ai-spokesgirl/AI_anchor/main.py

import openshot
import re
from PIL import Image,ImageDraw,ImageFont
import pandas as pd
import os
import cv2
import numpy as np
# import moviepy.editor as mp
import time
import pysrt
import shutil
import rpyc
import random
import string
import requests
from bs4 import BeautifulSoup
import zipfile

def cKey(r,g,b,fuzz):
    col=openshot.Color()
    col.red=openshot.Keyframe(r)
    col.green=openshot.Keyframe(g)
    col.blue=openshot.Keyframe(b)
    return openshot.ChromaKey(col, openshot.Keyframe(fuzz))

def video_writer_init(path):
    w = openshot.FFmpegWriter(path)
    w.SetAudioOptions(True, "aac", 44100, 2, openshot.LAYOUT_STEREO, 3000000)
    w.SetVideoOptions(True, "libx264", openshot.Fraction(30000, 1000), 1280, 720,
        openshot.Fraction(1, 1), False, False, 3000000)
    return w

def video_photo_clip(video=None,layer=None, position=None, end=None
    ,scale_x=1,scale_y=1,location_x=0,location_y=0,ck=None,audio=True):
    clip = openshot.Clip(video)
    clip.Layer(layer)
    clip.Position(position)
    clip.End(end)
    clip.scale_x=openshot.Keyframe(scale_x)
    clip.scale_y=openshot.Keyframe(scale_y)
    clip.location_x=openshot.Keyframe(location_x)
    clip.location_y=openshot.Keyframe(location_y)
    
    if ck!=None:
        clip.AddEffect(ck)
    if audio==True:
        clip.has_audio=openshot.Keyframe(1)
    else:
        clip.has_audio=openshot.Keyframe(0)
    return clip

def trim_punctuation(s):
    pat_block = u'[^\u4e00-\u9fff0-9a-zA-Z]+'
    pattern = u'([0-9]+{0}[0-9]+)|{0}'.format(pat_block)
    res = re.sub(pattern, lambda x: x.group(1) if x.group(1) else u" " ,s)
    return res


def randomString(stringLength=10):
    letters = string.ascii_lowercase
    return ''.join(random.choice(letters) for i in range(stringLength))


def mp3_to_anchor(fname):

    conn = rpyc.classic.connect("192.168.192.221",18812)
    fr=open(fname,'rb')
    ropen = conn.builtins.open
    randname=randomString(10)
    finalname=randomString(10)

    fw=ropen('/tmp/'+randname+'.mp4','wb')
    fw.write(fr.read())
    fw.close()
    ros = conn.modules.os
    ros.system('/root/to_video/p9.sh '+randname+".mp4 "+finalname+".mp4")
    return 'http://192.168.192.221/video/'+finalname+'.mp4'
#    conn.execute('import os')


def download_mp4(url,name):
    headers = {'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/90.0.4430.212 Safari/537.36'}
    with open('input_self/AI_girl/ai_spokesgirl_%s.mp4'%(name),'wb') as f:
        r = requests.get(url, headers=headers, stream=True)
        if r.status_code == 404:
            return False
        for chunk in r.iter_content(chunk_size=1024):
            if chunk:
                f.write(chunk)
    return True



                
   
 

def text_to_short_vedio_create(read_csv_use="導盲犬協會.csv",pwd_use="導盲犬協會影片素材2/",op="input_self/LOGO_OP_4.mp4",ed="input_self/LOGO_ED.mp4",bg="input_self/串場樣板1_long.mp4",input_zip="input/input_data.zip"):
    t = openshot.Timeline(1280, 720, openshot.Fraction(30000, 1000), 44100, 2, openshot.LAYOUT_STEREO)
    t.Open()

    files=os.listdir("input/")
    print(files)
    for i in files:
        if i[-4:]==".zip":
            input_zip="input/"+i
            break

