Remove 0s durations from silence splitting

This commit is contained in:
Vadim committed 2023-12-27 23:22:47 +01:00
1 parent dad84ebdea
commit 849c281544
3 files changed
+76 -9

No files matched your search

@@ -186,9 +186,61 @@ class SilenceProcessingTest(unittest.TestCase):
self.assertEqual(result[3].end, 8.0)
self.assertEqual(result[3].is_hyphen, False)
# todo
# | ** ** ** ***** | silence_parts_list
# | *00**00**0 | multi silence is inside
# |0 1 2 3 4 5 6 7 8 9 10 11 12| time
def test_remove_no_duration_silence2(self):
#
# | * * * * * * | silence_parts_list
# | 000 000 00000 | multi silence is inside
# |0 1 2 3 4 5 6 7 8 | time
# Arrange
silence_parts_list = [(2.0, 2.1), (3.09, 3.1), (4.0, 4.09), (5.0, 5.09), (6.0, 6.09), (7.0, 7.09)]
transcribed_data = [
TranscribedData({
"conf": 0.95,
"word": "Remove split ",
"end": 2.19,
"start": 1.0
}),
TranscribedData({
"conf": 0.95,
"word": "Is split ",
"end": 4.1,
"start": 3.0
}),
TranscribedData({
"conf": 0.95,
"word": "Is split2 ",
"end": 7.1,
"start": 5.0
}),
]
# Act
result = remove_silence2(silence_parts_list, transcribed_data)
# Assert
self.assertEqual(result[0].word, "Remove split ")
self.assertEqual(result[0].start, 1.0)
self.assertEqual(result[0].end, 2.0)
self.assertEqual(result[0].is_hyphen, None)
self.assertEqual(result[1].word, "Is split ")
self.assertEqual(result[1].start, 3.1)
self.assertEqual(result[1].end, 4.0)
self.assertEqual(result[1].is_hyphen, None)
self.assertEqual(len(result), 4) # "removed " is removed
self.assertEqual(result[2].word, "Is split2")
self.assertEqual(result[2].start, 5.09)
self.assertEqual(result[2].end, 6.0)
self.assertEqual(result[2].is_hyphen, True)
self.assertEqual(result[3].word, "~ ")
self.assertEqual(result[3].start, 6.09)
self.assertEqual(result[3].end, 7.0)
self.assertEqual(result[3].is_hyphen, False)
if __name__ == "__main__":
unittest.main()
+19 -4
View File
@@ -6,7 +6,6 @@ from pydub import AudioSegment, silence
from modules.console_colors import ULTRASINGER_HEAD
from modules.Speech_Recognition.TranscribedData import TranscribedData
def remove_silence_from_transcription_data(audio_path: str, transcribed_data: list[TranscribedData]) -> list[
TranscribedData]:
"""Remove silence from given transcription data"""
@@ -53,7 +52,7 @@ def remove_silence2(silence_parts_list: list[tuple[float, float]], transcribed_d
# | ** ** | silence
# | ********** | data
# |0 1 2 3 4 5 6 | time
if silence_start > data.start and silence_end < origin_end:
if silence_start >= data.start and silence_end <= origin_end:
next_index = silence_parts_list.index((silence_start, silence_end)) + 1
if next_index < len(silence_parts_list) and silence_parts_list[next_index][0] < origin_end:
split_end = silence_parts_list[next_index][0]
@@ -73,13 +72,29 @@ def remove_silence2(silence_parts_list: list[tuple[float, float]], transcribed_d
if not was_split:
data.end = silence_start
if data.end - data.start < 0.1:
data.start = silence_end
data.end = split_end
continue
if split_data.end - split_data.start <= 0.1:
continue
data.is_hyphen = True
# Remove last whitespace from the data.word
if data.word[-1] == " ":
data.word = data.word[:-1]
was_split = True
new_transcribed_data.append(split_data)
if split_data.end - split_data.start > 0.1:
was_split = True
new_transcribed_data.append(split_data)
elif split_word == "~ " and data.is_hyphen:
if new_transcribed_data[-1].word[-1] != " ":
new_transcribed_data[-1].word += " "
new_transcribed_data[-1].is_hyphen = False
continue
# | **** | silence
+1 -1
View File
@@ -70,7 +70,7 @@ def create_ultrastar_txt_from_automation(
ultrastar_bpm = real_bpm_to_ultrastar_bpm(bpm)
multiplication = get_multiplier(ultrastar_bpm)
ultrastar_bpm = ultrastar_bpm * get_multiplier(ultrastar_bpm)
silence_split_duration = get_eighth_note_second(bpm)
silence_split_duration = get_sixteenth_note_second(bpm)
with open(ultrastar_file_output, "w", encoding=FILE_ENCODING) as file:
gap = transcribed_data[0].start