diff --git a/videotrans/__init__.py b/videotrans/__init__.py index 9ee98843..595de5b5 100644 --- a/videotrans/__init__.py +++ b/videotrans/__init__.py @@ -1,4 +1,4 @@ # -*- coding: utf-8 -*- -VERSION = "v2.71" -VERSION_NUM = 120071 +VERSION = "v2.72" +VERSION_NUM = 120072 diff --git a/videotrans/configure/config.py b/videotrans/configure/config.py index f36bc920..88497c20 100644 --- a/videotrans/configure/config.py +++ b/videotrans/configure/config.py @@ -536,7 +536,7 @@ def getset_params(obj=None): "tts_type": 0, # 所选的tts顺序 "split_type": "all", - "model_name": "tiny", # 模型名 + "model_name": "medium" if Path(ROOT_DIR+"/models/models--Systran--faster-whisper-medium/snapshots").is_dir() else "tiny", # 模型名 "recogn_type": 0, # 语音识别方式，数字代表显示顺序 "voice_autorate": False, diff --git a/videotrans/mainwin/_main_win.py b/videotrans/mainwin/_main_win.py index 38bcd4b0..cb6f7624 100644 --- a/videotrans/mainwin/_main_win.py +++ b/videotrans/mainwin/_main_win.py @@ -109,34 +109,17 @@ def bind_action(self): self.voice_role.setCurrentText(config.params['voice_role']) self.win_action.show_listen_btn(config.params['voice_role']) - # tts_type 改变时，重设角色 - self.tts_type.currentIndexChanged.connect(self.win_action.tts_type_change) - self.translate_type.currentIndexChanged.connect(self.win_action.set_translate_type) - self.voice_role.currentTextChanged.connect(self.win_action.show_listen_btn) - self.target_language.currentTextChanged.connect(self.win_action.set_voice_role) + self.model_name.addItems(config.WHISPER_MODEL_LIST) + if config.params['model_name'] in config.WHISPER_MODEL_LIST: + self.model_name.setCurrentText(config.params['model_name']) + + try: + config.params['recogn_type'] = int(config.params['recogn_type']) + except Exception: + config.params['recogn_type'] = 0 + + self.recogn_type.setCurrentIndex(config.params['recogn_type']) - self.set_line_role.clicked.connect(self.win_action.set_line_role_fun) - self.proxy.textChanged.connect(self.win_action.change_proxy) - self.import_sub.clicked.connect(self.win_action.import_sub_fun) - self.export_sub.clicked.connect(self.win_action.export_sub_fun) - self.startbtn.clicked.connect(self.win_action.check_start) - self.btn_save_dir.clicked.connect(self.win_action.get_save_dir) - self.btn_get_video.clicked.connect(self.win_action.get_mp4) - self.stop_djs.clicked.connect(self.win_action.reset_timeid) - self.continue_compos.clicked.connect(self.win_action.set_djs_timeout) - self.listen_btn.clicked.connect(self.win_action.listen_voice_fun) - self.split_type.currentIndexChanged.connect(self.win_action.check_split_type) - self.model_name.currentTextChanged.connect(self.win_action.check_model_name) - self.recogn_type.currentIndexChanged.connect(self.win_action.recogn_type_change) - self.voice_rate.valueChanged.connect(self.win_action.voice_rate_changed) - self.voice_autorate.stateChanged.connect( - lambda: self.win_action.autorate_changed(self.voice_autorate.isChecked(), "voice")) - self.video_autorate.stateChanged.connect( - lambda: self.win_action.autorate_changed(self.video_autorate.isChecked(), "video")) - self.append_video.stateChanged.connect( - lambda: self.win_action.autorate_changed(self.video_autorate.isChecked(), "append_video")) - self.addbackbtn.clicked.connect(self.win_action.get_background) - self.enable_cuda.toggled.connect(self.win_action.check_cuda) self.moshis = { "biaozhun_jd": self.action_xinshoujandan, @@ -144,6 +127,10 @@ def bind_action(self): "tiqu": self.action_tiquzimu } + w = self.size().width() + h = self.size().height() + self.move(QPoint(int((self.width - w) / 2), int((self.height - h) / 2))) + def _bindsignal(self): try: from videotrans.task.check_update import CheckUpdateWorker @@ -174,9 +161,7 @@ def _set_cache_set(self): self.stop_djs.setStyleSheet("""background-color:#148CD2;color:#ffffff""") self.proxy.setText(config.params['proxy']) self.continue_compos.setToolTip(config.transobj['Click to start the next step immediately']) - self.split_type.addItems([config.transobj['whisper_type_all'], config.transobj['whisper_type_avg']]) - self.model_name.addItems(config.WHISPER_MODEL_LIST) self.export_sub.setText(config.transobj['Export srt']) self.subtitle_type.addItems( [ @@ -188,16 +173,11 @@ def _set_cache_set(self): ]) self.subtitle_type.setCurrentIndex(config.params['subtitle_type']) - try: - config.params['recogn_type'] = int(config.params['recogn_type']) - except Exception: - config.params['recogn_type'] = 0 - - self.recogn_type.setCurrentIndex(config.params['recogn_type']) if config.params['recogn_type'] > 1: self.model_name_help.setVisible(False) else: self.model_name_help.clicked.connect(self.win_action.show_model_help) + try: config.params['tts_type'] = int(config.params['tts_type']) except Exception: @@ -210,14 +190,19 @@ def _set_cache_set(self): if config.params['subtitle_type'] and int(config.params['subtitle_type']) > 0: self.subtitle_type.setCurrentIndex(int(config.params['subtitle_type'])) - if config.params['model_name'] in config.WHISPER_MODEL_LIST: - self.model_name.setCurrentText(config.params['model_name']) - try: self.voice_rate.setValue(int(config.params['voice_rate'].replace('%', ''))) except Exception: self.voice_rate.setValue(0) + self.voice_autorate.stateChanged.connect( + lambda: self.win_action.autorate_changed(self.voice_autorate.isChecked(), "voice")) + self.video_autorate.stateChanged.connect( + lambda: self.win_action.autorate_changed(self.video_autorate.isChecked(), "video")) + self.append_video.stateChanged.connect( + lambda: self.win_action.autorate_changed(self.video_autorate.isChecked(), "append_video")) + self.addbackbtn.clicked.connect(self.win_action.get_background) + self.split_type.setDisabled(True if config.params['recogn_type'] > 0 else False) self.voice_autorate.setChecked(bool(config.params['voice_autorate'])) self.video_autorate.setChecked(bool(config.params['video_autorate'])) @@ -227,9 +212,28 @@ def _set_cache_set(self): self.only_video.setChecked(True if config.params['only_video'] else False) self.is_separate.setChecked(True if config.params['is_separate'] else False) - w=self.size().width() - h=self.size().height() - self.move(QPoint(int((self.width - w) / 2), int((self.height - h) / 2))) + self.enable_cuda.toggled.connect(self.win_action.check_cuda) + # tts_type 改变时，重设角色 + self.tts_type.currentIndexChanged.connect(self.win_action.tts_type_change) + self.translate_type.currentIndexChanged.connect(self.win_action.set_translate_type) + self.voice_role.currentTextChanged.connect(self.win_action.show_listen_btn) + self.target_language.currentTextChanged.connect(self.win_action.set_voice_role) + + self.set_line_role.clicked.connect(self.win_action.set_line_role_fun) + self.proxy.textChanged.connect(self.win_action.change_proxy) + self.import_sub.clicked.connect(self.win_action.import_sub_fun) + self.export_sub.clicked.connect(self.win_action.export_sub_fun) + self.startbtn.clicked.connect(self.win_action.check_start) + self.btn_save_dir.clicked.connect(self.win_action.get_save_dir) + self.btn_get_video.clicked.connect(self.win_action.get_mp4) + self.stop_djs.clicked.connect(self.win_action.reset_timeid) + self.continue_compos.clicked.connect(self.win_action.set_djs_timeout) + self.listen_btn.clicked.connect(self.win_action.listen_voice_fun) + self.split_type.currentIndexChanged.connect(self.win_action.check_split_type) + self.model_name.currentTextChanged.connect(self.win_action.check_model_name) + self.recogn_type.currentIndexChanged.connect(self.win_action.recogn_type_change) + self.voice_rate.valueChanged.connect(self.win_action.voice_rate_changed) + def start_subform(self): self.import_sub.setCursor(Qt.PointingHandCursor) diff --git a/videotrans/recognition/_base.py b/videotrans/recognition/_base.py index 6f7b0eba..439212bd 100644 --- a/videotrans/recognition/_base.py +++ b/videotrans/recognition/_base.py @@ -179,8 +179,24 @@ def re_segment_sentences(self, data): Returns: 重新划分后的字幕数据，格式与输入相同。 """ + flags=r'[,?!，。？！]|(\. )' + if self.detect_language[:2] in ['zh', 'ja', 'ko']: + maxlen =config.settings['cjk_len'] + flags=r'[,?!，。？！]|(\. )' + else: + maxlen = config.settings['other_len'] + shound_rephase=False + for segment in data: + if segment['words'][0]['end']-segment['words'][0]['start']>15000: + shound_rephase=True + break + if len(segment['text'])>3*maxlen: + shound_rephase=True + break + print([f"{t['text']}\n" for t in data]) + new_data = [] - if not config.settings['rephrase']: + if not config.settings['rephrase'] or not shound_rephase: for segment in data: tmp = { "line": len(new_data) + 1, @@ -205,17 +221,12 @@ def re_segment_sentences(self, data): sentence = "" sentence_start = data[0]["words"][0]['start'] sentence_end = 0 - flags=r'[,?!，。？！]|(\. )' - if self.detect_language[:2] in ['zh', 'ja', 'ko']: - maxlen =config.settings['cjk_len'] - flags=r'[,?!，。？！]|(\. )' - else: - maxlen = config.settings['other_len'] + print("需要分词") + data_len=len(data) for seg_i,segment in enumerate(data): current_len=len(segment["words"]) - # print(f'\n\n{segment["words"]=}') for i, word_info in enumerate(segment["words"]): word = word_info["word"] start = word_info["start"] @@ -248,15 +259,16 @@ def re_segment_sentences(self, data): next2_word='' - if ( next_word and re.search(r'[,?!，。！？]|(\. )',next_word) ) or ( next2_word and re.search(r'[,?!，。！？]|(\. )',next2_word) ): + if len(sentence.strip()) < 1.2*maxlen and ( \ + ( next_word and re.search(flags,next_word) and len(next_word)<0.2*maxlen ) \ + or ( next2_word and re.search(flags,next2_word) and len(next2_word)<0.2*maxlen ) \ + ): continue if next_start> end: if next_start >= end+1000: is_insert=True - elif next_start>=end+250 and len(sentence.strip())>=0.2*maxlen: - is_insert=True - elif next_start >= end+50 and re.search(flags, word) and len(sentence.strip())>=0.3*maxlen: + elif next_start>=end+200 and len(sentence.strip())>=0.2*maxlen: is_insert=True elif re.search(flags, word) and len(sentence.strip())>=maxlen*0.5: is_insert=True @@ -265,14 +277,14 @@ def re_segment_sentences(self, data): is_insert=True if