Merge pull request #229 from Holzhaus/marytts-support

Add config options for TTS engine (+ MaryTTS support)

2 files changed, 159 insertions, 2 deletions

diff --git a/client/tts.py b/client/tts.py
index 55c33b0..ca9a5a0 100644
--- a/client/tts.py
+++ b/client/tts.py
@@ -15,9 +15,13 @@ import subprocess
 import pipes
 import logging
 import wave
+import urllib
+import urlparse
+import requests
 from abc import ABCMeta, abstractmethod
 
 import argparse
+import yaml
 try:
     import mad
     import gtts
@@ -25,6 +29,7 @@ except ImportError:
     pass
 
 import diagnose
+import jasperpath
 
 
 class AbstractTTSEngine(object):
@@ -34,6 +39,16 @@ class AbstractTTSEngine(object):
     __metaclass__ = ABCMeta
 
     @classmethod
+    def get_config(cls):
+        return {}
+
+    @classmethod
+    def get_instance(cls):
+        config = cls.get_config()
+        instance = cls(**config)
+        return instance
+
+    @classmethod
     @abstractmethod
     def is_available(cls):
         return diagnose.check_executable('aplay')
@@ -120,6 +135,27 @@ class EspeakTTS(AbstractTTSEngine):
         self.words_per_minute = words_per_minute
 
     @classmethod
+    def get_config(cls):
+        # FIXME: Replace this as soon as we have a config module
+        config = {}
+        # HMM dir
+        # Try to get hmm_dir from config
+        profile_path = jasperpath.config('profile.yml')
+        if os.path.exists(profile_path):
+            with open(profile_path, 'r') as f:
+                profile = yaml.safe_load(f)
+                if 'espeak-tts' in profile:
+                    if 'voice' in profile['espeak-tts']:
+                        config['voice'] = profile['espeak-tts']['voice']
+                    if 'pitch_adjustment' in profile['espeak-tts']:
+                        config['pitch_adjustment'] = \
+                            profile['espeak-tts']['pitch_adjustment']
+                    if 'words_per_minute' in profile['espeak-tts']:
+                        config['words_per_minute'] = \
+                            profile['espeak-tts']['words_per_minute']
+        return config
+
+    @classmethod
     def is_available(cls):
         return (super(cls, cls).is_available() and
                 diagnose.check_executable('espeak'))
@@ -249,6 +285,21 @@ class PicoTTS(AbstractTTSEngine):
         return (super(cls, cls).is_available() and
                 diagnose.check_executable('pico2wave'))
 
+    @classmethod
+    def get_config(cls):
+        # FIXME: Replace this as soon as we have a config module
+        config = {}
+        # HMM dir
+        # Try to get hmm_dir from config
+        profile_path = jasperpath.config('profile.yml')
+        if os.path.exists(profile_path):
+            with open(profile_path, 'r') as f:
+                profile = yaml.safe_load(f)
+                if 'pico-tts' in profile and 'language' in profile['pico-tts']:
+                    config['language'] = profile['pico-tts']['language']
+
+        return config
+
     @property
     def languages(self):
         cmd = ['pico2wave', '-l', 'NULL',
@@ -306,6 +357,22 @@ class GoogleTTS(AbstractMp3TTSEngine):
                 diagnose.check_python_import('gtts') and
                 diagnose.check_network_connection())
 
+    @classmethod
+    def get_config(cls):
+        # FIXME: Replace this as soon as we have a config module
+        config = {}
+        # HMM dir
+        # Try to get hmm_dir from config
+        profile_path = jasperpath.config('profile.yml')
+        if os.path.exists(profile_path):
+            with open(profile_path, 'r') as f:
+                profile = yaml.safe_load(f)
+                if ('google-tts' in profile and
+                   'language' in profile['google-tts']):
+                    config['language'] = profile['google-tts']['language']
+
+        return config
+
     @property
     def languages(self):
         langs = ['af', 'sq', 'ar', 'hy', 'ca', 'zh-CN', 'zh-TW', 'hr', 'cs',
@@ -328,6 +395,96 @@ class GoogleTTS(AbstractMp3TTSEngine):
         os.remove(tmpfile)
 
