Media
Module 02 · by AMLLC
Clip Grazing
Turn one long talking-head recording into a week of captioned vertical clips. You don't chop at random — you graze the whole thing for only the tight, self-contained moments, then cut them 9:16 with burned-in captions. Runs on your own machine, footage never leaves your disk.
Instant access — the reader opens right after payment and the skill file is yours to download. Or read chapter one free as a member.
The result
What you end up with
See it work
Real output from the pipeline
Three clips grazed from one 68-second recording. Each 9:16 clip sits in a 16:9 frame with a blurred side-fill — the trick that keeps every piece of media on your site the same shape.
The math
Stop renting the workflow.
Renting a clipping SaaS
Your own skill
Before you start
What you need
All free, all local. Have these installed and the pipeline runs uninterrupted.
brew install ffmpeg.GET IT →
03Whisper (via pipx)Local transcription with word timestamps. pipx install openai-whisper.GET IT →
04A talking-head recordingAny long-form video where someone is talking — an interview, a demo, a livestream.READ →
The skill file itself is included with the guide — download it from the top of the reader after purchase.
Inside the module
The pipeline, step by step
What grazing actually means
Walk the whole recording and keep only the tight, self-contained moments. The editorial test that separates a keeper from filler.
Get the audio out of the video
Extract a 16kHz mono WAV so transcription is fast and portable — never push the heavy video through a model.
Transcribe with word timestamps
Run Whisper locally, free and private. Why word-level timing is the raw material for every clean cut and caption.
Graze the transcript
Read it like an editor, snap picks to word edges, pad and tighten. Deliver a pick list before you cut anything.
Caption and cut the 9:16 clip
Word-synced captions burned in with libass, Lato lower-third. No-crop for vertical source, center-crop for 16:9.
Consistent media: the 16:9 blurred frame
Fit every vertical clip into a uniform 16:9 frame with a blurred fill so all your page media matches.
Ship it as a skill
Package the workflow as a .skill your assistant loads, triggered by hand — the folder convention and the one spoken command.
Gotchas from the real build
Blocked model downloads, cloud-only files that won't mount, ffmpeg eating stdin. The failures, and the one-line fixes.
Who it's for
Made for the person who records
You should read this if you make long-form video — interviews, demos, livestreams, podcasts with a camera — and want a steady stream of vertical clips without paying a SaaS or an editor every month.
Skip it if you don't record talking-head content, or you're happy uploading your raw footage to someone else's cloud and living inside their export caps.
Or skip the reading entirely — apply to have the whole pipeline and skill set up on your machine for you, from $500.
Two doors
Read it, or hand it to us.
Get the guide
The full 8-chapter reader with every command, plus the packaged skill file to download and load into your assistant. Yours to keep, updated as the pipeline improves.
Have it built
We set up the whole pipeline on your machine — ffmpeg, local Whisper, the skill, and your folder convention — and cut a first batch with you. Apply, we scope it, you approve a fixed price first.
Apply to have it built