def test(self):
url = "%s/list" % self.base_url
log.debug("LEXICON URL: %s" % url)
#print("LEXICON URL: %s" % url)
try:
r = requests.get(url)
response = r.text
response_json = json.loads(response)
#print(response_json)
exists = False
#print(response_json)
for lex in response_json:
if lex['name'] == self.lexicon_name:
exists = True
if not exists:
msg = "Lexicon does not exist: %s" % (self.lexicon_name)
log.error(msg)
raise LexiconException(msg)
except json.JSONDecodeError:
msg = "Unable to create lexicon client for %s. Response was: %s" % (
self.lexicon_name, response)
log.error(msg)
raise LexiconException(msg)
except Exception as e:
msg = "Unable to create lexicon client for %s at url %s. Reason: %s" % (
self.lexicon_name, url, e)
log.warning(msg)
raise LexiconException(msg)
def loadVoice(voice_config):
try:
v = Voice(voice_config)
voices.append(v)
except VoiceException as e:
log.error("Failed to load voice from %s. Reason:\n%s" %
(voice_config, e))
def mary2ws(maryxml, voice):
(lang, maryxml) = dropHeader(maryxml)
#lang = "sv"
#log.debug(maryxml)
root = ET.fromstring(maryxml.encode('utf-8'))
#root = ET.fromstring(maryxml)
paragraphs = []
utterance = {
"lang": lang,
"paragraphs": paragraphs
}
paragraph_elements = root.findall(".//p")
for paragraph_element in paragraph_elements:
sentences = []
paragraphs.append({"sentences": sentences})
sentence_elements = paragraph_element.findall("s")
for sentence_element in sentence_elements:
phrases = []
sentence = {"phrases": phrases}
sentences.append(sentence)
for sentence_child in sentence_element:
if sentence_child.tag == "phrase":
phrase_element = sentence_child
phrase = buildPhrase(phrase_element, lang, voice)
phrases.append(phrase)
elif sentence_child.tag == "prosody":
prosody_element = sentence_child
#can 'prosody' only contain exactly one 'phrase'?
phrase_element = prosody_element[0]
phrase = buildPhrase(phrase_element, lang, voice)
phrase = addIfExists(phrase, prosody_element, "pitch", prefix="prosody_")
phrase = addIfExists(phrase, prosody_element, "range", prefix="prosody_")
phrases.append(phrase)
elif sentence_child.tag == "t":
#This is a special case that happens (sometimes..) with single-word sentences
#doesn't work, produces error in synthesis
#TODO look at this again
#It only happens with the word "Hon." ...
phrase_element = ET.Element("phrase")
phrase_element.append(sentence_child)
phrase = buildPhrase(phrase_element, lang, voice)
phrases.append(phrase)
else:
log.error("sentence child should not have tag %s" % sentence_child.tag)
return utterance
def loadTextprocessor(tp_config):
try:
tp = Textprocessor(tp_config)
textprocessors.append(tp)
except TextprocessorException as e:
log.error("Failed to load textprocessor from %s. Reason:\n%s" %
(tp_config, e))
def test(self):
url = "%s/list" % self.base_url
log.debug("LEXICON URL: %s" % url)
print("LEXICON URL: %s" % url)
try:
r = requests.get(url)
response = r.text
response_json = json.loads(response)
print(response_json)
exists = False
print(response_json)
for lex in response_json:
if lex['name'] == self.lexicon_name:
exists = True
if not exists:
msg = "Lexicon does not exist: %s" % (self.lexicon_name)
log.error(msg)
raise LexiconException(msg)
except json.JSONDecodeError:
msg = "Unable to create lexicon client for %s. Response was: %s" % (self.lexicon_name, response)
log.error(msg)
raise LexiconException(msg)
except Exception as e:
msg = "Unable to create lexicon client for %s at url %s. Reason: %s" % (self.lexicon_name, url, e)
log.warning(msg)
raise LexiconException(msg)
def test_config():
log.debug("\nTEST CONFIG\n")
log.debug("Testing that audio_tmpdir exists and is writeable")
try:
tmpdir = config.config.get("Audio settings","audio_tmpdir")
log.debug("TMPDIR: %s" % tmpdir)
fh = NamedTemporaryFile(mode='w+b', dir=tmpdir, delete=False)
tmpfile = fh.name
fh.write("test".encode("utf-8"))
fh.close()
except:
log.error("audio_tmpdir does not exist or is not writeable")
raise
log.debug("Testing to make sure that config file contains url to lexicon server:")
try:
assert ( config.config.has_option("Services", "lexicon") == True )
log.debug("Services|lexicon = %s" % config.config.get("Services", "lexicon"))
log.debug("ok")
except:
log.error("Services|lexicon not found in config file\n")
raise
log.debug("\nEND TEST CONFIG\n")
def test_config():
log.debug("\nTEST CONFIG\n")
log.debug("Testing that audio_tmpdir exists and is writeable")
try:
tmpdir = config.config.get("Audio settings", "audio_tmpdir")
log.debug("TMPDIR: %s" % tmpdir)
fh = NamedTemporaryFile(mode='w+b', dir=tmpdir, delete=False)
tmpfile = fh.name
fh.write("test".encode("utf-8"))
fh.close()
except:
log.error("audio_tmpdir does not exist or is not writeable")
raise
log.debug(
"Testing to make sure that config file contains url to lexicon server:"
)
try:
assert (config.config.has_option("Services", "lexicon") == True)
log.debug("Services|lexicon = %s" %
config.config.get("Services", "lexicon"))
log.debug("ok")
except:
log.error("Services|lexicon not found in config file\n")
raise
log.debug("\nEND TEST CONFIG\n")
def mapperMapFromMary(trans, lang, voice):
log.info("mapperMapFromMary( %s , %s , %s )" % (trans, lang, voice))
if "mapper" in voice:
#Bad names.. It should be perhaps "external" and "internal" instead of "from" and "to"
to_symbol_set = voice["mapper"]["from"]
from_symbol_set = voice["mapper"]["to"]
else:
log.info("No marytts mapper defined for language %s" % lang)
return trans
## hl remove quote_plus 20170613
## hb "" b j A: rn . % 9 j doesn't work, testing to put quote back in 20170626
##url = mapper_url+"/mapper/map/%s/%s/%s" % (from_symbol_set, to_symbol_set, quote_plus(trans))
url = mapper_url+"/mapper/map/%s/%s/%s" % (from_symbol_set, to_symbol_set, quote(trans))
##url = mapper_url+"/mapper/map/%s/%s/%s" % (from_symbol_set, to_symbol_set, trans)
log.debug("MAPPER URL before requests: %s" % url)
r = requests.get(url)
log.debug("MAPPER URL: "+r.url)
response = r.text
#log.debug("RESPONSE: %s" % response)
try:
response_json = json.loads(response)
#log.debug("RESPONSE_JSON: %s" % response_json)
new_trans = response_json["Result"]
except:
log.error("unable to map %s, from %s to %s. response was %s" % (trans, from_symbol_set, to_symbol_set, response))
raise
#log.debug("NEW TRANS: %s" % new_trans)
return new_trans
def buildPhrase(phrase_element, lang, voice):
if phrase_element.tag != "phrase":
log.error("wrong type of element: %s\n%s" % (phrase_element.tag, phrase_element))
sys.exit(1)
tokens = []
phrase = {"tokens": tokens}
for phrase_child in phrase_element:
#log.debug("phrase_child: %s" % phrase_child.tag)
#no 'mtu'
