making-videos

Installation
SKILL.md

Making videos

A good narrated video is a short talk with pictures that move when the words do. The voice carries the story. The screen shows one idea at a time: the keyword, name, path, or line of code being talked about right now. When the idea changes, the picture changes with the sentence. When it doesn't, the picture stays. That is what makes it feel like a video and not a slideshow.

Write for the ear

Write the narration as speech, the way you would explain it to someone sitting next to you. Use short sentences, contractions, one idea per sentence. If a sentence would trip you up when you say it, rewrite it.

Let the story set the shape. Open with why it matters, build the idea in the order a listener needs it, and end on the one thing to remember or do. A fixed template of sections makes it read like a form.

Speech runs at about 150 words a minute. Most explainers fit in 45 to 120 seconds. Cut repetition before you cut substance.

Write say for the voice and show for the eye. In say, name things the way a person would say them out loud: "the webhook handler, around line forty". Don't spell out punctuation ("slash", "dot ts"); the captions show say as written. The exact path goes in show: handle-webhook.ts:40.

Pair every line with a picture

Each line is one spoken sentence, or two short ones, plus what appears at that moment. Put the compressed meaning on screen: the keyword, the claim, the number. Never put the sentence itself there; the captions already carry it. A line without show keeps the current picture.

Group lines into scenes of one to five lines. A new scene is a new beat in the story, and its label is the small kicker above it. Pick the scene kind by the shape of the idea:

Installs
1
GitHub Stars
3
First Seen
Oct 19, 2026