libpiper migration

Migrate now deprecated/archived piper-phonemize[1] based cxxpiper
implementation to libpiper[2] which was recently added to the upstream
piper repository.

Because we are now using the upstream libpiper C library directly, I
figured we should probably also remove the cxx prefix from the module name.

Changes made:

- Removed conf options that new libpiper no longer support
- Added peak normalization for parity with Python CLI
- Simplified voice handling and configuration

[1]: https://github.com/rhasspy/piper-phonemize
[2]: https://github.com/OHF-Voice/piper1-gpl/tree/main/libpiper

Author: unlimitedsola

INSTALL

@@ -152,22 +152,26 @@ chmod +x autogen.sh
 make
 make install
 
-Notes about cxxpiper support
-============================
+Notes about piper support
+=========================
 
-The cxxpiper module uses the Piper synthesis
-https://github.com/rhasspy/piper/
+The piper module uses the Piper neural TTS engine
+https://github.com/OHF-Voice/piper1-gpl
 
-Upstream piper however does not provide an API that can really be used nicely,
-so the build is quite convoluted. The following should be working:
+Speech Dispatcher links against the `libpiper` C shared library (and `librubberband`).
+To build and install `libpiper` from source:
 
-wget https://github.com/rhasspy/piper-phonemize/releases/download/2023.11.14-4/piper-phonemize_linux_x86_64.tar.gz
-sudo tar -C /opt -xvf piper-phonemize_linux_x86_64.tar.gz
-wget https://github.com/rhasspy/piper/archive/refs/tags/2023.11.14-2.tar.gz
-sudo tar -C /opt -xvf 2023.11.14-2.tar.gz
-CXXFLAGS=-I/opt/piper_phonemize/include LDFLAGS=-L/opt/piper_phonemize/lib ./configure --with-piper=/opt/piper-2023.11.14-2
-make
-make install
+  $ git clone https://github.com/OHF-Voice/piper1-gpl.git
+  $ cd piper1-gpl/libpiper
+  $ cmake -B build -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=/usr
+  $ cmake --build build -j$(nproc)
+  $ sudo cmake --install build
+
+Then configure Speech Dispatcher with:
+  $ ./configure --with-piper
+
+(If libpiper is installed in a non-standard prefix like /opt/piper, use:
+  $ ./configure --with-piper=/opt/piper)
 
 Continue building Speech Dispatcher
 ===================================

config/modules/Makefile.am

@@ -54,8 +54,8 @@ dist_moduleconforig_DATA += kali.conf
 endif
 
 if piper_support
-dist_moduleconf_DATA += cxxpiper.conf
-dist_moduleconforig_DATA += cxxpiper.conf
+dist_moduleconf_DATA += piper.conf
+dist_moduleconforig_DATA += piper.conf
 endif
 
 if ibmtts_support

config/modules/cxxpiper.conf

@@ -0,0 +1,56 @@
+#
+# Configuration for piper Speech Dispatcher output module.
+#
+
+# Debugging level (0 = disabled, 1 = errors only, 5 = full debug)
+Debug 0
+
+# -- Model Configuration --
+#
+# ModelPath is required. On Arch Linux (e.g. using the piper-voices-en-us package from AUR),
+# voices are installed to /usr/share/piper-voices/<lang>/<locale>/<voice>/<quality>/.
+#
+# en_US-lessac-medium is the recommended standard default voice.
+ModelPath "/usr/share/piper-voices/en/en_US/lessac/medium/en_US-lessac-medium.onnx"
+
+# ConfigPath (optional):
+# Path to the voice's JSON configuration file. If omitted, piper automatically
+# looks for <ModelPath>.json.
+# ConfigPath "/usr/share/piper-voices/en/en_US/lessac/medium/en_US-lessac-medium.onnx.json"
+
+# -- Multi-Speaker Voice Selection (Optional) --
+#
+# For single-speaker models (like lessac), a single voice named after the model stem is exposed.
+# For multi-speaker models (e.g., libritts_r, vctk, thorsten_emotional), each speaker defined in the
+# model's "speaker_id_map" is exposed as an individual voice.
+#
+# Listing available speaker names:
+#   1. Using the Speech Dispatcher CLI:
+#        spd-say -o piper -L
+#   2. Inspecting the model's JSON configuration file directly:
+#        jq '.speaker_id_map | keys' /path/to/model.onnx.json
+#
+# DefaultVoice (optional):
+# Sets the initial speaker for multi-speaker models using a speaker name from speaker_id_map.
+# DefaultVoice "8419"
+#
+# AddVoice (optional):
+# Maps standard Speech Dispatcher voice types (MALE1, FEMALE1, etc.) to specific speakers
+# within a language code.
+#
+# Examples for multi-speaker models:
+# AddVoice "en_US" "MALE1" "8419"
+# AddVoice "en_US" "FEMALE1" "4137"
+# AddVoice "de_DE" "MALE1" "neutral"
+# AddVoice "de_DE" "MALE2" "amused"
+
+# -- Sound Icons --
+# Folder containing sound icon audio files (path must end with a trailing slash '/').
+SoundIconFolder "/usr/share/sounds/sound-icons/"
+
+# -- ESpeak NG Data Directory --
+# Path to the espeak-ng-data directory used by Piper for phonemization.
+# On Arch Linux, this is installed by the 'espeak-ng' package to /usr/share/espeak-ng-data/.
+ESpeakNGDataDirPath "/usr/share/espeak-ng-data/"
+
+# End of piper.conf

