многопроцессорность с помощью Moviepy

Используя процессы, я сократил время всего на 15-20 секунд, потому что ffmpeg даже в одном процессе использовала почти полную мощность процессора, а у моего компьютера не было мощности для более быстрого запуска других процессов.

Сначала я сократил код, чтобы сделать его короче.

В коде try и except были похожие элементы, поэтому я переместил их в сторону try/except.

Далее я использовал

        if end > video_duration:
            end = video_duration

а мне вообще не нужен был try/except.

При использовании os.makedirs(..., exist_ok=True) мне не нужно запускать его в try/except

Тем временем я сократил время на 20 секунд, используя

        clip = VideoFileClip(filename).subclip(start, end)

вместо того

        temp_filename  = f"{base_folder}/cut_{number}.mp4"
        
        fmpeg_extract_subclip(filename, start, end, targetname=temp_filename)
        
        clip = VideoFileClip(temp_filename)

Таким образом, я не записываю подклип на диск, и мне не нужно читать его снова с диска.

from moviepy.video.io.ffmpeg_tools import ffmpeg_extract_subclip
from moviepy.editor import *
import time

def main():

    text = input("Enter Your text please: ") [::-1]
    #text = 'Hello World'
    
    base_folder = time.strftime("result_%Y_%m_%d_%H_%M_%S")    
    os.makedirs(f"{base_folder}/result_edit", exist_ok=True)
    
    filename = "video.mp4"
    #filename = "BigBuckBunny.mp4"
    
    video_duration = VideoFileClip(filename).duration

    number = 0 # instead of `cut_name_num` and `txt_part` because both had the same value

    time_start = time.time()

    for start in range(0, int(video_duration), 60):

        end = start + 60
        
        if end > video_duration:
            end = video_duration

        number += 1

        clip_duration = end - start
        print(f'[DEBUG] number: {number:2} | start: {start:6.2f} | end: {end:6.2f} | duration: {clip_duration:.2f}')
       
        final_text = f"{number} {text}"
    
        temp_filename  = f"{base_folder}/cut_{number}.mp4"
        final_filename = f"{base_folder}/result_edit/cut_{number}.mp4"
        
        #ffmpeg_extract_subclip(filename, start, end, targetname=temp_filename)
        
        #clip = VideoFileClip(temp_filename)
        clip = VideoFileClip(filename).subclip(start, end)
        clip = clip.volumex(2)
    
        txt_clip = TextClip(final_text, font="font/VarelaRound-Regular.ttf", fontsize=50, color='white')
        txt_clip = txt_clip.set_pos(("center","top")).set_duration(60)
        
        video = CompositeVideoClip([clip, txt_clip])
    
        video.write_videofile(final_filename)

    # - after loop -
    
    # because I use `number += 1` before loop so now `number` has number of subclips
    print('number of subclips:', number)

    time_end = time.time()
    
    diff = time_end - time_start
    print(f'time: {diff:.2f}s ({diff//60:02.0f}:{diff%60:02.2f})')
    
    
if __name__ == "__main__":
    main()

Затем я переместил код в функцию с аргументами my_process(filename, text, start, end, number, base_folder)

from moviepy.video.io.ffmpeg_tools import ffmpeg_extract_subclip
from moviepy.editor import *
import time

def my_process(filename, text, start, end, number, base_folder):

    clip_duration = end - start
    print(f'[DEBUG] number: {number:2} | start: {start:6.2f} | end: {end:6.2f} | duration: {clip_duration:.2f}')
    
    final_text = f"{number} {text}"

    temp_filename  = f"{base_folder}/cut_{number}.mp4"
    final_filename = f"{base_folder}/result_edit/cut_{number}.mp4"
    
    #print('[DEBUG] ffmpeg_extract_subclip')
    #ffmpeg_extract_subclip(filename, start, end, targetname=temp_filename)
    
    #print('[DEBUG] VideoClip')
    #clip = VideoFileClip(temp_filename)
    clip = VideoFileClip(filename).subclip(start, end)
    clip = clip.volumex(2)

    #print('[DEBUG] TextClip')
    txt_clip = TextClip(final_text, font="font/VarelaRound-Regular.ttf", fontsize=50, color='white')
    txt_clip = txt_clip.set_pos(("center","top")).set_duration(60)
    
