audio/piper: Fast and local text to speech system

Pull Request:	https://github.com/freebsd/freebsd-ports/pull/406
Co-authored-by:	Gleb Popov <arrowd@FreeBSD.org>

audio/Makefile

@@ -594,6 +594,7 @@
     SUBDIR += pianod2
     SUBDIR += picard
     SUBDIR += picard-plugins
+    SUBDIR += piper
     SUBDIR += piper-phonemize
     SUBDIR += pipewire-module-xrdp
     SUBDIR += pipewire-spa-oss

audio/piper/Makefile

@@ -0,0 +1,27 @@
+PORTNAME=	piper
+DISTVERSION=	2023.11.14
+DISTVERSIONSUFFIX=	-2
+CATEGORIES=	audio
+
+MAINTAINER=	kenrap@kennethraplee.com
+COMMENT=	Fast, local text to speech system
+WWW=		https://rhasspy.github.io/piper-samples/
+
+LICENSE=	MIT
+LICENSE_FILE=	${WRKSRC}/LICENSE.md
+
+LIB_DEPENDS=	libespeak-ng.so:audio/espeak-ng \
+		libfmt.so:devel/libfmt \
+		libonnxruntime.so:misc/onnxruntime \
+		libpiper_phonemize.so:audio/piper-phonemize \
+		libspdlog.so:devel/spdlog
+
+USES=		compiler:c++17-lang cmake localbase pkgconfig tar:xz xorg
+USE_XORG=	x11
+
+USE_GITHUB=	yes
+GH_ACCOUNT=	rhasspy
+
+PLIST_FILES=	bin/piper
+
+.include <bsd.port.mk>

audio/piper/distinfo

@@ -0,0 +1,3 @@
+TIMESTAMP = 1750142677
+SHA256 (rhasspy-piper-2023.11.14-2_GH0.tar.gz) = 8b684e102cfe23af097830db9555a8641f17735725853c531a1daeb34b6ed1a8
+SIZE (rhasspy-piper-2023.11.14-2_GH0.tar.gz) = 25602353

audio/piper/files/patch-CMakeLists.txt

@@ -0,0 +1,139 @@
+--- CMakeLists.txt.orig	2023-11-14 17:55:22 UTC
++++ CMakeLists.txt
+@@ -13,7 +13,7 @@ if(MSVC)
+   add_compile_options("$<$<CXX_COMPILER_ID:MSVC>:/utf-8>")
+ elseif(NOT APPLE)
+   # Linux flags
+-  string(APPEND CMAKE_CXX_FLAGS " -Wall -Wextra -Wl,-rpath,'$ORIGIN'")
++  string(APPEND CMAKE_CXX_FLAGS " -Wall -Wextra")
+   string(APPEND CMAKE_C_FLAGS " -Wall -Wextra")
+ endif()
+ 
+@@ -25,7 +25,7 @@ add_executable(test_piper src/cpp/test.cpp src/cpp/piper.cpp)
+ 
+ # ---- fmt ---
+ 
+-if(NOT DEFINED FMT_DIR)
++if(FALSE)
+   set(FMT_VERSION "10.0.0")
+   set(FMT_DIR "${CMAKE_CURRENT_BINARY_DIR}/fi")
+ 
+@@ -43,7 +43,7 @@ endif()
+ 
+ # ---- spdlog ---
+ 
+-if(NOT DEFINED SPDLOG_DIR)
++if(FALSE)
+   set(SPDLOG_DIR "${CMAKE_CURRENT_BINARY_DIR}/si")
+   set(SPDLOG_VERSION "1.12.0")
+   ExternalProject_Add(
+@@ -58,7 +58,13 @@ endif()
+ 
+ # ---- piper-phonemize ---
+ 
+-if(NOT DEFINED PIPER_PHONEMIZE_DIR)
++find_package(PkgConfig)
++pkg_check_modules(FMT REQUIRED fmt)
++pkg_check_modules(ONNXRUNTIME REQUIRED libonnxruntime)
++pkg_check_modules(PIPER_PHONEMIZE REQUIRED piper_phonemize)
++pkg_check_modules(SPDLOG REQUIRED spdlog)
++
++if(FALSE)
+   set(PIPER_PHONEMIZE_DIR "${CMAKE_CURRENT_BINARY_DIR}/pi")
+   ExternalProject_Add(
+     piper_phonemize_external
+@@ -74,9 +80,8 @@ endif()
+ 
+ if((NOT MSVC) AND (NOT APPLE))
+   # Linux flags
+-  string(APPEND CMAKE_CXX_FLAGS " -Wall -Wextra -Wl,-rpath,'$ORIGIN'")
++  string(APPEND CMAKE_CXX_FLAGS " -Wall -Wextra")
+   string(APPEND CMAKE_C_FLAGS " -Wall -Wextra")
+-  target_link_libraries(piper -static-libgcc -static-libstdc++)
+ 
+   set(PIPER_EXTRA_LIBRARIES "pthread")
+ endif()
+@@ -91,18 +96,18 @@ target_link_libraries(piper
+ )
+ 
+ target_link_directories(piper PUBLIC
+-  ${FMT_DIR}/lib
+-  ${SPDLOG_DIR}/lib
+-  ${PIPER_PHONEMIZE_DIR}/lib
++  ${FMT_LIBRARY_DIRS}
++  ${SPDLOG_LIBRARY_DIRS}
++  ${ONNXRUNTIME_LIBRARY_DIRS}
+ )
+ 
+ target_include_directories(piper PUBLIC
+-  ${FMT_DIR}/include
+-  ${SPDLOG_DIR}/include
+-  ${PIPER_PHONEMIZE_DIR}/include
++  ${FMT_INCLUDE_DIRS}
++  ${SPDLOG_INCLUDE_DIRS}
++  ${ONNXRUNTIME_INCLUDE_DIRS}
+ )
+ 
+-target_compile_definitions(piper PUBLIC _PIPER_VERSION=${piper_version})
++target_compile_definitions(piper PUBLIC _PIPER_VERSION="${piper_version}")
+ 
+ # ---- Declare test ----
+ include(CTest)
+@@ -116,16 +121,16 @@ target_compile_features(test_piper PUBLIC cxx_std_17)
+ 
+ target_include_directories(
+   test_piper PUBLIC
+-  ${FMT_DIR}/include
+-  ${SPDLOG_DIR}/include
+-  ${PIPER_PHONEMIZE_DIR}/include
++  ${FMT_INCLUDE_DIRS}
++  ${SPDLOG_INCLUDE_DIRS}
++  ${ONNXRUNTIME_INCLUDE_DIRS}
+ )
+ 
+ target_link_directories(
+   test_piper PUBLIC
+-  ${FMT_DIR}/lib
+-  ${SPDLOG_DIR}/lib
+-  ${PIPER_PHONEMIZE_DIR}/lib
++  ${FMT_LIBRARY_DIRS}
++  ${SPDLOG_LIBRARY_DIRS}
++  ${ONNXRUNTIME_LIBRARY_DIRS}
+ )
+ 
+ target_link_libraries(test_piper PUBLIC
+@@ -140,33 +145,5 @@ target_link_libraries(test_piper PUBLIC
+ 
+ install(
+   TARGETS piper
+-  DESTINATION ${CMAKE_INSTALL_PREFIX})
+-
+-# Dependencies
+-install(
+-  DIRECTORY ${PIPER_PHONEMIZE_DIR}/bin/
+-  DESTINATION ${CMAKE_INSTALL_PREFIX}
+-  USE_SOURCE_PERMISSIONS  # keep +x
+-  FILES_MATCHING
+-  PATTERN "piper_phonemize"
+-  PATTERN "espeak-ng"
+-  PATTERN "*.dll"
+-)
+-
+-install(
+-  DIRECTORY ${PIPER_PHONEMIZE_DIR}/lib/
+-  DESTINATION ${CMAKE_INSTALL_PREFIX}
+-  FILES_MATCHING
+-  PATTERN "*.dll"
+-  PATTERN "*.so*"
+-)
+-
+-install(
+-  DIRECTORY ${PIPER_PHONEMIZE_DIR}/share/espeak-ng-data
+-  DESTINATION ${CMAKE_INSTALL_PREFIX}
+-)
+-
+-install(
+-  FILES ${PIPER_PHONEMIZE_DIR}/share/libtashkeel_model.ort
+-  DESTINATION ${CMAKE_INSTALL_PREFIX}
++  DESTINATION ${CMAKE_INSTALL_PREFIX}/bin
+ )

