version}";
+    hash = "sha256-cMer7CSLOXv3jc9huVA3Oy5cjXjOX9XuEXpIWau1BNQ=";
+  };
+
+  nativeBuildInputs = [
+    cmake
+    pkg-config
+  ];
+
+  buildInputs = [
+    espeak-ng'
+    onnxruntime
+  ];
+
+  passthru = {
+    espeak-ng = espeak-ng';
+  };
+
+  meta = with lib; {
+    description = "C++ library for converting text to phonemes for Piper";
+    homepage = "https://github.com/rhasspy/piper-phonemize";
+    license = licenses.mit;
+    maintainers = with maintainers; [ hexa ];
+  };
+}
diff --git a/pkgs/development/libraries/piper-phonemize/espeak-mbrola.patch b/pkgs/development/libraries/piper-phonemize/espeak-mbrola.patch
new file mode 100644
index 0000000000000..9d3f0aeb4abe6
--- /dev/null
+++ b/pkgs/development/libraries/piper-phonemize/espeak-mbrola.patch
@@ -0,0 +1,26 @@
+diff --git a/src/libespeak-ng/mbrowrap.c b/src/libespeak-ng/mbrowrap.c
+index ae137873..9015cc01 100644
+--- a/src/libespeak-ng/mbrowrap.c
++++ b/src/libespeak-ng/mbrowrap.c
+@@ -206,7 +206,7 @@ static int start_mbrola(const char *voice_path)
+ 		signal(SIGTERM, SIG_IGN);
+ 
+ 		snprintf(charbuf, sizeof(charbuf), "%g", mbr_volume);
+-		execlp("mbrola", "mbrola", "-e", "-v", charbuf,
++		execlp("@mbrola/bin/mbrola", "mbrola", "-e", "-v", charbuf,
+ 		       voice_path, "-", "-.wav", (char *)NULL);
+ 		/* if execution reaches this point then the exec() failed */
+ 		snprintf(mbr_errorbuf, sizeof(mbr_errorbuf),
+diff --git a/src/libespeak-ng/synth_mbrola.c b/src/libespeak-ng/synth_mbrola.c
+index 734631b7..46d1f13e 100644
+--- a/src/libespeak-ng/synth_mbrola.c
++++ b/src/libespeak-ng/synth_mbrola.c
+@@ -85,7 +85,7 @@ espeak_ng_STATUS LoadMbrolaTable(const char *mbrola_voice, const char *phtrans,
+ 	if (!load_MBR())
+ 		return ENS_MBROLA_NOT_FOUND;
+ 
+-	sprintf(path, "%s/mbrola/%s", path_home, mbrola_voice);
++	sprintf(path, "@mbrola@/share/mbrola/voices/%s/%s", mbrola_voice, mbrola_voice);
+ #if PLATFORM_POSIX
+ 	// if not found, then also look in
+ 	//   usr/share/mbrola/xx, /usr/share/mbrola/xx/xx, /usr/share/mbrola/voices/xx
diff --git a/pkgs/development/python-modules/piper-phonemize/default.nix b/pkgs/development/python-modules/piper-phonemize/default.nix
new file mode 100644
index 0000000000000..cd09567a61e36
--- /dev/null
+++ b/pkgs/development/python-modules/piper-phonemize/default.nix
@@ -0,0 +1,34 @@
+{ buildPythonPackage
+, onnxruntime-native
+, piper-phonemize-native
+, pybind11
+, setuptools
+}:
+
+buildPythonPackage {
+  inherit (piper-phonemize-native) pname version src;
+  format = "pyproject";
+
+  nativeBuildInputs = [
+    pybind11
+    setuptools
+  ];
+
+  buildInputs = [
+    onnxruntime-native
+    piper-phonemize-native
+    piper-phonemize-native.espeak-ng
+  ];
+
+  pythonImportsCheck = [
+    "piper_phonemize"
+  ];
+
+  # no tests
+  doCheck = false;
+
+  meta = {
+    description = "Phonemization libary used by Piper text to speech system";
+    inherit (piper-phonemize-native.meta) homepage license maintainers;
+  };
+}
diff --git a/pkgs/tools/audio/piper/default.nix b/pkgs/tools/audio/piper/default.nix
index c3d8a7638304d..29fdb0705fec6 100644
--- a/pkgs/tools/audio/piper/default.nix
+++ b/pkgs/tools/audio/piper/default.nix
@@ -1,17 +1,24 @@
 { lib
 , stdenv
 , fetchFromGitHub
+
+# build time
 , cmake
 , pkg-config
-, espeak-ng
+
+# runtime
 , onnxruntime
 , pcaudiolib
+, piper-phonemize
+, spdlog
+
+# tests
 , piper-train
 }:
 
 let
   pname = "piper";
-  version = "0.0.2";
+  version = "1.2.0";
 in
 stdenv.mkDerivation {
   inherit pname version;
@@ -19,30 +26,27 @@ stdenv.mkDerivation {
   src = fetchFromGitHub {
     owner = "rhasspy";
     repo = "piper";
-    rev = "70afec58bc131010c8993c154ff02a78d1e7b8b0";
-    hash = "sha256-zTW7RGcV8Hh7G6Braf27F/8s7nNjAqagp7tmrKO10BY=";
+    rev = "refs/tags/v${version}";
+    hash = "sha256-6WNWqJt0PO86vnf+3iHaRRg2KwBOEj4aicmB+P2phlk=";
   };
 
   sourceRoot = "source/src/cpp";
 