config/modules/piper.conf

@@ -287,6 +287,7 @@ SymbolsPreprocFile "orca-math.dic"
 #AddModule "baratinoo"                "sd_baratinoo" "baratinoo.conf"
 #AddModule "rhvoice"                  "sd_rhvoice"   "rhvoice.conf"
 #AddModule "voxin"                    "sd_voxin"     "voxin.conf"
+#AddModule "piper"                    "sd_piper"     "piper.conf"
 
 # The output module testing doesn't actually connect to anything. It
 # outputs the requested commands to standard output and reads

config/speechd.conf

@@ -427,70 +427,59 @@ AS_IF([test $with_kali != no], [output_modules="${output_modules} kali"])
 
 # check for piper support
 AC_ARG_WITH([piper],
-	[AS_HELP_STRING([--with-piper=/path], [include Piper support])],
+	[AS_HELP_STRING([--with-piper[[=yes|no|check|/path]]], [include Piper TTS support])],
 	[],
-	[with_piper=no])
-AS_IF([test $with_espeak = yes -a $with_piper != no], [
-	# We currently need various headers of piper
-	PIPER_SRC_DIR="$with_piper/src/cpp"
+	[with_piper=check])
+
+AS_IF([test "$with_piper" != "no"], [
+	PIPER_CFLAGS=""
+	PIPER_LIBS=""
+	AS_IF([test "$with_piper" != "yes" -a "$with_piper" != "check"], [
+		PIPER_CFLAGS="-I$with_piper/include"
+		PIPER_LIBS="-L$with_piper/lib -lpiper"
+	])
 	OLDCXXFLAGS="$CXXFLAGS"
 	OLDLDFLAGS="$LDFLAGS"
-	CXXFLAGS="$CXXFLAGS -I$PIPER_SRC_DIR"
-
-	AS_IF([test $with_piper != no],
-		[PKG_CHECK_MODULES([ONNXRUNTIME], [libonnxruntime], [
-				CXXFLAGS="$CXXFLAGS $ONNXRUNTIME_CFLAGS"
-				LDFLAGS="$LDFLAGS $ONNXRUNTIME_LIBS"
-			], [AC_CHECK_LIB([onnxruntime], [OrtGetApiBase],
-				[ONNXRUNTIME_LIBS="-lonnxruntime"],
-				[with_piper=no])
-			])])
-
-	AS_IF([test $with_piper != no],
-		[PKG_CHECK_MODULES([RUBBERBAND], [rubberband], [
-				CXXFLAGS="$CXXFLAGS $RUBBERBAND_CFLAGS"
-				LDFLAGS="$LDFLAGS $RUBBERBAND_LIBS"
-			], [with_piper=no])])
+	OLDLIBS="$LIBS"
+	CXXFLAGS="$CXXFLAGS $PIPER_CFLAGS"
+	LDFLAGS="$LDFLAGS $PIPER_LIBS"
 