not is_insert: - if self.subtitle_type>0 and len(sentence.strip())>=maxlen*2: + if self.subtitle_type>0 and len(sentence.strip())>=maxlen*1.5: is_insert=True - elif self.subtitle_type==0 and len(sentence.strip())>=maxlen*2.8: + elif self.subtitle_type==0 and len(sentence.strip())>=maxlen*2: is_insert=True if not is_insert: continue - # print(f'{sentence=}') + tmp = { "line": len(new_data) + 1, "start_time": sentence_start, @@ -300,7 +312,6 @@ def re_segment_sentences(self, data): tmp["endraw"]=tools.ms_to_time_string(ms=tmp["end_time"]) tmp['time'] = f'{tmp["startraw"]} --> {tmp["endraw"]}' new_data.append(tmp) - # print(f'\n\n%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%{new_data=}') return new_data # True 退出 diff --git a/videotrans/ui/downmodel.py b/videotrans/ui/downmodel.py index 88bef3ae..b0404ccd 100644 --- a/videotrans/ui/downmodel.py +++ b/videotrans/ui/downmodel.py @@ -31,7 +31,13 @@ def setupUi(self, downmodel): self.hlayout_name = QHBoxLayout() self.hlayout_name.setObjectName(u"hlayout_name") self.label_name = QLabel() + self.proxy=QtWidgets.QLineEdit() + self.proxy.setMinimumWidth(150) + self.proxy.setPlaceholderText('如果无法连接到GitHub，请填写代理地址' if config.defaulelang == 'zh' else "proxy address") + self.proxy.setToolTip('如果无法连接到GitHub，请填写代理地址' if config.defaulelang == 'zh' else "proxy address") self.hlayout_name.addWidget(self.label_name) + self.hlayout_name.addStretch() + self.hlayout_name.addWidget(self.proxy) self.hlayout_url = QHBoxLayout() self.hlayout_url.setObjectName(u"hlayout_url") @@ -46,11 +52,22 @@ def setupUi(self, downmodel): self.hlayout_btn = QHBoxLayout() self.hlayout_btn.setObjectName(u"hlayout_btn") + self.online_btn = QPushButton() + self.online_btn.setObjectName(u"online_btn") + self.online_btn.setMinimumSize(QSize(200, 35)) + self.online_btn.setCursor(QCursor(Qt.PointingHandCursor)) + self.online_btn.setMouseTracking(False) + self.down_btn = QPushButton() self.down_btn.setObjectName(u"down_btn") self.down_btn.setMinimumSize(QSize(200, 35)) self.down_btn.setCursor(QCursor(Qt.PointingHandCursor)) self.down_btn.setMouseTracking(False) + + + + + self.hlayout_btn.addWidget(self.online_btn) self.hlayout_btn.addWidget(self.down_btn) self.verticalLayout.addLayout(self.hlayout_btn) @@ -69,7 +86,7 @@ def setupUi(self, downmodel): self.text_help = QtWidgets.QPlainTextEdit() self.text_help.setReadOnly(True) - self.text_help.setMinimumSize(QSize(0, 150)) + self.text_help.setMinimumSize(QSize(0, 50)) self.verticalLayout.addWidget(self.text_help) self.retranslateUi(downmodel) @@ -79,3 +96,4 @@ def setupUi(self, downmodel): def retranslateUi(self, downmodel): downmodel.setWindowTitle("下载模型" if config.defaulelang == 'zh' else 'Download Models') self.down_btn.setText("点击打开浏览器下载" if config.defaulelang == 'zh' else 'Click to open browser to download') + self.online_btn.setText("在线下载模型" if config.defaulelang=='zh' else 'Download Model Online') diff --git a/videotrans/winform/fn_downmodel.py b/videotrans/winform/fn_downmodel.py index aa457a5c..58b23992 100644 --- a/videotrans/winform/fn_downmodel.py +++ b/videotrans/winform/fn_downmodel.py @@ -1,9 +1,128 @@ +import json +import os +from pathlib import Path + +import requests +from PySide6.QtCore import QThread, Signal +from PySide6.QtWidgets import QMessageBox +from py7zr import SevenZipFile + from videotrans.configure import config from videotrans.recognition import OPENAI_WHISPER, FASTER_WHISPER +from videotrans.util import tools +# 分块下载 +class DownloadPartThread(QThread): + uito = Signal(str) -# 视频 字幕 音频 合并 -from videotrans.util import tools + def __init__(self, url, save_dir, local_filename, start_range, end_range, part_num, proxy=None): + super().