-  patches = [
-    ./fix-compilation-with-newer-onnxruntime.patch
-  ];
-
-  postPatch = ''
-    substituteInPlace CMakeLists.txt \
-      --replace "/usr/local/include/onnxruntime" "${onnxruntime}"
-  '';
-
   nativeBuildInputs = [
     cmake
     pkg-config
   ];
 
   buildInputs = [
-    espeak-ng
     onnxruntime
     pcaudiolib
+    piper-phonemize
+    piper-phonemize.espeak-ng
+    spdlog
+  ];
+
+  env.NIX_CFLAGS_COMPILE = builtins.toString [
+    "-isystem ${lib.getDev piper-phonemize}/include/piper-phonemize"
   ];
 
   installPhase = ''
diff --git a/pkgs/development/python-modules/piper-train/default.nix b/pkgs/tools/audio/piper/train.nix
index 5cf8cdea10df6..a385ebb85ad66 100644
--- a/pkgs/development/python-modules/piper-train/default.nix
+++ b/pkgs/tools/audio/piper/train.nix
@@ -1,38 +1,26 @@
-{ buildPythonPackage
-, piper-tts
-
-# build
-, cython
-, python
-
-# propagates
-, espeak-phonemizer
-, librosa
-, numpy
-, onnxruntime
-, pytorch-lightning
-, torch
+{ piper-tts
+, python3
 }:
 
-buildPythonPackage {
-  inherit (piper-tts) version src meta;
+let
+  python = python3.override {
+    packageOverrides = self: super: {
+    };
+  };
+in
+
+python.pkgs.buildPythonPackage {
+  inherit (piper-tts) version src;
 
   pname = "piper-train";
   format = "setuptools";
 
   sourceRoot = "source/src/python";
 
-  nativeBuildInputs = [
+  nativeBuildInputs = with python.pkgs; [
     cython
   ];
 
-  postPatch = ''
-    substituteInPlace requirements.txt \
-      --replace "onnxruntime~=1.11.0" "onnxruntime" \
-      --replace "pytorch-lightning~=1.7.0" "pytorch-lightning" \
-      --replace "torch~=1.11.0" "torch"
-  '';
-
   postBuild = ''
     make -C piper_train/vits/monotonic_align
   '';
@@ -43,11 +31,12 @@ buildPythonPackage {
     cp -v ./piper_train/vits/monotonic_align/piper_train/vits/monotonic_align/core.*.so $MONOTONIC_ALIGN/
   '';
 
-  propagatedBuildInputs = [
+  propagatedBuildInputs = with python.pkgs; [
     espeak-phonemizer
     librosa
     numpy
     onnxruntime
+    piper-phonemize
     pytorch-lightning
     torch
   ];
@@ -57,4 +46,9 @@ buildPythonPackage {
   ];
 
   doCheck = false; # no tests
+
+  meta = piper-tts.meta // {
+    # requires torch<2, pytorch-lightning~=1.7
+    broken = true;
+  };
 }
diff --git a/pkgs/top-level/all-packages.nix b/pkgs/top-level/all-packages.nix
index 798adc1832b00..bc20f46ac2c88 100644
--- a/pkgs/top-level/all-packages.nix
+++ b/pkgs/top-level/all-packages.nix
@@ -11716,7 +11716,8 @@ with pkgs;
 
   pim6sd = callPackage ../servers/pim6sd { };
 
-  piper-train = with python3Packages; toPythonApplication piper-train;
+  piper-phonemize = callPackage ../development/libraries/piper-phonemize { };
+  piper-train = callPackage ../tools/audio/piper/train.nix { };
   piper-tts = callPackage ../tools/audio/piper { };
 
   phosh = callPackage ../applications/window-managers/phosh { };
diff --git a/pkgs/top-level/python-aliases.nix b/pkgs/top-level/python-aliases.nix
index a6ef2838bcb68..63addfa58a038 100644
--- a/pkgs/top-level/python-aliases.nix
+++ b/pkgs/top-level/python-aliases.nix
@@ -184,7 +184,6 @@ mapAliases ({
   jupyter_server = jupyter-server; # added 2023-01-05
   Kajiki = kajiki; # added 2023-02-19
   Keras = keras; # added 2021-11-25
-  larynx-train = piper-train; # added 2023-06-09
   ldap = python-ldap; # added 2022-09-16
   lammps-cython = throw "lammps-cython no longer builds and is unmaintained"; # added 2021-07-04
   logilab_astng = throw "logilab-astng has not been released since 2013 and is unmaintained"; # added 2022-11-29
diff --git a/pkgs/top-level/python-packages.nix b/pkgs/top-level/python-packages.nix
index 83b45cde21b35..bf738d556a9fd 100644
--- a/pkgs/top-level/python-packages.nix
+++ b/pkgs/top-level/python-packages.nix
@@ -7965,7 +7965,10 @@ self: super: with self; {
 
   pipenv-poetry-migrate = callPackage ../development/python-modules/pipenv-poetry-migrate { };
 
-  piper-train = callPackage ../development/python-modules/piper-train { };
+  piper-phonemize = callPackage ../development/python-modules/piper-phonemize {
+    onnxruntime-native = pkgs.onnxruntime;
+    piper-phonemize-native = pkgs.piper-phonemize;
+  };
 
   pip-api = callPackage ../development/python-modules/pip-api { };