if phrase_child.tag == "t":
token_element = phrase_child
words = []
token = {
"words": words
}
tokens.append(token)
#HB 171109
#This is the case with arabic words vocalised with mishkal
if "orig" in token_element.attrib:
orth = token_element.attrib["orig"]
else:
orth = token_element.text
token["token_orth"] = orth
word = buildWord(token_element, lang, voice)
words.append(word)
#END no 'mtu'
#'mtu'
elif phrase_child.tag == "mtu":
mtu_element = phrase_child
words = []
token = {
"mtu": True,
"words": words
}
tokens.append(token)
token_orth = mtu_element.attrib["orig"]
token["token_orth"] = token_orth
token_elements = mtu_element.findall("t")
for token_element in token_elements:
word = buildWord(token_element, lang, voice)
words.append(word)
#END 'mtu'
elif phrase_child.tag == "boundary":
boundary = {}
boundary = addIfExists(boundary, phrase_child, "breakindex")
boundary = addIfExists(boundary, phrase_child, "tone")
phrase["boundary"] = boundary
else:
log.warn("phrase child should not have tag %s" % phrase_child.tag)
return phrase
def saveAndConvertAudio(audio_url):
global config
tmpdir = config.config.get("Audio settings","audio_tmpdir")
log.debug("TMPDIR: %s" % tmpdir)
if not os.path.isdir(tmpdir):
os.system("mkdir -p %s" % tmpdir)
fh = NamedTemporaryFile(mode='w+b', dir=tmpdir, delete=False)
tmpwav = fh.name
log.debug("audio_url:\n%s" % audio_url)
r = requests.get(audio_url)
log.debug(r.headers['content-type'])
audio_data = r.content
fh = NamedTemporaryFile(mode='w+b', dir=tmpdir, delete=False)
tmpwav = fh.name
fh.write(audio_data)
fh.close()
#tmpwav is now the synthesised wav file
#tmpopus = "%s/%s.opus" % (tmpdir, tmpfilename)
tmpopus = "%s.opus" % tmpwav
convertcmd = "opusenc %s %s" % (tmpwav, tmpopus)
log.debug("convertcmd: %s" % convertcmd)
if log.log_level != "debug":
convertcmd = "opusenc --quiet %s %s" % (tmpwav, tmpopus)
retval = os.system(convertcmd)
if retval != 0:
log.error("ERROR: opusenc was not found. You should probably run something like\nsudo apt install opus-tools\n")
#remove everything before the tmpdir, to build the external url
#HB problem with wikimedia usage?
#opus_url_suffix = re.sub("^.*/%s/" % tmpdir, "%s/" % tmpdir, tmpopus)
opus_url_suffix = re.sub("^.*/%s/" % tmpdir, "", tmpopus)
log.debug("opus_url_suffix: %s" % opus_url_suffix)
#return tmpopus
return_audio_data = True
if return_audio_data:
#audio_data = "data:audio/wav;base64,%s" % encode_audio(re.sub("^.*/", "wikispeech_server/tmp/", audio_file)).decode()
audio_data = "%s" % encode_audio(tmpwav).decode()
else:
audio_data = ""
return (opus_url_suffix, audio_data)
def saveAndConvertAudio(audio_url, presynth=False):
global config
log.debug("PRESYNTH: %s, type: %s" % (presynth, type(presynth)))
tmpdir = config.config.get("Audio settings", "audio_tmpdir")
log.debug("TMPDIR: %s" % tmpdir)
fh = NamedTemporaryFile(mode='w+b', dir=tmpdir, delete=False)
tmpwav = fh.name
if presynth:
fh.close()
#The "url" is actually a filename at this point
cmd = "mv %s %s" % (audio_url, tmpwav)
log.debug(cmd)
os.system(cmd)
else:
log.debug("audio_url:\n%s" % audio_url)
r = requests.get(audio_url)
log.debug(r.headers['content-type'])
audio_data = r.content
fh = NamedTemporaryFile(mode='w+b', dir=tmpdir, delete=False)
tmpwav = fh.name
fh.write(audio_data)
fh.close()
#tmpwav is now the synthesised wav file
#tmpopus = "%s/%s.opus" % (tmpdir, tmpfilename)
tmpopus = "%s.opus" % tmpwav
convertcmd = "opusenc %s %s" % (tmpwav, tmpopus)
log.debug("convertcmd: %s" % convertcmd)
if log.log_level != "debug":
convertcmd = convertcmd # +" &> /dev/null"
retval = os.system(convertcmd)
if retval != 0:
log.error(
"ERROR: opusenc was not found. You should probably run something like\nsudo apt install opus-tools\n"
)
#remove everything before the tmpdir, to build the external url
#HB problem with wikimedia usage?
#opus_url_suffix = re.sub("^.*/%s/" % tmpdir, "%s/" % tmpdir, tmpopus)
opus_url_suffix = re.sub("^.*/%s/" % tmpdir, "", tmpopus)
log.debug("opus_url_suffix: %s" % opus_url_suffix)
#return tmpopus
return opus_url_suffix
def buildPhrase(phrase_element, lang, voice):
if phrase_element.tag != "phrase":
log.error("wrong type of element: %s\n%s" %
(phrase_element.tag, phrase_element))
sys.exit(1)
tokens = []
phrase = {"tokens": tokens}
for phrase_child in phrase_element:
#log.debug("phrase_child: %s" % phrase_child.tag)
#no 'mtu'
if phrase_child.tag == "t":
token_element = phrase_child
words = []
token = {"words": words}
tokens.append(token)
#HB 171109
#This is the case with arabic words vocalised with mishkal
if "orig" in token_element.attrib:
orth = token_element.attrib["orig"]
else:
orth = token_element.text
token["token_orth"] = orth
word = buildWord(token_element, lang, voice)
words.append(word)
#END no 'mtu'
#'mtu'
elif phrase_child.tag == "mtu":
mtu_element = phrase_child
words = []
token = {"mtu": True, "words": words}
tokens.append(token)
token_orth = mtu_element.attrib["orig"]
token["token_orth"] = token_orth
token_elements = mtu_element.findall("t")
for token_element in token_elements:
word = buildWord(token_element, lang, voice)
words.append(word)
#END 'mtu'
elif phrase_child.tag == "boundary":
boundary = {}
boundary = addIfExists(boundary, phrase_child, "breakindex")
boundary = addIfExists(boundary, phrase_child, "tone")
phrase["boundary"] = boundary
else:
log.warn("phrase child should not have tag %s" % phrase_child.tag)
return phrase
def saveAndConvertAudio(audio_url,presynth=False):
global config
log.debug("PRESYNTH: %s, type: %s" % (presynth, type(presynth)) )
tmpdir = config.config.get("Audio settings","audio_tmpdir")
log.debug("TMPDIR: %s" % tmpdir)
fh = NamedTemporaryFile(mode='w+b', dir=tmpdir, delete=False)
tmpwav = fh.name
if presynth:
fh.close()
#The "url" is actually a filename at this point
cmd = "mv %s %s" % (audio_url, tmpwav)
log.debug(cmd)
os.system(cmd)
else:
log.debug("audio_url:\n%s" % audio_url)
r = requests.get(audio_url)
log.debug(r.headers['content-type'])
audio_data = r.content
fh = NamedTemporaryFile(mode='w+b', dir=tmpdir, delete=False)
tmpwav = fh.name
fh.write(audio_data)
fh.close()
#tmpwav is now the synthesised wav file
#tmpopus = "%s/%s.opus" % (tmpdir, tmpfilename)
tmpopus = "%s.opus" % tmpwav
convertcmd = "opusenc %s %s" % (tmpwav, tmpopus)
log.debug("convertcmd: %s" % convertcmd)
if log.log_level != "debug":
convertcmd = "opusenc --quiet %s %s" % (tmpwav, tmpopus)
retval = os.system(convertcmd)
if retval != 0:
log.error("ERROR: opusenc was not found. You should probably run something like\nsudo apt install opus-tools\n")
#remove everything before the tmpdir, to build the external url
#HB problem with wikimedia usage?
