A New 3D AI Builder vs Just Claude Code?
Chapters7
Introduces Spline 3D and sets up a comparison with Claude code and MA to see how it performs on a one shot prompt.
A hands-on compare of Spline 3D's new AI builder vs Claude Code, testing real-time 3D water orbs and design workflows.
Summary
DesignCourse host reviews Spline 3D's fresh AI builder by putting it to a practical test against Claude Code. The video centers on whether the new 3D-focused tool can outperform Claude’s code-generated outputs in a single-shot prompt scenario. He feeds the same water‑orb hero-section prompt to both tools and compares results, noting speed differences and visual quality. MA (the other AI tool in the test) produces a shader-like 3D orb that reacts to mouse movement but lacks convincing 3D spherical geometry. Claude Code, on the other hand, delivers a more convincingly 3D orb and a workable options panel, though its typography and layout aren’t perfect yet. Throughout, the host critiques the UX, noting the need for presets and a quick-access library of 3D shaders to speed up production. He also demonstrates iterating on the outputs with prompts to adjust size, radii, and animation speed, and he even refines the design manually for a more professional look. Overall, the comparison suggests Claude Code currently offers a stronger first-pass 3D result, but Spline 3D shows promise—especially if it adds presets and better 3D mesh handling. The takeaway: if you’re already on Claude, you’ll likely get to a polished result faster; if you’re evaluating 3D builders, expect early-stage tradeoffs and watch for roadmap features like presets and shader libraries.
Key Takeaways
- Claude Code produced a clearly 3D orb with a usable options panel, outperforming MA in the first run when tested on the exact prompt.
- MA generated a shader-like water orb that reacted to mouse movement but lacked convincing spherical geometry and depth on the first pass.
- The creator emphasizes the need for presets or a shader library next to the prompt area to accelerate 3D design within builders like Spline 3D.
- With iterative prompts, the Claude path yielded a more impressive 'Surfs Up' water simulation, including interactive color transitions and a visible response to mouse movement.
- The reviewer still prefers Claude’s output for now due to immediate 3D realism and ease of tweaking, suggesting Claude as the default for similar one-shot prompts.
- Both tools benefited from clarifying prompt tweaks (size, radius, animation speed, mouse influence) to tailor the 3D water orb to the designer’s taste.
Who Is This For?
Essential viewing for designers exploring AI-assisted 3D builders and current users evaluating whether Claude Code or MA-based prompts deliver faster, more convincing 3D assets. If you’re prototyping interactive hero sections, this video helps you gauge tool maturity and roadmap gaps.
Notable Quotes
""The faster the mouse moves, the more the water orb should react.""
—Shows the core interaction idea being tested for realism and responsiveness.
""This is actually 3D. It's a little bit too large.""
—First impression of Claude Code's 3D output versus MA's result.
""I really want to see predefined like 3D shapes, shader effects, and stuff you can plug into the builder.""
—The creator's critique and wishlist for future Spline 3D features.
""Two prompts to get to this point. This is really cool.""
—Highlights how quickly Claude Code delivered a visually strong result.
Questions This Video Answers
- How does Claude Code compare to MA for one-shot 3D prompts in web design tools?
- What features would make Spline 3D competitive with Claude Code for interactive hero sections?
- Can preset 3D shapes and shader libraries improve the speed of AI builders like Spline 3D?
- Is Claude Code faster or slower than MA when rendering complex 3D water simulations?
- What are practical tips to optimize 3D prompts for better realism in AI-generated assets?
Spline 3DClaude CodeMA (AI tool)3D web designInteractive hero sectionShader effectsPrompt engineeringFront-end design workflow
Full Transcript
So, the people behind Spline 3D released. Oh my god. Yes, another builder, right? Um, but it's really geared towards 3D as you can see here in their examples. So, they have a bunch of community examples, but the real test is going to be, okay, what does it actually look like in practice? I'm going to put this thing to the test because every time a new website builder comes out, I'm thinking to myself, can I just use Cloud Code? I mean, I have the max $200 a month subscription. Why can't I just use clawed code rather than this wrapper?
So, we're going to take this prompt right here and we're going to feed it into MA to see what it produces versus just my clawed code and we're going to see I mean is is it is this better for 3D or is it kind of the same? Let's find out. So, here's the actual prompt. I'll read it real quick. It's not too long. Build a hero section for a water park called Surfs Up. The background should be a dark desaturated blue blah blah blah. The hero section should be 100 viewport height. In the background of the hero section, there should be a large canvas area with a transparent background that has a spherical realistic water fluid simulation that reacts to the mouse position.
