video-content-watcher

Extract core content and generate structured analysis reports from YouTube, Bilibili, or local video files. Triggers when user asks to summarize, analyze, extract key points from a video, or assess video value. Supports Chinese/English video content analysis with MiniMax LLM.

Extract audio → transcribe → analyze → generate structured Markdown report.

Quick Start

bash

cd /home/ykl/.openclaw/workspace-code-dev/video-reader
export MINIMAX_API_KEY="your-key"
export WHISPER_MODE="local"   # or "api"
python3 << 'EOF'
import sys; sys.path.insert(0, 'src')
from video_reader_mcp import VideoReaderMCP
mcp = VideoReaderMCP()
result = mcp.process_url('VIDEO_URL')
print(result['report'])
EOF

Or use the CLI script:

bash

python3 /home/ykl/.openclaw/workspace/skills/video-content-watcher/scripts/video_reader.py "VIDEO_URL"

Workflow

1. Audio extraction: Uses yt-dlp for online videos / ffmpeg for local files
2. Transcription: Whisper (local openai-whisper or API mode)
3. Analysis: MiniMax LLM analyzes transcript → structured report

Configuration (env vars)

Variable	Required	Default	Description
MINIMAX_API_KEY	Yes	—	MiniMax API key
WHISPER_MODE	No	api	api or local
WHISPER_API_KEY	If mode=api	—	OpenAI API key for Whisper
LLM_BASE_URL	No	https://api.minimaxi.com/v1	MiniMax endpoint
LLM_MODEL	No	MiniMax-M2.7	Model name

Supported Sources

• YouTube: https://www.youtube.com/watch?v=xxx
• Bilibili: https://www.bilibili.com/video/BVxxx
• Local files: /path/to/video.mp4

Output

Returns a Markdown report with:

• Basic info (source, transcript length)
• Core content summary
• Key takeaways
• Main viewpoints
• Pros / cons
• Target audience

Scripts

• scripts/video_reader.py — CLI entry point
• scripts/analyze_text.py — Analyze raw text directly (skip video processing)