def inner(c_ext, cew_subprocess, cache):
if ofp is None:
handler, output_file_path = gf.tmp_file(suffix=".wav")
else:
handler = None
output_file_path = ofp
try:
rconf = RuntimeConfiguration()
rconf[RuntimeConfiguration.TTS] = self.TTS
rconf[RuntimeConfiguration.TTS_PATH] = self.TTS_PATH
rconf[RuntimeConfiguration.C_EXTENSIONS] = c_ext
rconf[RuntimeConfiguration.
CEW_SUBPROCESS_ENABLED] = cew_subprocess
rconf[RuntimeConfiguration.TTS_CACHE] = cache
tts_engine = self.TTS_CLASS(rconf=rconf)
anchors, total_time, num_chars = tts_engine.synthesize_multiple(
text_file, output_file_path, quit_after, backwards)
gf.delete_file(handler, output_file_path)
if cache:
tts_engine.clear_cache()
if zero_length:
self.assertEqual(total_time, 0.0)
else:
self.assertGreater(total_time, 0.0)
except (OSError, TypeError, UnicodeDecodeError, ValueError) as exc:
gf.delete_file(handler, output_file_path)
if (cache) and (tts_engine is not None):
tts_engine.clear_cache()
with self.assertRaises(expected_exc):
raise exc
def synthesize_multiple(self,
text_file,
ofp=None,
quit_after=None,
backwards=False,
zero_length=False):
if ofp is None:
handler, output_file_path = gf.tmp_file(suffix=".wav")
else:
handler = None
output_file_path = ofp
try:
rconf = RuntimeConfiguration()
rconf[RuntimeConfiguration.TTS] = u"festival"
rconf[RuntimeConfiguration.TTS_PATH] = u"text2wave"
tts_engine = FESTIVALWrapper(rconf=rconf)
anchors, total_time, num_chars = tts_engine.synthesize_multiple(
text_file, output_file_path, quit_after, backwards)
gf.delete_file(handler, output_file_path)
if zero_length:
self.assertEqual(total_time, 0.0)
else:
self.assertGreater(total_time, 0.0)
except (OSError, TypeError, UnicodeDecodeError, ValueError) as exc:
gf.delete_file(handler, output_file_path)
raise exc
def align(text_path, audio_path, align_out_path, word_align=True):
# create Task object
config_string = u"task_language=hi"
config_string += "|os_task_file_format=json"
rconf = None
if word_align:
config_string += "|os_task_file_levels=3"
config_string += "|is_text_type=mplain"
rconf = RuntimeConfiguration()
rconf[RuntimeConfiguration.MFCC_MASK_NONSPEECH] = True
rconf[RuntimeConfiguration.MFCC_MASK_NONSPEECH_L3] = True
else:
config_string += "|is_text_type=plain"
task = Task(config_string=config_string)
task.text_file_path_absolute = text_path
task.audio_file_path_absolute = audio_path
task.sync_map_file_path_absolute = align_out_path
# process Task
ExecuteTask(task, rconf=rconf).execute()
# output sync map to file
task.output_sync_map_file()
# Remove annoying unicode characters
with open(align_out_path, 'r', encoding='utf8') as f:
alignment = json.load(f)
with open(align_out_path, 'w', encoding='utf8') as f:
json.dump(alignment, f, ensure_ascii=False, indent=2)
def test_use_cache(self):
if self.TTS == u"":
return
rconf = RuntimeConfiguration()
rconf[RuntimeConfiguration.TTS_CACHE] = True
tts_engine = self.TTS_CLASS(rconf=rconf)
self.assertTrue(tts_engine.use_cache)
self.assertIsNotNone(tts_engine.cache)
def test_set_tts(self):
rconf = RuntimeConfiguration()
rconf.set_tts(level=1)
self.assertEqual(rconf.tts, "espeak")
self.assertEqual(rconf.tts_path, None)
rconf.set_tts(level=2)
self.assertEqual(rconf.tts, "espeak")
self.assertEqual(rconf.tts_path, None)
rconf.set_tts(level=3)
self.assertEqual(rconf.tts, "espeak")
self.assertEqual(rconf.tts_path, None)
def test_set_granularity(self):
rconf = RuntimeConfiguration()
rconf.set_granularity(level=1)
self.assertEqual(rconf.mmn, False)
self.assertEqual(rconf.mwl, TimeValue("0.100"))
self.assertEqual(rconf.mws, TimeValue("0.040"))
rconf.set_granularity(level=2)
self.assertEqual(rconf.mmn, False)
