Spaces:
Running
Running
File size: 8,395 Bytes
5690e11 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 |
#
# Copyright (c) Microsoft. All rights reserved.
# Licensed under the MIT license. See LICENSE.md file in the project root for full license information.
#
from datetime import time
from typing import List, Optional, Tuple
from services.alinls.speech_process import AliRecognitionResult
import azure.cognitiveservices.speech as speechsdk
from services.audio.faster_whisper_recognition_service import FasterWhisperRecognitionResult
from services.audio.sensevoice_whisper_recognition_service import SenseVoiceRecognitionResult
from services.audio.tencent_recognition_service import TencentRecognitionResult
from services.captioning import helper
class Caption(object):
def __init__(self, language: Optional[str], sequence: int, begin: time, end: time, text: str):
self.language = language
self.sequence = sequence
self.begin = begin
self.end = end
self.text = text
def get_captions(language: Optional[str], max_width: int, max_height: int, results: List[dict]) -> List[Caption]:
caption_helper = CaptionHelper(language, max_width, max_height, results)
return caption_helper.get_captions()
class CaptionHelper(object):
def __init__(self, language: Optional[str], max_width: int, max_height: int,
results: List[object]):
self._language = language
self._max_width = max_width
self._max_height = max_height
self._results = results
self._first_pass_terminators = ["?", "!", ",", ";"]
self._second_pass_terminators = [" ", "."]
self._captions: List[Caption] = []
# consider adapting to use http://unicode.org/reports/tr29/#Sentence_Boundaries
if self._language is not None:
iso639 = self._language.split('-')[0]
if "zh" == iso639.lower():
self._first_pass_terminators = ["，", "、", "；", "？", "！", "?", "!", ",", ";"]
self._second_pass_terminators = ["。", " "]
if helper.DEFAULT_MAX_LINE_LENGTH_SBCS == self._max_width:
self._max_width = helper.DEFAULT_MAX_LINE_LENGTH_MBCS
def get_captions(self) -> List[Caption]:
self.ensure_captions()
return self._captions
def ensure_captions(self) -> None:
if not self._captions:
self.add_captions_for_all_results()
def add_captions_for_all_results(self) -> None:
for result in self._results:
if (hasattr(result, 'offset') and result.offset <= 0) or not self.is_final_result(result):
continue
text = self.get_text_or_translation(result)
if not text:
continue
self.add_captions_for_final_result(result, text)
def get_text_or_translation(self, result: object) -> Optional[str]:
return result.text
def add_captions_for_final_result(self, result: object, text: str) -> None:
caption_starts_at = 0
caption_lines: List[str] = []
index = 0
while index < len(text):
index = self.skip_skippable(text, index)
line_length = self.get_best_width(text, index)
caption_lines.append(text[index:index + line_length].strip())
index += line_length
is_last_caption = index >= len(text)
max_caption_lines = len(caption_lines) >= self._max_height
add_caption = is_last_caption or max_caption_lines
if add_caption:
caption_text = '\n'.join(caption_lines)
caption_lines.clear()
caption_sequence = len(self._captions) + 1
is_first_caption = 0 == caption_starts_at
caption_begin_and_end: Tuple[time, time]
if is_first_caption and is_last_caption:
caption_begin_and_end = self.get_full_caption_result_timing(result)
else:
caption_begin_and_end = self.get_partial_result_caption_timing(result, text, caption_text,
caption_starts_at,
index - caption_starts_at)
self._captions.append(
Caption(self._language, caption_sequence, caption_begin_and_end[0], caption_begin_and_end[1],
caption_text))
caption_starts_at = index
def get_best_width(self, text: str, start_index: int) -> int:
# print("get_best_width:",text)
remaining = len(text) - start_index
best_width = remaining if remaining < self._max_width else self.find_best_width(self._first_pass_terminators,
text, start_index)
if best_width < 0:
best_width = self.find_best_width(self._second_pass_terminators, text, start_index)
if best_width < 0:
best_width = self._max_width
# print("best_width",best_width)
return best_width
def find_best_width(self, terminators: List[str], text: str, start_at: int) -> int:
remaining = len(text) - start_at
check_chars = min(remaining, self._max_width)
best_width = -1
for terminator in terminators:
index = text.rfind(terminator, start_at, start_at + check_chars)
width = index - start_at
if width > best_width:
best_width = width + len(terminator)
return best_width
def skip_skippable(self, text: str, start_index: int) -> int:
index = start_index
while len(text) > index and ' ' == text[index]:
index += 1
return index
def get_full_caption_result_timing(self, result: object) -> Tuple[time, time]:
if isinstance(result, speechsdk.RecognitionResult):
begin = helper.time_from_ticks(result.offset)
end = helper.time_from_ticks(result.offset + result.duration)
return begin, end
if isinstance(result, AliRecognitionResult) or isinstance(result, TencentRecognitionResult):
begin = helper.time_from_milliseconds(result.begin_time)
end = helper.time_from_milliseconds(result.end_time)
return begin, end
if isinstance(result, FasterWhisperRecognitionResult):
begin = helper.time_from_seconds(result.begin_time)
end = helper.time_from_seconds(result.end_time)
return begin, end
if isinstance(result, SenseVoiceRecognitionResult):
begin = helper.time_from_seconds(result.begin_time)
end = helper.time_from_seconds(result.end_time)
return begin, end
def get_partial_result_caption_timing(self, result: object, text: str, caption_text: str,
caption_starts_at: int, caption_length: int) -> Tuple[time, time]:
(result_begin, result_end) = self.get_full_caption_result_timing(result)
result_duration = helper.subtract_times(result_end, result_begin)
text_length = len(text)
partial_begin = helper.add_time_and_timedelta(result_begin, result_duration * caption_starts_at / text_length)
partial_end = helper.add_time_and_timedelta(result_begin, result_duration * (
caption_starts_at + caption_length) / text_length)
return partial_begin, partial_end
def is_final_result(self, result: object) -> bool:
if isinstance(result, speechsdk.RecognitionResult):
return speechsdk.ResultReason.RecognizedSpeech == result.reason or speechsdk.ResultReason.RecognizedIntent == result.reason or speechsdk.ResultReason.TranslatedSpeech == result.reason
if isinstance(result, AliRecognitionResult) or isinstance(result, TencentRecognitionResult) or isinstance(result, FasterWhisperRecognitionResult) or isinstance(result, SenseVoiceRecognitionResult):
return True
def lines_from_text(self, text: str) -> List[str]:
retval: List[str] = []
index = 0
while index < len(text):
index = self.skip_skippable(text, index)
line_length = self.get_best_width(text, index)
retval.append(text[index:index + line_length].strip())
index += line_length
return retval
|