尊重百度的api语音合成规则
屏幕显示字幕内容与形式 和 字幕的播放时长分离
： 去除标点符号的影响
# 设置分句的标志符号；可以根据实际需要进行修改
# cutlist = "。！？".decode('utf-8')
cutlist = ['\n', '\t', '。', '；', '？', '.', ';', '?', '...', '、、、', '：']
cutlist = ['\n', '\t', '。', '；', '？', '：']
# cutlist = [ '。', '；', '？', '.', ';', '?', '...', '、、、','：',':','，']
# cutlist = [ '。', '；', '？', '.', ';', '?', '...', '、、、',':','，','、']
# 检查某字符是否分句标志符号的函数；如果是，返回True，否则返回False
def FindToken(cutlist, char):
if char in cutlist:
return True
else:
return False
# 进行分句的核心函数
def Cut(cutlist, lines): # 参数1：引用分句标志符；参数2：被分句的文本，为一行中文字符
l = [] # 句子列表，用于存储单个分句成功后的整句内容，为函数的返回值
line = [] # 临时列表，用于存储捕获到分句标志符之前的每个字符，一旦发现分句符号后，就会将其内容全部赋给l，然后就会被清空
for i in lines: # 对函数参数2中的每一字符逐个进行检查 （本函数中，如果将if和else对换一下位置，会更好懂）
if FindToken(cutlist, i): # 如果当前字符是分句符号
line.append(i) # 将此字符放入临时列表中
l.append(''.join(line)) # 并把当前临时列表的内容加入到句子列表中
line = [] # 将符号列表清空，以便下次分句使用
else: # 如果当前字符不是分句符号，则将该字符直接放入临时列表中
line.append(i)
return l
r_s = []
# 以下为调用上述函数实现从文本文件中读取内容并进行分句。
# with open('mybaidu.parp.b.txt','r',encoding='utf-8') as fr :
# for lines in fr:
# l = Cut(list(cutlist), list(lines))
# for line in l:
# if len(line.replace(' ', '')) == 0:
# continue
# if line.strip() != "":
# line=line.strip()
# r_s.append(line)
#
# # li = line.strip().split()
# # for sentence in li:
# # r_s.append(sentence)
str_ = ''
# cutlist = [ '。', '；', '？', '.', ';', '?', '...', '、、、','：',':','，','\n']
with open('mybaidu.parp.b.txt', 'r', encoding='utf-8') as fr:
for lines in fr:
if len(lines.replace(' ', '')) == 0:
continue
# str_='{}{}'.format(str_,lines.replace('\n',''))
# if len(lines.replace(' ','').replace('\n',''))==0:
# continue
str_ = '{}{}'.format(str_, lines)
# l = Cut(list(cutlist), list(lines))
# for line in l:
# if line.strip() != "":
# line=line.strip()
from aip import AipSpeech
bd_k_l = ['11059852', '5Kk01GtG2fjCwpzEkwdn0mjw', 'bp6Wyx377Elq7RsCQZzTBgGUFzLm8G2A']
APP_ID, API_KEY, SECRET_KEY = bd_k_l
mp3_dir = 'C:\\Users\\sas\\PycharmProjects\\produce_video\\result_liukeyun\\'
client = AipSpeech(APP_ID, API_KEY, SECRET_KEY)
# result = client.synthesis(str_, 'zh', 1, {
# 'vol': 5,
# })
uid = 'liukeyuanCAKE_whole_para'
# 识别正确返回语音二进制 错误则返回dict 参照下面错误码
f_w = '{}{}{}{}{}'.format(mp3_dir, 'g3db', uid, 'g3uid', '.mp3')
#
# if not isinstance(result, dict):
# # f_w = '{}{}{}{}'.format(mp3_dir, 'g3uid', uid, '.mp3')
# f_w = '{}{}{}{}{}'.format(mp3_dir, 'g3db', uid, 'g3uid', '.mp3')
# # ,'g3db',uid,'g3uid'
# # with open('auido.b.mp3', 'wb') as f:
# with open(f_w, 'wb') as f:
# f.write(result)
sentence_l, sentence_l_chk = Cut(list(cutlist), list(str_)), []
for i in sentence_l:
chk_br = i.replace('\n', ' ')
# del sentence_l[sentence_l.index(i)]
if len(chk_br.replace(' ', '')) > 0:
sentence_l_chk.append(chk_br.replace(' ', ''))
bdmp3filter_l = ['：', '——', ',', '《', '》', '“', '”', '、', '（', '）', '.', ' '] # 注意空格
# 保留其他标点符号
mp3_str = ' '.join(sentence_l_chk)
