Implement text-to-speech support on Android, iOS, HTML5, Linux, macOS and Windows.

Implement TextServer word break method.

6 files changed, 148 insertions, 0 deletions

diff --git a/servers/display_server.cpp b/servers/display_server.cpp
index 67bfc75426..8d97cd2543 100644
--- a/servers/display_server.cpp
+++ b/servers/display_server.cpp
@@ -220,6 +220,81 @@ void DisplayServer::global_menu_clear(const String &p_menu_root) {
 	WARN_PRINT("Global menus not supported by this display server.");
 }
 
+bool DisplayServer::tts_is_speaking() const {
+	WARN_PRINT("TTS is not supported by this display server.");
+	return false;
+}
+
+bool DisplayServer::tts_is_paused() const {
+	WARN_PRINT("TTS is not supported by this display server.");
+	return false;
+}
+
+void DisplayServer::tts_pause() {
+	WARN_PRINT("TTS is not supported by this display server.");
+}
+
+void DisplayServer::tts_resume() {
+	WARN_PRINT("TTS is not supported by this display server.");
+}
+
+Array DisplayServer::tts_get_voices() const {
+	WARN_PRINT("TTS is not supported by this display server.");
+	return Array();
+}
+
+PackedStringArray DisplayServer::tts_get_voices_for_language(const String &p_language) const {
+	PackedStringArray ret;
+	Array voices = tts_get_voices();
+	for (int i = 0; i < voices.size(); i++) {
+		const Dictionary &voice = voices[i];
+		if (voice.has("id") && voice.has("language") && voice["language"].operator String().begins_with(p_language)) {
+			ret.push_back(voice["id"]);
+		}
+	}
+	return ret;
+}
+
+void DisplayServer::tts_speak(const String &p_text, const String &p_voice, int p_volume, float p_pitch, float p_rate, int p_utterance_id, bool p_interrupt) {
+	WARN_PRINT("TTS is not supported by this display server.");
+}
+
+void DisplayServer::tts_stop() {
+	WARN_PRINT("TTS is not supported by this display server.");
+}
+
+void DisplayServer::tts_set_utterance_callback(TTSUtteranceEvent p_event, const Callable &p_callable) {
+	ERR_FAIL_INDEX(p_event, DisplayServer::TTS_UTTERANCE_MAX);
+	utterance_callback[p_event] = p_callable;
+}
+
+void DisplayServer::tts_post_utterance_event(TTSUtteranceEvent p_event, int p_id, int p_pos) {
+	ERR_FAIL_INDEX(p_event, DisplayServer::TTS_UTTERANCE_MAX);
+	switch (p_event) {
+		case DisplayServer::TTS_UTTERANCE_STARTED:
+		case DisplayServer::TTS_UTTERANCE_ENDED:
+		case DisplayServer::TTS_UTTERANCE_CANCELED: {
+			if (utterance_callback[p_event].is_valid()) {
+				Variant args[1];
+				args[0] = p_id;
+				const Variant *argp[] = { &args[0] };
+				utterance_callback[p_event].call_deferred(argp, 1); // Should be deferred, on some platforms utterance events can be called from different threads in a rapid succession.
+			}
+		} break;
+		case DisplayServer::TTS_UTTERANCE_BOUNDARY: {
+			if (utterance_callback[p_event].is_valid()) {
+				Variant args[2];
+				args[0] = p_pos;
+				args[1] = p_id;
+				const Variant *argp[] = { &args[0], &args[1] };
+				utterance_callback[p_event].call_deferred(argp, 2); // Should be deferred, on some platforms utterance events can be called from different threads in a rapid succession.
+			}
+		} break;
+		default:
+			break;
+	}
+}
+
 void DisplayServer::mouse_set_mode(MouseMode p_mode) {
 	WARN_PRINT("Mouse is not supported by this display server.");
 }
@@ -478,6 +553,19 @@ void DisplayServer::_bind_methods() {
 	ClassDB::bind_method(D_METHOD("global_menu_remove_item", "menu_root", "idx"), &DisplayServer::global_menu_remove_item);
 	ClassDB::bind_method(D_METHOD("global_menu_clear", "menu_root"), &DisplayServer::global_menu_clear);
 
