Add support for SSML speech.

CMakeLists.txt

@@ -16,7 +16,7 @@ endif()
 project ("SRAL")
 add_library(${PROJECT_NAME}_obj OBJECT)
 target_sources(${PROJECT_NAME}_obj PRIVATE
-  "SRC/SRAL.cpp" "SRC/Engine.h")
+  "SRC/SRAL.cpp" "SRC/Engine.h" "SRC/Util.h" "SRC/Util.cpp")
 target_sources(${PROJECT_NAME}_obj PUBLIC
   FILE_SET HEADERS
   BASE_DIRS "${INCLUDES}"

Include/SRAL.h

@@ -51,7 +51,8 @@ extern "C" {
 		SUPPORTS_SPEECH_RATE = 512,
 		SUPPORTS_SPEECH_VOLUME = 1024,
 		SUPPORTS_SELECT_VOICE = 2048,
-		SUPPORTS_PAUSE_SPEECH = 4096
+		SUPPORTS_PAUSE_SPEECH = 4096,
+		SUPPORTS_SSML = 8192
 	};
 
 	/**

README.md

@@ -31,6 +31,7 @@ This enumeration defines the features supported by the various speech engines. T
 - `SUPPORTS_SPEECH_VOLUME = 1024`: The engine supports setting the speech volume.
 - `SUPPORTS_SELECT_VOICE = 2048`: The engine supports selecting a specific voice.
 - `SUPPORTS_PAUSE_SPEECH = 4096`: The engine supports pause and resume speech.
+- `SUPPORTS_SSML = 8192`: The engine supports the SSML tags for speak.
 
 
 ## Functions

SRC/NVDA.cpp

@@ -1,16 +1,16 @@
-#ifdef _WIN32
 #include "Encoding.h"
 #include "NVDA.h"
+#include "Util.h"
 #include<Windows.h>
 
-
 bool NVDA::Initialize() {
 	lib = LoadLibraryW(L"nvdaControllerClient.dll");
 	if (lib == nullptr)return false;
 	nvdaController_speakText = (NVDAController_speakText)GetProcAddress(lib, "nvdaController_speakText");
 	nvdaController_brailleMessage = (NVDAController_brailleMessage)GetProcAddress(lib, "nvdaController_brailleMessage");
 	nvdaController_cancelSpeech = (NVDAController_cancelSpeech)GetProcAddress(lib, "nvdaController_cancelSpeech");
 	nvdaController_testIfRunning = (NVDAController_testIfRunning)GetProcAddress(lib, "nvdaController_testIfRunning");
+	nvdaController_speakSsml = (NVDAController_speakSsml)GetProcAddress(lib, "nvdaController_speakSsml");
 	return true;
 }
 bool NVDA::Uninitialize() {
@@ -20,6 +20,7 @@ bool NVDA::Uninitialize() {
 	nvdaController_brailleMessage = nullptr;
 	nvdaController_cancelSpeech = nullptr;
 	nvdaController_testIfRunning = nullptr;
+	nvdaController_speakSsml = nullptr;
 	return true;
 }
 bool NVDA::GetActive() {
@@ -31,9 +32,20 @@ bool NVDA::Speak(const char* text, bool interrupt) {
 	if (!GetActive())return false;
 	if (interrupt)
 		nvdaController_cancelSpeech();
+	std::string text_str(text);
+	if (!IsSsml(text_str))AddSsml(text_str);
+
 	std::wstring out;
-	UnicodeConvert(text, out);
-	return nvdaController_speakText(out.c_str()) == 0;
+	UnicodeConvert(text_str, out);
+	error_status_t result = nvdaController_speakSsml(out.c_str(), -1, 0, 0);
+	if (result == 0)return true;
+	else if (result == 1717) {
+		RemoveSsml(text_str);
+		UnicodeConvert(text_str, out);
+		result = nvdaController_speakText(out.c_str());
+		return result == 0;
+	}
+	return false;
 }
 bool NVDA::Braille(const char* text) {
 	if (!GetActive())return false;
@@ -64,4 +76,3 @@ bool NVDA::PauseSpeech() {
 bool NVDA::ResumeSpeech() {
 	return PauseSpeech(); // Don't know how to do it
 }
-#endif

SRC/NVDA.h

@@ -19,7 +19,7 @@ class NVDA : public Engine {
 	bool Initialize()override;
 	bool Uninitialize()override;
 	int GetFeatures()override {
-		return SUPPORTS_SPEECH | SUPPORTS_BRAILLE | SUPPORTS_PAUSE_SPEECH;
+		return SUPPORTS_SPEECH | SUPPORTS_BRAILLE | SUPPORTS_PAUSE_SPEECH | SUPPORTS_SSML;
 	}
 	void SetVolume(uint64_t)override { return; }
 	uint64_t GetVolume()override { return 0; }
@@ -45,11 +45,13 @@ class NVDA : public Engine {
 	typedef error_status_t(__stdcall* NVDAController_brailleMessage)(const wchar_t*);
 	typedef error_status_t(__stdcall* NVDAController_cancelSpeech)();
 	typedef error_status_t(__stdcall* NVDAController_testIfRunning)();
+	typedef error_status_t(__stdcall* NVDAController_speakSsml)(const wchar_t*, int, int, int);
 
