\u751f\u6210\u89c6\u9891\u5b57\u5e55\u662f\u8bb8\u591a\u89c6\u9891\u5904\u7406\u4efb\u52a1\u7684\u6838\u5fc3\u9700\u6c42\u3002\u672c\u6587\u5c06\u6307\u5bfc\u4f60\u4f7f\u7528 OpenAI \u7684 Whisper \u6a21\u578b\u4e3a\u89c6\u9891\u6587\u4ef6（\u5982\u7535\u89c6\u5267\u300aNormal People\u300b\u6216\u7535\u5f71\u300a\u82b1\u6837\u5e74\u534e\u300b）\u751f\u6210\u5b57\u5e55（SRT \u683c\u5f0f）\u3002\u6211\u4eec\u5c06\u4ece\u63d0\u53d6\u97f3\u9891\u5f00\u59cb，\u9010\u6b65\u5b9e\u73b0\u5b57\u5e55\u751f\u6210，\u5e76\u63d0\u4f9b\u4e00\u4e2a Python \u811a\u672c\u5b9e\u73b0\u6279\u91cf\u5904\u7406\u3002\u6b64\u5916，\u6211\u4eec\u8fd8\u5c06\u63a2\u8ba8\u5982\u4f55\u5904\u7406\u975e\u82f1\u8bed\u97f3\u9891（\u5982\u4e2d\u6587）\u5e76\u4f18\u5316\u5b57\u5e55\u8d28\u91cf\u3002<\/p>\n
\u5728\u5f00\u59cb\u4e4b\u524d，\u8bf7\u786e\u4fdd\u5b89\u88c5\u4ee5\u4e0b\u5de5\u5177：<\/p>\n
1.\u00a0FFmpeg：\u7528\u4e8e\u4ece\u89c6\u9891\u63d0\u53d6\u97f3\u9891\u3002<\/p>\n
2.\u00a0Python 3.8+：\u7528\u4e8e\u8fd0\u884c\u811a\u672c\u548c Whisper\u3002<\/p>\n
3.\u00a0Whisper：OpenAI \u7684\u8bed\u97f3\u8f6c\u6587\u5b57\u6a21\u578b\u3002<\/p>\n
4.\u00a0uv（\u53ef\u9009）：\u7528\u4e8e\u7ba1\u7406 Python \u9879\u76ee\u73af\u5883\u3002<\/p>\n
5.\u00a0\u89c6\u9891\u6587\u4ef6：\u51c6\u5907 MP4 \u6216 MKV \u683c\u5f0f\u7684\u89c6\u9891\u6587\u4ef6（\u5982\u300aNormal People\u300b\u6216\u300a\u82b1\u6837\u5e74\u534e\u300b）\u3002<\/p>\n
\u7b2c\u4e00\u6b65\u662f\u4ece\u89c6\u9891\u6587\u4ef6\u4e2d\u63d0\u53d6\u97f3\u9891\u3002\u6211\u4eec\u4f7f\u7528 FFmpeg \u5c06\u89c6\u9891\u7684\u97f3\u9891\u6d41\u4fdd\u5b58\u4e3a AAC \u683c\u5f0f\u3002<\/p>\n
\u4e3a\u300aNormal People\u300b\u7b2c1\u5b63\u7b2c1\u96c6\u63d0\u53d6\u97f3\u9891：<\/p>\n
ffmpeg -i \/path\/to\/Normal.People.S01E01.mp4 -vn -acodec copy \/path\/to\/audio\/Normal.People.S01E01.aac <\/p>\n
\u4f7f\u7528 Whisper \u6a21\u578b\u5c06\u97f3\u9891\u6587\u4ef6\u8f6c\u6362\u4e3a SRT \u683c\u5f0f\u7684\u5b57\u5e55\u6587\u4ef6\u3002Whisper \u652f\u6301\u591a\u79cd\u6a21\u578b（\u5982\u00a0tiny\u3001base\u3001small\u3001medium\u3001large\u00a0\u548c\u00a0turbo），turbo\u00a0\u901f\u5ea6\u5feb，\u9002\u5408\u5feb\u901f\u6d4b\u8bd5\u3002<\/p>\n
\u4e3a\u63d0\u53d6\u7684\u97f3\u9891\u751f\u6210\u5b57\u5e55：<\/p>\n
whisper \/path\/to\/audio\/Normal.People.S01E01.aac –model turbo –output_format srt –output_dir \/path\/to\/generated_subs\/ <\/p>\n
\u751f\u6210\u7684\u524d\u51e0\u6761\u5b57\u5e55\u53ef\u80fd\u5982\u4e0b：<\/p>\n
1
\n00:00:00,000 –> 00:00:24,000
\nIt's a simple game. You have 15 players. Give one of them the ball.