self.assertEqual(rconf.mwl, TimeValue("0.050"))
self.assertEqual(rconf.mws, TimeValue("0.020"))
rconf.set_granularity(level=3)
self.assertEqual(rconf.mmn, False)
self.assertEqual(rconf.mwl, TimeValue("0.020"))
self.assertEqual(rconf.mws, TimeValue("0.005"))
def synthesize_single(self, text, language, ofp=None, zero_length=False):
if ofp is None:
handler, output_file_path = gf.tmp_file(suffix=".wav")
else:
handler = None
output_file_path = ofp
try:
rconf = RuntimeConfiguration()
rconf[RuntimeConfiguration.TTS] = u"festival"
rconf[RuntimeConfiguration.TTS_PATH] = u"text2wave"
tts_engine = FESTIVALWrapper(rconf=rconf)
result = tts_engine.synthesize_single(text, language,
output_file_path)
gf.delete_file(handler, output_file_path)
if zero_length:
self.assertEqual(result, 0)
else:
self.assertGreater(result, 0)
except (OSError, TypeError, UnicodeDecodeError, ValueError) as exc:
gf.delete_file(handler, output_file_path)
raise exc
def __adjust_durations(self, subs: List[SubRipItem], audio_file_path: str, stretch_in_lang: str) -> List[SubRipItem]:
from aeneas.executetask import ExecuteTask
from aeneas.task import Task
from aeneas.runtimeconfiguration import RuntimeConfiguration
from aeneas.logger import Logger as AeneasLogger
# Initialise a DTW alignment task
task_config_string = (
"task_language={}|os_task_file_format=srt|is_text_type=subtitles".format(stretch_in_lang)
)
runtime_config_string = "dtw_algorithm=stripe" # stripe or exact
task = Task(config_string=task_config_string)
try:
segment_path, _ = MediaHelper.extract_audio_from_start_to_end(
audio_file_path,
str(subs[0].start),
str(subs[len(subs) - 1].end),
)
# Create a text file for DTW alignments
root, _ = os.path.splitext(segment_path)
text_file_path = "{}.txt".format(root)
with open(text_file_path, "w", encoding="utf8") as text_file:
for sub_new in subs:
text_file.write(sub_new.text)
text_file.write(os.linesep * 2)
task.audio_file_path_absolute = segment_path
task.text_file_path_absolute = text_file_path
task.sync_map_file_path_absolute = "{}.srt".format(root)
tee = False
if Logger.VERBOSE:
tee = True
if Logger.QUIET:
tee = False
with self.__lock:
# Execute the task
ExecuteTask(
task=task,
rconf=RuntimeConfiguration(config_string=runtime_config_string),
logger=AeneasLogger(tee=tee),
).execute()
# Output new subtitle segment to a file
task.output_sync_map_file()
# Load the above subtitle segment
adjusted_subs = Subtitle.load(
task.sync_map_file_path_absolute
).subs
for index, sub_new_loaded in enumerate(adjusted_subs):
sub_new_loaded.index = subs[index].index
adjusted_subs.shift(
seconds=MediaHelper.get_duration_in_seconds(
start=None, end=str(subs[0].start)
)
)
return adjusted_subs
finally:
# Housekeep intermediate files
if task.audio_file_path_absolute is not None and os.path.exists(
task.audio_file_path_absolute
):
os.remove(task.audio_file_path_absolute)
if task.text_file_path_absolute is not None and os.path.exists(
task.text_file_path_absolute
):
os.remove(task.text_file_path_absolute)
if task.sync_map_file_path_absolute is not None and os.path.exists(task.sync_map_file_path_absolute):
os.remove(task.sync_map_file_path_absolute)
def test_config_string(self):
rconf = RuntimeConfiguration()
rconf.config_string
def test_tts(self):
rconf = RuntimeConfiguration()
self.assertEqual(rconf.tts, "espeak")
def test_tts_path(self):
rconf = RuntimeConfiguration()
self.assertEqual(rconf.tts_path, None)
def test_mmn(self):
rconf = RuntimeConfiguration()
self.assertEqual(rconf.mmn, False)
def test_mwl(self):
rconf = RuntimeConfiguration()
self.assertEqual(rconf.mwl, TimeValue("0.100"))
def run(self, arguments, show_help=True):
"""
Program entry point.