    #print('[DEBUG] CompositeVideoClip')
    video = CompositeVideoClip([clip, txt_clip])

    #print('[DEBUG] CompositeVideoClip write')
    video.write_videofile(final_filename)
    #print('[DEBUG] CompositeVideoClip end')
    

def main():

    text = input("Enter Your text please: ") [::-1]
    #text = 'Hello World'
    
    base_folder = time.strftime("result_%Y_%m_%d_%H_%M_%S")    
    os.makedirs(f"{base_folder}/result_edit", exist_ok=True)
    
    filename = "video.mp4"
    #filename = "BigBuckBunny.mp4"
    
    video_duration = VideoFileClip(filename).duration

    number = 0 # instead of `cut_name_num` and `txt_part` because both had the same value

    time_start = time.time()
    
    for start in range(0, int(video_duration), 60):

        end = start + 60
        
        if end > video_duration:
            end = video_duration
        
        number += 1
        
        my_process(filename, text, start, end, number, base_folder)
        
    # - after loop -
    
    # because I use `number += 1` before loop so now `number` has number of subclips
    print('number of subclips:', number)

    time_end = time.time()
    
    diff = time_end - time_start
    print(f'time: {diff:.2f}s ({diff//60:02.0f}:{diff%60:02.2f})')


if __name__ == "__main__":
    main()

И теперь я могу запускать функцию в отдельных процессах, используя стандартный модуль multiprocessing.

(или стандартные модули threading, concurrent.futures или внешние модули Joblib, Ray и т. д. .).

Начинает единый процесс

# it has to use named arguments`target=`, `args=`

p = multiprocessing.Process(target=my_process, args=(filename, text, start, end, number, base_folder))
p.start()  # start it

но если я использую его в цикле, я запускаю много процессов одновременно.

from moviepy.video.io.ffmpeg_tools import ffmpeg_extract_subclip
from moviepy.editor import *
import time
import multiprocessing

def my_process(filename, text, start, end, number, base_folder):

    clip_duration = end - start
    print(f'[DEBUG] number: {number:2} | start: {start:6.2f} | end: {end:6.2f} | duration: {clip_duration:.2f}')
    
    final_text = f"{number} {text}"

    temp_filename  = f"{base_folder}/cut_{number}.mp4"
    final_filename = f"{base_folder}/result_edit/cut_{number}.mp4"
    
    #print('[DEBUG] ffmpeg_extract_subclip')
    #ffmpeg_extract_subclip(filename, start, end, targetname=temp_filename)
    
    #print('[DEBUG] VideoClip')
    #clip = VideoFileClip(temp_filename)
    clip = VideoFileClip(filename).subclip(start, end)
    clip = clip.volumex(2)

    #print('[DEBUG] TextClip')
    txt_clip = TextClip(final_text, font="font/VarelaRound-Regular.ttf", fontsize=50, color='white')
    txt_clip = txt_clip.set_pos(("center","top")).set_duration(60)
    
    #print('[DEBUG] CompositeVideoClip')
    video = CompositeVideoClip([clip, txt_clip])

    #print('[DEBUG] CompositeVideoClip write')
    video.write_videofile(final_filename)
    #print('[DEBUG] CompositeVideoClip end')
    

def main():

    text = input("Enter Your text please: ") [::-1]
    #text = 'Hello World'
    
    base_folder = time.strftime("result_%Y_%m_%d_%H_%M_%S")    
    os.makedirs(f"{base_folder}/result_edit", exist_ok=True)
    
    filename = "video.mp4"
    #filename = "BigBuckBunny.mp4"
    
    video_duration = VideoFileClip(filename).duration

    number = 0 # instead of `cut_name_num` and `txt_part` because both had the same value

    time_start = time.time()
    
    all_processes = []
    for start in range(0, int(video_duration), 60):

        end = start + 60
        
        if end > video_duration:
            end = video_duration
        
        number += 1
        
        print("add process:", number)
        p = multiprocessing.Process(target=my_process, args=(filename, text, start, end, number, base_folder)) # it has to use `target=`, `args=`
        p.start()  # start it
        all_processes.append(p)  # keep it to use `join()`
            
    # - after loop -
    
    for p in all_processes:
        p.join()  # wait for the end of process
        
    # because I use `number += 1` before loop so now `number` has number of subclips
    print('number of subclips:', number)

    time_end = time.time()
    
    diff = time_end - time_start
    print(f'time: {diff:.2f}s ({diff//60:02.0f}:{diff%60:02.2f})')


if __name__ == "__main__":
    main()

Предыдущая версия для 11 подклипов запускает 11 процессов. Используя Pool(4), вы можете объединить все процессы в пул, и он будет запускать 4 процесса одновременно. Когда один процесс завершит задачу, он запустит следующий процесс с новыми аргументами.