mp3_str_bdmp3filter = mp3_str
for i in bdmp3filter_l:
mp3_str_bdmp3filter = mp3_str_bdmp3filter.replace(i, '')
import os, time, glob
import cv2
os_sep = os.sep
this_file_abspath = os.path.abspath(__file__)
this_file_dirname, this_file_name = os.path.dirname(this_file_abspath), os.path.abspath(__file__).split(os_sep)[
-1]
f_img_d = '{}{}{}{}{}'.format(this_file_dirname, os_sep, 'mypng', os_sep, '*.jpg')
imgs, img_size_d = glob.glob(f_img_d), {}
for i in imgs:
img = cv2.imread(i)
w_h_s = '{},{}'.format(img.shape[1], img.shape[0])
if w_h_s not in img_size_d:
img_size_d[w_h_s] = 1
else:
img_size_d[w_h_s] += 1
mode_img_size_wh = [int(i) for i in
sorted(img_size_d.items(), key=lambda mytuple: mytuple[1], reverse=True)[0][0].split(',')]
os_sep = os.sep
this_file_abspath = os.path.abspath(__file__)
this_file_dirname, this_file_name = os.path.dirname(this_file_abspath), os.path.abspath(__file__).split(os_sep)[-1]
import time, math
this_time = time.time()
import imageio
imageio.plugins.ffmpeg.download()
from moviepy.editor import VideoFileClip
f_mp3 = 'g3dbCAKEg3uidnoBRBlankLine.09.mp3'
import mutagen.id3
from mutagen.easyid3 import EasyID3
from mutagen.mp3 import MP3
EasyID3.valid_keys["comment"] = "COMM::'XXX'"
id3info = MP3(f_mp3, ID3=EasyID3)
t_spend = id3info.info.length
import cv2
import glob
'''
python+opencv视频图像相互转换 - CSDN博客 https://blog.csdn.net/m0_37733057/article/details/79023693
链接：https://www.zhihu.com/question/49558804/answer/343058915
OpenCV: Drawing Functions in OpenCV https://docs.opencv.org/3.1.0/dc/da5/tutorial_py_drawing_functions.html
'''
# 每秒传输帧数(Frames Per Second)
fps = 100 # 保存视频的FPS，可以适当调整 FPS是图像领域中的定义，是指画面每秒传输帧数，通俗来讲就是指动画或视频的画面数。FPS是测量用于保存、显示动态视频的信息数量。每秒钟帧数愈多，所显示的动作就会愈流畅。通常，要避免动作不流畅的最低是30。某些计算机视频格式，每秒只能提供15帧。
f_img_d = '{}{}{}{}{}'.format(this_file_dirname, os_sep, 'mypng', os_sep, '*.jpg')
imgs = glob.glob(f_img_d)
"""
用图片总数均分音频时间
"""
def resize_rescale_pilimg(img_f, w_h_tuple=(mode_img_size_wh[0], mode_img_size_wh[1]), mid_factor=1):
# print(img_f)
img_n, img_type = img_f.split('.')[-2], img_f.split('.')[-1]
# print(img_n)
img_n_resize_rescale_pilimg_dir = '{}{}{}'.format(os_sep.join(img_n.split(os_sep)[:-1]), 'resize_rescale_pilimg',
os_sep, img_n.split(os_sep)[-1], os_sep)
img_n_resize_rescale_pilimg = '{}{}{}'.format(img_n_resize_rescale_pilimg_dir, img_n.split(os_sep)[-1], '.PNG')
# print(img_n_resize_rescale_pilimg)
img_type = 'PNG'
# img_f_new = '{}{}{}{}'.format(img_n, int(time.time()), 'resize_rescale.', img_type)
img_f_new = img_n_resize_rescale_pilimg
mid_icon = Image.open(img_f)
mid_icon_w, mid_icon_h = w_h_tuple[0] * mid_factor, w_h_tuple[1] * mid_factor
mid_icon = mid_icon.resize((mid_icon_w, mid_icon_h), Image.ANTIALIAS)
mid_icon.save(img_n_resize_rescale_pilimg, img_type)
return img_f_new
from PIL import Image, ImageDraw, ImageFont
# myfont = ImageFont.truetype("simhei.ttf", 50, encoding="utf-8")
myfont = ImageFont.truetype("simhei.ttf", encoding="utf-8")
