def from_textgrid(voice):
""" Create aligned Utterances by synthesising to Segment level
from the orthography and simply copying label end times into
segment items as "end" feature.
"""
#Setup and create necessary dirs...
CWD = os.getcwd()
wav_dir = os.path.join(CWD, WAV_DIR)
uttwav_dir = os.path.join(CWD, UTTWAV_DIR)
transcr_location = os.path.join(CWD, ETC_DIR, TRANSCR_FILE)
textgrid_dir = os.path.join(CWD, TEXTGRID_DIR)
aligned_utts_dir = os.path.join(CWD, ALIGNED_UTT_DIR)
os.makedirs(aligned_utts_dir)
#update utts from textgrids...
transcriptions = load_transcriptions_schemefile(transcr_location)
alignments = sl.Corpus(textgrid_dir)
#################
for sc_utt, uttname, wavfilename in zip(
alignments.utterances, sorted(transcriptions),
sorted(glob(os.path.join(uttwav_dir, "*")))):
assert sc_utt.name == uttname, "Utterance missmatch..."
assert os.path.basename(wavfilename).startswith(
uttname), "Utterance missmatch..."
print("Synthesizing:", uttname)
utt = voice.synthesize(transcriptions[uttname], 'text-to-segments')
utt["file_id"] = uttname
utt = transplant_segtime_info(voice, sc_utt, utt)
#add waveform to utt:
utt["waveform"] = Waveform(wavfilename)
#save utt...
ttslab.tofile(
utt, os.path.join(aligned_utts_dir, ".".join([uttname, UTT_EXT])))
def from_textgrid(voice):
""" Create aligned Utterances by synthesising to Word level from
the orthography and filling in further SylStructure from the
TextGrid input...
Normalised orthography must match for this to be successful...
"""
#Setup and create necessary dirs...
CWD = os.getcwd()
wav_dir = os.path.join(CWD, WAV_DIR)
transcr_location = os.path.join(CWD, ETC_DIR, TRANSCR_FILE)
textgrid_dir = os.path.join(CWD, TEXTGRID_DIR)
aligned_utts_dir = os.path.join(CWD, ALIGNED_UTT_DIR)
os.makedirs(aligned_utts_dir)
#correct utterances from textgrids...
transcriptions = load_transcriptions_schemefile(transcr_location)
alignments = sl.Corpus(textgrid_dir)
make_aligned_utts(voice, transcriptions, alignments, wav_dir,
aligned_utts_dir)
def to_textgrid(voice):
""" Should pull Word and Segment info from set of utterances make
input compatible with HAlign2 (align from orthography), and
run alignment process to make a set of TextGrid files
including Syllables...
or
Should pull Word and Segment info from a special set of
utterances where alternative utts are embedded. Makes input
compatible with HAlign2 (align from orthography) with
dictionary containing variants, and run alignment process
(with re-alignment stage) to make a set of TextGrid files
including Syllables...
"""
#create necessary output dirs...
CWD = os.getcwd()
wav_dir = os.path.join(CWD, WAV_DIR)
transcr_location = os.path.join(CWD, ETC_DIR, TRANSCR_FILE)
halign_config_location = os.path.join(CWD, ETC_DIR, HALIGNCONF_FILE)
halign_working_dir = os.path.join(CWD, HALIGNWORK_DIR)
halign_input_transcr_dir = os.path.join(halign_working_dir,
HALIGNINPUT_SUBDIR,
HALIGNINPUTTRANSCR_DIR)
textgrid_dir = os.path.join(CWD, TEXTGRID_DIR)
os.makedirs(textgrid_dir)
os.makedirs(halign_input_transcr_dir)
#get silence phone..
silence_phone = voice.phoneset.features["silence_phone"]
#start alignment process..
transcriptions = load_transcriptions_schemefile(transcr_location)
utts = make_base_utts(voice, transcriptions)
halign_input_transcr_dir, pronundict_location = make_halign_input(
voice, utts, halign_working_dir)
# if not any("alt_utts" in u for u in utts):
# GenHAlign(halign_config_location,
# overrides={"SOURCE:ORTHOGRAPHIC_TRANSCRIPTIONS" : halign_input_transcr_dir,
# "SOURCE:PRONUNCIATION_DICTIONARY" : pronundict_location,
# "SOURCE:AUDIO" : wav_dir,
# "PARMS:WORKING_DIR" : halign_working_dir,
# "PARMS:SILENCE_PHONE" : silence_phone})
# else:
GenHAlignRealign(halign_config_location,
overrides={
"SOURCE:ORTHOGRAPHIC_TRANSCRIPTIONS":
halign_input_transcr_dir,
"SOURCE:PRONUNCIATION_DICTIONARY":
pronundict_location,
"SOURCE:AUDIO": wav_dir,
"PARMS:WORKING_DIR": halign_working_dir,
"PARMS:SILENCE_PHONE": silence_phone
})
alignments = sl.Corpus(os.path.join(halign_working_dir, "textgrids"))
add_sylls_to_textgrids(voice, alignments, textgrid_dir)