1. [add] 在输出表格中添加推荐字数;

detect_with_ocr.py

@@ -13,6 +13,9 @@ up_b, down_b = 0, 0
 # 初始化ocr工具
 ocr = PaddleOCR(use_angle_cls=True, lang="ch", show_log=False)
 
+# 正常语速为4字/秒
+normal_speed = 4
+
 
 def get_position(video_path, start_time):
     """
@@ -45,7 +48,7 @@ def get_position(video_path, start_time):
         log = []
         for x in res:
             rect, (txt, confidence) = x
-            font_size = rect[2][1] - rect[0][1]
+            # font_size = rect[2][1] - rect[0][1]
             mid = (rect[0][0] + rect[1][0]) / 2
             gradient = np.arctan(abs((rect[1][1] - rect[0][1]) / (rect[1][0] - rect[0][0])))
             # 可能是字幕的文本
@@ -156,7 +159,8 @@ def process_video(video_path, begin, end, state):
         cnt += 1
         # 每秒取4帧画面左右
         if cnt % int(fps / 4) == 0:
-            state[0] = float(cnt / video.get(cv2.CAP_PROP_FRAME_COUNT)) if state[0] < 0.99 else 0.99
+            state[0] = float((video.get(cv2.CAP_PROP_POS_MSEC) / 1000 - begin) / (end - begin)) \
+                if state[0] is None or state[0] < 0.99 else 0.99
             subTitle = detect_subtitle(frame)
             # 第一次找到字幕
             if lastSubTitle is None and subTitle is not None:
@@ -167,7 +171,9 @@ def process_video(video_path, begin, end, state):
                 res.append([start_time, end_time, lastSubTitle])
                 if len(res) == 1 or res[-1][0] - res[-2][1] >= 1:
                     print('--------------------------------------------------')
-                    narratage_recommend.append(['', '', '', '插入旁白'])
+                    recommend_lens = int(res[-1][0] * normal_speed) if len(res) == 1 else int(
+                        (res[-1][0] - res[-2][1]) * normal_speed)
+                    narratage_recommend.append(['', '', '', '插入旁白，推荐字数为%d' % recommend_lens])
                 print(start_time, end_time, lastSubTitle)
                 narratage_recommend.append([round(start_time, 2), round(end_time, 2), lastSubTitle, ''])
             # 两句话连在一起，但是两句话不一样
@@ -176,8 +182,10 @@ def process_video(video_path, begin, end, state):
                     end_time = video.get(cv2.CAP_PROP_POS_MSEC) / 1000
                     res.append([start_time, end_time, lastSubTitle])
                     if len(res) == 1 or res[-1][0] - res[-2][1] >= 1:
-                        narratage_recommend.append(['', '', '', '插入旁白'])
                         print('--------------------------------------------------')
+                        recommend_lens = int(res[-1][0] * normal_speed) if len(res) == 1 else int(
+                            (res[-1][0] - res[-2][1]) * normal_speed)
+                        narratage_recommend.append(['', '', '', '插入旁白，推荐字数为%d' % recommend_lens])
                     print(start_time, end_time, lastSubTitle)
                     narratage_recommend.append([round(start_time, 2), round(end_time, 2), lastSubTitle, ''])
                     start_time = end_time
@@ -189,7 +197,9 @@ def process_video(video_path, begin, end, state):
         if video.get(cv2.CAP_PROP_POS_MSEC) / 1000 > end:
             if video.get(cv2.CAP_PROP_POS_MSEC) / 1000 - end_time > 1:
                 print('--------------------------------------------------')
-                narratage_recommend.append(['', '', '', '插入旁白'])
+                recommend_lens = int(res[-1][0] * normal_speed) if len(res) == 1 else int(
+                    (res[-1][0] - res[-2][1]) * normal_speed)
+                narratage_recommend.append(['', '', '', '插入旁白，推荐字数为%d' % recommend_lens])
                 break
     return narratage_recommend
 