This is something that's been done before. I just kind of want to see what it does based on one shot. Now, this should be realistic as possible, but still run smoothly on a system with a GPU. The faster the mouse moves, the more the water orb should react. And there should be a large prominent headline centered on top of it. Same thing with a sub head. and then a CTA. Um, keep the water simulation fixed and add an empty div with like 600 viewport height so that we can really scroll down and we want the water simulation color to transition between, you know, different colors every 100 viewport height or so.
So, it's enough for it for us to really give it a test. And we're going to compare it here in just by pasting this in into their prompt right there. And then also with Claude code and we'll we'll compare them. All right, it is done. So, we could see here I have open Claude Code. Um, it took probably about five times as long to generate. So, MA was definitely faster. I saw a sneak peek of what it looked like. Let's check out MA first. All right. So, here's what it created initially. It it it is kind of a water orb.
Um, it does respond to it, but it's not as impressive as like it's not what I really imagined. I want it to be a little bit more realistic. This is not a 3D ball. It looks like it's like a shader of some sort. So, it's it's not really it's not what I wanted. But, of course, this is just a one shot. Now, the big question is is if I scroll down, um is it going to change colors? Okay, it did. Okay. So, for for for a first time, it's like, okay, it's like, okay, not the greatest thing I've ever seen, but it kind of got it.
Let's see what I And again, I have not checked it yet. Let's see what uh Cloud Code produced. Okay. Alrighty. Yeah. You know what I have to say? I mean, it is This is actually 3D. It's a little bit too large. Um, it does respond. Um, let me scroll down. Oh, and it changed it. Let me just prompt it real quick and just ask it to shrink the size a bit. Okay, here we go. So, you know, the typography, I hate this. This is like a crappy hero section layout-wise. Um, but the actual orb itself, I have to say it's better than what MA produced.
Um, at least for a first go. Um, what I would really want to do is now I'm going to ask it to go ahead and create an options panel that allows me to define the different parameters as a part of this little 3D water orb. I'm going to do the same thing for MA as well. Going to say it doesn't look like the orb is based on an actual 3D spherical mesh. Fix that. also give me the ability to update parameters in a calibration modal. I don't even know if I'm going to have enough um credits left after this, but we're going to give it a shot.
All right, so they're both finished. Let's check out uh the surfs up one. This time I did ask it to add like little like simulations of the water like splashing out based on mouse movement. So that's kind of cool. Sort of oneshotted it. Now the options area is right here. And we can see we could change the radius. All right. Amplitude. Maybe scale it down a little bit. Ah, that's better in my opinion. Speed. Okay. So, I went ahead and I kind of just customized it to my own liking. And look at this. This is freaking awesome.
Just literally like what? Two prompts to get to this point. This is really cool. Okay, so let's go ahead and check out what MA did. All right, here's the new one, which is based on 3D this time. Um, here's the options modal, which kind of sucks because I can't move it around. So, you can change the mesh detail, the sphere radius, noise amplitude, probably bring that down. Noise frequency, animation speed, and then we have mouse influence. Maybe we'll increase that. Splash intensity. All right, final strength. Okay, I have to say it it's not as impressive as the Claude option.
Now, of course, this is just one test. I don't want to crap on their product or anything, but for somebody like me who uses claude code, I mean, we always you have to ask yourself, what is the moat here? if I can just prompt this into existence myself. I feel like, you know, if you're building one of these tools and you specialize in 3D, which Spline does, you should probably have like these presets that you can just choose from, uh, for your 3D models, so that, you know, right next to the prompt area, the text field, maybe there's an area where you have like a drop down for like a library of, you know, different shader effects that just quickly can be infused so that I don't have to play around in cloud code and try to create it by prompting from scratch because that kind of looks like that's what's happening at Elma anyways.
So, I have to say, um, this is super cool. This one right here, I'm going to try to elevate this design, uh, because the the overall layout of it, I didn't use the front-end design skill. So, let's let's try to elevate the design of this one. Uh, and then see what we come up with. All right, here it is. Okay, so this is the updated design. It's a little bit more modern. It's a lot more modern than what it had before because it used the front-end designer skill. Um, and I have to say I overall there it what has produced is not convincing enough for me just to forego using Claude and creating another subscription.
So like I said, it really seems like I, you know, if you're going to build a tool like this, and of course it's very early going for them. you know, they just launched it, but I want to see predefined like 3D shapes, shader effects, and stuff like things that I can just plug into the context of the builder and be like, "Hey, make this object consistent with this preset." I think that would be sick. But for now, it's not convincing enough for me to, you know, change course here with Claude based on what it was able to produce with the same prompt essentially.
So, let me know what you all think here in the description.
More from DesignCourse
Get daily recaps from
DesignCourse
AI-powered summaries delivered to your inbox. Save hours every week while staying fully informed.