Please note that the first item in ``arguments`` is discarded,
as it is assumed to be the script/invocation name;
pass a "dumb" placeholder if you call this method with
an argument different that ``sys.argv``.
:param arguments: the list of arguments
:type arguments: list
:param show_help: if ``False``, do not show help on ``-h`` and ``--help``
:type show_help: bool
:rtype: int
"""
# convert arguments into Unicode strings
if self.use_sys:
# check that sys.stdin.encoding and sys.stdout.encoding are set to utf-8
if not gf.FROZEN:
if sys.stdin.encoding not in ["UTF-8", "UTF8"]:
self.print_warning(
u"The default input encoding is not UTF-8.")
self.print_warning(
u"You might want to set 'PYTHONIOENCODING=UTF-8' in your shell."
)
if sys.stdout.encoding not in ["UTF-8", "UTF8"]:
self.print_warning(
u"The default output encoding is not UTF-8.")
self.print_warning(
u"You might want to set 'PYTHONIOENCODING=UTF-8' in your shell."
)
# decode using sys.stdin.encoding
args = [gf.safe_unicode_stdin(arg) for arg in arguments]
else:
# decode using utf-8 (but you should pass Unicode strings as parameters anyway)
args = [gf.safe_unicode(arg) for arg in arguments]
if show_help:
if u"-h" in args:
return self.print_help(short=True)
if u"--help" in args:
return self.print_help(short=False)
if u"--version" in args:
return self.print_name_version()
# store formal arguments
self.formal_arguments_raw = arguments
self.formal_arguments = args
# to obtain the actual arguments,
# remove the first one and "special" switches
args = args[1:]
set_args = set(args)
# set verbosity, if requested
for flag in set([u"-v", u"--verbose"]) & set_args:
self.verbose = True
args.remove(flag)
for flag in set([u"-vv", u"--very-verbose"]) & set_args:
self.verbose = True
self.very_verbose = True
args.remove(flag)
# set RuntimeConfiguration string, if specified
for flag in [u"-r", u"--runtime-configuration"]:
rconf_string = self.has_option_with_value(flag,
actual_arguments=False)
if rconf_string is not None:
self.rconf = RuntimeConfiguration(rconf_string)
args.remove("%s=%s" % (flag, rconf_string))
# set log file path, if requested
log_path = None
for flag in [u"-l", u"--log"]:
log_path = self.has_option_with_value(flag, actual_arguments=False)
if log_path is not None:
args.remove("%s=%s" % (flag, log_path))
elif flag in set_args:
handler, log_path = gf.tmp_file(
suffix=u".log",
root=self.rconf[RuntimeConfiguration.TMP_PATH])
args.remove(flag)
if log_path is not None:
self.log_file_path = log_path
# if no actual arguments left, print help
if (len(args) < 1) and (show_help):
return self.print_help(short=True)
# store actual arguments
self.actual_arguments = args
# create logger
self.logger = Logger(tee=self.verbose,
tee_show_datetime=self.very_verbose)
self.log([u"Formal arguments: %s", self.formal_arguments])
self.log([u"Actual arguments: %s", self.actual_arguments])
self.log([u"Runtime configuration: '%s'", self.rconf.config_string()])
# perform command
exit_code = self.perform_command()
self.log([u"Execution completed with code %d", exit_code])
# output log if requested
if self.log_file_path is not None:
self.log([
u"User requested saving log to file '%s'", self.log_file_path
])
self.logger.write(self.log_file_path)
if self.use_sys:
