Video News Downloader with AI Subtitle Proofreading

Complete workflow for downloading daily news videos, processing subtitles, and serving them via HTTP with web players.

Overview

This skill automates:

1. Video Download: CBS Evening News + BBC News at Ten from YouTube
2. Subtitle Processing: Extract auto-captions and convert to VTT format
3. AI Proofreading: Use DeepSeek to fix speech recognition errors
4. HTTP Streaming: Serve videos with embedded web players
5. Scheduled Updates: Daily cron jobs at configurable times

Quick Start

1. Download Latest News

python3 scripts/video_download.py --cbs --bbc

2. Proofread Subtitles

python3 scripts/subtitle_proofreader.py /path/to/subtitle.vtt

Or use DeepSeek directly:

> "校对字幕文件 /path/to/subtitle.vtt"

3. Start HTTP Servers

bash scripts/setup_server.sh

4. Setup Daily Cron Jobs

bash scripts/setup_cron.sh

Commands

Video Download Script

Download CBS only:

python3 scripts/video_download.py --cbs

Download BBC only:

python3 scripts/video_download.py --bbc

Download both:

python3 scripts/video_download.py --cbs --bbc

With subtitle proofreading:

python3 scripts/video_download.py --cbs --bbc --proofread

Subtitle Proofreading

Proofread single file:

python3 scripts/subtitle_proofreader.py <vtt_file_path>

Auto-proofread all news subtitles:

python3 scripts/subtitle_proofreader.py --all

Server Management

Start servers:

bash scripts/setup_server.sh start

Check status:

bash scripts/setup_server.sh status

Stop servers:

bash scripts/setup_server.sh stop

File Structure

/workspace/
├── cbs-live-local/
│   ├── cbs_latest.mp4
│   ├── cbs_latest.en.vtt          # Original subtitle
│   ├── cbs_latest.en.vtt-backup   # Backup
│   ├── cbs_latest-corrected.txt   # DeepSeek corrected text
│   └── cbs_latest-corrections.md  # Error list
│
├── bbc-news-live/
│   ├── bbc_news_latest.mp4
│   ├── bbc_news_latest.en.vtt
│   ├── bbc_news_latest.en.vtt-backup
│   ├── bbc_news_latest-corrected.txt
│   └── bbc_news_latest-corrections.md
│
└── temp/                           # Temporary download files

HTTP Endpoints

Cron Jobs

Default Schedule (Beijing Time)

Manual Cron Setup

See references/cron-setup.md for detailed cron configuration.

DeepSeek Proofreading

What Gets Fixed

• Speech recognition errors (e.g., "noraster" → "nor'easter")
• Name errors (e.g., "trunk" → "Trump")
• Location name errors
• Professional terminology errors
• Obvious spelling mistakes

Output Files

For each subtitle file, generates:

1. -backup.vtt - Original subtitle (never modified)
2. -corrected.txt - AI-corrected plain text
3. -corrections.md - List of corrections made

Troubleshooting

Video Download Fails

• Check yt-dlp is installed: yt-dlp --version
• Check YouTube URL is accessible
• Try manual download first

Subtitle Extraction Fails

• Some videos don't have auto-captions
• Check if --list-subs shows available languages

Server Won't Start

• Check ports 8093/8095 are free: lsof -i :8093
• Check Python http.server is available

Proofreading Issues

• Ensure DeepSeek model is available
• Check subtitle file exists and is valid VTT format

See Also

• references/workflow.md - Detailed workflow documentation
• references/cron-setup.md - Cron job configuration guide