Big Idea

The argument in one line.

Claude Opus 5.5 renders motion graphics entirely through code rather than a video-generation model, and getting a polished result depends on moving past a single prompt into storyboarding and shot-level direction.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

You use Claude Code already and want a repeatable process for turning a brief into a polished motion-graphics or promo video.
You're a video editor or motion designer curious whether AI-generated animation has gotten good enough to compete with hand-built work.
You want concrete prompting techniques, asking for SVG output, storyboarding before rendering, timestamped shot feedback, rather than a generic claim that AI video is good now.

SKIP IF…

You're comparing generative video models like Sora or Veo. This video is about Claude writing code (HTML, CSS, JavaScript, Three.js, FFmpeg), not diffusion-based video generation.
You want a beginner introduction to what Claude Code is. The video assumes you already have it set up.

TL;DR

The full version, fast.

Claude Opus 5.5 can produce polished motion-graphics videos from a single prompt without touching a generative video model. Everything on screen, the text, the camera moves, the particle effects, is code the model writes and runs (HTML, CSS, JavaScript, Three.js, FFmpeg), which is why every element stays editable afterward. The creator lays out three levels for using this capability. Level 1 is one prompt in, one video out, good for quick showreels but unlikely to hit production quality. Level 2 has Claude storyboard the animation as stills plus written descriptions first, so direction can be corrected before paying the full render cost. Level 3 means reviewing the actual rendered video shot by shot and leaving timestamped notes until each scene lands.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →

Chapters

Where the time goes.

00:00 – 02:06

01 · Intro

Cold open showing two AI-generated motion-graphics showreels, a particle-burst logo intro and a 3D product showcase, credibility beats including a 300K-plus-view video also made with Opus, and a pitch to teach the underlying method.

02:06 – 04:11

02 · What happens under the hood

Opus 5.5 does not call a generative video model at all. Every visual, motion, and text element in the demo reel is code, HTML, CSS, JavaScript, Three.js, FFmpeg, that Claude writes and runs, which keeps every element individually selectable and editable afterward.

04:11 – 05:49

03 · Level 1: One-shot prompting

One prompt in, one finished video out. Useful as a capability benchmark across model releases and as a fast way to build a portfolio showreel, but unlikely to hit production quality on the first try.

05:49 – 08:46

04 · Level 2: Storyboarding

Have Claude generate stills plus written animation descriptions before rendering anything, so wrong creative direction gets caught while it's still cheap to redirect. Frame-anchored comments, not chat messages, speed up the correction loop.

08:46 – 11:45

05 · Level 3: Directing

Review the actual rendered video shot by shot, leaving timestamped comments, remove this scene, refocus this line, use this reference link, until the cut lands. The same technique applies to editing real, non-AI footage.

11:45 – 12:54

06 · Wrap up

Recap of the three levels and a closing note that the lead time for solo-produced, high-quality motion-graphics video has collapsed enough for one person to do it entirely with Opus 5.5.

Atomic Insights

Lines worth screenshotting.

Claude Opus 5.5 can generate a full motion-graphics video from one prompt using zero generative video models, it writes the HTML, CSS, JavaScript, Three.js, and FFmpeg that render every frame.
Because an AI-generated video is actually code, every image, text block, and animation timing is a separate, selectable, editable object after the fact.
A one-shot prompt for a showreel is a fast way to benchmark a new model release, since you can re-run the exact same prompt every time a model updates.
Skipping straight from prompt to final render risks burning the full render cost on a result that veers in the wrong creative direction.
Storyboarding an AI video means asking for stills plus written animation descriptions before rendering, turning an expensive mistake into a cheap one.
Feedback tied to a specific frame and timestamp gives an AI model more usable context than a general note typed into a chat window.
A polished AI-made video routinely needs three passes: one-shot generation, storyboard correction, and shot-by-shot directing on the final render.
Claude Code can identify and keep only the best takes from raw, non-AI footage, applying the same directing skill to real video as to generated video.
If you already know how to run Claude Code, you're in a small minority of the population that can produce agency-grade motion graphics without hiring anyone.
The more specific the shot-level feedback gets, the fewer correction rounds a model needs, sometimes only minor tweaks by the final pass.

