video-analyzer

Analyze short-form videos by extracting representative frames using scene-change sampling and optional audio transcription to summarize scene flow, tone, and visible details. This Skill processes local mp4s, Shorts, TikToks, reels, and screen recordings to provide frame-based insights and multi-part synthesis.

Extract key frames from an mp4 using a scene-change-first sampling strategy with an even-spacing fallback for dense or sparse cuts. This approach probes duration and codecs with ffprobe, then isolates representative frames to support visual summarization and scene-by-scene analysis.

Transcribe audio from local video files using optional pocketsphinx transcription when audio data is available. The Skill probes audio presence with ffprobe and integrates the transcription into a concise summary alongside frame analysis and uncertainty notes.

ffprobe is required to gather duration, codecs, resolution, and audio presence before frame extraction begins. The Skill implements an end-to-end workflow using ffprobe for metadata probing and scene-change sampling to isolate representative frames for summarization.

Limitations when analyzing screen recordings include explicit uncertainty notes in the synthesis and a fallback to even-spacing frame sampling when scene-change detection yields dense or sparse cuts. The Skill provides a deeper scene-by-scene analysis but acknowledges visual summarization constraints.

Summarize scene flow and tone from a local video file by fusing extracted frame insights and optional audio transcription into a concise summary with deeper scene-by-scene analysis. The Skill describes subject, scene progression, actions, tone, humor, visible text, and notable details.