A promo video, made without a video editor

Built with Claude Code. The whole clip came from one prompt:

Make an 8-second, 1080x1080 promo video for Economy of Words (a wall where you
buy words by the letter, $1 a letter). Black background, one word on screen at
a time, each typed on letter by letter like a typewriter with a click per
keystroke, then cleared for the next. Sequence: SAY, LESS, MEAN, MORE, $1, A,
LETTER, 7, DAYS, ON, THE, WALL, ECONOMYOFWORDS.COM, with LESS / MORE / WALL in
green. A small dollar counter in the top-right that climbs $1 per letter as it
types. Open on a blinking block cursor, not a black void. Do not use a video
editor: generate it frame by frame with a script, drawing each frame with
ImageMagick and assembling the clicks and final mp4 with FFmpeg.

This is the first entry in the Toolbox: a series that takes apart a real workflow we used to ship something, names every tool in the stack, and shows the actual commands. If you have never opened a terminal, you should still be able to follow this one.

What we made

Economy of Words is a sister project: a wall where people buy words by the letter, a dollar a letter, and the words ride the wall for a week.

It needed a promo clip. The version we shipped is eight seconds, square, black background. Words type onto the screen one at a time, letter by letter, like an old typewriter, with a click on every keystroke. A small dollar counter in the top corner climbs as the letters land, because the whole pitch is a dollar a letter. It reads like a short headline, one word at a time, then the price, then the web address. Swap the words and you have a different ad; the rest of the machine never changes. (There is a row of variants from this same recipe further down.)

Nothing about that is hard to picture. The interesting part is how it was built.

The idea: the video is a program

We did not open a video editor. There was no timeline, no keyframes, no dragging clips around. The video is the output of a small script. The script draws every single frame as an image, generates the sound, and hands both to a tool that stitches them into a movie file.

A video is just a stack of still images shown quickly (here, 30 images per second) with an audio track playing alongside. If you can produce the images and the audio, you can produce the video. That is the entire trick.

The payoff shows up the moment someone asks for a change. Over the course of building this, the requests were: make it twice as fast, remove one of the words, shrink the text by half, add the dollar counter, fix the opening frame. In a video editor each of those is manual work you do by hand and cannot easily repeat. Here, each was a one line change to the script followed by re-running it. The machine redraws all 400-odd frames the same way every time. That is the point of the Toolbox: workflows you can re-run, not artifacts you babysit.

The stack

Every tool here is free, runs in the terminal, and is already on most Macs or one install away. There are only four pieces.

1. A Bash script: the brain

Bash is the language the terminal speaks. Our script is the recipe. It holds the list of words and their colors, the timing (how many frames each letter and each pause lasts), the typewriter logic, the dollar counter, and the math that centers each word. It loops through the words, and for every step it asks the next tool to draw one frame.

2. ImageMagick: draws every frame

ImageMagick is a tool for making and editing images from the command line. We use its magick command to draw each frame from scratch: a black square, the current word in white or green, the blinking block cursor, and the dollar amount in the corner.

The typewriter effect is not animation in the fancy sense. It is just drawing slightly more text on each successive frame. One frame says W, the next WO, the next WOR, and so on, with a solid block standing in for the cursor. Played back fast, it looks like typing. A single frame is drawn like this:

magick -size 1080x1080 xc:black -gravity West -annotate +120+0 "WOR█" out.png

That command means: make a 1080 by 1080 black image, and write the text WOR followed by a block, anchored to the left. Do that a few hundred times with the text growing, and you have the body of the video.

3. FFmpeg: makes the sound and assembles the movie

FFmpeg is the workhorse for anything audio or video on the command line. It does three jobs here.

First, it inspects the original clip so we know its size, frame rate, and that its audio was effectively silent, with ffprobe, a tool that ships with it.

Second, it synthesizes the typewriter click. There was no click sound file. FFmpeg generates a short burst of noise, shapes it into a tick, and we line up one tick per keystroke so the clicks land exactly when letters appear.

Third, it takes the folder of frames and the assembled audio and encodes them into the final .mp4:

ffmpeg -framerate 30 -i frames/f_%05d.png -i audio.wav -pix_fmt yuv420p out.mp4

The "twice as fast" version is also FFmpeg, one command that speeds the picture and the sound together without making the audio sound chipmunked.

4. The fonts

Two fonts already on the machine. A heavy sans-serif (Arial Bold) for the words and cursor, and a monospaced font (SF Mono) for the dollar counter, so the number has that mechanical, ticker feel.