Takeaway

AI video editing now runs on code, not clips

PROMPT TO POLISH

A polished AI-made video comes from a three-step loop: storyboard cheap before rendering, then give timestamped feedback on the actual footage until each shot lands.

02What happens under the hood

Under a single prompt, Claude Opus 5.5 does not call a generative video model. It writes and runs code (HTML, CSS, JavaScript, Three.js, FFmpeg) that renders every motion, text element, and cut in the video.
Because the output is code rather than a rendered video file, every element stays a separate, selectable object afterward, so a single image, text block, or timing can change without regenerating the whole clip.
Asking a model to diagram its own tool choices, in this case an Excalidraw drawing of the render pipeline, is a way to learn what libraries it reaches for without having to understand those libraries yourself.

03Level 1: One-shot prompting

One-shot prompting means one prompt in, one finished video out, with no feedback loop in between.
A one-shot prompt is a fast way to benchmark a new model release: run the same prompt against each new version and compare the ceiling of what it can do.
One-shot output is unlikely to hit true production quality on the first try, because the model has to guess the entire creative direction with no correction.

04Level 2: Storyboarding

Storyboarding an AI video means asking for still frames plus written descriptions of each planned animation before anything renders, so wrong direction gets caught while it's still cheap to fix.
Skipping storyboarding and going straight from prompt to final render risks spending the full render cost on a result that's far from what was wanted.
Feedback anchored to a specific frame and timestamp gives a model more usable context than a general note typed into a chat window, which is why a visual review interface speeds up the correction loop.

05Level 3: Directing

Directing means reviewing the actual rendered video shot by shot and leaving timestamped comments, remove this scene, refocus this line, use this source link, rather than one global note.
The same directing technique applies to real, non-AI footage: feeding raw clips into Claude and asking it to identify and keep only the best takes.
The more specific the shot-level feedback gets, the fewer correction rounds a model needs; by the final pass only minor corrections were required because the model already matched the intended taste.

Glossary

Terms worth knowing.

One-shot prompting: A single instruction sent to an AI model that is expected to produce a finished result with no follow-up correction.
Storyboarding (AI video): Having a model generate still frames and written animation descriptions before rendering the full video, so the creative plan can be corrected cheaply.
SVG (Scalable Vector Graphics): An image format built from math-described shapes rather than pixels, so it resizes to any dimension without blurring and its colors can be edited directly in the file.
Three.js: A JavaScript library for building 3D graphics and animation that run inside a web browser.
Rubric: A video review tool built by the video's creator for leaving timestamped, shot-level comments on AI-generated video drafts.

Resources

Things they pointed at.

00:00toolRUBRIC (getrubric.app) ↗

01:01linkAnthropic's official Opus 5.5 prompting guide

03:49channelRoboNuggets community (Skool) ↗

05:33link3 Levels of AI Motion Graphics PDF (free starter prompts) ↗

Quotables

Lines you could clip.

00:10

“No generative AI video model was used, and I just sent a simple prompt like this where I ask it to make a dynamic motion graphics video that showcases its capabilities.”

states the whole premise in one line: no video model, just a prompt→ TikTok hook↗ Tweet quote

03:48

“You need to understand that there's still levels to this skill still.”

short, quotable framing line that sets up the entire back half of the video→ IG reel cold open↗ Tweet quote

09:22

“Directing really is just giving feedback to the final video.”

clean one-line definition of the hardest of the three levels→ newsletter pull-quote↗ Tweet quote

11:16

“I'm finding that I need to give less and less comments just because of how good it one shots these animations.”

concrete claim about model improvement from someone using it daily→ TikTok hook↗ Tweet quote

The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

analogy

Opus 5 .5 for video editing and motion graphics has gotten pretty unbelievable lately. And so these examples that you'll see are all created with Opus, and all I did is to give it access to my workspace. No generative AI video model was used, and I just sent a simple prompt like this where I ask it to make a dynamic motion graphics video that showcases its capabilities.

This is another similar example, but where I pushed it to use more 3D elements.