audio/piper/files/patch-src_cpp_piper.cpp

@@ -0,0 +1,61 @@
+--- src/cpp/piper.cpp.orig
++++ src/cpp/piper.cpp
+@@ -6,7 +6,7 @@
+ #include <stdexcept>
+ 
+ #include <espeak-ng/speak_lib.h>
+-#include <onnxruntime_cxx_api.h>
++#include <onnxruntime/onnxruntime_cxx_api.h>
+ #include <spdlog/spdlog.h>
+ 
+ #include "json.hpp"
+@@ -229,22 +229,6 @@ void initialize(PiperConfig &config) {
+     spdlog::debug("Initialized eSpeak");
+   }
+ 
+-  // Load onnx model for libtashkeel
+-  // https://github.com/mush42/libtashkeel/
+-  if (config.useTashkeel) {
+-    spdlog::debug("Using libtashkeel for diacritization");
+-    if (!config.tashkeelModelPath) {
+-      throw std::runtime_error("No path to libtashkeel model");
+-    }
+-
+-    spdlog::debug("Loading libtashkeel model from {}",
+-                  config.tashkeelModelPath.value());
+-    config.tashkeelState = std::make_unique<tashkeel::State>();
+-    tashkeel::tashkeel_load(config.tashkeelModelPath.value(),
+-                            *config.tashkeelState);
+-    spdlog::debug("Initialized libtashkeel");
+-  }
+-
+   spdlog::info("Initialized piper");
+ }
+ 
+@@ -447,15 +431,6 @@ void textToAudio(PiperConfig &config, Voice &voice, std::string text,
+         voice.synthesisConfig.sampleRate * voice.synthesisConfig.channels);
+   }
+ 
+-  if (config.useTashkeel) {
+-    if (!config.tashkeelState) {
+-      throw std::runtime_error("Tashkeel model is not loaded");
+-    }
+-
+-    spdlog::debug("Diacritizing text with libtashkeel: {}", text);
+-    text = tashkeel::tashkeel_run(text, *config.tashkeelState);
+-  }
+-
+   // Phonemes for each sentence
+   spdlog::debug("Phonemizing text: {}", text);
+   std::vector<std::vector<Phoneme>> phonemes;
+--- src/cpp/piper.hpp.orig
++++ src/cpp/piper.hpp
+@@ -8,7 +8,7 @@
+ #include <string>
+ #include <vector>
+ 
+-#include <onnxruntime_cxx_api.h>
++#include <onnxruntime/onnxruntime_cxx_api.h>
+ #include <piper-phonemize/phoneme_ids.hpp>
+ #include <piper-phonemize/phonemize.hpp>
+ #include <piper-phonemize/tashkeel.hpp>

audio/piper/pkg-descr

@@ -0,0 +1,2 @@
+A fast, local neural text to speech system that sounds great and is optimized
+for the Raspberry Pi 4. Piper is used in a variety of projects.