#opus_url_suffix = re.sub("^.*/%s/" % tmpdir, "%s/" % tmpdir, tmpopus)
opus_url_suffix = re.sub("^.*/%s/" % tmpdir, "", tmpopus)
log.debug("opus_url_suffix: %s" % opus_url_suffix)
#return tmpopus
return opus_url_suffix
def testVoice(config):
voice = config["espeak_voice"]
espeak = "espeak-ng"
#espeak = "espeak"
try:
retval = os.system("%s -v %s -q test" % (espeak, voice))
assert retval == 0
log.debug("Test successful for voice %s" % config["name"])
return True
except:
msg = "Failed command: '%s -v %s'" % (espeak, voice)
log.error(msg)
raise VoiceException(msg)
def map(self, string):
url = "%s/%s/%s/%s/%s" % (self.base_url, "map", self.from_symbol_set, self.to_symbol_set, string)
r = requests.get(url)
log.debug(r.url)
response = r.text
try:
response_json = json.loads(response)
new_string = response_json["Result"]
return new_string
except:
log.error("unable to map string '%s'from %s to %s. response was %s" % (string,self.from_symbol_set, self.to_symbol_set, response))
raise MapperException
def saveAndConvertAudio(audio_url):
global config
tmpdir = config.config.get("Audio settings", "audio_tmpdir")
log.debug("TMPDIR: %s" % tmpdir)
if not os.path.isdir(tmpdir):
os.system("mkdir -p %s" % tmpdir)
fh = NamedTemporaryFile(mode='w+b', dir=tmpdir, delete=False)
tmpwav = fh.name
log.debug("audio_url:\n%s" % audio_url)
r = requests.get(audio_url)
log.debug(r.headers['content-type'])
audio_data = r.content
#fh = NamedTemporaryFile(mode='w+b', dir=tmpdir, delete=False)
#tmpwav = fh.name
fh.write(audio_data)
fh.close()
tmpopus = "%s.opus" % tmpwav
convertcmd = "opusenc %s %s" % (tmpwav, tmpopus)
log.debug("convertcmd: %s" % convertcmd)
if log.log_level != "debug":
convertcmd = "opusenc --quiet %s %s" % (tmpwav, tmpopus)
retval = os.system(convertcmd)
if retval != 0:
log.error(
"ERROR: opusenc was not found. You should probably run something like\nsudo apt install opus-tools\n"
)
opus_url_suffix = re.sub("^.*/%s/" % tmpdir, "", tmpopus)
log.debug("opus_url_suffix: %s" % opus_url_suffix)
return_audio_data = True
if return_audio_data:
audio_data = "%s" % encode_audio(tmpopus).decode()
else:
audio_data = ""
#Removing any remaining files in the tmpdir
tmpfiles = glob.glob("%s/*" % tmpdir)
for f in tmpfiles:
os.unlink(f)
return (opus_url_suffix, audio_data)
def buildWord(token_element, lang, voice):
if token_element.tag != "t":
log.error("wrong type of element: %s\n%s" % (token_element.tag, token_element))
sys.exit(1)
orth = token_element.text
word = {
"orth": orth
}
word = addIfExists(word, token_element, "accent")
#g2p_method can be rules, lexicon, or not there if there is sampa in input
#change to a tag for input_ssml_transcription that can be used by the lexicon_client
if token_element.attrib["pos"] not in ["$PUNCT","."] and "g2p_method" not in token_element.attrib:
word["input_ssml_transcription"] = True
word = addIfExists(word, token_element, "pos")
#word = addIfExists(word, token_element, "ph")
if "ph" in token_element.attrib:
mary_trans = token_element.attrib.get("ph")
ws_trans = mapperMapFromMary(mary_trans, lang, voice)
word["trans"] = ws_trans
#In the new version with output_type INTONATION from marytts the following is no longer used
syllable_elements = token_element.findall("syllable")
if len(syllable_elements) > 0:
syllables = []
word["syllables"] = syllables
for syllable_element in syllable_elements:
phonemes = []
syllable = {
"phonemes": phonemes
}
syllable = addIfExists(syllable, syllable_element, "accent")
syllable = addIfExists(syllable, syllable_element, "ph")
syllable = addIfExists(syllable, syllable_element, "stress")
syllables.append(syllable)
phoneme_elements = syllable_element.findall("ph")
for phoneme_element in phoneme_elements:
symbol = phoneme_element.attrib["p"]
phonemes.append({"symbol": symbol})
return word
def test(self):
url = "%s/%s/%s/%s" % (self.base_url, "maptable", self.from_symbol_set, self.to_symbol_set)
log.debug(url)
try:
r = requests.get(url)
response = r.text
response_json = json.loads(response)
except json.JSONDecodeError:
msg = "Unable to create mapper from %s to %s. Response was: %s" % (self.from_symbol_set, self.to_symbol_set, response)
log.error(msg)
raise MapperException(msg)
except Exception as e:
msg = "Unable to create mapper at url %s. Reason: %s" % (url, e)
log.error(msg)
raise MapperException(msg)
def map(self, string):
url = "%s/%s/%s/%s/%s" % (self.base_url, "map", self.from_symbol_set,
self.to_symbol_set, string)
r = requests.get(url)
log.debug(r.url)
response = r.text
try:
response_json = json.loads(response)
new_string = response_json["Result"]
return new_string
except:
log.error(
"unable to map string '%s'from %s to %s. response was %s" %
(string, self.from_symbol_set, self.to_symbol_set, response))
raise MapperException
def mapperMapToMary(trans, lang, voice):
log.debug("mapperMapToMary( %s, %s, %s)" % (trans, lang, voice))
if "mapper" in voice:
to_symbol_set = voice["mapper"]["to"]
from_symbol_set = voice["mapper"]["from"]
log.info("marytts mapper defined for language %s\nFrom: %s\nTo: %s" % (lang, from_symbol_set, to_symbol_set))
else:
log.info("No marytts mapper defined for language %s" % lang)
return trans
## hl remove quote_plus 20170613
## hb "" b j A: rn . % 9 j doesn't work, using quote 20170626
## url = mapper_url+"/mapper/map/%s/%s/%s" % (from_symbol_set, to_symbol_set, quote_plus(trans))
url = mapper_url+"/mapper/map/%s/%s/%s" % (from_symbol_set, to_symbol_set, quote(trans))
##url = mapper_url+"/mapper/map/%s/%s/%s" % (from_symbol_set, to_symbol_set, trans)
log.debug("MAPPER URL before requests: %s" % url)
r = requests.get(url)
log.debug("MAPPER URL: %s" % r.url)
response = r.text
log.debug("MAPPER RESPONSE: %s" % response)
try:
response_json = json.loads(response)
except json.JSONDecodeError:
log.error("JSONDecodeError:")
log.error("RESPONSE: %s" % response)
raise
new_trans = response_json["Result"]
#Special cases for Swedish pre-r allophones that are not handled by the mapper (because mary uses an old version of the phoneme set that desn't distinguish between normal and r-coloured E/{ (always E) and 2/9 (always 9). This should change in mary later on.