And in fact, some of the videos I launched recently, like this one that already has 300 ,000 views, it was also edited using Opus 5 .5. So Opus 5 .5 just came out and Entropic just published an official comprehensive guide on how to prompt it. You can keep one agents .md as your single instructions file now instead of keeping a claude .md and an agents .md in sync.

One of the best things you can do is to ask for them in the SVG format. That stands for scalable vector graphics so that it scales to any size without going blurry. You can open the file and change the color by hand and you can even have Claude animate these icons.

So with SVGs... graphic that you get back from cloud is editable and animatable. So clearly I've been testing this model heavily for automated videos and in my view these are all a major step up versus what we had previously.

So today I'll teach you how to best use this capability of 5 .5 so that by the end of it you can also level up your video and motion graphic skills using AI and know exactly how I make videos like these that get hundreds of thousands of views even if they are AI edited. And if you're new my name is Jay. I spent over a decade working with brands you probably know have been in a AI since my master's in data science, and now I'm leading our AI business in one of the largest AI communities globally.

Let's dive into it. So before we go into the different levels by which you can use Opus's capability for video editing, I actually want to give you context of what really happens under the hood when you send a simple prompt like this to 5 .5. Because as I mentioned earlier, this whole video that it created for us, it didn't really use any generative AI video model for it.

All of the movements, all of the texts, and basically everything you see here, those are all created using code. And in fact, in this demo interface, you can clearly see I can select these individual elements, which are all just images in my workspace. And all they are really is Claude pointing to this particular photo in this case and using some code, in this case JavaScript, in order to place them in this exact position in this exact timestamp.

And everything that you see here, including the movements, the motion graphics, that is all using Opus 5 .5's taste. And if you're curious exactly what tools it used to achieve that video, I also just asked Opus to illustrate that here in Excalibur. And you can see that with that one -shot prompt that we sent it, it just used its native knowledge of HTML, CSS, JavaScript, and a couple of other libraries like 3JS and FFmpeg in order to achieve that look, which again, you don't even need to understand all of this.

Although in my view, it's at least good to be aware of them. But because of how powerful these models have become, you can pretty much just rely on 5 .5 to create you great showroom. feels like that using a single prompt.

I did want to call out, though, that for this particular test, we did not use HyperFrames or Remotion, which are these popular open source libraries for video editing using Cloud, which just illustrates for you how good natively Opus 5 .5 has become, because even without these plugins, it's able to create videos like that.

And by the way, if you want to learn how to build and sell AI systems that businesses actually pay for, then that's pretty much all we do over at the RoboNuggets community, where not only do you get access to the Cloud Living Masterclass, which we update every week and takes you from zero to mastery with the latest on AI, but you also get access to our Agents as a Service course, which walks you through how to actually get paid for all these AI skills that you are learning.

You also get to be part of a genuinely great community of AI builders. In fact, you can see just some of the recent wins our members are getting from the program right here. So if you want to start earning from AI, then check that just in the pinned comment below.

Now back to the video. But if you're serious about producing video using AI and using Cloud, you need to understand that there's still levels to this skill still. Because that example that we went through, that's the most basic level.

That's level one, where you send a single prompt and you get a good result. Now, in my view, I think this level is still useful, at least for you to try out. Because number one, it gives you a sense of the capability of these models.

So if you have a prompt like that, that pushes the ceiling of what a certain model can do, then when a new model gets released, you can just repeat that same prompt to test how good things are now. I also think that one -shot prompts that provide showcase reels like that are quite good in terms of building up a portfolio for you.

Because if you're a video editor and you're just looking for showreels to showcase your work, then this is a great and easy way for you to get that almost immediately. Because remember, and this is a good time to bring up this chart again, if you're watching this video and you know how to use cloud code, amongst the whole population of people, you are probably here in the red cells.

So you're in a very small fraction of people who actually know how good these models have become when it comes to video editing, at least. But for a huge portion of people and the customers that you may be serving, this type of capability and motion graphics is actually something that you would expect a top professional agency to be able to produce.