+	ClassDB::bind_method(D_METHOD("tts_is_speaking"), &DisplayServer::tts_is_speaking);
+	ClassDB::bind_method(D_METHOD("tts_is_paused"), &DisplayServer::tts_is_paused);
+	ClassDB::bind_method(D_METHOD("tts_get_voices"), &DisplayServer::tts_get_voices);
+	ClassDB::bind_method(D_METHOD("tts_get_voices_for_language", "language"), &DisplayServer::tts_get_voices_for_language);
+
+	ClassDB::bind_method(D_METHOD("tts_speak", "text", "voice", "volume", "pitch", "rate", "utterance_id", "interrupt"), &DisplayServer::tts_speak, DEFVAL(50), DEFVAL(1.f), DEFVAL(1.f), DEFVAL(0), DEFVAL(false));
+	ClassDB::bind_method(D_METHOD("tts_pause"), &DisplayServer::tts_pause);
+	ClassDB::bind_method(D_METHOD("tts_resume"), &DisplayServer::tts_resume);
+	ClassDB::bind_method(D_METHOD("tts_stop"), &DisplayServer::tts_stop);
+
+	ClassDB::bind_method(D_METHOD("tts_set_utterance_callback", "event", "callable"), &DisplayServer::tts_set_utterance_callback);
+	ClassDB::bind_method(D_METHOD("_tts_post_utterance_event", "event", "id", "char_pos"), &DisplayServer::tts_post_utterance_event);
+
 	ClassDB::bind_method(D_METHOD("mouse_set_mode", "mouse_mode"), &DisplayServer::mouse_set_mode);
 	ClassDB::bind_method(D_METHOD("mouse_get_mode"), &DisplayServer::mouse_get_mode);
 
@@ -621,6 +709,7 @@ void DisplayServer::_bind_methods() {
 	BIND_ENUM_CONSTANT(FEATURE_ORIENTATION);
 	BIND_ENUM_CONSTANT(FEATURE_SWAP_BUFFERS);
 	BIND_ENUM_CONSTANT(FEATURE_CLIPBOARD_PRIMARY);
+	BIND_ENUM_CONSTANT(FEATURE_TEXT_TO_SPEECH);
 
 	BIND_ENUM_CONSTANT(MOUSE_MODE_VISIBLE);
 	BIND_ENUM_CONSTANT(MOUSE_MODE_HIDDEN);
@@ -689,6 +778,11 @@ void DisplayServer::_bind_methods() {
 	BIND_ENUM_CONSTANT(DISPLAY_HANDLE);
 	BIND_ENUM_CONSTANT(WINDOW_HANDLE);
 	BIND_ENUM_CONSTANT(WINDOW_VIEW);
+
+	BIND_ENUM_CONSTANT(TTS_UTTERANCE_STARTED);
+	BIND_ENUM_CONSTANT(TTS_UTTERANCE_ENDED);
+	BIND_ENUM_CONSTANT(TTS_UTTERANCE_CANCELED);
+	BIND_ENUM_CONSTANT(TTS_UTTERANCE_BOUNDARY);
 }
 
 void DisplayServer::register_create_function(const char *p_name, CreateFunction p_function, GetRenderingDriversFunction p_get_drivers) {
diff --git a/servers/display_server.h b/servers/display_server.h
index 4961b07ba3..19efcbd3dd 100644
--- a/servers/display_server.h
+++ b/servers/display_server.h
@@ -121,6 +121,7 @@ public:
 		FEATURE_SWAP_BUFFERS,
 		FEATURE_KEEP_SCREEN_ON,
 		FEATURE_CLIPBOARD_PRIMARY,
+		FEATURE_TEXT_TO_SPEECH,
 	};
 
 	virtual bool has_feature(Feature p_feature) const = 0;
@@ -172,6 +173,40 @@ public:
 	virtual void global_menu_remove_item(const String &p_menu_root, int p_idx);
 	virtual void global_menu_clear(const String &p_menu_root);
 
+	struct TTSUtterance {
+		String text;
+		String voice;
+		int volume = 50;
+		float pitch = 1.f;
+		float rate = 1.f;
+		int id = 0;
+	};
+
+	enum TTSUtteranceEvent {
+		TTS_UTTERANCE_STARTED,
+		TTS_UTTERANCE_ENDED,
+		TTS_UTTERANCE_CANCELED,
+		TTS_UTTERANCE_BOUNDARY,
+		TTS_UTTERANCE_MAX,
+	};
+
+private:
+	Callable utterance_callback[TTS_UTTERANCE_MAX];
+
+public:
+	virtual bool tts_is_speaking() const;
+	virtual bool tts_is_paused() const;
+	virtual Array tts_get_voices() const;
+	virtual PackedStringArray tts_get_voices_for_language(const String &p_language) const;
+
+	virtual void tts_speak(const String &p_text, const String &p_voice, int p_volume = 50, float p_pitch = 1.f, float p_rate = 1.f, int p_utterance_id = 0, bool p_interrupt = false);
+	virtual void tts_pause();
+	virtual void tts_resume();
+	virtual void tts_stop();
+
+	virtual void tts_set_utterance_callback(TTSUtteranceEvent p_event, const Callable &p_callable);
+	virtual void tts_post_utterance_event(TTSUtteranceEvent p_event, int p_id, int p_pos = 0);
+
 	enum MouseMode {
 		MOUSE_MODE_VISIBLE,
 		MOUSE_MODE_HIDDEN,
@@ -431,5 +466,6 @@ VARIANT_ENUM_CAST(DisplayServer::WindowFlags)
 VARIANT_ENUM_CAST(DisplayServer::HandleType)
 VARIANT_ENUM_CAST(DisplayServer::CursorShape)
 VARIANT_ENUM_CAST(DisplayServer::VSyncMode)
+VARIANT_ENUM_CAST(DisplayServer::TTSUtteranceEvent)
 
