THREE KINDS OF PROMPT
Sort the prompts by length and three distinct styles fall out. They aren't just longer or shorter versions of the same thing; they hand Claude very different amounts of the creative work.
The one-liner (under 40 words)
180 of 277 prompts. Almost half of these are the viral brief: "make a dynamic 15-second motion graphics video that shows what an incredible motion designer you are, like it's your showreel for a résumé. go all out." Claude makes every creative decision.
The brief (40–150 words)
60 prompts. A subject, an audience and a feel: a SaaS launch video, a history explainer, a brand bumper. Claude still chooses the scenes, but it's aiming at something specific.
The spec (150–500 words)
24 prompts. Named sections, a stage size, a duration, a palette, often a list of scenes with timings. Reads like a creative director's notes.
The production document (500+ words)
13 prompts, up to about 4,000 words. Inputs to ask for, a full shot list, rules for typography and motion, rendering instructions. Closer to a program than a request.
WHAT THE NUMBERS SAY (AND DON'T)
| Prompt length | Prompts | Median views on X |
|---|---|---|
| Under 40 words | 180 | 358 |
| 40–150 words | 60 | 1,011 |
| 150–500 words | 24 | 6,675 |
| 500+ words | 13 | 1,050 |
Videos made from 150–500-word prompts got a median of about 6,700 views, nearly 20 times the one-liners. It's tempting to read that as "longer prompts make better videos". Don't, at least not from this. Views depend mostly on who's posting: the people writing careful specs are more often designers and builders with established audiences, and a detailed prompt is itself a signal that someone is serious about the result.
What the table does show is that the one-liner is saturated. 87 people posted a video from the same viral brief, and their median was 349 views. Being the 87th showreel is hard, however good it looks.
WHAT THE LONGER PROMPTS SPECIFY
Reading the specs side by side, the same handful of decisions come up again and again. These are the ones worth making yourself rather than leaving to Claude:
- Duration and pacing. 58% of all prompts state a length, usually 15 or 30 seconds. The specs go further and time each scene ("0:00–0:03 logo resolves from particles").
- Stage size. 1920×1080 for landscape, 1080×1920 for vertical. Without it you get whatever size the browser window happens to be.
- Format. "A single self-contained HTML file, no external assets" is the most reliable way to get something that opens and plays first time.
- Visual system. A palette, a typeface, a mood reference. 14% of prompts name brand colours or fonts.
- Technique. 14% name a technology: GSAP, canvas, SVG, Three.js, shaders. Naming one steers the look; a Three.js prompt produces a very different video from an SVG one.
- Inputs to ask for. The most-shared template prompts in the index open with an <inputs> block telling Claude what to ask the user first (brand name, UI states, photos) so the prompt can be reused.
A TEMPLATE TO START FROM
This combines the elements that recur in the longer prompts. Replace the bracketed parts; delete anything you don't care about and Claude will decide it for you.
Create a single self-contained HTML file (no external assets) that plays a
[15]-second motion graphics video on a fixed [1920x1080] stage, then holds
on the last frame.
Subject: [what the video is about, in one sentence]
Audience and tone: [who it's for; e.g. confident, playful, technical]
Visual system:
- Palette: [2–4 colours]
- Type: [font or style, e.g. heavy condensed sans for headlines]
- Motion: [e.g. snappy easing, no bounces, hard cuts between scenes]
Scenes:
1. 0:00–0:03 [opening]
2. 0:03–0:08 [middle]
3. 0:08–0:12 [payoff]
4. 0:12–0:15 [end card with (text)]
Use [GSAP / CSS animations / canvas]. Make every animation deterministic
so the video looks identical each time it plays.The first render is a draft. The creators with the best results sent notes back: "scene 2 is too busy", "hold the end card for another second", "the easing feels floaty". Opus 5.5 is good at taking that kind of direction.
EXAMPLES FROM THE INDEX





