video-processor_skill

Overview

This skill provides a single CLI for downloading and processing videos from YouTube and many other sites. It handles video download, audio extraction, format conversion to MP4 or WebM, and speech-to-text transcription with Whisper. Sensible defaults and dependency checks make it practical for ad-hoc or batch workflows.

How this skill works

The script uses yt-dlp to fetch videos or metadata, FFmpeg to extract and convert audio/video, and OpenAI Whisper for transcription. Commands include download, extract-audio, to-mp4, to-webm, and transcribe; the transcribe flow extracts audio with FFmpeg, runs Whisper, then emits txt/srt/vtt/json. Temporary files are managed and common errors (missing deps, bad inputs) are reported clearly.

When to use it

• Download a YouTube or web-hosted video for local editing or archiving
• Extract audio from a video for podcasts, analysis, or transcription
• Convert legacy or device formats to MP4 for universal playback
• Create web-optimized WebM files for website streaming
• Generate transcripts or subtitle files (SRT/VTT) for accessibility or content repurposing
• Batch process folders of videos for training material or publishing

Best practices

• Install and verify yt-dlp, FFmpeg, and Whisper before running the CLI
• Use 'base' or 'tiny' models for quick drafts; choose 'small' or 'medium' for production accuracy
• Extract audio first for very long videos, then transcribe in segments to reduce memory and runtime
• Prefer WAV or MP3 at 16kHz for optimal Whisper performance; the script handles conversion automatically
• Test WebM VP9 encoding on a short file to tune preset and codec for your speed/size trade-offs

Example use cases

• Convert an old AVI camera file to MP4 for editing and sharing
• Extract lecture audio and generate a plain-text transcript for note-taking
• Transcribe a Spanish interview with the medium model and output SRT for publication
• Convert multiple training videos to WebM and produce transcripts for each file
• Download a YouTube video with subtitles embedded and then generate a cleaned transcript