[Python] WhisperDesktop can only handle a single file at a time, so I tried to make a script...

WhisperDesktop is a OpenAI's Whisper models for speech recognition tasks.

WhisperDeskTop.exe on Windows10 GPU used

WhisperDeskTop.exe .mkv to .srt, but only one at a time

So, if I want to convert all videos, I could only keep clicking. Then, I tried to write a script.

In Power Shell, Python 3.8 installed, create a Python Virtual Environment by python -m venv thename('thename' is just a name of the environment).

Then, activate the environment,used .\Scripts\Activate.ps1. As the picture,

I also install NVIDIA CUDA, CUDNN, ffmpeg, pytorch, and download OpenAI model such as large, medium to my notebook.

All of them have been checked.

What's the problem?

import os
import subprocess
import time
from tqdm import tqdm
from concurrent.futures import ThreadPoolExecutor

video_directory = ""

ffmpeg_command = 'ffmpeg -i "{}" -f wav -vn "{}"'

whisper_command = 'whispercli -gpu -nt -m "D:\\Whisper\\model\\ggml-large-v3-turbo.bin" -l en -nt -osrt -f "{}"'

def convert_video_to_audio(video_path, audio_path, video_name):
ffmpeg_output = subprocess.check_output(
ffmpeg_command.format(video_path, audio_path),
shell=True,
stderr=subprocess.DEVNULL,
)


def gen_audio_srt(audio_path, video_name):
whisper_output = subprocess.check_output(
whisper_command.format(audio_path), shell=True, encoding="utf-8"
)


def process_video(video_file):
video_path = os.path.join(video_directory, video_file)
video_name = os.path.splitext(video_file)[0]
audio_path = os.path.join(video_directory, video_name + ".wav")
srt_path = os.path.join(video_directory, video_name + ".srt")

if os.path.exists(srt_path):
print(f"srt is exisits.")
return 0

convert_video_to_audio(video_path, audio_path, video_name)
gen_audio_srt(audio_path, video_name)
os.remove(audio_path)
return 1

def process_videos():
start_time = time.time()
n = 0
video_files = [
f
for f in os.listdir(video_directory)
if f.endswith((".mp4", ".avi", ".mkv", ".flv", ".mov"))
]

with ThreadPoolExecutor() as executor:
results = list(tqdm(executor.map(process_video, video_files), total=len(video_files), desc="processing... "))

n = sum(results)
end_time = time.time()
print("{:d}videos，token times: {:.2f}s".format(n, end_time - start_time))

if __name__ == "__main__":
path = ""
while True:
path = input("dir: ")
if os.path.exists(path):
break
else:
print(f"{path}path maybe wrong...")
video_directory = path
process_videos()


Please help me finish the Python script.

Continue reading...