На этот раз я использую цикл для создания списка с аргументами для всех процессов.

args_for_all_processes = []

for start in range(0, int(video_duration), 60):

    end = start + 60
    
    if end > video_duration:
        end = video_duration
    
    number += 1
    print("add process:", number)
    
    args_for_all_processes.append( (filename, text, start, end, number, base_folder) )

и я использую этот список с Pool, и он сделает все остальное.

# I have 4 CPU so I use Pool(4) - but without value it should automatically use `os.cpu_count()`
with multiprocessing.Pool(4) as pool:      
    results = pool.starmap(my_process, args_for_all_processes)
    #print(results)

Pool может запускать процессы в другом порядке, но если они используют return для отправки какого-либо результата, то Pool выдаст результаты в правильном порядке.

from moviepy.video.io.ffmpeg_tools import ffmpeg_extract_subclip
from moviepy.editor import *
import time
import multiprocessing

def my_process(filename, text, start, end, number, base_folder):

    clip_duration = end - start
    print(f'[DEBUG] number: {number:2} | start: {start:6.2f} | end: {end:6.2f} | duration: {clip_duration:.2f}')
    
    final_text = f"{number} {text}"

    temp_filename  = f"{base_folder}/cut_{number}.mp4"
    final_filename = f"{base_folder}/result_edit/cut_{number}.mp4"
    
    #print('[DEBUG] ffmpeg_extract_subclip')
    #ffmpeg_extract_subclip(filename, start, end, targetname=temp_filename)
    
    #print('[DEBUG] VideoClip')
    #clip = VideoFileClip(temp_filename)
    clip = VideoFileClip(filename).subclip(start, end)
    clip = clip.volumex(2)

    #print('[DEBUG] TextClip')
    txt_clip = TextClip(final_text, font="font/VarelaRound-Regular.ttf", fontsize=50, color='white')
    txt_clip = txt_clip.set_pos(("center","top")).set_duration(60)
    
    #print('[DEBUG] CompositeVideoClip')
    video = CompositeVideoClip([clip, txt_clip])

    #print('[DEBUG] CompositeVideoClip write')
    video.write_videofile(final_filename)
    #print('[DEBUG] CompositeVideoClip end')
    
    # return "OK"  # you can use `return` to send result/information to main process.
    
def main():

    text = input("Enter Your text please: ") [::-1]
    #text = 'Hello World'
    
    base_folder = time.strftime("result_%Y_%m_%d_%H_%M_%S")    
    os.makedirs(f"{base_folder}/result_edit", exist_ok=True)
    
    filename = "video.mp4"
    #filename = "BigBuckBunny.mp4"
    
    video_duration = VideoFileClip(filename).duration

    number = 0 # instead of `cut_name_num` and `txt_part` because both had the same value

    time_start = time.time()
    
    # first create list with arguments for all processes
    
    args_for_all_processes = []
    
    for start in range(0, int(video_duration), 60):

        end = start + 60
        
        if end > video_duration:
            end = video_duration
        
        number += 1
        print("add process:", number)
        
        args_for_all_processes.append( (filename, text, start, end, number, base_folder) )

    # - after loop -            
        
    # next put all processes to pool
        
    with multiprocessing.Pool(4) as pool:  # I have 4 CPU so I use Pool(4) - but it should use `os.cpu_count()` in `Pool()
    
        results = pool.starmap(my_process, args_for_all_processes)
        #print(results)
            
    # - after loop -
    
    # because I use `number += 1` before loop so now `number` has number of subclips
    print('number of subclips:', number)

    time_end = time.time()
    
    diff = time_end - time_start
    print(f'time: {diff:.2f}s ({diff//60:02.0f}:{diff%60:02.2f})')


if __name__ == "__main__":
    main()