if lang == "sv":
new_trans = re.sub("{ - ",r"E - ", new_trans)
new_trans = re.sub("{ ",r"E ", new_trans)
new_trans = re.sub("2(:? -) r? ",r"9\1 r", new_trans)
log.debug("NEW TRANS: %s" % new_trans)
return new_trans
def mapperMapToMary(trans, lang, voice):
log.debug("mapperMapToMary( %s, %s, %s)" % (trans, lang, voice))
if "mapper" in voice:
to_symbol_set = voice["mapper"]["to"]
from_symbol_set = voice["mapper"]["from"]
log.debug("marytts mapper defined for language %s\nFrom: %s\nTo: %s" %
(lang, from_symbol_set, to_symbol_set))
else:
log.debug("No marytts mapper defined for language %s" % lang)
return trans
## hl remove quote_plus 20170613
## hb "" b j A: rn . % 9 j doesn't work, using quote 20170626
## url = mapper_url+"/mapper/map/%s/%s/%s" % (from_symbol_set, to_symbol_set, quote_plus(trans))
url = mapper_url + "/mapper/map/%s/%s/%s" % (from_symbol_set,
to_symbol_set, quote(trans))
##url = mapper_url+"/mapper/map/%s/%s/%s" % (from_symbol_set, to_symbol_set, trans)
log.debug("MAPPER URL before requests: %s" % url)
r = requests.get(url)
log.debug("MAPPER URL: %s" % r.url)
response = r.text
log.debug("MAPPER RESPONSE: %s" % response)
try:
response_json = json.loads(response)
except json.JSONDecodeError:
log.error("JSONDecodeError:")
log.error("RESPONSE: %s" % response)
raise
new_trans = response_json["Result"]
#Special cases for Swedish pre-r allophones that are not handled by the mapper (because mary uses an old version of the phoneme set that desn't distinguish between normal and r-coloured E/{ (always E) and 2/9 (always 9). This should change in mary later on.
if lang == "sv":
new_trans = re.sub("{ - ", r"E - ", new_trans)
new_trans = re.sub("{ ", r"E ", new_trans)
new_trans = re.sub("2(:? -) r? ", r"9\1 r", new_trans)
log.debug("NEW TRANS: %s" % new_trans)
return new_trans
def test(self):
url = "%s/%s/%s/%s" % (self.base_url, "maptable", self.from_symbol_set,
self.to_symbol_set)
log.debug(url)
try:
r = requests.get(url)
response = r.text
response_json = json.loads(response)
except json.JSONDecodeError:
msg = "Unable to create mapper from %s to %s. Response was: %s" % (
self.from_symbol_set, self.to_symbol_set, response)
log.error(msg)
raise MapperException(msg)
except Exception as e:
msg = "Unable to create mapper at url %s. Reason: %s" % (url, e)
log.error(msg)
raise MapperException(msg)
def buildWord(token_element, lang, voice):
if token_element.tag != "t":
log.error("wrong type of element: %s\n%s" %
(token_element.tag, token_element))
sys.exit(1)
orth = token_element.text
word = {"orth": orth}
word = addIfExists(word, token_element, "accent")
#g2p_method can be rules, lexicon, or not there if there is sampa in input
#change to a tag for input_ssml_transcription that can be used by the lexicon_client
if token_element.attrib["pos"] not in [
"$PUNCT", "."
] and "g2p_method" not in token_element.attrib:
word["input_ssml_transcription"] = True
word = addIfExists(word, token_element, "pos")
#word = addIfExists(word, token_element, "ph")
if "ph" in token_element.attrib:
mary_trans = token_element.attrib.get("ph")
ws_trans = mapperMapFromMary(mary_trans, lang, voice)
word["trans"] = ws_trans
#In the new version with output_type INTONATION from marytts the following is no longer used
syllable_elements = token_element.findall("syllable")
if len(syllable_elements) > 0:
syllables = []
word["syllables"] = syllables
for syllable_element in syllable_elements:
phonemes = []
syllable = {"phonemes": phonemes}
syllable = addIfExists(syllable, syllable_element, "accent")
syllable = addIfExists(syllable, syllable_element, "ph")
syllable = addIfExists(syllable, syllable_element, "stress")
syllables.append(syllable)
phoneme_elements = syllable_element.findall("ph")
for phoneme_element in phoneme_elements:
symbol = phoneme_element.attrib["p"]
phonemes.append({"symbol": symbol})
return word
def lookup(self, string):
if string.strip() == "":
log.warning("LEXICON LOOKUP STRING IS EMPTY!")
return {}
encString = urllib.parse.quote(string)
url = "%s/%s?lexicons=%s&words=%s" % (self.base_url, "lookup", self.lexicon_name, encString)
r = requests.get(url)
log.debug("LEXICON LOOKUP URL: %s" % r.url)
response = r.text
try:
response_json = json.loads(response)
log.debug(response_json)
return response_json
except:
log.error("unable to lookup '%s' in %s. response was %s" % (string, self.lexicon_name, response))
raise LexiconException(response)
def lookup(self, string):
if string.strip() == "":
log.warning("LEXICON LOOKUP STRING IS EMPTY!")
return {}
encString = urllib.parse.quote(string)
url = "%s/%s?lexicons=%s&words=%s" % (self.base_url, "lookup",
self.lexicon_name, encString)
r = requests.get(url)
log.debug("LEXICON LOOKUP URL: %s" % r.url)
response = r.text
try:
response_json = json.loads(response)
log.debug(response_json)
return response_json
except:
log.error("unable to lookup '%s' in %s. response was %s" %
(string, self.lexicon_name, response))
raise LexiconException(response)
def testVoice(voice_config):
#from wikispeech_server.adapters.ahotts_adapter import socket_write_filelength_file,socket_read_filelength_file
#cwdir = os.getcwd()
#tmpdir = config.config.get("Audio settings","audio_tmpdir")
#ahotts_server_ip = config.config.get("Services", "ahotts_server_ip")
#ahotts_server_port = config.config.get("Services", "ahotts_server_port")
#ahotts_speed = config.config.get("Services", "ahotts_speed")
lang = voice_config["lang"]
name = voice_config["name"]
try:
response = requests.post("http://" + ahotts_server_ip + ":" +
ahotts_server_port + "/ahotts_getaudio",
data={
'text':
"Hasierako proba".encode('latin-1'),
'lang': lang,
'voice': name,
'speed': ahotts_speed
})
except:
msg = "Ahotts server not found at %s:%s" % (ahotts_server_ip,
ahotts_server_port)
log.error(msg)
raise VoiceException(msg)
if response.status_code == 200:
files = response.json()
wavfile = files['wav']
wrdfile = files['wrd']
response2 = requests.get("http://" + ahotts_server_ip + ":" +
ahotts_server_port +
"/ahotts_downloadfile?file=" + wavfile)
if response2.status_code == 200:
response3 = requests.get("http://" + ahotts_server_ip + ":" +
ahotts_server_port +
"/ahotts_downloadfile?file=" + wrdfile)
if response3.status_code != 200:
msg = "AhoTTS server error"
log.error(msg)
raise VoiceException(msg)
else:
msg = "AhoTTS server error"
log.error(msg)
raise VoiceException(msg)
else:
msg = "AhoTTS server error"
log.error(msg)
raise VoiceException(msg)
def testVoice(self):
log.info("Testing voice %s" % self.name)
if self.engine == "marytts":
voice_host = config.config.get("Services", "marytts")
url = re.sub("process","voices",voice_host)
log.debug("Calling url: %s" % url)
try:
r = requests.get(url)
except:
msg = "Marytts server not found at url %s" % (url)
