475 Prompts for Making Videos with Claude Opus 5.5, Straight from the People Who Made Them

You've probably seen those slick animated videos floating around X and wondered how someone built them in an afternoon. Turns out a lot of them came from asking Claude Opus 5.5 to write the animation as code—and now there's a place to find the exact prompts behind them. The awesome-opus5-5-videos repository collects 475 of these prompts, each paired with the creator's original post.

What It Does

This is a curated collection of viral videos people made with Claude Opus 5.5, organized so you can grab the prompt and try it yourself. Every video in the repo was created by asking the model to write the animation as code—HTML, Canvas, SVG, Three.js and friends—rather than using traditional video editing tools. The collection includes the prompt the creator shared publicly, or where no prompt was published, a link to their original post.

The prompts live in two places: a prompts/ directory with individual markdown files, and a data/videos.json file if you want to work with the data programmatically. The README highlights 100 entries, one per distinct prompt, spread across four categories: motion graphics (58), explainers (16), 3D scenes (14), and games and interactive pieces (12). Each entry links back to the creator's original post, so you can see the finished video in context. The most recent update, dated 2026-09-29, added 86 videos, mostly motion graphics and explainers.

Why It's Cool

The real value here isn't the videos themselves—it's the prompts. Most "look what AI made" showcases stop at the output, leaving you to guess at the input. This repo flips that.

• You get the actual prompt, not a description of it. Each entry links to a markdown file with the prompt the creator used. That means you can copy it, paste it into your own agent, and see what happens. The repo is explicit about this: pick one, copy its prompt, and try it with your own agent.

• It's organized by what you're trying to build. The four categories—motion graphics, explainers, 3D scenes, games and interactive—map to actual use cases. If you need a product explainer, you're not scrolling through 3D demos to find one. The category counts give you a sense of where the community's energy is going (motion graphics, by a wide margin).

• The tech stack is noted per entry. Each video lists what it was built with—Canvas, SVG, GSAP, Three.js, or combinations like "Canvas · SVG · GSAP." That's useful if you're trying to learn a specific library or want to match the approach to your own toolchain.

• Every entry credits the creator and links to the original post. This isn't a scraped dump. You can trace each video back to the person who made it, which matters if you want to follow their work or see the discussion around it.

• There's a live comparison view. The README links to a Skillry page where you can watch all 475 videos side by side with live remakes. That's a nice touch—it lets you see not just the original output but a working version you can inspect.

The collection also handles the awkward case where a creator didn't share their prompt. Instead of skipping those videos or making something up, it links to the original post. That's honest curation.

How to Try It

Getting started is about as simple as it gets, since there's nothing to install.

1. Head to the repository: github.com/yihui-dev/awesome-opus5-5-videos

2. Browse the categories in the README—motion graphics, explainers, 3D scenes, or games and interactive—and find something close to what you want to build.

3. Click the "Prompt" link next to any entry to open its markdown file. Copy the prompt.

4. Paste it into your own agent (the repo assumes you're using Claude Opus 5.5, since that's what the videos were made with) and iterate from there.

If you'd rather work with the full dataset, clone the repo and look at data/videos.json or the prompts/ directory directly. The JSON gives you structured access to all 475 entries, which is handy if you want to filter by tech stack, creator, or category.

For a quick overview of everything at once, the Skillry page shows all 475 videos side by side with live remakes, so you can see the originals and working versions together.

Final Thoughts

This is a reference collection, not a framework or a tool you install. Its value depends entirely on whether you're interested in making code-driven animations with an AI agent. If you are, it's a genuinely useful shortcut—instead of staring at a blank prompt wondering how to describe a motion graphic, you can start from something that already worked for someone else and adapt it. The organization and per-entry tech notes make it easy to find relevant examples, and the commitment to linking original posts keeps it grounded in the community that produced the work. If you've been curious about this style of video-making but didn't know where to start, this is a reasonable place to begin.

Follow @githubprojects for more developer tools and open source projects.