import cv2
import numpy as np
equal_str_l = []
le = len(mp3_str)
br_step = 34
br_times = math.ceil(le / br_step)
for i_br_loop in range(br_times):
s_p = mp3_str[i_br_loop * br_step:i_br_loop * br_step + br_step]
equal_str_l.append(s_p)
l = equal_str_l
char_loop_l_len = len(mp3_str) / len(equal_str_l)
char_loop_l_len = len(mp3_str_bdmp3filter)
'''
bdmp3filter_l = ['：', '——', ',', '《', '》', '“', '”', '、', '（', '）', '.', ' '] # 注意空格
# 保留其他标点符号
mp3_str = ' '.join(sentence_l_chk)
mp3_str_bdmp3filter = mp3_str
for i in bdmp3filter_l:
mp3_str_bdmp3filter = mp3_str_bdmp3filter.replace(i, '')
'''
def gen_video(os_delay_factor=0.245, mystep=0.01, bear_error_second=1, audio_spend=t_spend, step_para=1):
f_v = '{}{}{}{}{}{}{}'.format('D:\\myv\\', 'g3db', uid, 'g3uid', uid, int(time.time()), 'saveVideo.b.avi')
fps, fourcc = 15, cv2.VideoWriter_fourcc('M', 'J', 'P', 'G')
videoWriter = cv2.VideoWriter(f_v, fourcc, fps, (mode_img_size_wh[0], mode_img_size_wh[1]))
for i in l:
print(i)
i_index = l.index(i)
img_index = i_index % len(imgs)
imgname = imgs[img_index]
mystr, le = i, len(i)
this_time = time.time()
frame = cv2.imread(imgname)
if (frame.shape[1], frame.shape[0]) != (mode_img_size_wh[0], mode_img_size_wh[1]):
imgname = resize_rescale_pilimg(imgname)
frame = cv2.imread(imgname)
else:
pass
frame_cv2 = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
frame_pil = Image.fromarray(frame_cv2) # 转为PIL的图片格式
# font = ImageFont.truetype("simhei.ttf", 50, encoding="utf-8")
font = ImageFont.truetype("simhei.ttf", encoding="utf-8")
# ImageDraw.Draw(frame_pil).text((100, 20), mystr, (0, 0, 255), font)
f_x, f_y = math.floor((mode_img_size_wh[0]) * 0.15), math.floor(mode_img_size_wh[1] * 0.85)
# ImageDraw.Draw(frame_pil).text((30, mode_img_size_wh[1]-30), mystr, (0, 0, 255), font)
ImageDraw.Draw(frame_pil).text((f_x, f_y), mystr, (0, 0, 255), font)
frame_cv2 = cv2.cvtColor(np.array(frame_pil), cv2.COLOR_RGB2BGR)
img = frame_cv2
line_bdmp3filter = i
for bdmp3filter in bdmp3filter_l:
line_bdmp3filter = line_bdmp3filter.replace(bdmp3filter, '')
myinterval = t_spend / (char_loop_l_len * 1) * os_delay_factor * len(line_bdmp3filter)
while time.time() - this_time < myinterval:
videoWriter.write(img)
videoWriter.release()
time.sleep(1)
video_playtime = VideoFileClip(f_v).duration
if video_playtime - audio_spend > bear_error_second:
# os_delay_factor -= mystep
os_delay_factor *= t_spend / video_playtime
gen_video(os_delay_factor=os_delay_factor, mystep=0.005, audio_spend=t_spend)
elif audio_spend - video_playtime > bear_error_second:
# os_delay_factor += mystep
os_delay_factor *= t_spend / video_playtime
gen_video(os_delay_factor=os_delay_factor, mystep=0.005, audio_spend=t_spend)
else:
os._exit(123)
'''
326
'''
mycom = 0.9 * 335 / 1801
gen_video(os_delay_factor=0.48, mystep=0.003, bear_error_second=0.5, audio_spend=t_spend)
