Fix TTS logging spam

Testing shows that it is more stable than the proactor loop. Dunno why

pyproject.toml

@@ -8,21 +8,20 @@ authors = [
 dependencies = [
     "click",
     "kdl-py",
-    "quart==0.18.*",
-    "werkzeug==2.3.7",
+    "quart",
+    "werkzeug",
     "hypercorn",
-    "websockets==11.0.3",
-    "aioprocessing",
+    "websockets",
     "aioscheduler",
-    "pyaudio==0.2.*",
-    "librosa==0.8.*",
+    "pyaudio",
+    "librosa",
     "pytsmod",
     "numpy",
     "multipledispatch",
     "blessed",
     "appdirs",
     "maya",
-    "httpx>=0.28.1",
+    "httpx",
 ]
 requires-python = ">=3.10,<3.11"
 readme = "README.md"
@@ -30,18 +29,17 @@ license = {text = "GPLv2"}
 
 [project.optional-dependencies]
 tts = [
-  "TTS==0.9.*",
-  "torch==1.13.*",
+  "coqui-tts",
 ]
 phrasecounter = ["num2words"]
-jail = ["owoify-py==2.*"]
+jail = ["owoify-py"]
 twitch = ["miniirc"]
 midi = [
   "mido",
   "python-rtmidi",
 ]
 obs = ["simpleobsws"]
-osc = ["python-osc>=1.9.0"]
+osc = ["python-osc"]
 yt-dlp = ["yt-dlp"]
 
 [build-system]

src/ovtk_audiencekit/cli/group.py

@@ -61,6 +61,10 @@ def cli(loglevel, show_time=False):
     logging.getLogger('hypercorn.access').setLevel(logging.WARN)
     logging.getLogger('httpx').setLevel(logging.WARN)
     logging.getLogger('httpcore').setLevel(logging.INFO)
+    logging.getLogger('torio._extension.utils').setLevel(logging.WARN)
+    logging.getLogger('matplotlib').setLevel(logging.INFO)
+    logging.getLogger('fsspec').setLevel(logging.INFO)
+    logging.getLogger('TTS').setLevel(logging.INFO if loglevel == logging.DEBUG else logging.WARN)
     # Quiet warnings
     if loglevel > logging.DEBUG:
         warnings.filterwarnings("ignore")

src/ovtk_audiencekit/core/Audio.py

@@ -8,7 +8,6 @@ import pyaudio as pya
 import librosa
 import pytsmod as tsm
 import soundfile
-from aioprocessing import AioEvent
 
 # HACK: Redirect stderr to /dev/null to silence portaudio boot
 devnull = os.open(os.devnull, os.O_WRONLY)
@@ -53,9 +52,9 @@ class Clip:
 
     def stretch(self, speed, keep_pitch=True):
         if keep_pitch:
-            stretched = tsm.wsola(self._stereo_transpose(self.raw), speed)
+            stretched = tsm.wsola(self._stereo_transpose(self.raw), 1 / speed)
         else:
-            stretched = librosa.resample(self._stereo_transpose(self.raw), self.samplerate * (1 / speed), self.samplerate, fix=False, scale=True)
+            stretched = librosa.resample(self._stereo_transpose(self.raw), self.samplerate * speed, self.samplerate, fix=False, scale=True)
         self.raw = np.ascontiguousarray(self._stereo_transpose(stretched), dtype='float32')
 
     def save(self, filename):
@@ -67,7 +66,8 @@ class Stream:
         self.clip = clip
         self.pos = 0
         self.playing = False
-        self._end_event = AioEvent()
+        self.loop = asyncio.get_event_loop()
+        self._end_event = asyncio.Event()
         self._stream = pyaudio.open(
             output_device_index=output_index,
             format=pya.paFloat32,
@@ -85,16 +85,11 @@ class Stream:
         if not self._stream.is_active():
             self._stream.start_stream()
 
-    def play(self):
-        self._end_event.clear()
-        self._play()
-        self._end_event.wait(timeout=self.clip.length)
-
-    async def aplay(self):
+    async def play(self):
         self._end_event.clear()
         self._play()
         try:
-            await self._end_event.coro_wait(timeout=self.clip.length)
+            await self._end_event.wait()
         except asyncio.CancelledError:
             self.playing = False
             self._stream.stop_stream()
@@ -117,7 +112,7 @@ class Stream:
 
             if self.pos >= self.clip.raw.shape[0]:
                 self.playing = False
-                self._end_event.set()
+                self.loop.call_soon_threadsafe(self._end_event.set)
 
         return buffer, pya.paContinue
 

src/ovtk_audiencekit/core/Config.py

@@ -139,7 +139,7 @@ def parse_kdl_deep(path, relativeto=None):
     if relativeto:
         path = os.path.normpath(os.path.join(relativeto, path))
 
-    with open(path, 'r') as f:
+    with open(path, 'r', encoding='utf-8') as f:
         try:
             config = kdl.parse(f.read(), kdl_parse_config)
             for node in config.nodes:

src/ovtk_audiencekit/core/MainProcess.py

@@ -3,7 +3,6 @@ import asyncio
 from datetime import datetime, timedelta
 import logging
 import os
-import os.path
 import pathlib
 import sys
 import signal
@@ -67,6 +66,9 @@ class MainProcess:
         # Save sys.path since some config will clobber it
         self._initial_syspath = sys.path
 