    # try:
    #     os.remove("./templates/index.html")
    # except:
    #     pass
    with zipfile.ZipFile(input_zip, 'r') as zf:
        for fn in zf.namelist():

            right_fn = fn.encode('cp437').decode('big5')  # 將檔名正確編碼
            check_p=right_fn.split("/")
            if right_fn[-1]=="/" :
                os.mkdir("input/"+right_fn[:-1])
                pwd_use=right_fn
                break
            if len(check_p)==2:
                os.mkdir("input/"+check_p[0])
                pwd_use=check_p[0]+"/"
                break

    with zipfile.ZipFile(input_zip, 'r') as zf:
        for fn in zf.namelist():
            right_fn = fn.encode('cp437').decode('big5')  # 將檔名正確編碼
            check_p=right_fn.split("/")
            if right_fn[-1]=="/" :
                continue
            if right_fn[-4:]==".csv":
                read_csv_use = right_fn
            with open("input/"+right_fn, 'wb') as output_file:  # 建立並開啟新檔案
                with zf.open(fn, 'r') as origin_file:  # 開啟原檔案
                    shutil.copyfileobj(origin_file, output_file)  # 將原檔案內容複製到新檔案





    # 去背參數
    # ck=cKey(0,254,0,270)
    ck=cKey(0,255,0,320)
    ck_anchor=None
    #時間
    time_= 0
    
	
    csv_use=pd.read_csv("input/"+read_csv_use)
    csv_use=csv_use.dropna(how='all')
    csv_use.reset_index(inplace=True)

    anchor_op = openshot.FFmpegReader(op)
    anchor_op.Open()
    anchor_clip_op = video_photo_clip(video=anchor_op,layer=2,scale_x=1,scale_y=1,
            location_x=0,location_y=0,position=time_, end=anchor_op.info.duration,ck=ck_anchor,audio=True)
    
    t.AddClip(anchor_clip_op)
    anchor_op.Close()
    time_+=anchor_op.info.duration


    for i in range(len(csv_use)):
    # for i in range(3):
        pwd_=str(csv_use.loc[i,['音檔']].values[0])

        
        locals()['anchor_music'+str(i)] = openshot.FFmpegReader("input/"+pwd_use+pwd_)
        locals()['anchor_music'+str(i)].Open()
        locals()['anchor_music'+str(i)+'clip'] = video_photo_clip(video=locals()['anchor_music'+str(i)],layer=3,scale_x=0,scale_y=0,
                   location_x=0,location_y=0,position=time_, end=locals()['anchor_music'+str(i)].info.duration,ck=ck_anchor,audio=True)
        t.AddClip(locals()['anchor_music'+str(i)+'clip'])
        locals()['anchor_music'+str(i)].Close()

        if str(csv_use.loc[i,['是否要場景']].values[0])=="是":
            scale_x_use = 0.59
            scale_y_use = 0.59
            t1 = openshot.Timeline(1280, 720, openshot.Fraction(30000, 1000), 44100, 2, openshot.LAYOUT_STEREO)
            t1.Open()
            pwd_=str(csv_use.loc[i,['音檔']].values[0])

        
            locals()['anchor_music_t'+str(i)] = openshot.FFmpegReader("input/"+pwd_use+pwd_)
            locals()['anchor_music_t'+str(i)].Open()
            locals()['anchor_music_t'+str(i)+'clip'] = video_photo_clip(video=locals()['anchor_music_t'+str(i)],layer=3,scale_x=0,scale_y=0,
                   location_x=0,location_y=0,position=0, end=locals()['anchor_music_t'+str(i)].info.duration,ck=ck_anchor,audio=True)
            t1.AddClip(locals()['anchor_music_t'+str(i)+'clip'])
            locals()['anchor_music_t'+str(i)].Close()

            w = video_writer_init("input_self/tmp1/%s.mp4"%('anchor_music'+str(i)))
            w.Open()
            frames = int(t1.info.fps)*int(locals()['anchor_music_t'+str(i)].info.duration)
            for n in range(frames):
                f=t1.GetFrame(n)
                w.WriteFrame(f)
            w.Close()
            t1.Close()

            fname=mp3_to_anchor("input_self/tmp1/%s.mp4"%('anchor_music'+str(i)))
            # print(fname)
            time.sleep(180)
            while True:
                result = download_mp4(fname,str(i))
                if result:
                    break
                print('等待...')
                time.sleep(60)