__init__() + self.url = url + self.save_dir = save_dir + self.local_filename = local_filename + self.start_range = start_range + self.end_range = end_range + self.part_num = part_num + self.proxy = proxy + + def run(self): + try: + headers = {'Range': f'bytes={self.start_range}-{self.end_range}'} + response = requests.get(self.url, headers=headers, stream=True, proxies=self.proxy) + progress = 0 + part_filename = f"{self.local_filename}.part{self.part_num}" + with open(part_filename, 'wb') as part_file: + for data in response.iter_content(1024): # 每次下载 1KB + part_file.write(data) + progress += len(data) + self.uito.emit(f'jindu:{self.part_num}:{progress}') + except Exception as e: + self.uito.emit(json.dumps({ + "type": "error", + "text": f"Part {self.part_num} : {str(e)}" + })) +# 下载线程，在此分块 4个线程下载 +class Down(QThread): + uito = Signal(str) + + def __init__(self, url, save_dir, proxy=None, parent=None): + super().__init__(parent=parent) + self.url = url + self.save_dir = save_dir + self.proxy = {"http": proxy, 'https': proxy} if proxy else None + self.total_size = 0 + self.progress_size = 0 + self.part_nums = {"0": 0, "1": 0, "2": 0, "3": 0} + + def post(self, d): + # 获取总进度 + if d.startswith("jindu:"): + d_split = d.split(':') + self.part_nums[d_split[1]] = int(d_split[2]) + if self.total_size > 0: + self.uito.emit(json.dumps({ + "type": "progress", + "text": f"{sum(self.part_nums.values()) / self.total_size * 100:.2f}%" + })) + return + self.uito.emit(d) + + def run(self): + try: + # 获取文件名 + local_filename = os.path.join(self.save_dir, self.url.split('/')[-1]) + # 获取文件大小 + # 通过 Range 请求来获取文件大小 + response = requests.get(self.url, headers={'Range': 'bytes=0-1'}, stream=True, proxies=self.proxy) + if 'Content-Range' not in response.headers: + self.post(json.dumps({"type":"error","text":"无法通过 Range 请求获取文件大小，服务器可能不支持" if config.defaulelang=='zh' else "Failed to get file size via Range request, server may not support"})) + return + total_size = int(response.headers['Content-Range'].split('/')[1]) + self.total_size = total_size + + # 计算每个线程下载的块大小 + num_threads = 4 + block_size = total_size // num_threads + + # 启动多个线程下载不同的部分 + threads = [] + for i in range(num_threads): + start_range = i * block_size + # 最后一块可能比其他块稍大 + end_range = (i + 1) * block_size - 1 if i < num_threads - 1 else total_size - 1 + thread = DownloadPartThread(self.url, self.save_dir, local_filename, start_range, end_range, i, self.proxy) + self.part_nums[str(i)] = 0 + thread.uito.connect(self.post) + threads.append(thread) + thread.start() + + # 等待所有线程下载完成 + for thread in threads: + thread.wait() + self.post(json.dumps({"type": 'progress', "text": '下载完毕，合并中' if config.defaulelang=='zh' else 'Download complete & Merger in progress'})) + # 合并所有下载的部分 + with open(local_filename, 'wb') as final_file: + for i in range(num_threads): + part_file = f"{local_filename}.part{i}" + with open(part_file, 'rb') as pf: + final_file.write(pf.read()) + try: + os.remove(part_file) # 删除部分文件 + except: + pass + + + # 