path: root/client/tts.py

# -*- coding: utf-8-*-
"""
A Speaker handles audio output from Jasper to the user

Speaker methods:
    say - output 'phrase' as speech
    play - play the audio in 'filename'
    is_available - returns True if the platform supports this implementation
"""
import os
import platform
import re
import tempfile
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
except ImportError:
    pass

import diagnose
import jasperpath


class AbstractTTSEngine(object):
    """
    Generic parent class for all speakers
    """
    __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')

    def __init__(self, **kwargs):
        self._logger = logging.getLogger(__name__)

    @abstractmethod
    def say(self, phrase, *args):
        pass

    def play(self, filename):
        # FIXME: Use platform-independent audio-output here
        # See issue jasperproject/jasper-client#188
        cmd = ['aplay', '-D', 'hw:1,0', str(filename)]
        self._logger.debug('Executing %s', ' '.join([pipes.quote(arg)
                                                     for arg in cmd]))
        with tempfile.TemporaryFile() as f:
            subprocess.call(cmd, stdout=f, stderr=f)
            f.seek(0)
            output = f.read()
            if output:
                self._logger.debug("Output was: '%s'", output)


class AbstractMp3TTSEngine(AbstractTTSEngine):
    """
    Generic class that implements the 'play' method for mp3 files
    """
    @classmethod
    def is_available(cls):
        return (super(AbstractMp3TTSEngine, cls).is_available() and
                diagnose.check_python_import('mad'))

    def play_mp3(self, filename):
        mf = mad.MadFile(filename)
        with tempfile.NamedTemporaryFile(suffix='.wav') as f:
            wav = wave.open(f, mode='wb')
            wav.setframerate(mf.samplerate())
            wav.setnchannels(1 if mf.mode() == mad.MODE_SINGLE_CHANNEL else 2)
            # 4L is the sample width of 32 bit audio
            wav.setsampwidth(4L)
            frame = mf.read()
            while frame is not None:
                wav.writeframes(frame)
                frame = mf.read()
            wav.close()
            self.play(f.name)


class DummyTTS(AbstractTTSEngine):
    """
    Dummy TTS engine that logs phrases with INFO level instead of synthesizing
    speech.
    """

    SLUG = "dummy-tts"

    @classmethod
    def is_available(cls):
        return True

    def say(self, phrase):
        self._logger.info(phrase)

    def play(self, filename):
        self._logger.debug("Playback of file '%s' requested")
        pass


class EspeakTTS(AbstractTTSEngine):
    """
    Uses the eSpeak speech synthesizer included in the Jasper disk image
    Requires espeak to be available
    """

    SLUG = "espeak-tts"

    def __init__(self, voice='default+m3', pitch_adjustment=40,
                 words_per_minute=160):
        super(self.__class__, self).__init__()
        self.voice = voice
        self.pitch_adjustment = pitch_adjustment
        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'))

    def say(self, phrase):
        self._logger.debug("Saying '%s' with '%s'", phrase, self.SLUG)
        with tempfile.NamedTemporaryFile(suffix='.wav', delete=False) as f:
            fname = f.name
        cmd = ['espeak', '-v', self.voice,
                         '-p', self.pitch_adjustment,
                         '-s', self.words_per_minute,
                         '-w', fname,
                         phrase]
        cmd = [str(x) for x in cmd]
        self._logger.debug('Executing %s', ' '.join([pipes.quote(arg)
                                                     for arg in cmd]))
        with tempfile.TemporaryFile() as f:
            subprocess.call(cmd, stdout=f, stderr=f)
            f.seek(0)
            output = f.read()
            if output:
                self._logger.debug("Output was: '%s'", output)
        self.play(fname)
        os.remove(fname)


class FestivalTTS(AbstractTTSEngine):
    """
    Uses the festival speech synthesizer
    Requires festival (text2wave) to be available
    """

    SLUG = 'festival-tts'