@@ -208,10 +218,12 @@ def write_excel_xlsx(path, sheet_name, value):
     sheet.title = sheet_name
     # 将字幕对应的那一列扩宽一些
     sheet.column_dimensions['C'].width = 50
+    sheet.column_dimensions['D'].width = 30
     for i in range(0, index):
         for j in range(0, len(value[i])):
             sheet.cell(row=i + 1, column=j + 1, value=str(value[i][j]))
-            if value[i][j] == '' or value[i][j] == '插入旁白' or value[i][j] == '翻译':
+            print(value[i][j])
+            if value[i][j] == '' or '插入旁白' in str(value[i][j]) or value[i][j] == '翻译':
                 sheet.cell(row=i + 1, column=j + 1).fill = PatternFill(fill_type='solid', fgColor='ffff00')
     workbook.save(path)
 
@@ -222,7 +234,7 @@ def detect_with_ocr(video_path, book_path, start_time, end_time, state):
 
     # 获取字幕在画面中的上下边界，方便在后续视频遍历过程中直接对字幕对应区域进行分析
     global up_b, down_b
-    up_b, down_b = get_position(video_path, start_time)
+    up_b, down_b = get_position(video_path, start_time + 30)
 
     # 获取并构建输出信息
     table_head = [["起始时间", "终止时间", "字幕", '建议', '解说脚本']]
@@ -236,4 +248,4 @@ def detect_with_ocr(video_path, book_path, start_time, end_time, state):
 if __name__ == '__main__':
     video_path = "D:/heelo/hysxm_1.mp4"
     book_path = '何以笙箫默.xlsx'
-    detect_with_ocr(video_path, book_path, 0, 300)
+    detect_with_ocr(video_path, book_path, 0, 300, [None])

judge_subtitle.py

@@ -64,6 +64,9 @@ def detect_movie(video_path, start, interval):
     """
     video = cv2.VideoCapture(video_path)
     fps = np.ceil(video.get(cv2.CAP_PROP_FPS))
+    end_time = video.get(cv2.CAP_PROP_FRAME_COUNT) / fps
+    if start + interval * 3 > end_time:
+        interval = int((end_time - start) / 3)
     start = start * fps
     interval = interval * fps
     random_number = 50

narratage_detection.py

@@ -18,7 +18,7 @@ def trans_to_seconds(timepoint):
     return time_in_seconds
 
 
-def detect(video_path, start_time, end_time, book_path, state):
+def detect(video_path, start_time, end_time, book_path, state, subtitle=None):
     print("开始检测")
     if book_path is None:
         book_path = os.path.basename(video_path).split('.')[0] + ".xlsx"
@@ -28,7 +28,14 @@ def detect(video_path, start_time, end_time, book_path, state):
     start_time = trans_to_seconds(start_time)
     end_time = trans_to_seconds(end_time)
 
-    has_subtitle = detect_movie(video_path, start_time, 60)
+    # 根据用户的选择来确定电影是否有字幕，如果“未知”，则自动检测
+    if subtitle == 0:
+        has_subtitle = detect_movie(video_path, start_time, 180)
+    elif subtitle == 1:
+        has_subtitle = True
+    else:
+        has_subtitle = False
+
     if has_subtitle:
         detect_with_ocr(video_path, book_path, start_time, end_time, state)
     else:

speech_synthesis.py

@@ -9,6 +9,9 @@ import openpyxl
 
 tmp_file = 'tmp.wav'
 
+normal_speed = 4
+normal_interval = 0.1
+
 
 def speech_synthesis(text, output_file, speed):
     """
@@ -38,7 +41,7 @@ def speech_synthesis(text, output_file, speed):
         change_speed(output_file, speed)
 
 
-def change_speed(wav_path, speed=1.5):
+def change_speed(wav_path, speed=1.0):
     """
     调整语速
     :param wav_path: 原音频路径
@@ -72,10 +75,11 @@ def read_sheet(book_path, sheet_name=None):
     return sheet_content
 
 
-def get_narratage_text(sheet_content):
+def get_narratage_text(sheet_content, speed):
     """
     根据从表格中获取到的内容，分析得到解说文本+对应开始时间
     :param sheet_content: dict，keys=["起始时间","终止时间","字幕","建议","解说脚本"]
+    :param speed: float, 旁白语速
     :return: narratage_text: list, 旁白文本，
              narratage_start_time: list, 旁白对应开始时间
     """
@@ -97,7 +101,7 @@ def get_narratage_text(sheet_content):
                 for x in text_split:
                     narratage_text.append(x)
                     narratage_start_time.append(cur_start)
-                    cur_start = cur_start + len(x) / 4.5 + 0.5
+                    cur_start = cur_start + (len(x) / normal_speed + normal_interval) / speed
     return narratage_text, narratage_start_time
 