log.error(msg)
raise VoiceException(msg)
response = r.text
log.debug("Response:\n%s" % response)
marytts_voicenames = self.getMaryttsVoicenames(response)
if not self.name in marytts_voicenames:
msg = "Voice %s not found at url %s" % (self.name, url)
log.error(msg)
raise VoiceException(msg)
else:
log.info("Voice found at url %s" % url)
def testVoice(voice_config):
voice_host = config.config.get("Services", "marytts")
url = re.sub("process", "voices", voice_host)
name = voice_config["name"]
log.debug("Calling url: %s" % url)
try:
r = requests.get(url)
except:
msg = "Marytts server not found at url %s" % (url)
log.error(msg)
raise VoiceException(msg)
response = r.text
log.debug("Response:\n%s" % response)
marytts_voicenames = getMaryttsVoicenames(response)
log.debug("marytts_voicenames: %s" % marytts_voicenames)
if not name in marytts_voicenames:
msg = "Voice %s not found at url %s" % (name, url)
log.error(msg)
raise VoiceException(msg)
else:
log.info("Voice found at url %s" % url)
def mapperMapFromMary(trans, lang, voice):
log.debug("mapperMapFromMary( %s , %s , %s )" % (trans, lang, voice))
if "mapper" in voice:
#Bad names.. It should be perhaps "external" and "internal" instead of "from" and "to"
to_symbol_set = voice["mapper"]["from"]
from_symbol_set = voice["mapper"]["to"]
else:
log.debug("No marytts mapper defined for language %s" % lang)
return trans
## hl remove quote_plus 20170613
## hb "" b j A: rn . % 9 j doesn't work, testing to put quote back in 20170626
##url = mapper_url+"/mapper/map/%s/%s/%s" % (from_symbol_set, to_symbol_set, quote_plus(trans))
url = mapper_url + "/mapper/map/%s/%s/%s" % (from_symbol_set,
to_symbol_set, quote(trans))
##url = mapper_url+"/mapper/map/%s/%s/%s" % (from_symbol_set, to_symbol_set, trans)
log.debug("MAPPER URL before requests: %s" % url)
r = requests.get(url)
log.debug("MAPPER URL: " + r.url)
response = r.text
#log.debug("RESPONSE: %s" % response)
try:
response_json = json.loads(response)
#log.debug("RESPONSE_JSON: %s" % response_json)
new_trans = response_json["Result"]
except:
log.error("unable to map %s, from %s to %s. response was %s" %
(trans, from_symbol_set, to_symbol_set, response))
raise
#log.debug("NEW TRANS: %s" % new_trans)
return new_trans
def testVoice(self):
log.info("Testing voice %s" % self.name)
if self.engine == "marytts":
voice_host = config.config.get("Services", "marytts")
url = re.sub("process","voices",voice_host)
log.debug("Calling url: %s" % url)
try:
r = requests.get(url)
except:
msg = "Marytts server not found at url %s" % (url)
log.error(msg)
raise VoiceException(msg)
response = r.text
log.debug("Response:\n%s" % response)
marytts_voicenames = self.getMaryttsVoicenames(response)
if not self.name in marytts_voicenames:
msg = "Voice %s not found at url %s" % (self.name, url)
log.error(msg)
raise VoiceException(msg)
else:
log.info("Voice found at url %s" % url)
elif self.engine == "ahotts":
cwdir = os.getcwd()
tmpdir = config.config.get("Audio settings","audio_tmpdir")
ahotts_dir = config.config.get("Services", "ahotts_dir")
ahotts_server_ip = config.config.get("Services", "ahotts_server_ip")
ahotts_server_port = config.config.get("Services", "ahotts_server_port")
ahotts_command = "cd %s/bin; echo \"Hasierako proba\" > ahotts_test.txt; ./tts_client -IP=%s -Port=%s -InputFile=ahotts_test.txt -OutputFile=ahotts_test.wav ; mv ahotts_test.wav %s/%s/ahotts_test.wav ; rm ahotts_test.txt" % (ahotts_dir, ahotts_server_ip, ahotts_server_port, cwdir, tmpdir)
os.system(ahotts_command)
try:
wavfile=open('%s/%s/ahotts_test.wav'%(cwdir, tmpdir),'r')
wavfile.close()
os.remove('%s/%s/ahotts_test.wav'%(cwdir, tmpdir))
except:
msg = "AhoTTS server not found at IP %s and Port %s" % (ahotts_server_ip,ahotts_server_port)
log.error(msg)
raise VoiceException(msg)
if __name__ == "__main__":
log.log_level = "debug" #debug, info, warning, error
tp_config = {
"name":"wikitextproc_sv",
"lang":"sv",
"components":[
{
"module":"adapters.marytts_adapter",
"call":"marytts_preproc",
"mapper": {
"from":"sv-se_ws-sampa",
"to":"sv-se_sampa_mary"
},
},
{
"module":"adapters.lexicon_client",
"call":"lexLookup",
"lexicon":"wikispeech_testdb:sv"
}
]
}
try:
tp = Textprocessor(tp_config)
log.info("Created textprocessor %s from %s" % (tp, tp_config))
except TextprocessorException as e:
log.error("Failed to create textprocessor for %s\nException message was:\n%s" % (tp_config, e))
def synthesise(lang, voice, utterance, hostname=None):
log.debug("Utterance: %s" % utterance)
input = utterance['original_text']
log.debug("Text: %s" % input)
words = get_orth(utterance)
log.debug("Words: %s" % words)
hashstring = input + '&Lang=' + lang + '&Voice=' + voice['name']
try:
hash_object = hashlib.md5(hashstring.encode('latin-1'))
except:
try:
hash_object = hashlib.md5(hashstring.encode('utf-8'))
except:
hash_object = hashlib.md5(hashstring.encode())
hashnumber = hash_object.hexdigest()
""" Call to tts_client, only works if ahotts is installed in same server as wikispeech-server. Better to make socket calls over the network.
inputfile=open("%s/bin/tts_%s.txt" % (ahotts_dir, hashnumber),"wb")
inputfile.write(input.encode('latin-1')+'\n'.encode('latin-1'))
inputfile.close()
ahotts_command = "cd %s/bin ; ./tts_client -SetDur=y -Speed=%s -IP=%s -Port=%s -InputFile=tts_%s.txt -OutputFile=tts_%s.wav -WordFile=tts_%s.wrd -PhoFile=tts_%s.pho ; mv tts_%s.wav %s/%s/tts_%s.wav ; rm tts_%s.txt" % (ahotts_dir, ahotts_speed, ahotts_server_ip, ahotts_server_port, hashnumber, hashnumber, hashnumber, hashnumber, hashnumber, cwdir, tmpdir, hashnumber, hashnumber)
log.info("Ahotts command: %s" % ahotts_command)
os.system(ahotts_command)
audio_url = "%s%s/%s" % (hostname, "audio",'tts_%s.wav' % hashnumber)
words_times_file=open(ahotts_dir+'/tts_'+hashnumber+'.wrd','r')
words_times=words_times_file.readlines()
words_times_file.close()
os.remove(ahotts_dir+'/tts_'+hashnumber+'.wrd')
os.remove(ahotts_dir+'/tts_'+hashnumber+'.pho')
log.info(str(words))
audio_url = "%s%s/%s" % (hostname, "audio",'tts_%s.wav' % hashnumber)
words_times_file=open(wrdfilename,'r')
words_times=words_times_file.readlines()
words_times_file.close()
os.remove(inputfilename)
os.remove(wrdfilename)
"""
""" Call to socket, does not work properly in docker compose environment
# Write text to file
inputfilename="%s/%s/tts_%s.txt" % (cwdir,tmpdir,hashnumber)
inputfile=open(inputfilename,"wb")
inputfile.write(input.encode('latin-1')+'\n'.encode('latin-1'))
inputfile.close()
# Open socket
socketa=socket.socket(socket.AF_INET,socket.SOCK_STREAM)
ipa=socket.gethostbyname(ahotts_server_ip)
socketa.connect((ipa,int(ahotts_server_port)))
# Write options
options_struct=struct.Struct('4s 4s 4s 1024s 1024s 1024s ?')
