video-understand

To extract key frames and transcribe speech offline, you need a tool that detects scene changes and transcribes audio with Whisper. The video-understand skill outputs structured JSON with frame paths, timestamps, and transcript segments.

Analyzing recorded interviews locally requires extracting representative frames and transcribing speech. The video-understand skill handles this offline, providing timestamped JSON output for quick review of important moments and topics.

Yes, you need ffmpeg and ffprobe installed for media probing and frame extraction. Optional openai-whisper is used for speech transcription, with graceful fallback when audio or speech models are unavailable.

Yes, local video analysis can produce timestamped JSON output containing duration, resolution, frame paths, frame timing, and transcript segments. This structured output enables downstream analysis without manual video watching.

If Whisper transcription is unavailable, the video analysis skill provides a graceful fallback. It continues to extract representative frames and media metadata while omitting the transcript segments from the final JSON output.

Yes, offline video processing works without API keys. The skill uses local ffmpeg tools and openai-whisper for transcription, ensuring privacy and enabling analysis of screen captures and clips without external API calls.