import sys
import os
import argparse
from dotenv import load_dotenv
from genSrt import transcribe_video, download_video
from translator import Translator
def setup(translate_to_lang: str = 'none'):
load_dotenv() # 環境変数を読み込む
api_key = os.getenv("OPENAI_API_KEY")
translator = Translator(api_key, translate_to_lang)
return translator
def main():
parser = argparse.ArgumentParser(description="Transcribe and translate videos.")
parser.add_argument('input', type=str, help='URL or local path of the video')
parser.add_argument('--device', type=str, choices=['cuda', 'cpu'], default='cuda', help='Device to use for inference (default: cuda)')
parser.add_argument('--model', type=str, default='large-v3', help='Whisper model size (default: large-v3)')
parser.add_argument('--input-lang', type=str, default=None, help='Input language code (default: auto-detect)')
parser.add_argument('--output-lang', type=str, default=None, help='Output language code for translation (default: no translation)')
# 精度向上オプション
parser.add_argument('--initial-prompt', type=str, default=None, help='Initial prompt with hints for specialized terms')
parser.add_argument('--condition-on-previous-text', action='store_true', help='Enable conditioning on previous text (default: disabled to prevent hallucination)')
parser.add_argument('--temperature', type=float, default=0.0, help='Temperature for sampling (default: 0.0)')
parser.add_argument('--no-speech-threshold', type=float, default=0.6, help='No speech threshold (default: 0.6)')
parser.add_argument('--compression-ratio-threshold', type=float, default=2.4, help='Compression ratio threshold (default: 2.4)')
parser.add_argument('--no-vad-filter', action='store_true', help='Disable VAD filter (default: enabled)')
args = parser.parse_args()
input_arg = args.input
device = args.device
model_size = args.model
input_lang = args.input_lang
output_lang = args.output_lang
initial_prompt = args.initial_prompt
condition_on_previous_text = args.condition_on_previous_text
temperature = args.temperature
no_speech_threshold = args.no_speech_threshold
compression_ratio_threshold = args.compression_ratio_threshold
vad_filter = not args.no_vad_filter
translator = None
if output_lang is not None:
translator = setup(output_lang)
# URLで始まる場合
if input_arg.startswith("https://"):
output_path = 'output'
original_title, video_extension, timestamp = download_video(input_arg, output_path)
downloaded_file_path = os.path.join(output_path, f"{timestamp}.{video_extension}")
# ローカルファイルの場合
else:
downloaded_file_path = os.path.abspath(input_arg)
output_path = os.path.dirname(downloaded_file_path)
original_title = os.path.splitext(os.path.basename(downloaded_file_path))[0]
video_extension = os.path.splitext(downloaded_file_path)[1][1:]
print(f"output_path: {output_path}")
transcribe_video(
downloaded_file_path,
output_path,
translator,
input_lang,
output_lang,
device,
model_size,
initial_prompt,
condition_on_previous_text,
temperature,
no_speech_threshold,
compression_ratio_threshold,
vad_filter,
)
if __name__ == "__main__":
main()
