a11y: linux tts with espeak-ng (#5428)

* a11y: linux tts with espeak-ng

* dlopen

* Show setting on Linux

.github/workflows/apt-deps.txt

@@ -1 +1 @@
-libusb-dev libusb-1.0-0-dev libsdl2-dev libsdl2-net-dev libpng-dev libglew-dev nlohmann-json3-dev libtinyxml2-dev libspdlog-dev ninja-build 
+libusb-dev libusb-1.0-0-dev libsdl2-dev libsdl2-net-dev libpng-dev libglew-dev nlohmann-json3-dev libtinyxml2-dev libspdlog-dev libespeak-ng-dev ninja-build 

soh/CMakeLists.txt

@@ -142,16 +142,21 @@ endif()
 
 # handle Network removals
 if (!BUILD_REMOTE_CONTROL)
-    list(FILTER soh__ EXCLUDE REGEX "soh/Enhancements/crowd-control/*")
+    list(FILTER soh__ EXCLUDE REGEX "soh/Enhancements/crowd-control/")
 endif()
 
 # handle speechsynthesizer removals
 if (CMAKE_SYSTEM_NAME STREQUAL "Windows")
-    list(FILTER soh__ EXCLUDE REGEX "soh/Enhancements/speechsynthesizer/Darwin*")
+    list(FILTER soh__ EXCLUDE REGEX "soh/Enhancements/speechsynthesizer/Darwin")
 elseif (CMAKE_SYSTEM_NAME STREQUAL "Darwin")
-    list(FILTER soh__ EXCLUDE REGEX "soh/Enhancements/speechsynthesizer/SAPI*")
+    list(FILTER soh__ EXCLUDE REGEX "soh/Enhancements/speechsynthesizer/SAPI")
 else()
-    list(FILTER soh__ EXCLUDE REGEX "soh/Enhancements/speechsynthesizer/(Darwin|SAPI).*")
+    list(FILTER soh__ EXCLUDE REGEX "soh/Enhancements/speechsynthesizer/(Darwin|SAPI)")
+endif()
+
+find_library(ESPEAK espeak-ng)
+if (NOT ESPEAK)
+    list(FILTER soh__ EXCLUDE REGEX "soh/Enhancements/speechsynthesizer/ESpeak")
 endif()
 
 # soh/Extractor {{{
@@ -177,12 +182,12 @@ file(GLOB_RECURSE src__ RELATIVE ${CMAKE_CURRENT_SOURCE_DIR} "src/*.c" "src/*.h"
 set_source_files_properties(${src__} PROPERTIES COMPILE_OPTIONS "${WARNING_OVERRIDE}")
 
 list(APPEND src__ ${CMAKE_CURRENT_SOURCE_DIR}/Resource.rc)
-list(FILTER src__ EXCLUDE REGEX "src/dmadata/*")
+list(FILTER src__ EXCLUDE REGEX "src/dmadata/")
-list(FILTER src__ EXCLUDE REGEX "src/elf_message/*")
+list(FILTER src__ EXCLUDE REGEX "src/elf_message/")
-list(FILTER src__ EXCLUDE REGEX "src/libultra/io/*")
+list(FILTER src__ EXCLUDE REGEX "src/libultra/io/")
-list(FILTER src__ EXCLUDE REGEX "src/libultra/libc/*")
+list(FILTER src__ EXCLUDE REGEX "src/libultra/libc/")
-list(FILTER src__ EXCLUDE REGEX "src/libultra/os/*")
+list(FILTER src__ EXCLUDE REGEX "src/libultra/os/")
-list(FILTER src__ EXCLUDE REGEX "src/libultra/rmon/*")
+list(FILTER src__ EXCLUDE REGEX "src/libultra/rmon/")
 list(APPEND src__ "src/libultra/libc/sprintf.c")
 list(REMOVE_ITEM src__ "src/libultra/gu/cosf.c")
 list(REMOVE_ITEM src__ "src/libultra/gu/lookat.c")
@@ -292,6 +297,10 @@ if (BUILD_REMOTE_CONTROL)
     endif()
 endif()
 
+if (ESPEAK)
+    add_compile_definitions(ESPEAK=1)
+endif()
+
 target_include_directories(${PROJECT_NAME} PRIVATE assets
 	${CMAKE_CURRENT_SOURCE_DIR}/include/
 	${CMAKE_CURRENT_SOURCE_DIR}/src/