self.print_info(u"Log written to file '%s'" %
self.log_file_path)
return self.exit(exit_code)
def test_dtw_margin(self):
rconf = RuntimeConfiguration()
self.assertEqual(rconf.dtw_margin, TimeValue("60.000"))
def __init__(self, logger=None, rconf=None):
self.logger = logger if logger is not None else Logger()
self.rconf = rconf if rconf is not None else RuntimeConfiguration()
def test_clone(self):
rconf = RuntimeConfiguration()
rconf2 = rconf.clone()
self.assertNotEqual(id(rconf), id(rconf2))
self.assertEqual(rconf.config_string, rconf2.config_string)
if lang not in ["eng", "hi", "hin"]:
print("only hi and eng allowed for language")
exit(1)
from aeneas.executetask import ExecuteTask
from aeneas.task import Task
from aeneas.runtimeconfiguration import RuntimeConfiguration
config_string = u"task_language=" + lang + u"|is_text_type=subtitles|os_task_file_format=srt"
tempout, tempfilename = tempfile.mkstemp()
task = Task(config_string=config_string)
task.audio_file_path_absolute = args.audio
task.text_file_path_absolute = args.txt
task.sync_map_file_path_absolute = tempfilename
rconf = RuntimeConfiguration()
# This option ignores the non-word sounds in the audio
rconf[RuntimeConfiguration.MFCC_MASK_NONSPEECH] = True
rconf[RuntimeConfiguration.MFCC_MASK_LOG_ENERGY_THRESHOLD] = 2.5
# To use a different Text-to-Speech engine
#rconf[RuntimeConfiguration.TTS] = "festival"
# process Task
ExecuteTask(task, rconf=rconf).execute()
# output sync map to file
task.output_sync_map_file()
f = open(args.out, "w")
f.writelines("WEBVTT\n")
def test_loggable_rconf(self):
rconf = RuntimeConfiguration()
loggable = Loggable(rconf=rconf)
self.assertEqual(rconf, loggable.rconf)
self.assertIsNotNone(loggable.logger)
def test_loggable_rconf_logger(self):
logger = Logger()
rconf = RuntimeConfiguration()
loggable = Loggable(rconf=rconf, logger=logger)
self.assertEqual(rconf, loggable.rconf)
self.assertEqual(logger, loggable.logger)
def test_set_rconf_string(self):
params = [
(u"aba_nonspeech_tolerance=0.040", "aba_nonspeech_tolerance",
TimeValue("0.040")),
(u"aba_no_zero_duration=0.040", "aba_no_zero_duration",
TimeValue("0.040")),
(u"allow_unlisted_languages=True", "allow_unlisted_languages",
True),
(u"c_extensions=False", "c_extensions", False),
(u"cdtw=False", "cdtw", False),
(u"cew=False", "cew", False),
(u"cmfcc=False", "cmfcc", False),
(u"cew_subprocess_enabled=True", "cew_subprocess_enabled", True),
(u"cew_subprocess_path=/foo/bar/python", "cew_subprocess_path",
"/foo/bar/python"),
(u"downloader_sleep=5.000", "downloader_sleep",
TimeValue("5.000")),
(u"downloader_retry_attempts=5", "downloader_retry_attempts", 5),
(u"dtw_algorithm=exact", "dtw_algorithm", "exact"),
(u"dtw_margin=100", "dtw_margin", TimeValue("100")),
(u"ffmpeg_path=/foo/bar/ffmpeg", "ffmpeg_path", "/foo/bar/ffmpeg"),
(u"ffmpeg_sample_rate=8000", "ffmpeg_sample_rate", 8000),
(u"ffprobe_path=/foo/bar/ffprobe", "ffprobe_path",
"/foo/bar/ffprobe"),
(u"job_max_tasks=10", "job_max_tasks", 10),
(u"mfcc_filters=100", "mfcc_filters", 100),
(u"mfcc_size=20", "mfcc_size", 20),
(u"mfcc_fft_order=256", "mfcc_fft_order", 256),
(u"mfcc_lower_frequency=120.0", "mfcc_lower_frequency", 120.0),
(u"mfcc_upper_frequency=5000.0", "mfcc_upper_frequency", 5000.0),