    @classmethod
    def is_available(cls):
        if (super(cls, cls).is_available() and
           diagnose.check_executable('text2wave') and
           diagnose.check_executable('festival')):

            logger = logging.getLogger(__name__)
            cmd = ['festival', '--pipe']
            with tempfile.SpooledTemporaryFile() as out_f:
                with tempfile.SpooledTemporaryFile() as in_f:
                    logger.debug('Executing %s', ' '.join([pipes.quote(arg)
                                                           for arg in cmd]))
                    subprocess.call(cmd, stdin=in_f, stdout=out_f,
                                    stderr=out_f)
                    out_f.seek(0)
                    output = out_f.read().strip()
                    if output:
                        logger.debug("Output was: '%s'", output)
                    return ('No default voice found' not in output)
        return False

    def say(self, phrase):
        self._logger.debug("Saying '%s' with '%s'", phrase, self.SLUG)
        cmd = ['text2wave']
        with tempfile.NamedTemporaryFile(suffix='.wav') as out_f:
            with tempfile.SpooledTemporaryFile() as in_f:
                in_f.write(phrase)
                in_f.seek(0)
                with tempfile.SpooledTemporaryFile() as err_f:
                    self._logger.debug('Executing %s',
                                       ' '.join([pipes.quote(arg)
                                                 for arg in cmd]))
                    subprocess.call(cmd, stdin=in_f, stdout=out_f,
                                    stderr=err_f)
                    err_f.seek(0)
                    output = err_f.read()
                    if output:
                        self._logger.debug("Output was: '%s'", output)
            self.play(out_f.name)


class MacOSXTTS(AbstractTTSEngine):
    """
    Uses the OS X built-in 'say' command
    """

    SLUG = "osx-tts"

    @classmethod
    def is_available(cls):
        return (platform.system().lower() == 'darwin' and
                diagnose.check_executable('say') and
                diagnose.check_executable('afplay'))

    def say(self, phrase):
        self._logger.debug("Saying '%s' with '%s'", phrase, self.SLUG)
        cmd = ['say', str(phrase)]
        self._logger.debug('Executing %s', ' '.join([pipes.quote(arg)
                                                     for arg in cmd]))
        with tempfile.TemporaryFile() as f:
            subprocess.call(cmd, stdout=f, stderr=f)
            f.seek(0)
            output = f.read()
            if output:
                self._logger.debug("Output was: '%s'", output)

    def play(self, filename):
        cmd = ['afplay', str(filename)]
        self._logger.debug('Executing %s', ' '.join([pipes.quote(arg)
                                                     for arg in cmd]))
        with tempfile.TemporaryFile() as f:
            subprocess.call(cmd, stdout=f, stderr=f)
            f.seek(0)
            output = f.read()
            if output:
                self._logger.debug("Output was: '%s'", output)


class PicoTTS(AbstractTTSEngine):
    """
    Uses the svox-pico-tts speech synthesizer
    Requires pico2wave to be available
    """

    SLUG = "pico-tts"

    def __init__(self, language="en-US"):
        super(self.__class__, self).__init__()
        self.language = language

    @classmethod
    def is_available(cls):
        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',
                            '-w', os.devnull,
                            'NULL']
        with tempfile.SpooledTemporaryFile() as f:
            subprocess.call(cmd, stderr=f)
            f.seek(0)
            output = f.read()
        pattern = re.compile(r'Unknown language: NULL\nValid languages:\n' +
                             r'((?:[a-z]{2}-[A-Z]{2}\n)+)')
        matchobj = pattern.match(output)
        if not matchobj:
            raise RuntimeError("pico2wave: valid languages not detected")
        langs = matchobj.group(1).split()
        return langs

    def say(self, phrase):
        self._logger.debug("Saying '%s' with '%s'", phrase, self.SLUG)
        with tempfile.NamedTemporaryFile(suffix='.wav', delete=False) as f:
            fname = f.name
        cmd = ['pico2wave', '--wave', fname]
        if self.language not in self.languages:
                raise ValueError("Language '%s' not supported by '%s'",
                                 self.language, self.SLUG)
        cmd.extend(['-l', self.language])
        cmd.append(phrase)
        self._logger.debug('Executing %s', ' '.join([pipes.quote(arg)
                                                     for arg in cmd]))
        with tempfile.TemporaryFile() as f:
            subprocess.call(cmd, stdout=f, stderr=f)
            f.seek(0)
            output = f.read()
            if output:
                self._logger.debug("Output was: '%s'", output)
        self.play(fname)
        os.remove(fname)


class GoogleTTS(AbstractMp3TTSEngine):
    """
    Uses the Google TTS online translator
    Requires pymad and gTTS to be available
    """

