* Fixed some bugs related to voice variants

* Fixed voice manager shows invalid options for some voices

3.0-beta.json

@@ -3,7 +3,7 @@
 	"displayName": "Sonata Neural Voices",
 	"URL": "",
 	"description": "Neural voices for NVDA based on Sonata",
-	"sha256": "63e667de02ed7ee4aa63354e66b5b781a1c601de2a62594c9589d4cd5af4a4c6",
+	"sha256": "9f00146ef8d2ee07ff42d99e8aaf5f20aa4412e32249c27a6da2aaa59831f6ef",
 	"homepage": null,
 	"addonVersionName": "3.0-beta",
 	"addonVersionNumber": {

addon/globalPlugins/sonata_tts_global_plugin/voice_download.py

@@ -174,12 +174,7 @@ def get_preview_url(self, speaker_idx=0):
     def get_rt_variant_download_url(self):
         if not self.has_rt_variant:
             raise ValueError(f"Voice `{self.key}` has no RT variant")
-        lang, name, quality = self.key.split("-")
-        rt_voice_key = "-".join([
-            lang,
-            f"{name}+RT",
-            quality
-        ])
+        ___, rt_voice_key = SonataTextToSpeechSystem.get_voice_variants(self.key)
         return RT_VOICE_DOWNLOAD_URL_PREFIX + rt_voice_key + ".tar.gz"
 
 
@@ -490,23 +485,13 @@ def get_available_voices(force_online=False):
             installed_voice_keys = {voice.key for voice in installed_voices}
             not_installed = []
             for (key, value) in voices.items():
-                value["standard_variant_installed"] = False
-                value["fast_variant_installed"] = False
-                lang, name, quality = key.split("-")
-                rt_key = "-".join([
-                    lang,
-                    f"{name}+RT",
-                    quality
-                ])
-                if (key not in installed_voice_keys) or (rt_key not in installed_voice_keys):
-                    not_installed.append(value)
-                    continue
-                if key in installed_voice_keys:
-                    value["standard_variant_installed"] = True
-                if rt_key in installed_voice_keys:
-                    value["fast_variant_installed"] = True
+                std_key, rt_key = SonataTextToSpeechSystem.get_voice_variants(key)
+                value["standard_variant_installed"] = std_key in installed_voice_keys
+                value["fast_variant_installed"] = rt_key in installed_voice_keys
                 if value["standard_variant_installed"] and value["fast_variant_installed"]:
                     continue
+                if value["standard_variant_installed"] and not value["has_rt_variant"]:
+                    continue
                 not_installed.append(value)
             voice_objs = PiperVoice.from_list_of_dicts(not_installed)
             return voice_objs

addon/globalPlugins/sonata_tts_global_plugin/voice_manager.py

@@ -102,10 +102,7 @@ def invalidate_cache(self):
                 synth.__init__()
 
     def _get_installed_voice_name(self, voice):
-        name = voice.name
-        if "+RT" in name:
-            return name.replace("+RT", "") + " (fast)"
-        return name + " (standard)"
+        return f"{voice.name} ({voice.variant})"
 
     def on_model_card(self, event):
         selected = self.voices_list.get_selected()
@@ -328,7 +325,9 @@ def on_voice_selected(self, event):
         if selected_voice is None:
             return
         self.download_std_btn.Enable(not selected_voice.standard_variant_installed)
-        self.download_rt_btn.Enable(not selected_voice.fast_variant_installed)
+        self.download_rt_btn.Enable(
+            selected_voice.has_rt_variant and not selected_voice.fast_variant_installed
+        )
         if selected_voice.num_speakers > 1:
             self.speaker_choice.Enable(True)
             speakers = list(selected_voice.speaker_id_map.keys())

addon/installTasks.py

@@ -14,24 +14,24 @@
 
 
 _DIR = os.path.abspath(os.path.dirname(__file__))
-_PIPER_SYNTH_DIR = os.path.join(_DIR, "synthDrivers", "piper_neural_voices")
+_PIPER_SYNTH_DIR = os.path.join(_DIR, "synthDrivers", "sonata_neural_voices")
 LIB_DIR = os.path.join(_PIPER_SYNTH_DIR, "lib")
 BIN_DIR = os.path.join(_PIPER_SYNTH_DIR, "bin")
 del _DIR, _PIPER_SYNTH_DIR
 
 
 def onUninstall():
     with _temporary_import_psutil() as psutil:
-        force_kill_piper_grpc_server(psutil)
+        force_kill_sonata_grpc_server(psutil)
 
 
-def force_kill_piper_grpc_server(psutil):
+def force_kill_sonata_grpc_server(psutil):
     log.debug("Trying to force kill GRPC server process")
     grpc_server_processes = list(filter(
-        lambda p: "piper-grpc" in p.name().lower(),
+        lambda p: "sonata-grpc" in p.name().lower(),
         psutil.process_iter(attrs=["name", "exe"])
     ))
-    grpc_server_exe = os.path.join(BIN_DIR, "piper-grpc.exe")
+    grpc_server_exe = os.path.join(BIN_DIR, "sonata-grpc.exe")
     for proc in grpc_server_processes:
         if os.path.samefile(proc.exe(), grpc_server_exe):
             proc.kill()

addon/synthDrivers/sonata_neural_voices/__init__.py

@@ -234,11 +234,7 @@ def __init__(self):
         self.availableLanguages = {v.language for v in self.voices}
         self._voice_map = {v.key: v for v in self.voices}
         self._standard_voice_map = {v.standard_variant_key: v for v in self.voices}
-        (
-            self.availableVoices,
-            self.__voice_variants,
-            self.__default_variants,
-        ) = self._get_voices_and_variants()
+        self.availableVoices = self._get_valid_voices()
         self.__voice = None
 
     def terminate(self):
@@ -518,11 +514,12 @@ def _set_voice(self, value):
             del self._availableVariants
         with suppress(AttributeError):
             del self._availableSpeakers
+        self.tts.voice = self._standard_voice_map[value].key
         if value in SonataConfig:
             variant = SonataConfig[value].get("variant", self.variant)
             speaker = SonataConfig[value].get("speaker")
         else:
-            variant = self.__default_variants[value]
+            variant = self._standard_voice_map[value].variant
             speaker = None
         self._set_variant(variant)
 