(u"mfcc_emphasis_factor=1.0", "mfcc_emphasis_factor", 1.0),
(u"mfcc_mask_nonspeech=True", "mfcc_mask_nonspeech", True),
(u"mfcc_window_length=0.360", "mfcc_window_length",
TimeValue("0.360")),
(u"mfcc_window_shift=0.160", "mfcc_window_shift",
TimeValue("0.160")),
(u"dtw_margin_l1=100", "dtw_margin_l1", TimeValue("100")),
(u"mfcc_mask_nonspeech_l1=True", "mfcc_mask_nonspeech_l1", True),
(u"mfcc_window_length_l1=0.360", "mfcc_window_length_l1",
TimeValue("0.360")),
(u"mfcc_window_shift_l1=0.160", "mfcc_window_shift_l1",
TimeValue("0.160")),
(u"dtw_margin_l2=30", "dtw_margin_l2", TimeValue("30")),
(u"mfcc_mask_nonspeech_l2=True", "mfcc_mask_nonspeech_l2", True),
(u"mfcc_window_length_l2=0.360", "mfcc_window_length_l2",
TimeValue("0.360")),
(u"mfcc_window_shift_l2=0.160", "mfcc_window_shift_l2",
TimeValue("0.160")),
(u"dtw_margin_l3=10", "dtw_margin_l3", TimeValue("10")),
(u"mfcc_mask_nonspeech_l3=True", "mfcc_mask_nonspeech_l3", True),
(u"mfcc_window_length_l3=0.360", "mfcc_window_length_l3",
TimeValue("0.360")),
(u"mfcc_window_shift_l3=0.160", "mfcc_window_shift_l3",
TimeValue("0.160")),
(u"mfcc_mask_extend_speech_after=1",
"mfcc_mask_extend_speech_after", 1),
(u"mfcc_mask_extend_speech_before=1",
"mfcc_mask_extend_speech_before", 1),
(u"mfcc_mask_log_energy_threshold=0.750",
"mfcc_mask_log_energy_threshold", 0.750),
(u"mfcc_mask_min_nonspeech_length=5",
"mfcc_mask_min_nonspeech_length", 5),
(u"nuance_tts_api_id=foo", "nuance_tts_api_id", "foo"),
(u"nuance_tts_api_key=bar", "nuance_tts_api_key", "bar"),
(u"safety_checks=False", "safety_checks", False),
(u"task_max_audio_length=1000", "task_max_audio_length",
TimeValue("1000")),
(u"task_max_text_length=1000", "task_max_text_length", 1000),
(u"tmp_path=/foo/bar", "tmp_path", "/foo/bar"),
(u"tts=festival", "tts", "festival"),
(u"tts_path=/foo/bar/festival", "tts_path", "/foo/bar/festival"),
(u"tts_api_sleep=5.000", "tts_api_sleep", TimeValue("5.000")),
(u"tts_api_retry_attempts=3", "tts_api_retry_attempts", 3),
(u"tts_voice_code=ru", "tts_voice_code", "ru"),
(u"tts_cache=True", "tts_cache", True),
(u"tts_l1=festival", "tts_l1", "festival"),
(u"tts_path_l1=/foo/bar/festival", "tts_path_l1",
"/foo/bar/festival"),
(u"tts_l2=festival", "tts_l2", "festival"),
(u"tts_path_l2=/foo/bar/festival", "tts_path_l2",
"/foo/bar/festival"),
(u"tts_l3=festival", "tts_l3", "festival"),
(u"tts_path_l3=/foo/bar/festival", "tts_path_l3",
"/foo/bar/festival"),
(u"vad_extend_speech_after=1.000", "vad_extend_speech_after",
TimeValue("1.000")),
(u"vad_extend_speech_before=1.000", "vad_extend_speech_before",
TimeValue("1.000")),
(u"vad_log_energy_threshold=0.750", "vad_log_energy_threshold",
0.750),
(u"vad_min_nonspeech_length=0.500", "vad_min_nonspeech_length",
TimeValue("0.500")),
]
for string, key, value in params:
rconf = RuntimeConfiguration(string)
self.assertEqual(rconf[key], value)
def test_safety_checks(self):
rconf = RuntimeConfiguration()
self.assertEqual(rconf.safety_checks, True)
def test_convert_rc(self):
rc = RuntimeConfiguration(u"ffmpeg_sample_rate=44100")
for f in self.FILES:
self.convert(f["path"], runtime_configuration=rc)
def test_sample_rate(self):
rconf = RuntimeConfiguration()
self.assertEqual(rconf.sample_rate, 16000)