 #endif // DISPLAY_SERVER_H
diff --git a/servers/text/text_server_extension.cpp b/servers/text/text_server_extension.cpp
index 001706bb6f..005cb68302 100644
--- a/servers/text/text_server_extension.cpp
+++ b/servers/text/text_server_extension.cpp
@@ -293,6 +293,8 @@ void TextServerExtension::_bind_methods() {
 
 	GDVIRTUAL_BIND(strip_diacritics, "string");
 
+	GDVIRTUAL_BIND(string_get_word_breaks, "string", "language");
+
 	GDVIRTUAL_BIND(string_to_upper, "string", "language");
 	GDVIRTUAL_BIND(string_to_lower, "string", "language");
 
@@ -1503,6 +1505,14 @@ Array TextServerExtension::parse_structured_text(StructuredTextParser p_parser_t
 	return Array();
 }
 
+PackedInt32Array TextServerExtension::string_get_word_breaks(const String &p_string, const String &p_language) const {
+	PackedInt32Array ret;
+	if (GDVIRTUAL_CALL(string_get_word_breaks, p_string, p_language, ret)) {
+		return ret;
+	}
+	return PackedInt32Array();
+}
+
 TextServerExtension::TextServerExtension() {
 	//NOP
 }
diff --git a/servers/text/text_server_extension.h b/servers/text/text_server_extension.h
index ce781097f3..7b7fc61ed7 100644
--- a/servers/text/text_server_extension.h
+++ b/servers/text/text_server_extension.h
@@ -485,6 +485,9 @@ public:
 	virtual String strip_diacritics(const String &p_string) const override;
 	GDVIRTUAL1RC(String, strip_diacritics, const String &);
 
+	virtual PackedInt32Array string_get_word_breaks(const String &p_string, const String &p_language = "") const override;
+	GDVIRTUAL2RC(PackedInt32Array, string_get_word_breaks, const String &, const String &);
+
 	virtual String string_to_upper(const String &p_string, const String &p_language = "") const override;
 	virtual String string_to_lower(const String &p_string, const String &p_language = "") const override;
 	GDVIRTUAL2RC(String, string_to_upper, const String &, const String &);
diff --git a/servers/text_server.cpp b/servers/text_server.cpp
index d66e769e3c..7d9945f5d7 100644
--- a/servers/text_server.cpp
+++ b/servers/text_server.cpp
@@ -439,6 +439,8 @@ void TextServer::_bind_methods() {
 	ClassDB::bind_method(D_METHOD("parse_number", "number", "language"), &TextServer::parse_number, DEFVAL(""));
 	ClassDB::bind_method(D_METHOD("percent_sign", "language"), &TextServer::percent_sign, DEFVAL(""));
 
+	ClassDB::bind_method(D_METHOD("string_get_word_breaks", "string", "language"), &TextServer::string_get_word_breaks, DEFVAL(""));
+
 	ClassDB::bind_method(D_METHOD("strip_diacritics", "string"), &TextServer::strip_diacritics);
 
 	ClassDB::bind_method(D_METHOD("string_to_upper", "string", "language"), &TextServer::string_to_upper, DEFVAL(""));
diff --git a/servers/text_server.h b/servers/text_server.h
index 7e7f26b32d..b08aa26917 100644
--- a/servers/text_server.h
+++ b/servers/text_server.h
@@ -431,6 +431,9 @@ public:
 	virtual String parse_number(const String &p_string, const String &p_language = "") const = 0;
 	virtual String percent_sign(const String &p_language = "") const = 0;
 
+	// String functions.
+	virtual PackedInt32Array string_get_word_breaks(const String &p_string, const String &p_language = "") const = 0;
+
 	virtual String strip_diacritics(const String &p_string) const;
 
 	// Other string operations.