And so bottom line is, if you want to create production level videos using Claude, you probably won't get there with one shot, but there's clearly still benefits to doing this. And so to make it easy for you to copy those prompts and for you to test it, I just made this nine page PDF, which you can just send to Claude and pick out all of the starter prompts that I will mention in this video, including those one shot prompts that I used in order to create those demo reels.

And you can just grab this PDF for free down below in the description. Now, like I said, level one, it's useful, but you probably won't get to the end result that you want for production level quality videos. And so the way that you do that is to move towards level two, which is taking you from one shot prompting.

to actually storyboarding your videos. Now, the great thing about this is you don't actually have to sketch it yourself. Even though you can do that, you can actually have Claude storyboard your videos for you.

So why is storyboarding even important in the first place? Well, usually for most beginners, if they send, let's say, a prompt. and they already have an idea of the video that they want, depending on the complexity of this final video, it's more often the case that this prompt that you start with isn't really detailed enough to give you that final result in one shot.

And so what ends up happening is that your AI model, Claude in this case, will actually veer away towards the wrong direction and produce for you a final video that has all of the animations and has all of the tokens already spent and burnt, but producing the wrong result, which is very far away from what you want. And so the reason why Storyboard are important is that if you ask the AI model to create them first before going all out on this wrong result that it's about to make, if you're able to preview what it's about to do at less cost and at much faster turnaround times, because this is just a storyboard anyway, then you can steer this model towards the right direction to give you the video that you want.

Now, I'll just show you the interface that I use and which we're developing as part of the toolkit for our community. But at its core, to do storyboarding effectively, you just need two things. You need to be able to tell Claude that you would want first a storyboard or a of images of the animations that it's about to do as well as a description of what actually happens in this particular scene and so you can see here when i ask it to create a storyboard for me for the rubric app it already created all the html elements for me but without the animations yet but instead what it created for me are these screenshots as well as the descriptions of the animations that it's about to do so it's saying here that the camera whips to the first panel it then continues across the panels and so on and so forth and so at this point it hasn't rendered the video yet but you get a clear sense of what it's about to do.

Now, the second element when it comes to storyboarding is a way for you to give feedback to it. Now, obviously you can just open up your chat window and give feedback that way. But I just think with a lot of video and design work, it is much more effective if you can give feedback in the context of the visuals that you're seeing.

And so in this interface, let's say I want this to be less text heavy. I can just click on this particular frame. It will add a comment here and I can just give that feedback for this particular frame.

And the great thing about this is when this comment is posted to Claude, it will also give the exact timestamp and even the exact point that I made the comment on in this particular frame. So it's much more directed and in my view is much more effective as well in terms of steering your agent towards what you want. And so what I usually do is I just run through this whole storyboard that it created for us.

I give all of my comments and when I'm done with that, I just copy all of these comments and then just paste all of those comments to Claude, either ask for another storyboard or just go directly to the video, depending on how close Claude is already to the video that I actually want. Now let's take this one level up. Because if you want to be really good at steering your AI agent, in this case, Opus 5 .5, into creating videos that are really stunning, then you want to get to level three, where you're actually directing it based on the video that it already created.

Because if we go back to our illustration earlier, if you give feedback on the storyboard and you ask Opus 5 .5 for the video that you want, I often find that that takes you around, let's say, 80 to 90 % there in terms of the final polished quality output that you actually really want. And so if you get really good at directing AI to create the final video, output for you, then you have reached level three.

And at the very core of it, directing really is just giving feedback to the final video. And similar to the previous level, you can always do this by speaking to Claude via the chat window, giving your feedback there via natural language. But often what I find is when it comes to reviewing video, it's much easier, again, if you actually have an interface on an operating system yourself of how you can give feedback to AI.

Just to show you one example for my work. This is, again, another beta version of the toolkit that we're developing for our community. So it's not yet the final look of this piece of software.

But this rubric editor basically lets you play the video, give comments on each of these frames. And more importantly, you can also give feedback on the audio that it used, which isn't really something that you will hear during the storyboard process, right? So I'm not sure if you can hear it.

You can pretty much hear the sound effects that it used and give comments there. So let's say after viewing the video, we decided that this particular frame or scene is no longer important. Then we can just say to remove this scene.