 
@@ -152,7 +156,7 @@ def ss_and_export(sheet_path, output_dir, speed, caption_file, state):
 
     # 读取表格，并获取旁白及对应插入位置
     sheet_content = read_sheet(book_path)
-    narratages, start_timepoint = get_narratage_text(sheet_content)
+    narratages, start_timepoint = get_narratage_text(sheet_content, speed)
     export_caption(sheet_content, caption_file)
     print("已导出旁白文件")
 
@@ -164,6 +168,8 @@ def ss_and_export(sheet_path, output_dir, speed, caption_file, state):
         print("目前正在处理{}".format(wav_path))
         state[0] = float((i + 1) / len(narratages))
 
+    # 合成总音频，并入原视频音频中
+
     # 删除临时语音文件
     if os.path.exists(tmp_file):
         time.sleep(1)

split_wav.py

@@ -7,8 +7,6 @@ os.environ["CUDA_VISIBLE_DEVICES"] = "-1"
 import numpy as np
 from moviepy.editor import *
 from pydub import AudioSegment
-from spleeter.audio.adapter import AudioAdapter
-from spleeter.separator import Separator
 
 audio_path = r'D:\AddCaption\disappearedBullet.wav'
 folder_name = r'D:\AddCaption\hysxm'
@@ -55,6 +53,8 @@ def split_audio():
 
 # 从音频中提取人声
 def extrac_speech():
+    from spleeter.audio.adapter import AudioAdapter
+    from spleeter.separator import Separator
     separator = Separator('spleeter:2stems', multiprocess=False)
     audio_adapter = AudioAdapter.default()
 

try_with_gui.py

@@ -18,8 +18,10 @@ def open_video_file():
     打开文件
     :return:
     """
-    video_path = filedialog.askopenfilename(title=u'选择文件', initialdir=(os.path.expanduser('D:/heelo')))
-    if len(video_path) != 0 and is_video(video_path):
+    video_path = filedialog.askopenfilename(title=u'选择文件',
+                                            filetype=[("视频文件", ".avi"), ("视频文件", ".mp4"), ("视频文件", ".rmvb"),
+                                                      ("视频文件", ".mkv")])
+    if len(video_path) != 0:
         print('打开文件：', video_path)
         inputFilePath.set(video_path)
         # 获取视频的时长等信息，初始化开始结束时间
@@ -37,8 +39,6 @@ def open_video_file():
                 k_l = str.lower(k)
                 if 'duration' in k_l:
                     endTime.set(vs['tags'][k])
-    if len(video_path) != 0 and not is_video(video_path):
-        messagebox.showwarning('警告', "请选择正确的视频格式，能够处理的视频格式如下所示：\n'.mkv', '.rmvb', '.mp4', '.avi'")
 
 
 def find_save_file():
@@ -54,26 +54,12 @@ def find_save_file():
     book_path = filedialog.asksaveasfilename(title=u'保存文件至',
                                              initialdir=os.path.dirname(video_path),
                                              initialfile=defaultName,
-                                             filetype=[('excel文件', ".xlsx")])
-    if len(book_path) != 0 and ".xlsx" not in book_path:
-        book_path += ".xlsx"
+                                             filetype=[('Excel工作簿', ".xlsx"), ('Excel 97工作簿', ".xls")],
+                                             defaultextension='.xlsx')
     print('保存文件至：', book_path)
     outputFilePath.set(book_path)
 
