利用Python做一個小姐姐詞云跳舞視頻
一、前言
大家好，歡迎來到 Crossin的編程教室 ！
B站上會跳舞的漂亮小姐姐真的好多好多。今天我們來帶大家看一個不一樣的效果：利用視頻中的彈幕文字來制作一個漂亮小姐姐詞云跳舞視頻，一起來看看吧。
二、實現(xiàn)思路
1. you-get下載視頻
pip install you-get -i http://pypi.douban.com/simple --trusted-host pypi.douban.com視頻鏈接：
https://www.bilibili.com/video/BV1rD4y1Q7jc?from=search&seid=10634574434789745619
you-get -i https://www.bilibili.com/video/BV1rD4y1Q7jc?from=search&seid=10634574434789745619
you-get -o 本地保存路徑 視頻鏈接
更多關(guān)于 you-get 的詳細(xì)使用，可以參考官方文檔：
2. 爬取彈幕內(nèi)容
寫 python 爬蟲，解析網(wǎng)頁、提取彈幕數(shù)據(jù)保存到txt，注意構(gòu)造 URL 參數(shù)和偽裝請求頭。
import requests
import pandas as pd
import re
import time
import random
from concurrent.futures import ThreadPoolExecutor
import datetime
from fake_useragent import UserAgent
# 隨機(jī)產(chǎn)生請求頭
ua = UserAgent(verify_ssl=False, path='fake_useragent.json')
start_time = datetime.datetime.now()
爬取彈幕數(shù)據(jù)
def Grab_barrage(date):
# 偽裝請求頭
headers = {
"sec-fetch-dest": "empty",
"sec-fetch-mode": "cors",
"sec-fetch-site": "same-site",
"accept-encoding": "gzip",
"origin": "https://www.bilibili.com",
"referer": "https://www.bilibili.com/video/BV1rD4y1Q7jc?from=search&seid=10634574434789745619",
"user-agent": ua.random,
"cookie": "_uuid=0EBFC9C8-19C3-66CC-4C2B-6A5D8003261093748infoc; buvid3=4169BA78-DEBD-44E2-9780-B790212CCE76155837infoc; sid=ae7q4ujj; rpdid=|(JJmlY|YukR0J'ulmumY~u~m; LIVE_BUVID=AUTO4315952457375679; CURRENT_QUALITY=80; blackside_state=1; CURRENT_FNVAL=80; PVID=1; fingerprint3=89f3acebeacc72a0a25afa9c05f6d87c; fingerprint=2c691e81ffde16dfbb39b8f6d20eb5f0; fingerprint_s=99dc5d2a438924de14f663d6a4cf9cc8; buivd_fp=4169BA78-DEBD-44E2-9780-B790212CCE76155837infoc; bp_video_offset_501048197=472333401972842834; buvid_fp=4169BA78-DEBD-44E2-9780-B790212CCE76155837infoc; buvid_fp_plain=0AE6F247-D84F-48C7-87A8-F8F35A578544184985infoc; bfe_id=1e33d9ad1cb29251013800c68af42315; DedeUserID=501048197; DedeUserID__ckMd5=1d04317f8f8f1021; SESSDATA=2ae431c2%2C1625306326%2Ca312a*11; bili_jct=d4edec1bd2ab84e0eb453adb3971b19a"
}
# 構(gòu)造url訪問 需要用到的參數(shù) 爬取指定日期的彈幕
params = {
'type': 1,
'oid': '206344228',
'date': date
}
# 發(fā)送請求 獲取響應(yīng)
response = requests.get(url, params=params, headers=headers)
# print(response.encoding) 重新設(shè)置編碼
response.encoding = 'utf-8'
# print(response.text)
# 正則匹配提取數(shù)據(jù) 轉(zhuǎn)成集合去除重復(fù)彈幕
comment = set(re.findall('<d p=".*?">(.*?)</d>', response.text))
# 將每條彈幕數(shù)據(jù)寫入txt
with open('bullet.txt', 'a+') as f:
for con in comment:
f.write(con + '\n')