        else:
            scale_x_use = 1
            scale_y_use = 1

        choose=str(csv_use.loc[i,['素材']].values[0]).split(".")[-1]
        pwd_p1=str(csv_use.loc[i,['素材']].values[0])




        if choose == 'mp4':
            locals()['anchor'+str(i)] = openshot.FFmpegReader("input/"+pwd_use+pwd_p1)
            locals()['anchor'+str(i)].Open()
            locals()['anchor'+str(i)+'clip'] = video_photo_clip(video=locals()['anchor'+str(i)],layer=4,scale_x=scale_x_use,scale_y=scale_y_use,
               location_x=-0.04,location_y=-0.04,position=time_, end=locals()['anchor_music'+str(i)].info.duration,ck=ck_anchor,audio=False)
        
            t.AddClip(locals()['anchor'+str(i)+'clip'])
            locals()['anchor'+str(i)].Close()

            if str(csv_use.loc[i,['是否要場景']].values[0])=="是":
                locals()['anchor_ad'+str(i)] = openshot.FFmpegReader('input_self/AI_girl/ai_spokesgirl_%s.mp4'%(str(i)))
                locals()['anchor_ad'+str(i)].Open()
                locals()['anchor_clip_ad'+str(i)] = video_photo_clip(video=locals()['anchor_ad'+str(i)],layer=6,scale_x=0.8,scale_y=0.8,
                   location_x=0.38,location_y=0.35,position=time_, end=locals()['anchor_music'+str(i)].info.duration,ck=ck,audio=False)
                t.AddClip(locals()['anchor_clip_ad'+str(i)])
                locals()['anchor_ad'+str(i)].Close()

                locals()['anchor_bg'+str(i)] = openshot.FFmpegReader(bg)
                locals()['anchor_bg'+str(i)].Open()
                locals()['anchor_clip_bg'+str(i)] = video_photo_clip(video=locals()['anchor_bg'+str(i)],layer=2,scale_x=1,scale_y=1,
                   location_x=0,location_y=0,position=time_, end=locals()['anchor_music'+str(i)].info.duration,ck=ck_anchor,audio=False)
                t.AddClip(locals()['anchor_clip_bg'+str(i)])
                locals()['anchor_bg'+str(i)].Close()
         

        elif choose == 'jpg' or choose == 'png':
            locals()['anchor'+str(i)] = openshot.QtImageReader("input/"+pwd_use+pwd_p1)
            locals()['anchor'+str(i)].Open()
            locals()['anchor'+str(i)+'clip'] = video_photo_clip(video=locals()['anchor'+str(i)],layer=4,scale_x=scale_x_use,scale_y=scale_y_use,
               location_x=-0.04,location_y=-0.04,position=time_, end=locals()['anchor_music'+str(i)].info.duration,ck=ck_anchor,audio=False)
            t.AddClip(locals()['anchor'+str(i)+'clip'])
            locals()['anchor'+str(i)].Close()
            if str(csv_use.loc[i,['是否要場景']].values[0])=="是":
                locals()['anchor_ad'+str(i)] = openshot.FFmpegReader("input_self/AI_girl/ai_spokesgirl_%s.mp4"%(str(i)))
                locals()['anchor_ad'+str(i)].Open()
                locals()['anchor_clip_ad'+str(i)] = video_photo_clip(video=locals()['anchor_ad'+str(i)],layer=6,scale_x=0.8,scale_y=0.8,
                   location_x=0.38,location_y=0.35,position=time_, end=locals()['anchor_music'+str(i)].info.duration,ck=ck,audio=False)
                t.AddClip(locals()['anchor_clip_ad'+str(i)])
                locals()['anchor_ad'+str(i)].Close()

                locals()['anchor_bg'+str(i)] = openshot.FFmpegReader(bg)
                locals()['anchor_bg'+str(i)].Open()
                locals()['anchor_clip_bg'+str(i)] = video_photo_clip(video=locals()['anchor_bg'+str(i)],layer=2,scale_x=1,scale_y=1,
                   location_x=0,location_y=0,position=time_, end=locals()['anchor_music'+str(i)].info.duration,ck=ck_anchor,audio=False)
                t.AddClip(locals()['anchor_clip_bg'+str(i)])
                locals()['anchor_bg'+str(i)].Close()

        time_+=locals()['anchor_music'+str(i)].info.duration





            
 
