Video

Requirements

Required:

• ffmpeg / ffprobe — core video processing

Optional:

• whisper — local transcription for captions
• realesrgan — AI upscaling

Quick Reference

Situation	Load
Platform specs (YouTube, TikTok, Instagram)	platforms.md
FFmpeg commands by task	commands.md
Quality/compression settings	quality.md
Workflow by use case	workflows.md

Core Capabilities

Task	Method
Convert/compress	FFmpeg (see commands.md)
Generate captions	Whisper → SRT/VTT
Change aspect ratio	Crop, pad, or smart reframe
Clean audio	Normalize, denoise, enhance
Batch operations	Process entire folders in one run

Execution Pattern

1. Clarify target — What platform? What format? File size limit?
2. Check source — ffprobe for codec, resolution, duration, audio
3. Process — FFmpeg for transformation
4. Verify — Confirm output meets specs before delivering
5. Deliver — Provide file to user

Common Requests → Actions

User says	Agent does
"Make this work for TikTok"	Reframe to 9:16, check duration ≤3min, compress
"Add subtitles"	Whisper → SRT → burn-in or deliver separately
"Compress for WhatsApp"	Target <64MB, H.264, AAC
"Extract audio"	-vn -acodec mp3 or -acodec copy
"Make a GIF"	Extract frames, optimize palette, loop
"Split into clips"	Cut at timestamps with -ss and -t

Quality Rules

• Always re-encode audio to AAC for maximum compatibility
• Use -movflags +faststart for web playback
• CRF 23 is good default for H.264 (lower = better, bigger)
• Check before delivering — verify duration, file size, playability

Platform Quick Reference

Platform	Aspect	Max Duration	Max Size
TikTok	9:16	3 min	287MB
Instagram Reels	9:16	90s	250MB
YouTube Shorts	9:16	60s	No limit
YouTube	16:9	12h	256GB
WhatsApp	Any	3 min	64MB

Scope

This skill:

• Processes video files user explicitly provides
• Runs FFmpeg commands on user request
• Does NOT access files without user instruction
• Does NOT upload to external services automatically