\nGet it into the net. <\/p>\n
2
\n00:00:24,000 –> 00:00:26,000
\nVery simple. Isn't it? <\/p>\n
\u624b\u52a8\u4e3a\u591a\u4e2a\u89c6\u9891\u751f\u6210\u5b57\u5e55\u6548\u7387\u4f4e\u4e0b\u3002\u4ee5\u4e0b Python \u811a\u672c\u81ea\u52a8\u5904\u7406\u76ee\u5f55\u4e2d\u7684\u6240\u6709\u89c6\u9891\u6587\u4ef6，\u63d0\u53d6\u97f3\u9891\u5e76\u751f\u6210\u5b57\u5e55\u3002<\/p>\n
import os
\nimport subprocess
\nimport argparse <\/p>\n
defextract_audio(input_dir, output_dir):
\n """Extract audio from video files in input_dir and save to output_dir."""
\n ifnot os.path.exists(output_dir):
\n os.makedirs(output_dir)
\n for filename in os.listdir(input_dir):
\n if filename.endswith(('.mp4', '.mkv')):
\n input_path = os.path.join(input_dir, filename)
\n audio_filename = os.path.splitext(filename)[0] + '.aac'
\n output_path = os.path.join(output_dir, audio_filename)
\n command = [
\n 'ffmpeg', '-i', input_path, '-vn', '-acodec', 'copy', output_path
\n ]
\n print(f"Extracting audio: {command}")
\n try:
\n subprocess.run(command, check=True)
\n except subprocess.CalledProcessError as e:
\n print(f"Error extracting audio from {filename}: {e}") <\/p>\n
defgenerate_subtitles(input_dir, output_dir):
\n """Generate subtitles for audio files using Whisper."""
\n ifnot os.path.exists(output_dir):
\n os.makedirs(output_dir)
\n for filename in os.listdir(input_dir):
\n if filename.endswith('.aac'):
\n input_path = os.path.join(input_dir, filename)
\n command = [
\n 'whisper', input_path, '–model', 'turbo',
\n '–output_format', 'srt', '–output_dir', output_dir
\n ]
\n print(f"Generating subtitles: {command}")
\n try:
\n subprocess.run(command, check=True)
\n except subprocess.CalledProcessError as e:
\n print(f"Error generating subtitles for {filename}: {e}") <\/p>\n
if __name__ == "__main__":
\n parser = argparse.ArgumentParser(description="Extract audio and generate subtitles.")
\n parser.add_argument("input_dir", help="Directory containing video files.")
\n parser.add_argument("audio_dir", help="Directory to save extracted audio files.")
\n parser.add_argument("subtitle_dir", help="Directory to save generated subtitles.")
\n args = parser.parse_args()
\n extract_audio(args.input_dir, args.audio_dir)
\n generate_subtitles(args.audio_dir, args.subtitle_dir) <\/p>\n
python generate_subtitles.py \/path\/to\/videos \/path\/to\/audio \/path\/to\/generated_subs <\/p>\n
\u751f\u6210\u7684\u5b57\u5e55\u53ef\u80fd\u5b58\u5728\u4ee5\u4e0b\u95ee\u9898，\u6211\u4eec\u63d0\u4f9b\u4f18\u5316\u65b9\u6cd5：<\/p>\n
import pysrt
\nsubs = pysrt.open('subtitles.srt')
\nfor sub in subs:
\n if sub.start.seconds < 18:
\n sub.shift(seconds=18)
\nsubs.save('adjusted_subtitles.srt') <\/p>\n
import nltk
\nnltk.download('punkt')
\nfrom nltk.tokenize import sent_tokenize <\/p>\n
def split_long_subtitle(text):
\n return sent_tokenize(text) <\/p>\n
long_text = "It's a simple game. You have 15 players. Give one of them the ball."