options_struct_packed=options_struct.pack(lang.encode('utf-8'),"".encode('utf-8'),ahotts_speed.encode('utf-8'),"".encode('utf-8'),("wav/tts_%s.pho" % hashnumber).encode('utf-8'),("wav/tts_%s.wrd" % hashnumber).encode('utf-8'),True)
totalsent=0
while totalsent<len(options_struct_packed):
sent=socketa.send(options_struct_packed[totalsent:])
if sent==0:
raise RuntimeError("socket connection broken")
totalsent=totalsent+sent
# Write text file length + text file
socket_write_filelength_file(socketa,inputfilename)
# Read wav file
wavfilename="%s/%s/tts_%s.wav" % (cwdir,tmpdir,hashnumber)
wavfile=open(wavfilename,"wb")
wavfile.write(socket_read_filelength_file(socketa))
wavfile.close()
# Read pho file
socket_read_filelength_file(socketa)
# Read wrd file
wrdfilename="%s/%s/tts_%s.wrd" % (cwdir,tmpdir,hashnumber)
wrdfile=open(wrdfilename,"wb")
wrdfile.write(socket_read_filelength_file(socketa))
wrdfile.close()
# Close socket
socketa.close()
audio_url = "%s%s/%s" % (hostname, "audio",'tts_%s.wav' % hashnumber)
words_times_file=open(wrdfilename,'r')
words_times=words_times_file.readlines()
words_times_file.close()
os.remove(inputfilename)
os.remove(wrdfilename)
"""
response = requests.post("http://" + ahotts_server_ip + ":" +
ahotts_server_port + "/ahotts_getaudio",
data={
'text':
input.encode('latin-1') +
'\n'.encode('latin-1'),
'lang':
lang,
'voice':
voice,
'speed':
ahotts_speed
})
url = "http://" + ahotts_server_ip + ":" + ahotts_server_port + "/ahotts_getaudio"
data = {
'text': input.encode('latin-1') + '\n'.encode('latin-1'),
'lang': lang,
'voice': voice,
'speed': ahotts_speed
}
if response.status_code == 200:
files = response.json()
wavfile = files['wav']
wrdfile = files['wrd']
response2 = requests.get("http://" + ahotts_server_ip + ":" +
ahotts_server_port +
"/ahotts_downloadfile?file=" + wavfile)
if response2.status_code == 200:
wavfilename = "%s/%s/tts_%s.wav" % (cwdir, tmpdir, hashnumber)
wavfile = open(wavfilename, "wb")
for chunk in response2.iter_content(1024):
wavfile.write(chunk)
wavfile.close()
response3 = requests.get("http://" + ahotts_server_ip + ":" +
ahotts_server_port +
"/ahotts_downloadfile?file=" + wrdfile)
if response3.status_code == 200:
wrdfilename = "%s/%s/tts_%s.wrd" % (cwdir, tmpdir, hashnumber)
wrdfile = open(wrdfilename, "wb")
for chunk in response3.iter_content(1024):
wrdfile.write(chunk)
wrdfile.close()
else:
msg = "AhoTTS server error"
log.error(msg)
raise VoiceException(msg)
else:
msg = "AhoTTS server error"
log.error(msg)
raise VoiceException(msg)
else:
msg = "AhoTTS server error"
log.error(msg)
raise VoiceException(msg)
audio_url = "%s%s/%s" % (hostname, "audio", 'tts_%s.wav' % hashnumber)
words_times_file = open(wrdfilename, 'r')
words_times = words_times_file.readlines()
words_times_file.close()
os.remove(wrdfilename)
words_times = list(map(lambda x: x[:-1].split(' ')[1], words_times))
tokens = []
starttime = 0.0
lastendtime = 0.0
for word_ind in range(len(words)):
word = words[word_ind]
if word_ind > len(words_times) - 1:
endtime = lastendtime
else:
endtime = float(words_times[word_ind]) / 1000
tokens.append({
"orth": word,
"starttime": starttime,
"endtime": endtime
})
starttime = endtime
lastendtime = endtime
return (audio_url, tokens)
def maryxml2tokensET(maryxmlstring):
try:
root = ET.fromstring(maryxmlstring)
except:
log.error("ERROR IN PARSING XML:\n%s" % maryxmlstring)
#raise with no argument re-raises the last exception
raise
#root = doc.getroot()
lang = root.attrib['{http://www.w3.org/XML/1998/namespace}lang']
#lang = utt2["maryxml"]["@xml:lang"]
ps = root.findall(".//{http://mary.dfki.de/2002/MaryXML}p")
utt = []
output_tokens = []
endtime = 0
for p in ps:
ss = p.findall(".//{http://mary.dfki.de/2002/MaryXML}s")
paragraph = []
utt.append(paragraph)
for s in ss:
sentence = []
paragraph.append(sentence)
#log.debug "S:",s
phrases = s.findall(".//{http://mary.dfki.de/2002/MaryXML}phrase")
for phrase in phrases:
#log.debug "PHRASE:", phrase
for child in phrase:
#log.debug "CHILD:", child.tag
expanded = None
if child.tag == "{http://mary.dfki.de/2002/MaryXML}t":
#orth = child.text.strip()
orth = "".join(child.itertext()).strip()
#log.debug "ORTH:", orth
tokendur = 0
for ph in child.findall(
".//{http://mary.dfki.de/2002/MaryXML}ph"):
#log.debug ph.attrib
#tokendur += ph.attrib['{http://mary.dfki.de/2002/MaryXML}d']
tokendur += int(ph.attrib['d'])
endtime += tokendur
#endtime_seconds = endtime/1000.0
#token = (orth,endtime_seconds)
elif child.tag == "{http://mary.dfki.de/2002/MaryXML}mtu":
log.debug("Found mtu token")
log.debug(child.attrib)
#The expanded words of the mtu
expanded = "".join(child.itertext()).strip()
log.debug("expanded: %s" % expanded)
#The original orthography
#TODO return both
orth = child.attrib["orig"]
log.debug("orig orth: %s" % orth)
tokendur = 0
for ph in child.findall(
".//{http://mary.dfki.de/2002/MaryXML}ph"):
#log.debug ph.attrib
#tokendur += ph.attrib['{http://mary.dfki.de/2002/MaryXML}d']
tokendur += int(ph.attrib['d'])
endtime += tokendur
#endtime_seconds = endtime/1000.0
#token = (orth,endtime_seconds)
elif child.tag == "{http://mary.dfki.de/2002/MaryXML}boundary":
orth = "PAUSE"
orth = ""
#log.debug "ORTH:", orth
#log.debug child.attrib
#endtime += child.attrib['{http://mary.dfki.de/2002/MaryXML}duration']
if "duration" in child.attrib:
endtime += int(child.attrib['duration'])
#endtime_seconds = endtime/1000.0
#token = (orth,endtime_seconds)
orth = re.sub("\s+", " ", orth)
endtime_seconds = endtime / 1000.0
token = {"orth": orth, "endtime": endtime_seconds}
if expanded:
expanded = re.sub("\s+", " ", expanded)
token["expanded"] = expanded
sentence.append(token)
output_tokens.append(token)
#log.debug utt
#return (output_tokens, lang)
return output_tokens
def test_wikispeech():
sent = "apa"
trans = {}
trans["apa"] = '" A: - p a'
lang = "sv"
try:
tmp = textproc(lang,"default_textprocessor", sent)
res = synthesise(lang,"default_voice",tmp,"markup","json")
except FileNotFoundError:
log.error("Failed to do wikispeech test.\nError type: %s\nError info:%s" % (sys.exc_info()[0], sys.exc_info()[1]))
import traceback
log.debug("Stacktrace:")
traceback.print_tb(sys.exc_info()[2])
log.debug("END stacktrace")