@@ -543,14 +540,14 @@ def _set_language(self, value):
         self.tts.language = value
 
     def _get_variant(self):
-        return self._voice_map[self.tts.voice].variant
+        return self.tts.speech_options.voice.variant
 
     def _set_variant(self, value):
         variant = value.lower()
         if variant == "standard":
-            voice_key = self._standard_voice_map[self.__voice].standard_variant_key
+            voice_key = self.tts.speech_options.voice.standard_variant_key
         elif variant == "fast":
-            voice_key = self._standard_voice_map[self.__voice].fast_variant_key
+            voice_key = self.tts.speech_options.voice.fast_variant_key
         else:
             log.info(f"Unknown voice variant: {variant}")
             return
@@ -562,30 +559,26 @@ def _set_variant(self, value):
         self._player = self._get_or_create_player(voice.sample_rate)
 
     def _getAvailableVariants(self):
+        std_key, rt_key = SonataTextToSpeechSystem.get_voice_variants(self.__voice)
         rv = OrderedDict()
-        for (variant, vinfo) in self.__voice_variants[self.voice].items():
-            rv[variant] = VoiceInfo(variant, variant.title(), vinfo.language)
+        if std_key in self._voice_map:
+            rv["standard"] = VoiceInfo("standard", "Standard", self.language)
+        if rt_key in self._voice_map:
+            rv["fast"] = VoiceInfo("fast", "Fast", self.language)
         return rv
 
     def _get_variant_independent_voice_id(self, voice_key):
-        return voice_key.replace("+RT", "")
+        return SonataTextToSpeechSystem.get_voice_variants(voice_key)[0]
 
-    def _get_voices_and_variants(self):
+    def _get_valid_voices(self):
         all_voices = OrderedDict()
-        all_variants = OrderedDict()
-        default_variants = OrderedDict()
         for voice in self.voices:
             voice_id = self._get_variant_independent_voice_id(voice.key)
             quality = voice.properties["quality"]
             lang = languageHandler.normalizeLanguage(voice.language).replace("_", "-")
             display_name = f"{voice.name} ({lang}) - {quality}"
-            variant = voice.variant
             all_voices[voice_id] = VoiceInfo(voice_id, display_name, voice.language)
-            all_variants.setdefault(voice_id, {})[variant] = VoiceInfo(
-                variant, variant.title(), voice.language
-            )
-            default_variants.setdefault(voice_id, variant)
-        return all_voices, all_variants, default_variants
+        return all_voices
 
     def _get_speaker(self):
         return self.tts.speaker

addon/synthDrivers/sonata_neural_voices/grpc_client/__init__.py

@@ -84,8 +84,6 @@ async def initialize():
     port = SONATA_GRPC_SERVER_PORT
     CHANNEL = grpc.aio.insecure_channel(f"localhost:{port}")
     SONATA_GRPC_SERVICE = sonata_grpcStub(CHANNEL)
-    retval = await bench()
-    log.info(f"Bench: {retval}")
 
 
 @atexit.register

addon/synthDrivers/sonata_neural_voices/tts_system.py

@@ -89,7 +89,7 @@ def from_path(cls, path):
             raise ValueError(f"Invalid voice path: {path}")
         return cls(
             key=key,
-            name=name,
+            name=name.replace("+RT", ""),
             language=normalizeLanguage(lang),
             description="",
             location=path,
@@ -170,17 +170,11 @@ def variant(self):
 
     @property
     def standard_variant_key(self):
-        return self.key.replace("+RT", "")
+        return SonataTextToSpeechSystem.get_voice_variants(self.key)[0]
 
     @property
     def fast_variant_key(self):
-        key = self.key.replace("+RT", "")
-        lang, name, quality = key.split("-")
-        return "-".join([
-            lang,
-            f"{name}+RT",
-            quality
-        ])
+        return SonataTextToSpeechSystem.get_voice_variants(self.key)[1]
 
     async def synthesize(self, text, rate, volume, pitch, sentence_silence_ms):
         if (len(text) < 10) and (set(text.strip()).issubset(IGNORED_PUNCS)):
@@ -363,6 +357,12 @@ def get_speakers(self):
                 FALLBACK_SPEAKER_NAME,
             ]
 
+    @staticmethod
+    def get_voice_variants(voice_key):
+        std_key = voice_key.replace("+RT", "")
+        lang, name, quality = std_key.split("-")
+        rt_key = f"{lang}-{name}+RT-{quality}"
+        return std_key, rt_key
     def create_speech_provider(self, text):
         return SpeechProvider(text, self.speech_options.copy())
 

buildVars.py

@@ -25,7 +25,7 @@ def _(arg):
 	# Translators: Long description to be shown for this add-on on add-on information from add-ons manager
 	"addon_description": _("""Neural voices for NVDA based on Sonata"""),
 	# version
-	"addon_version": "3.0-beta",
+	"addon_version": "3.0-beta.2",
 	# Author(s)
 	"addon_author": "Musharraf Omer <[email protected]>",
 	# URL for the add-on documentation support