print(con)
time.sleep(random.randint(1, 3)) # 休眠
def main():
# 開多線程爬取 提高爬取效率
with ThreadPoolExecutor(max_workers=4) as executor:
executor.map(Grab_barrage, date_list)
# 計算所用時間
delta = (datetime.datetime.now() - start_time).total_seconds()
print(f'用時：{delta}s -----------> 彈幕數(shù)據(jù)成功保存到本地txt')
if __name__ == '__main__':
# 目標(biāo)url
url = "https://api.bilibili.com/x/v2/dm/history"
start = '20201201'
end = '20210128'
# 生成時間序列
date_list = [x for x in pd.date_range(start, end).strftime('%Y-%m-%d')]
print(date_list)
count = 0
# 調(diào)用主函數(shù)
main()
結(jié)果如下：
3. 從視頻中提取圖片
import cv2
# ============================ 視頻處理 分割成一幀幀圖片 =======================================
cap = cv2.VideoCapture(r"beauty.flv")
num = 1
while True:
# 逐幀讀取視頻 按順序保存到本地文件夾
ret, frame = cap.read()
if ret:
if 88 <= num < 888:
cv2.imwrite(f"./pictures/img_{num}.jpg", frame) # 保存一幀幀的圖片
print(f'========== 已成功保存第{num}張圖片 ==========')
num += 1
else:
break
cap.release() # 釋放資源
結(jié)果如下：
4. 利用百度AI進(jìn)行人像分割
查看人像分割的Python SDK 文檔，熟悉它的基本使用。
# -*- coding: UTF-8 -*-
"""
@File ：人像分割.py
@Author ：葉庭云
@CSDN ：https://yetingyun.blog.csdn.net/
@百度AI ：https://ai.baidu.com/tech/body/seg
"""
import cv2
import base64
import numpy as np
import os
from aip import AipBodyAnalysis
import time
import random
# 利用百度AI的人像分割服務(wù) 轉(zhuǎn)化為二值圖 有小姐姐身影的蒙版
# 百度云中已創(chuàng)建應(yīng)用的 APP_ID API_KEY SECRET_KEY
APP_ID = '23485847'
API_KEY = 'VwGY053Y1A8ow3CFBTFrK0Pm'
SECRET_KEY = '**********************************'
client = AipBodyAnalysis(APP_ID, API_KEY, SECRET_KEY)
# 保存圖像分割后的路徑
path = './mask_img/'
# os.listdir 列出保存到圖片名稱
img_files = os.listdir('./pictures')
print(img_files)
for num in range(88, len(img_files) + 1):
# 按順序構(gòu)造出圖片路徑
img = f'./pictures/img_{num}.jpg'
img1 = cv2.imread(img)
height, width, _ = img1.shape
# print(height, width)
# 二進(jìn)制方式讀取圖片
with open(img, 'rb') as fp:
img_info = fp.read()
# 設(shè)置只返回前景 也就是分割出來的人像
seg_res = client.bodySeg(img_info)
labelmap = base64.b64decode(seg_res['labelmap'])
nparr = np.frombuffer(labelmap, np.uint8)
labelimg = cv2.imdecode(nparr, 1)
labelimg = cv2.resize(labelimg, (width, height), interpolation=cv2.INTER_NEAREST)
new_img = np.where(labelimg == 1, 255, labelimg)
mask_name = path + 'mask_{}.png'.format(num)
# 保存分割出來的人像
cv2.imwrite(mask_name, new_img)
print(f'======== 第{num}張圖像分割完成 ========')
time.sleep(random.randint(1,2))
5. 小姐姐跳舞詞云生成
# -*- coding: UTF-8 -*-
"""
@File ：詞云.py
@Author ：葉庭云
@CSDN ：https://yetingyun.blog.csdn.net/
"""