 
-def is_video(filename):
-    """
-    检查视频格式是否正确
-    :param filename: 视频文件名
-    :return: True or False
-    """
-    video_tag = ['.mkv', '.rmvb', '.mp4', '.avi']
-    for x in video_tag:
-        if x in filename:
-            return True
-    return False
-
-
 def trans_to_seconds(timepoint):
     time_in_seconds = 0
     timepoint = timepoint.split(':')
@@ -120,9 +106,6 @@ def start_detect():
     if len(video_path) == 0:
         messagebox.showwarning('警告', "请输入视频文件路径")
         return
-    elif not is_video(video_path):
-        messagebox.showwarning('警告', "视频文件格式错误，请重新确认")
-        return
     if len(book_path) == 0:
         messagebox.showwarning("警告", "请输入表格存放路径")
         return
@@ -142,7 +125,8 @@ def start_detect():
     state = [None]
     threads = [
         threading.Thread(target=start_process, args=(progressbar_1, progress_1, state, 100000), name="startProgress1"),
-        threading.Thread(target=detect, args=(video_path, startTime.get(), endTime.get(), book_path, state),
+        threading.Thread(target=detect,
+                         args=(video_path, startTime.get(), endTime.get(), book_path, state, hasSubtitle.get()),
                          name="detect")]
     for t in threads:
         t.start()
@@ -179,8 +163,8 @@ def open_sheet_file():
     :return:
     """
     sheet_path = filedialog.askopenfilename(title=u'选择文件',
-                                            initialdir=(os.path.expanduser('D:/AddCaption/test/detectSubtitle/电影旁白打标')))
-    if len(sheet_path) != 0 and ".xlsx" in sheet_path:
+                                            filetype=[('Excel工作簿', ".xlsx"), ('Excel 97工作簿', ".xls")], )
+    if len(sheet_path) != 0:
         print("打开表格", sheet_path)
         narratagePath.set(sheet_path)
 
@@ -209,6 +193,14 @@ def set_caption_file():
     captionPath.set(caption_path)
 
 
+def confirm_video_path():
+    # 仅能打开mp4\rmvb\avi\mkv格式的文件
+    video_path = filedialog.askopenfilename(title=u'选择文件',
+                                            filetype=[("视频文件", ".avi"), ("视频文件", ".mp4"), ("视频文件", ".rmvb"),
+                                                      ("视频文件", ".mkv")])
+    videoPath.set(video_path)
+
+
 def start_synthesis():
     """
     开始合成语音
@@ -216,7 +208,7 @@ def start_synthesis():
     """
     audio_dir = audioDir.get()
     sheet_path = narratagePath.get()
-    speed = audio_speed.get()
+    speed = float(audio_speed.get().split('(')[0])
     caption_path = captionPath.get()
 
     # 判断各个变量的合理性
@@ -241,9 +233,10 @@ def start_synthesis():
 
     # 多线程同时实现语音合成+字幕导出、进度条
     state = [None]
-    threads = [threading.Thread(target=start_process, args=(progressbar_2, progress_2, state), name="startProgress2"),
-               threading.Thread(target=ss_and_export,
-                                args=(sheet_path, audio_dir, speed, caption_path, state), name="ssAndExport")]
+    threads = [
+        threading.Thread(target=start_process, args=(progressbar_2, progress_2, state, 100000), name="startProgress2"),
+        threading.Thread(target=ss_and_export,
+                         args=(sheet_path, audio_dir, speed, caption_path, state), name="ssAndExport")]
     for t in threads:
         t.start()
     for t in threads:
@@ -343,6 +336,16 @@ endTime = tk.StringVar(value="23:59:59")
 endTime_entered = ttk.Entry(video_info, width=11, textvariable=endTime)
 endTime_entered.grid(column=1, row=2, sticky="W")
 
+subtitle_label = ttk.Label(video_info, text="是否有字幕")
+subtitle_label.grid(column=0, row=3)
+hasSubtitle = tk.IntVar()
+choice_1 = tk.Radiobutton(video_info, text="未知", variable=hasSubtitle, value=0)
+choice_1.grid(column=1, row=3, sticky="W")
+choice_2 = tk.Radiobutton(video_info, text="是", variable=hasSubtitle, value=1)
+choice_2.grid(column=2, row=3, sticky="W")
+choice_3 = tk.Radiobutton(video_info, text="否", variable=hasSubtitle, value=2)
+choice_3.grid(column=3, row=3, sticky="W")
+
 """
     检测步骤相关内容，包含以下内容：
     - 输出表格路径|输出表格路径文本框|打开文件夹
@@ -378,37 +381,47 @@ stopDetection.config(state=tk.DISABLED)
 """
     语音相关设置，包含以下内容：
     - 旁白脚本表格|表格路径|上传文件按钮
+    - 原视频|视频路径|上传文件按钮
+    - 旁白语速选择
 """
 audio_info = ttk.LabelFrame(tab2, text=" 语音相关设置 ")
 audio_info.place(relx=0.05, rely=0.05, relwidth=0.9, relheight=0.4)
 