+        if os.name == 'nt':
+            asyncio.set_event_loop_policy(asyncio.WindowsSelectorEventLoopPolicy())
+
     def _unload_plugin(self, plugin_name):
         plugin = self.plugins[plugin_name]
         plugin.close()
@@ -244,12 +246,15 @@ class MainProcess:
         sys.path = self._initial_syspath
 
     async def _discount_repl(self):
+        loop = asyncio.get_event_loop()
         # REVIEW: Not a good UX at the moment (as new logs clobber the terminal entry)
-        async for line in reader:
+        while True:
+            line = await loop.run_in_executor(None, sys.stdin.readline)
             line = line.strip()
-            if line == b'reload':
+            logger.debug(f'Got terminal input: {line}')
+            if line == 'reload':
                 self.reload_ev.set()
-            elif line == b'quit':
+            elif line == 'quit':
                 self.shutdown_ev.set()
 
     async def run(self):
@@ -261,8 +266,6 @@ class MainProcess:
         try:
             # System setup
             ## Make stdin handler
-            reader = asyncio.StreamReader()
-            await loop.connect_read_pipe(lambda: asyncio.StreamReaderProtocol(reader), sys.stdin)
             self.cli_task = loop.create_task(self._discount_repl())
             ## Init websocket server (external end of the event bus)
             self.bus_server = WebsocketServerProcess(self.event_queue, *self.bus_conf)

src/ovtk_audiencekit/core/Plugins.py

@@ -29,6 +29,17 @@ class OvtkBlueprint(quart.Blueprint):
             endpoint = self.name + endpoint
         return quart.url_for(endpoint, *args, **kwargs)
 
+    def render(self, name, **kwargs):
+        """render_template that prefers the plugin-specific templates"""
+        full = self.template_folder / name
+        if os.path.exists(full):
+            template_string = None
+            with open(full, 'r') as template_file:
+                template_string = template_file.read()
+            return quart.render_template_string(template_string, **kwargs)
+        else:
+            return quart.render_template(name, **kwargs)
+
 
 class PluginBase(ABC):
     plugins = {}

src/ovtk_audiencekit/plugins/AudioAlert/AudioAlert.py

@@ -54,9 +54,9 @@ class AudioAlert(PluginBase):
 
 
         if wait:
-            await stream.aplay()
+            await stream.play()
         else:
-            task = asyncio.create_task(stream.aplay())
+            task = asyncio.create_task(stream.play())
             task.add_done_callback(self.tasks.discard)
             self.tasks.add(task)
 

src/ovtk_audiencekit/plugins/TTS/TTS.py

@@ -52,8 +52,8 @@ class TextToSpeechPlugin(PluginBase):
             config_path = override_conf_path
 
         self.synthesizer = Synthesizer(
-            model_path,
-            config_path,
+            tts_checkpoint=model_path,
+            tts_config_path=config_path,
             vocoder_checkpoint=vocoder_path,
             vocoder_config=vocoder_config_path,
             use_cuda=self.cuda,
@@ -66,22 +66,26 @@ class TextToSpeechPlugin(PluginBase):
             task.cancel()
 
     def make_tts_wav(self, text, filename=None):
+        # Force punctuation (keeps the models from acting unpredictably)
+        text = text.strip()
+        if not any([text.endswith(punc) for punc in '.!?:']):
+            text += '.'
+
         if filename is None:
             filename = os.path.join(self.cache_dir, f'{uuid.uuid1()}.wav')
 
+        self.logger.info(f'Generating TTS "{text}"...')
         if self.speaker_wav:
             wav = self.synthesizer.tts(text, None, 'en', self.speaker_wav)
         else:
             wav = self.synthesizer.tts(text)
 
         self.synthesizer.save_wav(wav, filename)
+        self.logger.info(f'Done - saved as {filename}')
         return filename
 
     async def run(self, text, *args, _ctx={}, wait=False, **kwargs):
         try:
-            # Force punctuation (keep AI from spinning off into random noises)
-            if not any([text.endswith(punc) for punc in '.!?:']):
-                text += '.'
             # Do TTS processing in a thread to avoid blocking main loop
             filename = await asyncio.get_running_loop().run_in_executor(None, self.make_tts_wav, text)
 
@@ -90,7 +94,7 @@ class TextToSpeechPlugin(PluginBase):
             stream = Stream(clip, self.output_index)
             async def play():
                 try:
-                    await stream.aplay()
+                    await stream.play()
                 finally:
                     stream.close()
                     os.remove(os.path.join(self.cache_dir, filename))

src/ovtk_audiencekit/plugins/builtins/Scene/Plugin.py

@@ -34,7 +34,7 @@ class ScenePlugin(PluginBase):
 
         self.blueprint.add_url_rule('/', 'ctrlpanel', self.ui_ctrlpanel)
         self.blueprint.add_url_rule('/<name>/<cmd>', 'api-sceneset', self.ui_setscene)
-        self.blueprint.add_url_rule('/monitor', 'monitor', self.ui_monitor_ws, is_websocket=True)
+        self.blueprint.add_url_rule('/monitor', 'monitor', self.ui_monitor_ws, websocket=True)
 
     async def run(self, name, _children=None, _ctx={}, active=None, group=None, oneshot=False, **kwargs):
         if _children is None:
@@ -138,7 +138,7 @@ class ScenePlugin(PluginBase):
 
     async def ui_ctrlpanel(self):
         groups = self._get_state()
-        return await quart.render_template('index.html', init_state=json.dumps(groups))
+        return await self.blueprint.render('index.html', init_state=json.dumps(groups))
 
     async def ui_setscene(self, name=None, cmd=None):
         active = cmd == 'activate'

src/ovtk_audiencekit/plugins/builtins/Scene/templates/index.html

@@ -2,7 +2,7 @@
 <html lang="en" dir="ltr">
   <head>
     <meta charset="utf-8">
-    <title>Test page</title>
+    <title>Scene control</title>
     <script type="importmap">
       {
         "imports": { "vue": "https://unpkg.com/vue@3/dist/vue.esm-browser.js" }