# 设置分句的标志符号；可以根据实际需要进行修改
# cutlist = "。！？".decode('utf-8')
cutlist = ['\n', '\t', '。', '；', '？', '.', ';', '?', '...', '、、、', '：']
cutlist = ['\n', '\t', '。', '；', '？', '：']
# cutlist = [ '。', '；', '？', '.', ';', '?', '...', '、、、','：',':','，']
# cutlist = [ '。', '；', '？', '.', ';', '?', '...', '、、、',':','，','、']
# 检查某字符是否分句标志符号的函数；如果是，返回True，否则返回False
def FindToken(cutlist, char):
if char in cutlist:
return True
else:
return False
# 进行分句的核心函数
def Cut(cutlist, lines): # 参数1：引用分句标志符；参数2：被分句的文本，为一行中文字符
l = [] # 句子列表，用于存储单个分句成功后的整句内容，为函数的返回值
line = [] # 临时列表，用于存储捕获到分句标志符之前的每个字符，一旦发现分句符号后，就会将其内容全部赋给l，然后就会被清空
for i in lines: # 对函数参数2中的每一字符逐个进行检查 （本函数中，如果将if和else对换一下位置，会更好懂）
if FindToken(cutlist, i): # 如果当前字符是分句符号
line.append(i) # 将此字符放入临时列表中
l.append(''.join(line)) # 并把当前临时列表的内容加入到句子列表中
line = [] # 将符号列表清空，以便下次分句使用
else: # 如果当前字符不是分句符号，则将该字符直接放入临时列表中
line.append(i)
return l
r_s = []
# 以下为调用上述函数实现从文本文件中读取内容并进行分句。
# with open('mybaidu.parp.b.txt','r',encoding='utf-8') as fr :
# for lines in fr:
# l = Cut(list(cutlist), list(lines))
# for line in l:
# if len(line.replace(' ', '')) == 0:
# continue
# if line.strip() != "":
# line=line.strip()
# r_s.append(line)
#
# # li = line.strip().split()
# # for sentence in li:
# # r_s.append(sentence)
str_ = ''
# cutlist = [ '。', '；', '？', '.', ';', '?', '...', '、、、','：',':','，','\n']
with open('mybaidu.parp.b.txt', 'r', encoding='utf-8') as fr:
for lines in fr:
if len(lines.replace(' ', '')) == 0:
continue
# str_='{}{}'.format(str_,lines.replace('\n',''))
# if len(lines.replace(' ','').replace('\n',''))==0:
# continue
str_ = '{}{}'.format(str_, lines)
# l = Cut(list(cutlist), list(lines))
# for line in l:
# if line.strip() != "":
# line=line.strip()
from aip import AipSpeech
bd_k_l = ['11059852', '5Kk01GtG2fjCwpzEkwdn0mjw', 'bp6Wyx377Elq7RsCQZzTBgGUFzLm8G2A']
APP_ID, API_KEY, SECRET_KEY = bd_k_l
mp3_dir = 'C:\\Users\\sas\\PycharmProjects\\produce_video\\result_liukeyun\\'
client = AipSpeech(APP_ID, API_KEY, SECRET_KEY)
# result = client.synthesis(str_, 'zh', 1, {
# 'vol': 5,
# })
uid = 'liukeyuanCAKE_whole_para'
# 识别正确返回语音二进制 错误则返回dict 参照下面错误码
f_w = '{}{}{}{}{}'.format(mp3_dir, 'g3db', uid, 'g3uid', '.mp3')
#
# if not isinstance(result, dict):
# # f_w = '{}{}{}{}'.format(mp3_dir, 'g3uid', uid, '.mp3')
# f_w = '{}{}{}{}{}'.format(mp3_dir, 'g3db', uid, 'g3uid', '.mp3')
# # ,'g3db',uid,'g3uid'
# # with open('auido.b.mp3', 'wb') as f:
# with open(f_w, 'wb') as f:
# f.write(result)
sentence_l, sentence_l_chk = Cut(list(cutlist), list(str_)), []
for i in sentence_l:
chk_br = i.replace('\n', ' ')
# del sentence_l[sentence_l.index(i)]
if len(chk_br.replace(' ', '')) > 0:
sentence_l_chk.append(chk_br.replace(' ', ''))
bdmp3filter_l = ['：', '——', ',', '《', '》', '“', '”', '、', '（', '）', '.', ' '] # 注意空格
# 保留其他标点符号
mp3_str = ' '.join(sentence_l_chk)
mp3_str_bdmp3filter = mp3_str
for i in bdmp3filter_l:
mp3_str_bdmp3filter = mp3_str_bdmp3filter.replace(i, '')
import os, time, glob
import cv2
os_sep = os.sep
this_file_abspath = os.path.abspath(__file__)
this_file_dirname, this_file_name = os.path.dirname(this_file_abspath), os.path.abspath(__file__).split(os_sep)[
-1]
f_img_d = '{}{}{}{}{}'.format(this_file_dirname, os_sep, 'mypng', os_sep, '*.jpg')
imgs, img_size_d = glob.glob(f_img_d), {}
for i in imgs:
img = cv2.imread(i)
w_h_s = '{},{}'.format(img.shape[1], img.shape[0])
if w_h_s not in img_size_d:
img_size_d[w_h_s] = 1
else:
img_size_d[w_h_s] += 1
mode_img_size_wh = [int(i) for i in
sorted(img_size_d.items(), key=lambda mytuple: mytuple[1], reverse=True)[0][0].split(',')]
os_sep = os.sep
this_file_abspath = os.path.abspath(__file__)
this_file_dirname, this_file_name = os.path.dirname(this_file_abspath), os.path.abspath(__file__).split(os_sep)[-1]
import time, math
this_time = time.time()
import imageio
imageio.plugins.ffmpeg.download()
from moviepy.editor import VideoFileClip
f_mp3 = 'g3dbCAKEg3uidnoBRBlankLine.09.mp3'
import mutagen.id3
from mutagen.easyid3 import EasyID3
from mutagen.mp3 import MP3
EasyID3.valid_keys["comment"] = "COMM::'XXX'"
id3info = MP3(f_mp3, ID3=EasyID3)
t_spend = id3info.info.length
import cv2
import glob
'''
python+opencv视频图像相互转换 - CSDN博客 https://blog.csdn.net/m0_37733057/article/details/79023693
链接：https://www.zhihu.com/question/49558804/answer/343058915
OpenCV: Drawing Functions in OpenCV https://docs.opencv.org/3.1.0/dc/da5/tutorial_py_drawing_functions.html
'''
# 每秒传输帧数(Frames Per Second)
fps = 100 # 保存视频的FPS，可以适当调整 FPS是图像领域中的定义，是指画面每秒传输帧数，通俗来讲就是指动画或视频的画面数。FPS是测量用于保存、显示动态视频的信息数量。每秒钟帧数愈多，所显示的动作就会愈流畅。通常，要避免动作不流畅的最低是30。某些计算机视频格式，每秒只能提供15帧。
f_img_d = '{}{}{}{}{}'.format(this_file_dirname, os_sep, 'mypng', os_sep, '*.jpg')
imgs = glob.glob(f_img_d)
"""
用图片总数均分音频时间
"""
def resize_rescale_pilimg(img_f, w_h_tuple=(mode_img_size_wh[0], mode_img_size_wh[1]), mid_factor=1):
# print(img_f)
img_n, img_type = img_f.split('.')[-2], img_f.split('.')[-1]
# print(img_n)
img_n_resize_rescale_pilimg_dir = '{}{}{}'.format(os_sep.join(img_n.split(os_sep)[:-1]), 'resize_rescale_pilimg',
os_sep, img_n.split(os_sep)[-1], os_sep)
img_n_resize_rescale_pilimg = '{}{}{}'.format(img_n_resize_rescale_pilimg_dir, img_n.split(os_sep)[-1], '.PNG')
# print(img_n_resize_rescale_pilimg)
img_type = 'PNG'
# img_f_new = '{}{}{}{}'.format(img_n, int(time.time()), 'resize_rescale.', img_type)
img_f_new = img_n_resize_rescale_pilimg
mid_icon = Image.open(img_f)
mid_icon_w, mid_icon_h = w_h_tuple[0] * mid_factor, w_h_tuple[1] * mid_factor
mid_icon = mid_icon.resize((mid_icon_w, mid_icon_h), Image.ANTIALIAS)
mid_icon.save(img_n_resize_rescale_pilimg, img_type)
return img_f_new
from PIL import Image, ImageDraw, ImageFont
# myfont = ImageFont.truetype("simhei.ttf", 50, encoding="utf-8")
myfont = ImageFont.truetype("simhei.ttf", encoding="utf-8")
