videoanalyzer_skill

Overview

This skill automates downloading, transcribing, and capturing frames from videos to accelerate analysis and archiving. It combines yt-dlp for downloads, ffmpeg for audio/frames, and Whisper for transcripts, producing video, audio, subtitle, plain-text, and frame outputs ready for review.

How this skill works

Point the tool at a video URL and it downloads the source using yt-dlp, extracts audio with ffmpeg, and runs Whisper to generate a plain transcript and SRT subtitles. It also captures periodic screenshots (configurable interval) to create a visual index of the video for quick scanning and documentation.

When to use it

• Archiving or backing up important online video content for research or compliance
• Generating searchable transcripts and subtitles for tutorials, lectures, or podcasts
• Extracting visual highlights from long videos to create quick previews
• Preparing due-diligence or meeting records that require text and frame evidence
• Automating input for downstream summarization or QA workflows

Best practices

• Install yt-dlp, ffmpeg, and Whisper (or openai-whisper) before running the scripts
• Tune the frame interval to balance coverage and storage (e.g., 30s for long talks)
• Choose a Whisper model that matches your accuracy and CPU/GPU limits (small for speed, medium/large for quality)
• Store outputs in a structured directory per video for easy indexing and retrieval
• Post-process transcripts (clean timestamps, remove filler) before automated summarization

Example use cases

• Download a YouTube tutorial, extract a transcript, and capture frames to build documentation
• Archive investor or DD presentations with synchronized subtitles and visual proof points
• Turn long lecture recordings into searchable text and thumbnail timelines for students
• Create podcast episode transcripts for SEO and show notes generation
• Capture meeting recordings and produce a text record plus key visual stills for compliance