 
+class MaryTTS(AbstractTTSEngine):
+    """
+    Uses the MARY Text-to-Speech System (MaryTTS)
+    MaryTTS is an open-source, multilingual Text-to-Speech Synthesis platform
+    written in Java.
+    Please specify your own server instead of using the demonstration server
+    (http://mary.dfki.de:59125/) to save bandwidth and to protect your privacy.
+    """
+
+    SLUG = "mary-tts"
+
+    def __init__(self, server="mary.dfki.de", port="59125", language="en_GB",
+                 voice="dfki-spike"):
+        super(self.__class__, self).__init__()
+        self.server = server
+        self.port = port
+        self.netloc = '{server}:{port}'.format(server=self.server,
+                                               port=self.port)
+        self.language = language
+        self.voice = voice
+        self.session = requests.Session()
+
+    @property
+    def languages(self):
+        r = self.session.get(self._makeurl('/locales'))
+        r.raise_for_status()
+        return r.text.splitlines()
+
+    @property
+    def voices(self):
+        r = self.session.get(self._makeurl('/voices'))
+        r.raise_for_status()
+        return [line.split()[0] for line in r.text.splitlines()]
+
+    @classmethod
+    def get_config(cls):
+        # FIXME: Replace this as soon as we have a config module
+        config = {}
+        # HMM dir
+        # Try to get hmm_dir from config
+        profile_path = jasperpath.config('profile.yml')
+        if os.path.exists(profile_path):
+            with open(profile_path, 'r') as f:
+                profile = yaml.safe_load(f)
+                if 'mary-tts' in profile:
+                    if 'server' in profile['mary-tts']:
+                        config['server'] = profile['mary-tts']['server']
+                    if 'port' in profile['mary-tts']:
+                        config['port'] = profile['mary-tts']['port']
+                    if 'language' in profile['mary-tts']:
+                        config['language'] = profile['mary-tts']['language']
+                    if 'voice' in profile['mary-tts']:
+                        config['voice'] = profile['mary-tts']['voice']
+
+        return config
+
+    @classmethod
+    def is_available(cls):
+        return (super(cls, cls).is_available() and
+                diagnose.check_network_connection())
+
+    def _makeurl(self, path, query={}):
+        query_s = urllib.urlencode(query)
+        urlparts = ('http', self.netloc, path, query_s, '')
+        return urlparse.urlunsplit(urlparts)
+
+    def say(self, phrase):
+        self._logger.debug("Saying '%s' with '%s'", phrase, self.SLUG)
+        if self.language not in self.languages:
+            raise ValueError("Language '%s' not supported by '%s'"
+                             % (self.language, self.SLUG))
+
+        if self.voice not in self.voices:
+            raise ValueError("Voice '%s' not supported by '%s'"
+                             % (self.voice, self.SLUG))
+        query = {'OUTPUT_TYPE': 'AUDIO',
+                 'AUDIO': 'WAVE_FILE',
+                 'INPUT_TYPE': 'TEXT',
+                 'INPUT_TEXT': phrase,
+                 'LOCALE': self.language,
+                 'VOICE': self.voice}
+
+        r = self.session.get(self._makeurl('/process', query=query))
+        with tempfile.NamedTemporaryFile(suffix='.wav', delete=False) as f:
+            f.write(r.content)
+            tmpfile = f.name
+        self.play(tmpfile)
+        os.remove(tmpfile)
+
+
 def get_default_engine_slug():
     return 'osx-tts' if platform.system() == 'darwin' else 'espeak-tts'
 
@@ -401,5 +558,5 @@ if __name__ == '__main__':
     print("")
     for i, engine in enumerate(available_engines, start=1):
         print("%d. Testing engine '%s'..." % (i, engine.SLUG))
-        engine().say("This is a test.")
+        engine.get_instance().say("This is a test.")
     print("Done.")
diff --git a/jasper.py b/jasper.py
index 24396cf..13c3ca9 100755
--- a/jasper.py
+++ b/jasper.py
@@ -100,7 +100,7 @@ class Jasper(object):
         tts_engine_class = tts.get_engine_by_slug(tts_engine_slug)
 
         # Initialize Mic
-        self.mic = Mic(tts_engine_class(),
+        self.mic = Mic(tts_engine_class.get_instance(),
                        stt.PocketSphinxSTT(**stt.PocketSphinxSTT.get_config()),
                        stt.newSTTEngine(stt_engine_type, api_key=api_key))