initial_whisper_script

create_srt.py

import subprocess as sp
import whisper

#replace this with the path to the video.
sourcefile = r'Z:\completeTorrents\Ai Uehara MEGAPACK\SHKD525.FHD\SHKD525.FHD.wmv'

start_time_minutes = 0.0

model = whisper.load_model("small")
slice_size = 30

i=0
with open('subs.srt','w') as subsfile:
    subsfile.write('\n')
    for st in range(int(start_time_minutes*60), 900000, slice_size):
        proc = sp.Popen(['ffmpeg','-y','-ss',str(st),'-i',sourcefile, '-t', str(slice_size), 'out.wav'], stderr=sp.DEVNULL)
        proc.communicate()

        result = model.transcribe("out.wav",language='Japanese', task='translate')
        for sv in result['segments']:

            i+=1
            
            time_start = st+sv['start']
            hours, minutes, seconds = int(time_start/3600), (time_start/60) % 60, (time_start) % 60
            timestamp_start = "%02d:%02d:%06.3f" % (hours, minutes, seconds)
            timestamp_start = timestamp_start.replace('.', ',')

            time_end = st+sv['end']
            hours, minutes, seconds = int(time_end/3600), (time_end/60) % 60, (time_end) % 60
            timestamp_end = "%02d:%02d:%06.3f" % (hours, minutes, seconds)
            timestamp_end = timestamp_end.replace('.', ',')

            subsline = "{i}\n{s} --> {e}\n{text}".format(i=i, s=timestamp_start, e=timestamp_end, text=sv['text'].strip())
            print(subsline)
            try:
                subsfile.write(subsline+'\n\n')
            except Exception as e:
                pass

        subsfile.flush()