\nsentences = split_long_subtitle(long_text) # \u8f93\u51fa：['It's a simple game.', 'You have 15 players.', …] <\/p>\n
import spacy
\nnlp = spacy.load("en_core_web_sm")
\ntext = "It's a simple game You have 15 players"
\ndoc = nlp(text)
\npunctuated_text = " ".join(token.text_with_ws for token in doc) # \u8f93\u51fa：It's a simple game. You have 15 players. <\/p>\n
\u751f\u6210\u4e2d\u6587\u5b57\u5e55\u5e76\u7ffb\u8bd1\u4e3a\u82f1\u6587：<\/p>\n
whisper \/path\/to\/In.the.Mood.for.Love.mp4 –model large –output_format srt –output_dir \/path\/to\/generated_subs –language zh –task transcribe <\/p>\n
ffmpeg -i input.mp4 -af "afftdn" -vn -acodec copy output.aac <\/p>\n
\u901a\u8fc7 FFmpeg \u548c Whisper，\u53ef\u4ee5\u8f7b\u677e\u4e3a\u89c6\u9891\u751f\u6210\u9ad8\u8d28\u91cf\u5b57\u5e55\u3002\u6279\u91cf\u5904\u7406\u811a\u672c\u81ea\u52a8\u5316\u4e86\u63d0\u53d6\u97f3\u9891\u548c\u751f\u6210\u5b57\u5e55\u7684\u8fc7\u7a0b，\u4f18\u5316\u65f6\u95f4\u6233\u3001\u5b57\u5e55\u957f\u5ea6\u548c\u6807\u70b9\u7684\u65b9\u6cd5\u8fdb\u4e00\u6b65\u63d0\u5347\u4e86\u5b57\u5e55\u8d28\u91cf\u3002\u5bf9\u4e8e\u975e\u82f1\u8bed\u97f3\u9891（\u5982\u4e2d\u6587），\u4f7f\u7528 large \u6a21\u578b\u3001\u9884\u5904\u7406\u97f3\u9891\u548c\u5206\u79bb\u8f6c\u5f55\u7ffb\u8bd1\u662f\u5173\u952e\u3002 \u00a0<\/p>\n","protected":false},"excerpt":{"rendered":"
\u6587\u7ae0\u6d4f\u89c8\u9605\u8bfb424\u6b21\uff0c\u70b9\u8d5e4\u6b21\uff0c\u6536\u85cf11\u6b21\u3002\u672c\u6587\u5c06\u6307\u5bfc\u4f60\u4f7f\u7528 OpenAI \u7684 Whisper \u6a21\u578b\u4e3a\u89c6\u9891\u6587\u4ef6\uff08\u5982\u7535\u89c6\u5267\u300aNormal People\u300b\u6216\u7535\u5f71\u300a\u82b1\u6837\u5e74\u534e\u300b\uff09\u751f\u6210\u5b57\u5e55\uff08SRT \u683c\u5f0f\uff09\u3002\u6211\u4eec\u5c06\u4ece\u63d0\u53d6\u97f3\u9891\u5f00\u59cb\uff0c\u9010\u6b65\u5b9e\u73b0\u5b57\u5e55\u751f\u6210\uff0c\u5e76\u63d0\u4f9b\u4e00\u4e2a Python \u811a\u672c\u5b9e\u73b0\u6279\u91cf\u5904\u7406\u3002\u6279\u91cf\u5904\u7406\u811a\u672c\u81ea\u52a8\u5316\u4e86\u63d0\u53d6\u97f3\u9891\u548c\u751f\u6210\u5b57\u5e55\u7684\u8fc7\u7a0b\uff0c\u4f18\u5316\u65f6\u95f4\u6233\u3001\u5b57\u5e55\u957f\u5ea6\u548c\u6807\u70b9\u7684\u65b9\u6cd5\u8fdb\u4e00\u6b65\u63d0\u5347\u4e86\u5b57\u5e55\u8d28\u91cf\u3002\u4ee5\u4e0b Python \u811a\u672c\u81ea\u52a8\u5904\u7406\u76ee\u5f55\u4e2d\u7684\u6240\u6709\u89c6\u9891\u6587\u4ef6\uff0c\u63d0\u53d6\u97f3\u9891\u5e76\u751f\u6210\u5b57\u5e55\u3002\uff1a\u51c6\u5907 MP4 \u6216 MKV \u683c\u5f0f\u7684\u89c6\u9891\u6587\u4ef6\uff08\u5982\u300aNormal People\u300b\u6216\u300a\u82b1\u6837\u5e74\u534e\u300b\uff09\u3002<\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[66,3065,50,3356,202,214],"topic":[],"class_list":["post-38361","post","type-post","status-publish","format-standard","hentry","category-server","tag-ai","tag-whisper","tag-50","tag-3356","tag-202","tag-214"],"yoast_head":"\n