from wordcloud import WordCloud
import collections
import jieba
import re
from PIL import Image
import matplotlib.pyplot as plt
import numpy as np
# 讀取數(shù)據(jù)
with open('bullet.txt') as f:
data = f.read()
# 文本預(yù)處理 去除一些無用的字符 只提取出中文出來
new_data = re.findall('[\u4e00-\u9fa5]+', data, re.S)
new_data = "/".join(new_data)
# 文本分詞
seg_list_exact = jieba.cut(new_data, cut_all=True)
result_list = []
with open('stop_words.txt', encoding='utf-8') as f:
con = f.read().split('\n')
stop_words = set()
for i in con:
stop_words.add(i)
for word in seg_list_exact:
# 設(shè)置停用詞并去除單個詞
if word not in stop_words and len(word) > 1:
result_list.append(word)
# 篩選后統(tǒng)計詞頻
word_counts = collections.Counter(result_list)
path = './wordcloud/'
for num in range(88, 888):
img = f'./mask_img/mask_{num}'
# 獲取蒙版圖片
mask_ = 255 - np.array(Image.open(img))
# 繪制詞云
plt.figure(figsize=(8, 5), dpi=200)
my_cloud = WordCloud(
background_color='black', # 設(shè)置背景顏色 默認(rèn)是black
mask=mask_, # 自定義蒙版
mode='RGBA',
max_words=500,
font_path='simhei.ttf', # 設(shè)置字體 顯示中文
).generate_from_frequencies(word_counts)
# 顯示生成的詞云圖片
plt.imshow(my_cloud)
# 顯示設(shè)置詞云圖中無坐標(biāo)軸
plt.axis('off')
word_cloud_name = path + 'wordcloud_{}.png'.format(num)
my_cloud.to_file(word_cloud_name) # 保存詞云圖片
print(f'======== 第{num}張詞云圖生成 ========')
結(jié)果如下：
6. 合成跳舞視頻
# -*- coding: UTF-8 -*-
"""
@File ：生成視頻.py
@Author ：葉庭云
@CSDN ：https://yetingyun.blog.csdn.net/
"""
import cv2
import os
# 輸出視頻的保存路徑
video_dir = 'result.mp4'
# 幀率
fps = 30
# 圖片尺寸
img_size = (1920, 1080)
fourcc = cv2.VideoWriter_fourcc('M', 'P', '4', 'V') # opencv3.0 mp4會有警告但可以播放
videoWriter = cv2.VideoWriter(video_dir, fourcc, fps, img_size)
img_files = os.listdir('./wordcloud')
for i in range(88, 888):
img_path = './wordcloud/' + 'wordcloud_{}.png'.format(i)
frame = cv2.imread(img_path)
frame = cv2.resize(frame, img_size) # 生成視頻 圖片尺寸和設(shè)定尺寸相同
videoWriter.write(frame) # 寫進(jìn)視頻里
print(f'======== 按照視頻順序第{i}張圖片合進(jìn)視頻 ========')
videoWriter.release() # 釋放資源
效果如下：
7. 視頻插入音頻
漂亮小姐姐跳舞那么好看，再加上自己喜歡的背景音樂，豈不美哉。
# -*- coding: UTF-8 -*-
"""
@File ：插入音頻.py
@Author ：葉庭云
@CSDN ：https://yetingyun.blog.csdn.net/
"""
import moviepy.editor as mpy
# 讀取詞云視頻
my_clip = mpy.VideoFileClip('result.mp4')
# 截取背景音樂
audio_background = mpy.AudioFileClip('song.mp4').subclip(17, 44)
audio_background.write_audiofile('vmt.mp3')
# 視頻中插入音頻
final_clip = my_clip.set_audio(audio_background)
# 保存為最終的視頻 動聽的音樂！漂亮小姐姐詞云跳舞視頻！
final_clip.write_videofile('final_video.mp4')
_往期文章推薦_
評論
圖片
表情