And you can see this comment comes with a timestamp where I place this feedback. And it knows exactly what graphic I am referring to. And so similar to storyboarding, you can actually just make a lot of comments on this reel, copy all of those comments, and then just send it back to Claude.

And just to give you an example from an actual video. So that Opus 5 .5 tips video, which... the one that has already 300 ,000 plus views.

This is actually the rubric editor where I work with Opus 5 .5 to produce that final video. So essentially the difference with this one is that if you can see here, since this is coming from a raw footage, I asked Claude to identify the specific takes that I want to keep in this video. And I made a separate lesson on this if you want to learn how to do that.

And then if I play this, which right now it's at... 2x speed which is you know a much faster way for me to just review its animations in there you can see that when i reviewed this i just made this comment for this animation where i just said to focus on the title of the page and i gave the link to that cloud platform documentation there which is why it came back with this animation And then I just did that across all clips because at least for this video, I did want it to pass my own review and I wanted to match the level of quality that I am envisioning.

The great thing with Opus is that I'm finding that I need to give less and less comments just because of how good it one shots these animations. Like it really gets the stuff that I want to be shown. And sometimes when I'm directing this particular video, I just need to provide minor corrections here and there.

And then I copy all of the comments for all of the clips and I send it back to Claude. Wait for it to update this rubric editor. and then ask it to render the final video when it's done.

And so there you go. That is, in my view, how you maximize the capability of Opus 5 .5 when it comes to automated video editing. As you can see, there's still a lot of human review involved, which I think is how it should be anyway, because in a world where everyone can pretty much one -shot show reels like what you saw in this video, then the real differentiator that you would have in terms of creating videos would be your own taste, plus also the operating system by which you work and give feedback to your AI models.

Now, to be clear, even if you are storyboarding and even if you are directing on your own, I still think that the lead time to create quality videos have shrunk so much that you can pretty much have just one person do an entire motion graphics video entirely on their own, plus Opus 5 .5. And as I mentioned, if you need starter prompts for everything that I showed in this video, you can just grab this PDF and send it to your agent.

Hopefully that was useful. As always, thank you for watching until the end and consider tapping subscribe below if it is because that helps me a lot to put out more educational stuff like this. I'll see you all next time.

Cheers.

The Hook

The bait, then the rug-pull.

A single Claude Opus 5.5 prompt can now produce a finished motion-graphics reel with no generative video model involved. The creator breaks down what's actually happening under the hood, then walks through the three levels of skill that separate a lucky one-shot output from a genuinely directed, production-ready video.

Frameworks

Named ideas worth stealing.

04:12model

The 3 Levels of AI Motion Graphics

Level 1: One-shot prompting, one prompt in, one video out
Level 2: Storyboarding, sketch first, comment, then build
Level 3: Directing, direct every scene until it lands

A maturity ladder for using Claude to produce motion graphics, from a single blind prompt to full shot-by-shot creative direction.

Steal forany AI-assisted video, motion graphics, or design workflow where the first output rarely matches the brief

02:39list

Under-the-hood render stack

HTML/CSS
JavaScript
Three.js (3D)
FFmpeg (compositing/export)

The specific native code libraries Opus 5.5 reaches for to build a motion-graphics video without any generative video model.

Steal forknowing what to ask an AI coding model for when you want editable, non-generative video output

CTA Breakdown

How they asked for the click.

VERBAL ASK

03:50product

“if you want to learn how to build and sell AI systems that businesses actually pay for, then that's pretty much all we do over at the RoboNuggets community... check that just in the pinned comment below”

soft mid-video pitch for his paid community, delivered without breaking from the topic, then an explicit 'back to the video' transition; the description also carries a separate free PDF lead magnet as the primary funnel entry

MENTIONED ON CAMERA

00:00toolRUBRIC (getrubric.app) ↗

03:49channelRoboNuggets community (Skool) ↗

05:33link3 Levels of AI Motion Graphics PDF (free starter prompts) ↗

FROM THE DESCRIPTION

PRIMARY CTAWhere the creator wants you to go next.

AFFILIATECommission earned if you click.

OTHER LINKSAlso linked in the description.