下载完成后判断是否为 .7z 文件并解压 + if local_filename.endswith('.7z'): + self.post( + json.dumps({"type": 'extract', "text": ''})) + with SevenZipFile(local_filename, 'r') as archive: + archive.extractall(path=self.save_dir) + Path(local_filename).unlink(missing_ok=True) + + self.post(json.dumps({"type": 'end', "text": ''})) + except Exception as e: + self.post(json.dumps({"type": 'error', "text": str(e)})) def openwin(model_name=None, recogn_type=None): @@ -11,25 +130,54 @@ def openwin(model_name=None, recogn_type=None): return from videotrans.component import DownloadModelForm + + def feed(info): + info = json.loads(info) + if info['type'] == 'end': + winobj.online_btn.setText("下载完毕" if config.defaulelang == 'zh' else 'Download Complete') + winobj.online_btn.setDisabled(False) + elif info['type'] == 'error': + QMessageBox.critical(winobj, config.transobj['anerror'], info['text']) + winobj.online_btn.setText("下载失败" if config.defaulelang == 'zh' else 'Download fail') + winobj.online_btn.setDisabled(False) + elif info['type'] == 'extract': + winobj.online_btn.setText('正在解压' if config.defaulelang == 'zh' else 'Extracting') + else: + winobj.online_btn.setText(info['text']) + + def start_down(): + url = winobj.url.text() + print(f'{url=}') + winobj.online_btn.setDisabled(True) + winobj.online_btn.setText('开始下载模型...' if config.defaulelang=='zh' else 'Start downloading model...') + proxy = winobj.proxy.text().strip() + task = Down(url, config.ROOT_DIR + '/models', proxy if proxy else None, winobj) + task.uito.connect(feed) + task.start() + try: winobj = DownloadModelForm() config.child_forms['down_win'] = winobj if recogn_type == OPENAI_WHISPER: - name = f'OpenAI Whisper: {model_name}' + name = f'OpenAI Whisper: {model_name} {"模型" if config.defaulelang=="zh" else "Model" }' url = config.MODELS_DOWNLOAD['openai'][model_name] text_help = f'请下载 {model_name}.pt 后将该文件复制到 {config.ROOT_DIR}/models 文件夹内' if config.defaulelang == 'zh' else f'Please download {model_name}.pt and copy the file to {config.ROOT_DIR}/models folder.' else: - name = f'Faster Whisper: {model_name}' + name = f'Faster Whisper: {model_name} {"模型" if config.defaulelang=="zh" else "Model" }' url = config.MODELS_DOWNLOAD['faster'][model_name] zipname = url.split('/')[-1].replace('?download=true', '') folder_name = f'models--Systran--faster-whisper-{model_name}' if model_name.startswith('distil'): folder_name = f'models--Systran--faster-{model_name}' - text_help = f'请下载 {zipname} 后将该压缩包内的文件夹 {folder_name} 复制到 {config.ROOT_DIR}/models 文件夹内' if config.defaulelang == 'zh' else f'Please download {zipname}, open the zip file, and copy the folder {folder_name} into {config.ROOT_DIR}/models folder.' + text_help = f'如果在线下载失败，请点击打开浏览器下载，下载 {zipname} 后将该压缩包内的文件夹 {folder_name} 复制到 {config.ROOT_DIR}/models 文件夹内' if config.defaulelang == 'zh' else f'Please download {zipname}, open the zip file, and copy the folder {folder_name} into {config.ROOT_DIR}/models folder.' + winobj.label_name.setText(name) winobj.url.setText(url) winobj.text_help.setPlainText(text_help) - winobj.down_btn.clicked.connect(lambda:tools.open_url(url=url)) + winobj.down_btn.clicked.connect(lambda: tools.open_url(url=url)) + winobj.online_btn.clicked.connect(start_down) + if config.params['proxy']: + winobj.proxy.setText(config.params['proxy']) winobj.show() except Exception as e: print(e)