log.error("wikispeech test failure")
log.error("Is the audio_tmpdir %s correctly configured?" % config.config.get("Audio settings", "audio_tmpdir"))
raise
except:
log.error("Failed to do wikispeech test.\nError type: %s\nError info:%s" % (sys.exc_info()[0], sys.exc_info()[1]))
import traceback
log.debug("Stacktrace:")
traceback.print_tb(sys.exc_info()[2])
log.debug("END stacktrace")
log.error("wikispeech test failure")
log.error("No running marytts server found at %s" % config.config.get("Services","marytts"))
raise
#TODO Better with exception than return value
if type(res) == type("") and res.startswith("No voice available"):
log.error("Failed to do wikispeech test")
log.error(res)
log.error("wikispeech test failure")
raise
log.debug("%s --> %s" % (sent,res))
log.debug("SUCCESS: wikispeech test")
def test_textproc():
sent = "apa"
try:
res = textproc("sv","default_textprocessor", sent)
except:
log.error("Failed to do textprocessing.\nError type: %s\nError info:%s" % (sys.exc_info()[0], sys.exc_info()[1]))
import traceback
log.debug("Stacktrace:")
traceback.print_tb(sys.exc_info()[2])
log.debug("END stacktrace")
log.error("textprocessing test failure")
log.error("No running marytts server found at %s" % config.config.get("Services","marytts"))
raise
#TODO Better with exception than return value
if type(res) == type("") and res.startswith("ERROR:"):
log.error("Failed to do textprocessing")
log.error(res)
log.error("textprocessing test failure")
raise
log.debug("%s --> %s" % (sent,res))
log.debug("SUCCESS: textprocessing test")
def test_lexicon_client():
lexicon = "wikispeech_testdb:sv"
sent = "apa hund färöarna"
trans = {}
trans["apa"] = '"" A: . p a'
trans["hund"] = '" h u0 n d'
trans["färöarna"] = '"" f {: . % r 2: . a . rn a'
try:
lexicon_client.loadLexicon(lexicon)
lex = lexicon_client.getLookupBySentence(sent, lexicon)
log.debug("LEX: %s" % lex)
except:
log.error("Failed to do lexicon lookup.\nError type: %s\nError info:%s" % (sys.exc_info()[0], sys.exc_info()[1]))
import traceback
log.debug("Stacktrace:")
if log.log_level == "debug":
traceback.print_tb(sys.exc_info()[2])
log.debug("END stacktrace")
log.error("lexicon lookup test failure")
log.error("No running lexserver found at %s" % config.config.get("Services","lexicon"))
raise
for word in sent.split(" "):
try:
if lex[word] != trans[word]:
log.error("lexicon lookup test failure")
log.error("word %s, found %s, expected %s" % (word, lex[word], trans[word]))
raise
except KeyError:
log.error("Lexicon lookup test failure: Word %s not found in lexicon %s" % (word, lexicon))
raise
log.debug("SUCCESS: lexicon lookup test")
import wikispeech_server.log as log
from wikispeech_server.textprocessor import Textprocessor, TextprocessorException
from wikispeech_server.voice import Voice, VoiceException
#################
#
# Test opusenc before anything else
#
################
log.debug("\nOPUSENC\n\nChecking that opusenc is installed on your system..")
retval = os.system("opusenc -V")
if retval != 0:
os.system("opusenc -V")
log.error(
"ERROR: opusenc was not found. You should probably run something like\nsudo apt install opus-tools\n"
)
sys.exit(1)
else:
log.debug("opusenc found.\n\nEND OPUSENC\n")
###############
#
# Load textprocessors and voices
#
###############
def loadTextprocessor(tp_config):
try:
tp = Textprocessor(tp_config)
import pytz
from pytz import timezone
import subprocess
#################
#
# Test opusenc before anything else
#
################
log.info("\nOPUSENC\n\nChecking that opusenc is installed on your system..")
retval = os.system("opusenc -V")
if retval != 0:
os.system("opusenc -V")
log.error("ERROR: opusenc was not found. You should probably run something like\nsudo apt install opus-tools\n")
sys.exit(1)
else:
log.info("opusenc found.\n\nEND OPUSENC\n")
###############
#
# Load textprocessors and voices
#
###############
textprocessors = []
def loadTextprocessor(tp_config):
try:
def maryxml2tokensET(maryxmlstring):
try:
root = ET.fromstring(maryxmlstring)
except:
log.error("ERROR IN PARSING XML:\n%s" % maryxmlstring)
#raise with no argument re-raises the last exception
raise
#root = doc.getroot()
lang = root.attrib['{http://www.w3.org/XML/1998/namespace}lang']
#lang = utt2["maryxml"]["@xml:lang"]
ps = root.findall(".//{http://mary.dfki.de/2002/MaryXML}p")
utt = []
output_tokens = []
endtime = 0
for p in ps:
ss = p.findall(".//{http://mary.dfki.de/2002/MaryXML}s")
paragraph = []
utt.append(paragraph)
for s in ss:
sentence = []
paragraph.append(sentence)
#log.debug "S:",s
phrases = s.findall(".//{http://mary.dfki.de/2002/MaryXML}phrase")
for phrase in phrases:
#log.debug "PHRASE:", phrase
for child in phrase:
#log.debug "CHILD:", child.tag
expanded = None
if child.tag == "{http://mary.dfki.de/2002/MaryXML}t":
#orth = child.text.strip()
orth = "".join(child.itertext()).strip()
#log.debug "ORTH:", orth
tokendur = 0
for ph in child.findall(".//{http://mary.dfki.de/2002/MaryXML}ph"):
#log.debug ph.attrib
#tokendur += ph.attrib['{http://mary.dfki.de/2002/MaryXML}d']
tokendur += int(ph.attrib['d'])
endtime += tokendur
#endtime_seconds = endtime/1000.0
#token = (orth,endtime_seconds)
elif child.tag == "{http://mary.dfki.de/2002/MaryXML}mtu":
log.debug("Found mtu token")
log.debug(child.attrib)
#The expanded words of the mtu
expanded = "".join(child.itertext()).strip()
log.debug("expanded: %s" % expanded)
#The original orthography
#TODO return both
orth = child.attrib["orig"]
log.debug("orig orth: %s" % orth)
tokendur = 0
for ph in child.findall(".//{http://mary.dfki.de/2002/MaryXML}ph"):
#log.debug ph.attrib
#tokendur += ph.attrib['{http://mary.dfki.de/2002/MaryXML}d']
tokendur += int(ph.attrib['d'])
endtime += tokendur
#endtime_seconds = endtime/1000.0
#token = (orth,endtime_seconds)
elif child.tag == "{http://mary.dfki.de/2002/MaryXML}boundary":
orth = "PAUSE"
orth = ""
#log.debug "ORTH:", orth
#log.debug child.attrib
#endtime += child.attrib['{http://mary.dfki.de/2002/MaryXML}duration']
if "duration" in child.attrib:
endtime += int(child.attrib['duration'])
#endtime_seconds = endtime/1000.0
#token = (orth,endtime_seconds)
orth = re.sub("\s+", " ", orth)