    

    
    anchor_ed = openshot.FFmpegReader(ed)
    anchor_ed.Open()
    anchor_clip_ed = video_photo_clip(video=anchor_ed,layer=2,scale_x=1,scale_y=1,
            location_x=0,location_y=0,position=time_, end=anchor_ed.info.duration,ck=ck_anchor,audio=True)
    time_+=anchor_ed.info.duration
    t.AddClip(anchor_clip_ed)
    anchor_ed.Close()



  













    
    
    w = video_writer_init("input_self/tmp1/test.mp4")
    w.Open()
    
    frames = int(t.info.fps)*int(time_)
    for n in range(frames):
        f=t.GetFrame(n)
        w.WriteFrame(f)

    t.Close()
    w.Close()
    shutil.rmtree('input')
    os.mkdir('input')
    shutil.rmtree('input_self/AI_girl')
    os.mkdir('input_self/AI_girl')
#文字轉圖片
def txt2image(content, save_target,lang='zh',size=26,fon="input_self/font/DFT_B7.ttc"):
    unicode_text = trim_punctuation(content)
    font = ''
    if lang=='zh':
        font = ImageFont.truetype(font=fon, size=size)
    else :
        font = ImageFont.truetype(font="input_self/font/arial.ttf", size=size)
    
    W, H = (1280,500)
    canvas = Image.new('RGB', (W, H), "#00FF00")
    draw = ImageDraw.Draw(canvas)
    
    text= content
    if "\n" in text:
        w, h = draw.textsize(text.split("\n")[0],font = font)
        #draw.text(((W-w)/2,0), text[0:18],'black', font)
        text_border(draw,(W-w)/2,0,text.split("\n")[0],font,'black','white')
        w, h = draw.textsize(text.split("\n")[1],font = font)
        #draw.text(((W-w)/2,h+2), text[18:],'black', font)
        text_border(draw,(W-w)/2,h+2,text.split("\n")[1],font,'black','white')
    else:
        w, h = draw.textsize(content,font = font)
        #draw.text(((W-w)/2,0), text,'black', font)
        text_border(draw,(W-w)/2,0,text,font,'black','white')
    canvas.save(save_target, "PNG")

def text_border(draw,x,y,text,font,shadowcolor,fillcolor):
    draw.text((x-1, y), text, font=font, fill=shadowcolor)
    draw.text((x+1, y), text, font=font, fill=shadowcolor)
    draw.text((x, y-1), text, font=font, fill=shadowcolor)
    draw.text((x, y+1), text, font=font, fill=shadowcolor)

    draw.text((x-1, y+1), text, font=font, fill=shadowcolor)
    draw.text((x+1, y-1), text, font=font, fill=shadowcolor)
    draw.text((x-1, y-1), text, font=font, fill=shadowcolor)
    draw.text((x+1, y+1), text, font=font, fill=shadowcolor)
    # thicker border
    draw.text((x-2, y-2), text, font=font, fill=shadowcolor)
    draw.text((x+2, y-2), text, font=font, fill=shadowcolor)
    draw.text((x-2, y+2), text, font=font, fill=shadowcolor)
    draw.text((x+2, y+2), text, font=font, fill=shadowcolor)

    # now draw the text over it
    draw.text((x, y), text, font=font, fill=fillcolor)

def srt_to_csv(srt_file):
    subs = pysrt.open(srt_file)
    csv_file = srt_file.split('.')[0] + ".csv"
    with open(csv_file, 'w', newline='') as csvfile:
        # 建立 CSV 檔寫入器
        writer = csv.writer(csvfile)
        for context in subs:
            writer.writerow([context.index, context.start,context.end, context.text])
    return csv_file

def csv_to_text(csv_file,text_font):
    text_form = []
    with open(csv_file, newline='') as csvfile:

        # 讀取 CSV 檔案內容
        rows = csv.reader(csvfile)

        # 以迴圈輸出每一列
        for row in rows:
            start = datetime.strptime(row[1], "%H:%M:%S,%f")
            end = datetime.strptime(row[2], "%H:%M:%S,%f") - datetime.strptime(row[1], "%H:%M:%S,%f")
            end_timeStamp=end.seconds+0.000001*end.microseconds
            start_timeStamp=start.minute*60+start.second+ 0.000001*start.microsecond
            text_form.append({'text':row[3],'start':start_timeStamp,'end':end_timeStamp,'size':36,'font':text_font})
    
    return text_form
def text_to_short_vedio(mp4_file ,sound_file,output_filename,text_font):
    t = openshot.Timeline(1280, 720, openshot.Fraction(30000, 1000), 44100, 2, openshot.LAYOUT_STEREO)
    t.Open()
    
    # 去背參數
    ck = cKey(0, 254, 0, 270)
    ck_anchor = cKey(0, 255, 0, 320)

    anchor = openshot.FFmpegReader(mp4_file)
    anchor.Open()
    anchor_clip = video_photo_clip(video=anchor,layer=2,scale_x=1,scale_y=1,
            location_x=0,location_y=0,position=0, end=anchor.info.duration,audio=True)
    t.AddClip(anchor_clip)
    anchor.Close()
    number = 0

    sound_srt_file = ""
    #音檔自動產生srt(逐字稿)
    if ".srt" in sound_file:
        sound_srt_file = sound_file
    elif not sound_file is None:
        cmd = "autosub -S zh-TW -D zh-TW " + sound_file
        os.system(cmd)
        sound_srt_file = sound_file.split('.')[0] + ".srt"
    csv_file = srt_to_csv(sound_srt_file)
    text_form = csv_to_text(csv_file,text_font)
    print(sound_srt_file)
    #開啟srt檔
    try:
        # subs = pysrt.open(sound_srt_file)
        # text_form = []
        # for context in subs:
        #     #print(context.start.minutes*60+context.start.seconds+ 0.001*context.start.milliseconds)
        #     end = context.end-context.start
        #     end_timeStamp=(end.minutes*60+end.seconds+ 0.001*end.milliseconds)
        #     start_timeStamp=(context.start.minutes*60+context.start.seconds+ 0.001*context.start.milliseconds)
        #     text_form.append({'text':context.text,'start':start_timeStamp,'end':end_timeStamp,'size':36,'font':text_font})
        number = 0
        for text_tmp in text_form:
            file_name = "input_self/tmp/save_target_" + str(number) + ".png"
            txt2image(text_tmp['text'], file_name,lang='zh',size = text_tmp['size'],fon = text_tmp['font'])
            exec('text_anchor_{} = openshot.QtImageReader("input_self/tmp/save_target_{}.png")'.format(number,number))
            exec('text_anchor_{}.Open()'.format(number))
            exec('text_anchor_{}.Open()'.format(number))
            exec('text_anchor_clip_{} = video_photo_clip(video=text_anchor_{},layer=4,scale_x=1,scale_y=1,\
                    location_x=0,location_y=0.67,position=text_tmp["start"], end=text_tmp["end"],ck=ck_anchor,audio=True)'.format(number,number))
            exec('t.AddClip(text_anchor_clip_{})'.format(number))
            exec('text_anchor_{}.Close()'.format(number))
            number = number+1
    except:
        print("無法開啟srt檔案(字幕產生失敗)")

    w = video_writer_init(output_filename)
    w.Open()

    frames = int(t.info.fps)*int(anchor.info.duration)
    for n in range(frames):
        f=t.GetFrame(n)
        w.WriteFrame(f)

    t.Close()
    w.Close()

    #刪除暫存檔案
    shutil.rmtree('input_self/tmp')
    os.mkdir('input_self/tmp')
    shutil.rmtree('input_self/tmp1')
    os.mkdir('input_self/tmp1')


if __name__ == '__main__':
	text_to_short_vedio_create()
	text_to_short_vedio(mp4_file = "input_self/tmp1/test.mp4",sound_file ='input_self/tmp1/test.mp4',
        output_filename="output/demo.mp4",text_font ="input_self/font/DFT_R7.ttc")