 	NVDAController_speakText nvdaController_speakText = nullptr;
 	NVDAController_brailleMessage nvdaController_brailleMessage = nullptr;
 	NVDAController_cancelSpeech nvdaController_cancelSpeech = nullptr;
 	NVDAController_testIfRunning nvdaController_testIfRunning = nullptr;
+	NVDAController_speakSsml nvdaController_speakSsml = nullptr;
 };
 #endif
 #endif

SRC/SAPI.cpp

@@ -1,8 +1,11 @@
 #ifdef _WIN32
 #include "SAPI.h"
+#include "Util.h"
 #include <cstdio>
 #include<string>
 #include<thread>
+
+
 static char* trim(char* data, unsigned long* size, WAVEFORMATEX* wfx, int threshold) {
 	int channels = wfx->nChannels;
 	int bytesPerSample = wfx->wBitsPerSample / 8;
@@ -128,9 +131,10 @@ bool SAPI::Speak(const char* text, bool interrupt) {
 		this->Uninitialize();
 		this->Initialize();
 	}
-
+	std::string text_str(text);
+	if (!IsSsml(text_str))AddSsml(text_str);
 	unsigned long bytes;
-	char* audio_ptr = blastspeak_speak_to_memory(instance, &bytes, text);
+	char* audio_ptr = blastspeak_speak_to_memory(instance, &bytes, text_str.c_str());
 	if (audio_ptr == nullptr)
 		return false;
 

SRC/SAPI.h

@@ -20,7 +20,7 @@ class SAPI : public Engine {
 	bool Initialize()override;
 	bool Uninitialize()override;
 	int GetFeatures()override {
-		return SUPPORTS_SPEECH | SUPPORTS_SPEECH_RATE | SUPPORTS_SPEECH_VOLUME | SUPPORTS_SELECT_VOICE | SUPPORTS_PAUSE_SPEECH;
+		return SUPPORTS_SPEECH | SUPPORTS_SPEECH_RATE | SUPPORTS_SPEECH_VOLUME | SUPPORTS_SELECT_VOICE | SUPPORTS_PAUSE_SPEECH | SUPPORTS_SSML;
 	}
 	void SetVolume(uint64_t value)override;
 	uint64_t GetVolume()override;

SRC/SRAL.cpp

@@ -1,6 +1,7 @@
 #define SRAL_EXPORT
 #include "../Include/SRAL.h"
 #include "Engine.h"
+#include "Util.h"
 #if defined(_WIN32)
 #define UNICODE
 #include "NVDA.h"
@@ -366,7 +367,9 @@ extern "C" SRAL_API bool SRAL_Output(const char* text, bool interrupt) {
 	if (g_currentEngine == nullptr)return false;
 	speech_engine_update();
 	const bool speech = SRAL_Speak(text, interrupt);
-	const bool braille = SRAL_Braille(text);
+	std::string braille_str(text);
+	if (IsSsml(braille_str))RemoveSsml(braille_str);
+	const bool braille = SRAL_Braille(braille_str.c_str());
 	return speech || braille;
 }
 extern "C" SRAL_API bool SRAL_StopSpeech(void) {

SRC/Util.cpp

@@ -0,0 +1,26 @@
+#include "Util.h"
+#include <algorithm>
+#include <regex>
+
+
+bool IsSsml(const std::string& str) {
+	std::string cpy_str = str;
+	cpy_str.erase(std::remove_if(cpy_str.begin(), cpy_str.end(), ::isspace), cpy_str.end());
+	size_t pos = cpy_str.find("<speak>");
+	if (pos == 0 && cpy_str.find("</speak>") != std::string::npos) {
+		return true;
+	}
+	return false;
+}
+bool AddSsml(std::string& str) {
+	if (IsSsml(str))return true;
+	str = "<speak>" + str + "</speak>";
+	return IsSsml(str);
+}
+bool RemoveSsml(std::string& str) {
+	if (!IsSsml(str))return true;
+	std::regex ssml_tags("<[^>]+>");
+	str = std::regex_replace(str, ssml_tags, ""); // Replace SSML tags with an empty string
+	return true;
+}
+

SRC/Util.h

@@ -0,0 +1,7 @@
+#ifndef UTIL_H_
+#define UTIL_H_
+#include <string>
+bool IsSsml(const std::string& str);
+bool AddSsml(std::string& str);
+bool RemoveSsml(std::string& str);
+#endif