Build a Video to Shorts Pipeline

Step 1 of 9 · 20 minutes, once

Set the Weekly Output Contract

Decide what one recording owes you before you touch a video file. Without a written contract you will ship whatever the week allows, which is usually nothing.

What you get

A written contract that every later step is measured against, and a folder convention that keeps episodes from turning into a pile of loose mp4 files.

Everything downstream is built to satisfy a contract. If you skip this, the idea miner has no quota to hit, the scoring step has nothing to fail against, and you will quietly lower the bar every week you are busy.

The contract we run

Every recording produces, without exception:

ArtifactQuantityGate
Longform episode1Lightly edited, cold open that holds, scores 85 or better
Vertical shorts10 or moreEach one a complete idea, each scores 85 or better on its own
Thumbnails216:9 and 1:1, one design, two crops
Per platform copy1 fileTitles, descriptions, captions, hashtags
Schedule file1Machine readable, feeds whatever you publish with
The quota rule that makes this work. To ship 10 shorts you mine 14 to 16 candidate ideas. You hit quota by mining more, never by lowering the bar. The moment you pad a weak idea to reach a number, the whole pipeline is worth less than posting nothing.

Give every asset one job

An asset that serves two goals serves neither. Assign exactly one to each, and hold a ratio across the week. Ours:

  • Two thirds community. The call itself is the product. The CTA is to come to the next one.
  • One third services. The paid offer. The CTA is to hire you.

Write your own two down now. The copywriting step later reads this ratio and will refuse to guess.

Write your content rules before you have to enforce them

These are ours. Yours will differ, but write them now, because you will meet the edge case at 11pm while rendering.

  1. Never criticize or name a competitor in published content. A clip that does gets recut around it, or killed.
  2. Only approved people appear. Face or voice. On a recorded group call that means everyone who is not on the approved list gets cut out, including their name in captions. If an idea cannot survive without them, kill it and note it for a reshoot.
  3. 85 or nothing. Every artifact passes its own gate, maximum three fix passes, then it is killed or flagged. Never pad a weak idea to ship it.
  4. Kill beats pad. Two great shorts beat six acceptable ones, every time.

The folder convention

One folder per recording, named by date. Finished work lives in episodes/, scratch lives in work/ and is never committed.

The layout every later step expects
episodes/
  <date>/
    shorts/            finished vertical mp4 files
    thumbnails/        thumb-16x9.jpg, thumb-1x1.jpg
    magnets/           lead magnet docs, one per keyword
    shorts-slate.md    the mined ideas and their segments
    copy.md            per platform copy for human review
    schedule.json      machine readable publish plan

work/                  downloads, transcripts, renders (gitignored)
Paste into Claude Code
Write my weekly output contract to docs/CONTRACT.md.

Contract:
- 1 longform episode, lightly edited, cold open, scores 85+
- 10+ vertical shorts, each a complete standalone idea, each scores 85+
- 2 thumbnails: 16:9 and 1:1, one design, two crops
- 1 copy.md with per platform copy
- 1 schedule.json

Quota rule: mine 14-16 candidate ideas so 10+ survive the gate. Never pad a
weak idea to hit the number.

Goal split: [YOUR SPLIT, e.g. two thirds community / one third services]

Content rules:
[PASTE YOUR RULES]

Also create the episodes/ and work/ folder convention described in the file,
and add an episodes/<date>/ template folder I can copy each week.
Check it worked
You have docs/CONTRACT.md committed, and a template episode folder. Read the contract out loud. If any line is vague enough that two people would interpret it differently, tighten it now.
Where this bites
The contract is worthless if it lives in your head. The scoring step later reads it, and so do you at the end of a long week. Commit it to the repo where the pipeline can see it.

Questions

Why 10 shorts and not 3?

Because the gate kills a lot of them. Mining 14 to 16 ideas and killing the weak ones is what produces 10 good shorts. If you set the target at 3 you will mine 4, and you will ship all 4 including the bad one.

What if my recording is only 20 minutes?

Lower the quota, not the gate. A 20 minute call might honestly yield 4 or 5 real ideas. Ship 4 great ones. The bar never moves, only the count.

Talk it through

Stuck on this step?

Free 30 minutes. Bring the error, the transcript, or the clip that will not pass the gate, and we will work through it.