Opus 5.5 is really good at creating explainer videos. wild!
I shipped this launch video for a startup working on agentic engineering and agent-native infrastructure in about 20 minutes, for roughly $40.
Opus 5.5 writes the film, and an agent renders it.
No video generation model was used here. No additional libraries. Just JavaScript, Playwright, and ffmpeg. Claude coded and drew every frame of this animation, then took a video of it.
This could be a new medium for communication.
People are turning IKEA assembly manuals into narrated 3D instructions and research papers into eight-minute, 3Blue1Brown-style explainers. Someone made a generative experience by imagining and describing Atlantis or Altissia from Final Fantasy XV. Within seconds, you could explore its cities, discover its stories, and shape what happens next through your choices. We’re on the verge of a new kind of media where you describe a world and then step inside it. This goes far beyond gaming.
This is pure code for everything. Music. Art. Animation. insane.
The educational examples still need checking. A convincing animation can explain something incorrectly. But the ability to go from “I want to explain this” to a watchable first version in minutes is a big deal.
There are plenty of things I’d never spend days making a video about. Twenty minutes and $40? I’ll try it.
Opus 5.5 can create rich, interactive motion graphics.
Here’s a prompt that gets at what makes this model interesting:
Make a dynamic 15-second motion graphics video that shows what an incredible motion designer you are, like it’s your showreel for a resume. Go all out.
That leaves a lot for the model to decide. Typography, composition, pacing, transitions. It has to figure out what would actually look good.
Using Claude regularly again has made one of the biggest differences between Anthropic and OpenAI models obvious to me: taste.
GPT, Astra included, just doesn’t really compete in the motion design examples I’ve seen. Sure, you can get similar-ish quality with enough prompting, time, and effort. But Opus 5.5 combines taste, intuition, and vision in a way I haven’t felt from a model before.
It makes choices that feel deliberate. The elements belong together. The movement has rhythm. There’s an idea behind the thing.
And because it’s code, you can keep working with it. Change the timing, rewrite a scene, or make part of it interactive.
Opus 5.5 feels like a bigger deal than Opus 4.6 did. This thing is so god damn addictive.
Yeah, explainer video and motion graphics are falling into the LLM black hole.
The music videos and explainer demos are fun. But how is it at talking, coding, and all the stuff we’ll actually be doing next week?
Opus 5.5 is a really good model. It’s been my daily driver the last few days.
It communicates clearly. The intelligence feels like Fable 5.1, and getting through a task feels faster and cheaper than with Opus 5.
It doesn’t tell me to go to sleep nearly as much as Fable 5.1.
A cool thing about Opus 5.5 is that it’s nice to use at any effort level. Use low when you want speed, or xhigh when you want it to think harder. I use medium for most work.
It also feels usage-light in my Claude subscription, something I can’t say about Astra.
I still prefer Astra and GPT-6 Sol for much of my work. But Opus 5.5 is the first time since 4.6 that I’ve wanted to heavily use Claude.
It has great taste, and it’s useful beyond the spectacular demos.
I started by wanting a launch video. Now I’m looking at things I’d normally explain in a document and wondering which ones deserve an animation.
Receipt