Sync/tts cache #15
Sync/tts cache #15
Conversation
JarbasAl
force-pushed
the
sync/tts_cache
branch
from
77439f1
to
aea5cea
Compare
JarbasAl
force-pushed
the
sync/tts_cache
branch
from
82855b1
to
6dbd4fa
Compare
JarbasAl
force-pushed
the
sync/tts_cache
branch
4 times, most recently
from
d91f288
to
7f52c9e
Compare
port and improve the new TTS cache from mycroft-core refactor the playback thread to make it more readable solve TODO for pause/resume functionality
JarbasAl
force-pushed
the
sync/tts_cache
branch
from
2b52bbd
to
fe55018
Compare
bumps ovos plugin manager to 0.0.3a1 and removes duplicted code latest mycroft-core tts cache implementation ported in OpenVoiceOS/ovos-plugin-manager#15 psutil is no longer a mandatory requirement authored-by: jarbasai <jarbasai@mailfence.com>
| try: | ||
| if len(self._now_playing) == 5: | ||
| # new mycroft style | ||
| snd_type, data, visemes, ident, listen = self._now_playing |
There was a problem hiding this comment.
Is there a reason this is a tuple and not a dict or another more explicit structure?
There was a problem hiding this comment.
its what mycroft uses and we need backwards compat, i can improve the code but the user facing api can't change only be augmented
| self.lang = lang or config.get("lang") or 'en-us' | ||
| self.config = config or {} | ||
| self.validator = validator or TTSValidator(self) | ||
| self.phonetic_spelling = phonetic_spelling | ||
| self.audio_ext = audio_ext | ||
| self.ssml_tags = ssml_tags or [] | ||
| self.log_timestamps = self.config.get("log_timestamps", False) |
There was a problem hiding this comment.
What timestamps? Could this be annotated or renamed for clarity?
There was a problem hiding this comment.
this is the default metrics handler, all TTS plugins have this option
see these debug logs for example (debugging TTS not happening, note the timestamps in playback step)
2021-11-17 20:42:08.042 - OVOS - mycroft.audio.speech:mute_and_speak:130 - INFO - Speak: Please wait a moment as I finish booting up.
2021-11-17 20:42:08.043 - OVOS - ovos_plugin_manager.templates.tts:handle_metric:249 - DEBUG - time delta: 0 metric: {'metric_type': 'tts.ssml.validated'}
2021-11-17 20:42:08.049 - OVOS - ovos_plugin_manager.templates.tts:handle_metric:249 - DEBUG - time delta: 0.00035881996154785156 metric: {'metric_type': 'tts.preprocessed', 'n_chunks': 1}
2021-11-17 20:42:08.050 - OVOS - ovos_plugin_manager.templates.tts:handle_metric:249 - DEBUG - time delta: 0.0017445087432861328 metric: {'metric_type': 'tts.synth.start'}
2021-11-17 20:42:08.242 - OVOS - ovos_plugin_manager.templates.tts:handle_metric:249 - DEBUG - time delta: 0.194044828414917 metric: {'metric_type': 'tts.synth.finished'}
2021-11-17 20:42:08.243 - OVOS - ovos_plugin_manager.templates.tts:handle_metric:249 - DEBUG - time delta: 0.19539237022399902 metric: {'metric_type': 'tts.synth.cached'}
2021-11-17 20:42:08.246 - OVOS - ovos_plugin_manager.templates.tts:handle_metric:249 - DEBUG - time delta: 0.19644379615783691 metric: {'metric_type': 'tts.queued'}
2021-11-17 20:42:08.249 - OVOS - ovos_plugin_manager.templates.tts:handle_metric:249 - DEBUG - time delta: 0.20165657997131348 metric: {'metric_type': 'tts.start'}
2021-11-17 20:42:08.254 - OVOS - ovos_plugin_manager.templates.tts:handle_metric:249 - DEBUG - time delta: 0.20674896240234375 metric: {'metric_type': 'tts.end'}
| self.handle_metric({"metric_type": "tts.queued"}) | ||
|
|
||
| def _determine_ext(self, audio_file): | ||
| # determine audio_ext on the fly |
There was a problem hiding this comment.
A potentially better approach here is to read it from file headers in case the plugin incorrectly specifies (i.e. a remote TTS implementation changes)
There was a problem hiding this comment.
i dont think plugins should specify it at all to be honest, but that needs a larger refactor. and mycroft is also talking about the same so lets wait and see.... mycroft intends to deprecate allowing TTS to return file path, so there will be breakage either way....
the aim here is to allow plugins to return different extensions per request, i had this happening when i added a permanent cache to polly that was in .wav format, but the plugin returns mp3.
| except: | ||
| return self.audio_ext | ||
|
|
||
| def _synth(self, sentence, sentence_hash=None, **kwargs): |
port the new handling of TTS cache from mycroft-core
I am not exactly a fan of the new approach (1 class was enough.....) but for backwards compat everything was ported, I slightly augmented the api (and hopefully usefulness) of the new classes and removed everything mimic2 specific
Cleaned up the TTS class as much as possible,
_executewas getting huge so i split it into logical smaller functions. Previously existing cache methods were marked for deprecation in mycroft-core but here are integrated with the new code and won't be deprecatedAlso ported the RemoteTTS class from ovos-core, it is not very useful but allows for full deprecation of tts module in ovos-core
Creating the persistent cache files is out of scope and depends on the engine
companion PR OpenVoiceOS/ovos-core#16