soh/soh/Enhancements/speechsynthesizer/ESpeakSpeechSynthesizer.cpp

@@ -0,0 +1,49 @@
+#include "ESpeakSpeechSynthesizer.h"
+#include <dlfcn.h>
+#include <libultraship/libultraship.h>
+
+ESpeakSpeechSynthesizer::ESpeakSpeechSynthesizer() {
+}
+
+bool ESpeakSpeechSynthesizer::DoInit() {
+    void* espeak = dlopen("libespeak-ng.so", RTLD_LAZY | RTLD_LOCAL);
+    if (espeak != NULL) {
+        this->Initialize = (speak_Initialize)dlsym(espeak, "espeak_Initialize");
+        this->SetVoiceByProperties = (speak_SetVoiceByProperties)dlsym(espeak, "espeak_SetVoiceByProperties");
+        this->Synth = (speak_Synth)dlsym(espeak, "espeak_Synth");
+        this->Terminate = (speak_Terminate)dlsym(espeak, "espeak_Terminate");
+        if (this->Initialize == NULL || this->SetVoiceByProperties == NULL || this->Synth == NULL ||
+            this->Terminate == NULL) {
+            lusprintf(__FILE__, __LINE__, 2, "Failed to load espeak-ng");
+            dlclose(espeak);
+            return false;
+        } else {
+            this->espeak = espeak;
+            return this->Initialize(AUDIO_OUTPUT_PLAYBACK, 100, NULL, 0) != -1;
+        }
+    }
+    return true;
+}
+
+void ESpeakSpeechSynthesizer::DoUninitialize() {
+    if (this->espeak != NULL) {
+        this->Terminate();
+        dlclose(this->espeak);
+        this->espeak = NULL;
+    }
+}
+
+void ESpeakSpeechSynthesizer::Speak(const char* text, const char* language) {
+    if (this->espeak == NULL) {
+        lusprintf(__FILE__, __LINE__, 2, "Spoken Text (%s): %s", language, text);
+    } else {
+        if (language != this->mLanguage) {
+            espeak_VOICE voice = { .languages = language };
+            if (this->SetVoiceByProperties(&voice)) {
+                return;
+            }
+            this->mLanguage = language;
+        }
+        this->Synth(text, 100, 0, POS_CHARACTER, 0, espeakCHARS_UTF8, NULL, NULL);
+    }
+}

soh/soh/Enhancements/speechsynthesizer/ESpeakSpeechSynthesizer.h

@@ -0,0 +1,34 @@
+#pragma once
+
+#include "SpeechSynthesizer.h"
+
+extern "C" {
+#include <espeak-ng/speak_lib.h>
+
+// C23 typeof could help here
+typedef ESPEAK_API int (*speak_Initialize)(espeak_AUDIO_OUTPUT output, int buflength, const char* path, int options);
+typedef ESPEAK_API espeak_ERROR (*speak_Terminate)(void);
+typedef ESPEAK_API espeak_ERROR (*speak_SetVoiceByProperties)(espeak_VOICE* voice_spec);
+typedef ESPEAK_API espeak_ERROR (*speak_Synth)(const void* text, size_t size, unsigned int position,
+                                               espeak_POSITION_TYPE position_type, unsigned int end_position,
+                                               unsigned int flags, unsigned int* unique_identifier, void* user_data);
+}
+
+class ESpeakSpeechSynthesizer : public SpeechSynthesizer {
+  public:
+    ESpeakSpeechSynthesizer();
+
+    void Speak(const char* text, const char* language);
+
+  protected:
+    bool DoInit(void);
+    void DoUninitialize(void);
+
+  private:
+    const char* mLanguage = NULL;
+    void* espeak = NULL;
+    speak_Initialize Initialize = NULL;
+    speak_SetVoiceByProperties SetVoiceByProperties = NULL;
+    speak_Synth Synth = NULL;
+    speak_Terminate Terminate = NULL;
+};

soh/soh/Enhancements/speechsynthesizer/SpeechSynthesizer.h

@@ -35,6 +35,8 @@ class SpeechSynthesizer {
 #include "SAPISpeechSynthesizer.h"
 #elif defined(__APPLE__)
 #include "DarwinSpeechSynthesizer.h"
+#elif ESPEAK
+#include "ESpeakSpeechSynthesizer.h"
 #endif
 
 #include "SpeechLogger.h"

soh/soh/OTRGlobals.cpp

@@ -1229,14 +1229,14 @@ extern "C" void InitOTR() {
     ActorDB::Instance = new ActorDB();
 #ifdef __APPLE__
     SpeechSynthesizer::Instance = new DarwinSpeechSynthesizer();
-    SpeechSynthesizer::Instance->Init();
 #elif defined(_WIN32)
     SpeechSynthesizer::Instance = new SAPISpeechSynthesizer();
-    SpeechSynthesizer::Instance->Init();
+#elif ESPEAK
+    SpeechSynthesizer::Instance = new ESpeakSpeechSynthesizer();
 #else
     SpeechSynthesizer::Instance = new SpeechLogger();
-    SpeechSynthesizer::Instance->Init();
 #endif
+    SpeechSynthesizer::Instance->Init();
 
 #ifdef ENABLE_REMOTE_CONTROL
     CrowdControl::Instance = new CrowdControl();

soh/soh/SohGui/SohMenuSettings.cpp

@@ -190,7 +190,7 @@ void SohMenu::AddMenuSettings() {
                      .ComboMap(languages)
                      .DefaultIndex(LANGUAGE_ENG));
     AddWidget(path, "Accessibility", WIDGET_SEPARATOR_TEXT);
-#if defined(_WIN32) || defined(__APPLE__)
+#if defined(_WIN32) || defined(__APPLE__) || defined(ESPEAK)
     AddWidget(path, "Text to Speech", WIDGET_CVAR_CHECKBOX)
         .CVar(CVAR_SETTING("A11yTTS"))
         .RaceDisable(false)