Video

Requirements

Required:

ffmpeg / ffprobe — core video processing

Optional:

whisper — local transcription for captions realesrgan — AI upscaling Quick Reference Situation Load 平台 specs (YouTube, TikTok, Instagram) 平台s.md FFmpeg commands by task commands.md 质量/压缩ion 设置tings 质量.md 工作流 by use case 工作流s.md Core Capabilities Task Method Convert/压缩 FFmpeg (see commands.md) 生成 captions Whisper → SRT/VTT Change aspect ratio Crop, pad, or smart reframe 清理 audio Normalize, denoise, enhance Batch operations Process entire folders in one 运行 Execution Pattern Clarify tar获取 — What 平台? What 格式化? File size limit? 检查 source — ffprobe for codec, resolution, duration, audio Process — FFmpeg for trans格式化ion 验证 — Confirm 输出 meets specs before delivering Deliver — Provide file to user Common 请求s → Actions User says 代理 does "Make this work for TikTok" Reframe to 9:16, 检查 duration ≤3min, 压缩 "添加 subtitles" Whisper → SRT → burn-in or deliver separately "压缩 for Whats应用" Tar获取 <64MB, H.264, AAC "提取 audio" -vn -acodec mp3 or -acodec copy "Make a GIF" 提取 frames, 优化 palette, loop "Split into 命令行工具ps" Cut at timestamps with -ss and -t 质量 Rules Always re-encode audio to AAC for maximum compatibility Use -movflags +fast启动 for 网页 playback CRF 23 is good default for H.264 (lower = better, bigger) 检查 before delivering — 验证 duration, file size, playability 平台 Quick Reference 平台 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 Whats应用 Any 3 min 64MB Scope

This 技能:

Processes video files user explicitly provides 运行s FFmpeg commands on user 请求 Does NOT 访问 files without user instruction Does NOT 上传 to external 服务s automatically