    SLUG = "google-tts"

    def __init__(self, language='en'):
        super(self.__class__, self).__init__()
        self.language = language

    @classmethod
    def is_available(cls):
        return (super(cls, cls).is_available() and
                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',
                 'da', 'nl', 'en', 'eo', 'fi', 'fr', 'de', 'el', 'ht', 'hi',
                 'hu', 'is', 'id', 'it', 'ja', 'ko', 'la', 'lv', 'mk', 'no',
                 'pl', 'pt', 'ro', 'ru', 'sr', 'sk', 'es', 'sw', 'sv', 'ta',
                 'th', 'tr', 'vi', 'cy']
        return langs

    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)
        tts = gtts.gTTS(text=phrase, lang=self.language)
        with tempfile.NamedTemporaryFile(suffix='.mp3', delete=False) as f:
            tmpfile = f.name
        tts.save(tmpfile)
        self.play_mp3(tmpfile)
        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):
        try:
            r = self.session.get(self._makeurl('/locales'))
            r.raise_for_status()
        except requests.exceptions.RequestException:
            self._logger.critical("Communication with MaryTTS server at %s " +
                                  "failed.", self.netloc)
            raise
        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().lower() == 'darwin' else 'espeak-tts'


def get_engine_by_slug(slug=None):
    """
    Returns:
        A speaker implementation available on the current platform

    Raises:
        ValueError if no speaker implementation is supported on this platform
    """

    if not slug or type(slug) is not str:
        raise TypeError("Invalid slug '%s'", slug)

    selected_engines = filter(lambda engine: hasattr(engine, "SLUG") and
                              engine.SLUG == slug, get_engines())
    if len(selected_engines) == 0:
        raise ValueError("No TTS engine found for slug '%s'" % slug)
    else:
        if len(selected_engines) > 1:
            print("WARNING: Multiple TTS engines found for slug '%s'. " +
                  "This is most certainly a bug." % slug)
        engine = selected_engines[0]
        if not engine.is_available():
            raise ValueError(("TTS engine '%s' is not available (due to " +
                              "missing dependencies, etc.)") % slug)
        return engine


def get_engines():
    def get_subclasses(cls):
        subclasses = set()
        for subclass in cls.__subclasses__():
            subclasses.add(subclass)
            subclasses.update(get_subclasses(subclass))
        return subclasses
    return [tts_engine for tts_engine in
            list(get_subclasses(AbstractTTSEngine))
            if hasattr(tts_engine, 'SLUG') and tts_engine.SLUG]

if __name__ == '__main__':
    parser = argparse.ArgumentParser(description='Jasper TTS module')
    parser.add_argument('--debug', action='store_true',
                        help='Show debug messages')
    args = parser.parse_args()

    logging.basicConfig()
    if args.debug:
        logger = logging.getLogger(__name__)
        logger.setLevel(logging.DEBUG)

    engines = get_engines()
    available_engines = []
    for engine in get_engines():
        if engine.is_available():
            available_engines.append(engine)
    disabled_engines = list(set(engines).difference(set(available_engines)))
    print("Available TTS engines:")
    for i, engine in enumerate(available_engines, start=1):
        print("%d. %s" % (i, engine.SLUG))

    print("")
    print("Disabled TTS engines:")

    for i, engine in enumerate(disabled_engines, start=1):
        print("%d. %s" % (i, engine.SLUG))

    print("")
    for i, engine in enumerate(available_engines, start=1):
        print("%d. Testing engine '%s'..." % (i, engine.SLUG))
        engine.get_instance().say("This is a test.")
    print("Done.")