Extract frames, contact sheets, and transcripts from video files or URLs for visual analysis.
npx skills add https://github.com/timikalo7/Execute --skill watch-video-timikalo7
Please help me install this Agent Skill. Skill: watch-video Source: https://github.com/timikalo7/Execute/tree/main/.claude/skills/watch-video Command: npx skills add https://github.com/timikalo7/Execute --skill watch-video-timikalo7
High-intent search queries and answers about installing and using this skill.
Run the watch.py script with the video path or URL and a frame count. It extracts evenly-spaced frames and a labelled contact sheet as images, which the agent reads directly, plus an optional transcript with the --transcribe flag.
The script locates ffmpeg via imageio-ffmpeg, computes the video duration, then runs ffmpeg with -ss at evenly-spaced timestamps to save scaled JPEG frames. No system ffmpeg installation or root access is required.
Yes, TikTok and Instagram reels download end-to-end through yt-dlp in this pipeline, and Instagram failures are often transient so retrying once is recommended. YouTube metadata works but media downloads may return 403 from datacenter IPs.
Transcription requires the openai-whisper package installed via pip. Without it, the script still extracts the audio to a WAV file and reports that no whisper backend is available.
A traceback such as a pyo3_runtime PanicException from secretstorage indicates an environment bug, typically an incompatible system cryptography package. Upgrading cryptography with pip install --user --upgrade cryptography resolves it; retrying the URL will not.
yikartCrawl social media URLs to download videos and extract metadata.
scottopellExplore web applications and generate bug reports with screenshots and videos.
joelklaboDiagnose Chrome DevTools errors and UX regressions in the upload-to-convert flow.
ahostbrDownload YouTube videos, extract frames, and caption them with a vision model.
