Short-Form Video Clip Pipeline — Skill

You can automate extracting viral short clips by using an AI pipeline that downloads long YouTube videos, transcribes the audio, and uses Claude to select 30–60 second segments with high standalone value. The pipeline then renders these moments as vertical short clips.

To convert 16:9 YouTube videos to 9:16 vertical format with burned-in captions, the pipeline applies layout-aware cropping and uses Whisper word timestamps to generate TikTok-style word-highlighted ASS captions. FFmpeg is then used to burn these captions directly into the video output.

Yes, you need the Claude API to generate and verify short-form video clips because it handles AI segmentation by selecting segments based on hook strength. A second Claude pass runs cut verification to ensure each clip ends on a complete thought.

You can batch process YouTube channel episodes into TikTok and Reels by scanning a local YouTube knowledge base for new weekly episodes. The pipeline automatically processes these episodes on a per-channel basis to generate ready-to-post vertical clips.

Running a short-form video clipping pipeline requires yt-dlp and ffmpeg for video downloading and transcoding, openai-whisper for word timestamps, and the anthropic API for AI segmentation. MediaPipe and OpenCV are also required for layout-aware vertical reformatting.