endtime_seconds = endtime/1000.0
token = {"orth":orth,"endtime":endtime_seconds}
if expanded:
expanded = re.sub("\s+", " ", expanded)
token["expanded"] = expanded
sentence.append(token)
output_tokens.append(token)
#log.debug utt
#return (output_tokens, lang)
return output_tokens
def test_lexicon_client():
lexicon = "wikispeech_testdb:sv"
sent = "apa hund färöarna"
trans = {}
trans["apa"] = '"" A: . p a'
trans["hund"] = '" h u0 n d'
trans["färöarna"] = '"" f {: . % r 2: . a . rn a'
try:
lexicon_client.loadLexicon(lexicon)
lex = lexicon_client.getLookupBySentence(sent, lexicon)
log.debug("LEX: %s" % lex)
except:
log.error(
"Failed to do lexicon lookup.\nError type: %s\nError info:%s" %
(sys.exc_info()[0], sys.exc_info()[1]))
import traceback
log.debug("Stacktrace:")
if log.log_level == "debug":
traceback.print_tb(sys.exc_info()[2])
log.debug("END stacktrace")
log.error("lexicon lookup test failure")
log.error("No running lexserver found at %s" %
config.config.get("Services", "lexicon"))
raise
for word in sent.split(" "):
try:
if lex[word] != trans[word]:
log.error("lexicon lookup test failure")
log.error("word %s, found %s, expected %s" %
(word, lex[word], trans[word]))
raise
except KeyError:
log.error(
"Lexicon lookup test failure: Word %s not found in lexicon %s"
% (word, lexicon))
raise
log.debug("SUCCESS: lexicon lookup test")
def mary2ws(maryxml, voice):
(lang, maryxml) = dropHeader(maryxml)
#lang = "sv"
#log.debug(maryxml)
root = ET.fromstring(maryxml.encode('utf-8'))
#root = ET.fromstring(maryxml)
paragraphs = []
utterance = {"lang": lang, "paragraphs": paragraphs}
paragraph_elements = root.findall(".//p")
for paragraph_element in paragraph_elements:
sentences = []
paragraphs.append({"sentences": sentences})
sentence_elements = paragraph_element.findall("s")
for sentence_element in sentence_elements:
phrases = []
sentence = {"phrases": phrases}
sentences.append(sentence)
for sentence_child in sentence_element:
if sentence_child.tag == "phrase":
phrase_element = sentence_child
phrase = buildPhrase(phrase_element, lang, voice)
phrases.append(phrase)
elif sentence_child.tag == "prosody":
prosody_element = sentence_child
#can 'prosody' only contain exactly one 'phrase'?
phrase_element = prosody_element[0]
phrase = buildPhrase(phrase_element, lang, voice)
phrase = addIfExists(phrase,
prosody_element,
"pitch",
prefix="prosody_")
phrase = addIfExists(phrase,
prosody_element,
"range",
prefix="prosody_")
phrases.append(phrase)
elif sentence_child.tag == "t":
#This is a special case that happens (sometimes..) with single-word sentences
#doesn't work, produces error in synthesis
#TODO look at this again
#It only happens with the word "Hon." ...
phrase_element = ET.Element("phrase")
phrase_element.append(sentence_child)
phrase = buildPhrase(phrase_element, lang, voice)
phrases.append(phrase)
else:
log.error("sentence child should not have tag %s" %
sentence_child.tag)
return utterance
def test_textproc():
sent = "apa"
try:
res = textproc("sv", "default_textprocessor", sent)
except:
log.error(
"Failed to do textprocessing.\nError type: %s\nError info:%s" %
(sys.exc_info()[0], sys.exc_info()[1]))
import traceback
log.debug("Stacktrace:")
traceback.print_tb(sys.exc_info()[2])
log.debug("END stacktrace")
log.error("textprocessing test failure")
log.error("No running marytts server found at %s" %
config.config.get("Services", "marytts"))
raise
#TODO Better with exception than return value
if type(res) == type("") and res.startswith("ERROR:"):
log.error("Failed to do textprocessing")
log.error(res)
log.error("textprocessing test failure")
raise
log.debug("%s --> %s" % (sent, res))
log.debug("SUCCESS: textprocessing test")
def test_wikispeech():
sent = "apa"
trans = {}
trans["apa"] = '" A: - p a'
lang = "sv"
try:
tmp = textproc(lang, "default_textprocessor", sent)
res = synthesise(lang, "default_voice", tmp, "markup", "json")
except FileNotFoundError:
log.error(
"Failed to do wikispeech test.\nError type: %s\nError info:%s" %
(sys.exc_info()[0], sys.exc_info()[1]))
import traceback
log.debug("Stacktrace:")
traceback.print_tb(sys.exc_info()[2])
log.debug("END stacktrace")
log.error("wikispeech test failure")
log.error("Is the audio_tmpdir %s correctly configured?" %
config.config.get("Audio settings", "audio_tmpdir"))
raise
except:
log.error(
"Failed to do wikispeech test.\nError type: %s\nError info:%s" %
(sys.exc_info()[0], sys.exc_info()[1]))
import traceback
log.debug("Stacktrace:")
traceback.print_tb(sys.exc_info()[2])
log.debug("END stacktrace")
log.error("wikispeech test failure")
log.error("No running marytts server found at %s" %
config.config.get("Services", "marytts"))
raise
#TODO Better with exception than return value
if type(res) == type("") and res.startswith("No voice available"):
log.error("Failed to do wikispeech test")
log.error(res)
log.error("wikispeech test failure")
raise
log.debug("%s --> %s" % (sent, res))
log.debug("SUCCESS: wikispeech test")
return "{name:%s, lang:%s}" % (self.name, self.lang)
def __str__(self):
return {"name":self.name, "lang":self.lang}
#return "{name:%s, lang:%s}" % (self.name, self.lang)
if __name__ == "__main__":
log.log_level = "debug" #debug, info, warning, error
voice_config = {
"lang":"sv",
"name":"stts_sv_nst-hsmm",
"engine":"marytts",
"adapter":"adapters.marytts_adapter",
"mapper": {
"from":"sv-se_ws-sampa",
"to":"sv-se_sampa_mary"
}
}
try:
v = Voice(voice_config)
log.info("Created voice %s from %s" % (v, voice_config))
except VoiceException as e:
log.error("Failed to create voice for %s\nException message was:\n%s" % (voice_config, e))
l.append("config_file: %s" % (self.config["config_file"]))
return "{%s}" % ", ".join(l)
def __str__(self):
return self.__repr__()
if __name__ == "__main__":
log.log_level = "debug" #debug, info, warning, error
voice_config = {
"lang":"sv",
"name":"stts_sv_nst-hsmm",
"engine":"marytts",
"adapter":"adapters.marytts_adapter",
"mapper": {
"from":"sv-se_ws-sampa",
"to":"sv-se_sampa_mary"
}
}
try:
v = Voice(voice_config)
log.info("Created voice %s from %s" % (v, voice_config))
except VoiceException as e:
log.error("Failed to create voice for %s\nException message was:\n%s" % (voice_config, e))