 	AC_LANG_PUSH(C++)
-	AS_IF([test $with_piper != no],
-		[AC_CHECK_LIB([piper_phonemize], [_ZN5piper19DEFAULT_PHONEME_MAPB5cxx11E],
-			[:],
-			[with_piper=no])])
-
-	AS_IF([test $with_piper != no],
-		[AC_CHECK_HEADER([json.hpp],
-			[],
-			[with_piper=no])])
-	AS_IF([test $with_piper != no],
-		[AC_CHECK_HEADER([piper.hpp],
-			[],
-			[with_piper=no])])
-	AS_IF([test $with_piper != no],
-		[AC_CHECK_HEADER([utf8.h],
-			[],
-			[with_piper=no])])
-	AS_IF([test $with_piper != no],
-		[AC_CHECK_HEADER([wavfile.hpp],
-			[],
-			[with_piper=no])])
-
-	AS_IF([test $with_piper != no],
-		# We need espeak with https://github.com/espeak-ng/espeak-ng/pull/2127 applied
-		[AC_CHECK_LIB([espeak-ng], [espeak_TextToPhonemesWithTerminator],
-			[:],
-			[with_piper=no],
-			[$ESPEAK_NG_LIBS])])
-
+	# Check for piper.h header
+	AC_CHECK_HEADER([piper.h], [have_piper_h=yes], [have_piper_h=no])
+	# Check for libpiper library
+	AC_CHECK_LIB([piper], [piper_create], [have_libpiper=yes], [have_libpiper=no])
+	# Check for nlohmann json
+	AC_CHECK_HEADER([nlohmann/json.hpp], [have_json_h=yes], [
+		AC_CHECK_HEADER([json.hpp], [have_json_h=yes], [have_json_h=no])
+	])
 	AC_LANG_POP(C++)
+
 	CXXFLAGS="$OLDCXXFLAGS"
 	LDFLAGS="$OLDLDFLAGS"
+	LIBS="$OLDLIBS"
+
+	AS_IF([test "$have_piper_h" = "yes" -a "$have_libpiper" = "yes" -a "$have_json_h" = "yes"], [
+		PKG_CHECK_MODULES([RUBBERBAND], [rubberband], [
+			AS_IF([test -z "$PIPER_LIBS"], [PIPER_LIBS="-lpiper"])
+			with_piper=yes
+		], [
+			AS_IF([test "$with_piper" = "yes"], [
+				AC_MSG_FAILURE([rubberband is required for piper])
+			])
+			with_piper=no
+		])
+	], [
+		AS_IF([test "$with_piper" = "yes"], [
+			AC_MSG_FAILURE([libpiper (piper.h, libpiper, and json.hpp) is not available])
+		])
+		with_piper=no
+	])
 ])
-AM_CONDITIONAL([piper_support], [test $with_piper != no])
-AS_IF([test $with_piper != no], [output_modules="${output_modules} cxxpiper"])
-AC_SUBST([PIPER_SRC_DIR])
-AC_SUBST([ONNXRUNTIME_CFLAGS])
-AC_SUBST([ONNXRUNTIME_LIBS])
+AM_CONDITIONAL([piper_support], [test "$with_piper" = "yes"])
+AS_IF([test "$with_piper" = "yes"], [output_modules="${output_modules} piper"])
+AC_SUBST([PIPER_CFLAGS])
+AC_SUBST([PIPER_LIBS])
 AC_SUBST([RUBBERBAND_CFLAGS])
 AC_SUBST([RUBBERBAND_LIBS])
 

configure.ac

@@ -295,16 +295,14 @@ endif
 endif
 
 #
-#  cxxpiper
+#  piper
 #
 if piper_support
-modulebin_PROGRAMS += sd_cxxpiper
-sd_cxxpiper_SOURCES = cxxpiper.cpp module_utils_addvoice.c module_utils_play.c $(common_SOURCES)
-sd_cxxpiper_CPPFLAGS = -I$(PIPER_SRC_DIR) $(ONNXRUNTIME_CFLAGS) $(RUBBERBAND_CFLAGS) $(AM_CPPFLAGS)
-sd_cxxpiper_LDADD = $(top_builddir)/src/common/libcommon.la \
-	-lpiper_phonemize \
-	$(ONNXRUNTIME_LIBS) \
-	-lespeak-ng \
+modulebin_PROGRAMS += sd_piper
+sd_piper_SOURCES = piper.cpp module_utils_addvoice.c module_utils_play.c $(common_SOURCES)
+sd_piper_CPPFLAGS = $(PIPER_CFLAGS) $(RUBBERBAND_CFLAGS) $(AM_CPPFLAGS)
+sd_piper_LDADD = $(top_builddir)/src/common/libcommon.la \
+	$(PIPER_LIBS) \
 	$(RUBBERBAND_LIBS) \
 	$(SNDFILE_LIBS) \
 	$(common_LDADD)