javimosch Video Analyzer 🎥
A tool to download, transcribe, and analyze videos from any platform using a smart two-tier system (yt-dlp for fast subtitles, local whisper-cpp for robust fallback).
How to Use
When the user asks you to summarize, transcribe, or download a video/audio from a URL, use the bundled python script:
uv run {baseDir}/scripts/analyze_video.py --action <ACTION> --url "<URL>" [--quality <normal|max>] [--lang <en|it|etc>]
Supported Actions:
transcript: Extracts the text with timestamps. Use this when the user asks for a summary or transcript.download-video: Downloads the video as MP4 to the Desktop.download-audio: Downloads the audio as M4A/MP3 to the Desktop.
Analyzing a Video (IMPORTANT)
If the user asks for a summary, analysis, or key moments:
- Run the script with
--action transcript. - The script will output the path to a
.txtfile containing the timestamped transcript. - Read that file.
- Output your response EXACTLY in this Markdown format:
## 📝 TL;DR
[A punchy 3-sentence summary of the video's core message]
## ⏱️ Key Moments
- [MM:SS] [Brief description of what is discussed]
- [MM:SS] [Brief description of what is discussed]
- [MM:SS] [Brief description of what is discussed]
*(Extract 3 to 7 key moments depending on video length)*
## 💡 Actionable Insights
1. [Practical takeaway 1]
2. [Practical takeaway 2]
3. [Practical takeaway 3]
---
Local Whisper Quality
If the script needs to fall back to Whisper (e.g., for X/Twitter videos), it uses normal by default:
normal: Fast (~1 min for 30 min video) — Defaultmax: Best quality (~5 min for 30 min video) — use--quality maxwhen accuracy is critical
Multilingual Support
All Whisper models are multilingual by default. The skill can transcribe videos in any language (Italian, Spanish, Japanese, etc.).
IMPORTANT: Always respond to the user in THEIR language, not the video's language. If the user speaks Italian but sends an English video, give them the summary in Italian.
Finding specific moments
The transcript includes precise timestamps like [05:53] text.... If the user asks "When do they talk about X?", grep the transcript and return the exact timestamp from the file.