html_content += f"<p>{entry['time_sec']} </p> <br><br>"

app.py

@@ -8,6 +8,12 @@ from openai import OpenAI
 import json
 from youtube_transcript_api import YouTubeTranscriptApi
 
+from moviepy.editor import VideoFileClip
+from pytube import YouTube
+import os
+
+OUTPUT_PATH = 'videos'
+
 
 OPEN_AI_KEY = os.getenv("OPEN_AI_KEY")
 client = OpenAI(api_key=OPEN_AI_KEY)
@@ -74,7 +80,8 @@ def process_youtube_link(link):
             "start_time": start_time,
             "end_time": end_time,
             "text": entry['text'],
-            "embed_url": embed_url
+            "embed_url": embed_url,
+            "time_sec": entry['start']
         }
         formatted_transcript.append(line)
     
@@ -95,13 +102,62 @@ def format_transcript_to_html(formatted_transcript):
     for entry in formatted_transcript:
         html_content += f"<h3>{entry['start_time']} - {entry['end_time']}</h3>"
         html_content += f"<p>{entry['text']}</p>"
-        html_content += f"<iframe width='560' height='315' src='{entry['embed_url']}' frameborder='0' allowfullscreen></iframe><br><br>"
+        html_content += f"<p>{entry['time_sec']} </p>  <br><br>"
     return html_content
 
 def get_embedded_youtube_link(video_id, start_time):
     embed_url = f"https://www.youtube.com/embed/{video_id}?start={start_time}&autoplay=1"
     return embed_url
 
+def download_youtube_video(youtube_id, output_path=OUTPUT_PATH):
+    # Construct the full YouTube URL
+    youtube_url = f'https://www.youtube.com/watch?v={youtube_id}'
+
+    # Create the output directory if it doesn't exist
+    if not os.path.exists(output_path):
+        os.makedirs(output_path)
+
+    # Download the video
+    yt = YouTube(youtube_url)
+    video_stream = yt.streams.filter(progressive=True, file_extension='mp4').order_by('resolution').desc().first()
+    video_stream.download(output_path=output_path, filename=youtube_id+".mp4")
+
+    print(f"Video downloaded successfully: {output_path}/{youtube_id}.mp4")
+
+
+def screenshot_youtube_video(youtube_id, snapshot_sec):
+    # 由于在 Hugging Face Spaces 中直接下载 YouTube 视频可能不可行，
+    # 您可能需要调整这部分代码，例如允许用户上传视频。
+    # ... 代码 ...
+
+    # 这里假设视频已经在适当的位置
+    video_path = f'{OUTPUT_PATH}/{youtube_id}.mp4'
+
+    # Load the video and take a screenshot
+    with VideoFileClip(video_path) as video:
+        screenshot_path = f'{OUTPUT_PATH}/screenshots/{youtube_id}_{snapshot_sec}.jpg'
+        video.save_frame(screenshot_path, snapshot_sec)
+    
+    return screenshot_path
+
+def process_video(youtube_id):
+    download_youtube_video(youtube_id)
+    video_path = f'{OUTPUT_PATH}/{youtube_id}.mp4'
+    video = VideoFileClip(video_path)
+    duration = int(video.duration)
+    output_path = f'{OUTPUT_PATH}/screenshots/{youtube_id}'
+    os.makedirs(output_path, exist_ok=True)
+
+    #  fake duration
+    duration = 10
+
+    screenshot_paths = []
+    for i in range(1, duration):
+        screenshot_path = screenshot_youtube_video(youtube_id, i)
+        screenshot_paths.append(screenshot_path)
+
+    return screenshot_paths
+
 def get_screenshot_from_video(video_link, start_time):
     # 实现从视频中提取帧的逻辑
     # 由于这需要服务器端处理，你可能需要一种方法来下载视频，

requirements.txt

@@ -4,4 +4,6 @@ openai>=1.0.0
 requests
 beautifulsoup4
 python-docx
-youtube-transcript-api
+youtube-transcript-api
+moviepy
+pytube