import cv2
import numpy as np
equal_str_l = []
le = len(mp3_str)
br_step = 34
br_times = math.ceil(le / br_step)
for i_br_loop in range(br_times):
s_p = mp3_str[i_br_loop * br_step:i_br_loop * br_step + br_step]
equal_str_l.append(s_p)
l = equal_str_l
char_loop_l_len = len(mp3_str) / len(equal_str_l)
char_loop_l_len = len(mp3_str_bdmp3filter)
'''
bdmp3filter_l = ['：', '——', ',', '《', '》', '“', '”', '、', '（', '）', '.', ' '] # 注意空格
# 保留其他标点符号
mp3_str = ' '.join(sentence_l_chk)
mp3_str_bdmp3filter = mp3_str
for i in bdmp3filter_l:
mp3_str_bdmp3filter = mp3_str_bdmp3filter.replace(i, '')
'''
def gen_video(os_delay_factor=0.245, mystep=0.01, bear_error_second=1, audio_spend=t_spend, step_para=1):
f_v = '{}{}{}{}{}{}{}'.format('D:\\myv\\', 'g3db', uid, 'g3uid', uid, int(time.time()), 'saveVideo.b.avi')
fps, fourcc = 15, cv2.VideoWriter_fourcc('M', 'J', 'P', 'G')
videoWriter = cv2.VideoWriter(f_v, fourcc, fps, (mode_img_size_wh[0], mode_img_size_wh[1]))
for i in l:
# print(i)
i_index = l.index(i)
img_index = i_index % len(imgs)
imgname = imgs[img_index]
mystr, le = i, len(i)
frame = cv2.imread(imgname)
if (frame.shape[1], frame.shape[0]) != (mode_img_size_wh[0], mode_img_size_wh[1]):
imgname = resize_rescale_pilimg(imgname)
frame = cv2.imread(imgname)
else:
pass
frame_cv2 = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
frame_pil = Image.fromarray(frame_cv2) # 转为PIL的图片格式
# font = ImageFont.truetype("simhei.ttf", 50, encoding="utf-8")
font = ImageFont.truetype("simhei.ttf", encoding="utf-8")
# ImageDraw.Draw(frame_pil).text((100, 20), mystr, (0, 0, 255), font)
f_x, f_y = math.floor((mode_img_size_wh[0]) * 0.15), math.floor(mode_img_size_wh[1] * 0.85)
# ImageDraw.Draw(frame_pil).text((30, mode_img_size_wh[1]-30), mystr, (0, 0, 255), font)
ImageDraw.Draw(frame_pil).text((f_x, f_y), mystr, (0, 0, 255), font)
frame_cv2 = cv2.cvtColor(np.array(frame_pil), cv2.COLOR_RGB2BGR)
img = frame_cv2
line_bdmp3filter = i
for bdmp3filter in bdmp3filter_l:
line_bdmp3filter = line_bdmp3filter.replace(bdmp3filter, '')
myinterval = t_spend / (len(mp3_str_bdmp3filter) * 1) * os_delay_factor * len(line_bdmp3filter)
print(myinterval,'---------------',mystr)
this_time = time.time()
while time.time() - this_time < myinterval:
videoWriter.write(img)
videoWriter.release()
time.sleep(1)
video_playtime = VideoFileClip(f_v).duration
if video_playtime - audio_spend > bear_error_second:
# os_delay_factor -= mystep
os_delay_factor *= t_spend / video_playtime
gen_video(os_delay_factor=os_delay_factor, mystep=0.005, audio_spend=t_spend)
elif audio_spend - video_playtime > bear_error_second:
# os_delay_factor += mystep
os_delay_factor *= t_spend / video_playtime
gen_video(os_delay_factor=os_delay_factor, mystep=0.005, audio_spend=t_spend)
else:
os._exit(123)
'''
326
'''
gen_video(os_delay_factor=0.045, mystep=0.003, bear_error_second=0.5, audio_spend=t_spend)