 narratage_label = ttk.Label(audio_info, text="旁白脚本表格")
-narratage_label.grid(column=0, row=0)
+narratage_label.grid(column=0, row=1)
 narratagePath = tk.StringVar()
 narratagePath_input = ttk.Entry(audio_info, width=30, textvariable=narratagePath)
-narratagePath_input.grid(column=1, row=0)
+narratagePath_input.grid(column=1, row=1)
 upload_button_2 = ttk.Button(audio_info, text="上传文件", command=open_sheet_file)
-upload_button_2.grid(column=2, row=0)
+upload_button_2.grid(column=2, row=1)
 
 speed_label = ttk.Label(audio_info, text="旁白语速")
-speed_label.grid(column=0, row=1)
+speed_label.grid(column=0, row=2)
 audio_speed = tk.StringVar()
 speedChosen = ttk.Combobox(audio_info, width=12, textvariable=audio_speed)
-speedChosen['values'] = ("1.00", "1.10", "1.25", "1.50", "1.75", "2.00", "2.50")
+speedChosen['values'] = (
+    "1.00(4字/秒)", "1.10(4.5字/秒)", "1.25(5字/秒)", "1.50(6字/秒)", "1.75(7字/秒)", "2.00(8字/秒)", "2.50(10字/秒)")
 speedChosen.current(0)
-speedChosen.grid(column=1, row=1, sticky="W")
+speedChosen.grid(column=1, row=2, sticky="W")
 
+video_label = ttk.Label(audio_info, text="原视频")
+video_label.grid(column=0, row=0)
+videoPath = tk.StringVar()
+videoPath_input = ttk.Entry(audio_info, width=30, textvariable=videoPath)
+videoPath_input.grid(column=1, row=0)
+upload_button_3 = ttk.Button(audio_info, text="上传文件", command=confirm_video_path)
+upload_button_3.grid(column=2, row=0)
+
+"""
+    语音合成步骤，包含以下内容：
+    - 输出音频存放于|路径文本框|打开文件夹
+    - 输出字幕文件|路径文本框|打开文件夹
+    - 开始合成按钮|当前检测状态提示文本|任务进度条|进度条百分比
+    - 停止合成按钮
+"""
 synthesis_command = ttk.LabelFrame(tab2, text=" 语音合成步骤 ")
 synthesis_command.place(relx=0.05, rely=0.55, relwidth=0.9, relheight=0.4)
 
-caption_label = ttk.Label(synthesis_command, text="输出字幕文件")
-caption_label.grid(column=0, row=1)
-captionPath = tk.StringVar()
-captionPath_input = ttk.Entry(synthesis_command, width=30, textvariable=captionPath)
-captionPath_input.grid(column=1, row=1)
-save_button_2 = ttk.Button(synthesis_command, text="打开文件夹", command=set_caption_file)
-save_button_2.grid(column=2, row=1)
-
 audioDir_label = ttk.Label(synthesis_command, text="输出音频存放于")
 audioDir_label.grid(column=0, row=0)
 audioDir = tk.StringVar()
@@ -417,6 +430,14 @@ audioDir_input.grid(column=1, row=0)
 save_button_2 = ttk.Button(synthesis_command, text="打开文件夹", command=find_save_dir)
 save_button_2.grid(column=2, row=0)
 
+caption_label = ttk.Label(synthesis_command, text="输出字幕文件")
+caption_label.grid(column=0, row=1)
+captionPath = tk.StringVar()
+captionPath_input = ttk.Entry(synthesis_command, width=30, textvariable=captionPath)
+captionPath_input.grid(column=1, row=1)
+save_button_2 = ttk.Button(synthesis_command, text="打开文件夹", command=set_caption_file)
+save_button_2.grid(column=2, row=1)
+
 startSynthesis = ttk.Button(synthesis_command, text="开始合成",
                             command=lambda: thread_it(start_synthesis, name="startSynthesis"))
 startSynthesis.grid(column=0, row=2)