Post your design related challenges and I can help!

Hi everyone. I finished the course and decided to try the method right away. I would love your eyes on it. Willka is a fictional eco-lodge on the Bolivian altiplano. Two pods and a tour. The film is the page: the gate at dawn, the sun blinding you, an orbit around the lodge, a walk to the door, and then the salt fields. A hummingbird rides the whole route as a three.js layer over the frames. Every camera move was created in Blender first, then generated on Seedance, and the stills came from Midjourney. Link: [willka-spectralmoon.netlify.app] (desktop first, phone second) One question I keep going back and forth on. Right now there are two rest points where the film freezes and a card sits on it, plus one page section with glass cards that slide over the film before the salt. I was tempted to add a wide section of plain text, or white cards, so it reads more like a normal website. My worry is that it would cut the flow. How do you handle information on sites like this without breaking the spell? Any honest feedback on the scroll feel is very welcome too. The website is far from done, but I just wanted to show what I've done and get some feedback :D

First of all, personally I hate long courses that give you no practical advice. I really tried my best to put the actual client work, actual examples, and real builds here. Second of all, there are millions of courses on Figma and Claude already, so I tried to make this course different by actually teaching you how to create scroll-stopping animations and ideas versus boring ones like a flying sandwich or animated soda drink parallax. Third, everybody's websites look the same, especially when done with Claude designs, so I've added practical tools on how to avoid that look. What's inside: Your creative input diet. Where I actually pull from, and why most people's references are the reason their work looks the same as everyone else's. Why websites look cheap. The specific, nameable decisions that give it away. Once you can see them you can't unsee them. Breaking the rules on purpose. One break per page. It only reads as intentional when everything around it is controlled. Motion design. GSAP and ScrollTrigger, game feel, and how I use Claude as a coding partner so the motion work doesn't stall on syntax. Full build walkthroughs. A 3D butterfly and a Buddha temple, start to finish. Two real site builds, recorded. Not demos. Actual client work, actual decisions, actual moments where I changed my mind. Optimization. Making it fast without gutting the thing you built. Plus my complete tech stack. Every tool, what I use it for, and what I've stopped using. And the part nobody's teaching: how to make your work not look AI-generated. Everyone has the same tools now. The output all looks the same. The difference is taste, and taste is learnable.

Spline shipped their own 3D generation, so I ran it against Higgsfield today. The difference is crazy :D Spline shipped their own 3D generation, so I ran both against the same subject today: a cotton branch. Spline got a detailed text prompt and ran it as an agent. Sixteen or so steps: grew the woody stem, added the bolls, built bark and cotton materials, lit the scene, set the branch in a vase, staged loose bolls on a linen mat, framed a camera, cleaned up the shadows. Higgsfield got a reference image and generated from that. First pass came back untextured, so I resubmitted with texturing on. Not a clean A/B, and I want to say that up front. Text to 3D versus image to 3D is a real advantage to Higgsfield on surface detail. But the results are still worth putting next to each other, because they failed and succeeded in completely different places. What Spline gave me A scene. Vase, mat, floor, lighting, camera angle, named material assets I can go edit, and an export button that puts it on the web. From three words of intent it built a whole set. The surface is where it falls apart. The bracts under each boll are flat dark triangles, basically paper cutouts. And the fiber pass broke outright. Two of its own steps were "learning about hair" and "growing cotton fibers on bolls," and what came out was a white wireframe sphere floating over the middle of the branch. It applied hair to a volume instead of to the boll surfaces. Sixteen steps in, and the step meant to add texture is the one that failed. What Higgsfield gave me One object, on black, correctly textured. Real tonal range in the husks from dark brown to tan. The bolls separate into lobes with shading between them. The stem reads as wood. What are you using for 3D on the web at the moment?

Welcome, glad you found this. This is a space for people who care about web design and want to get better at it. What I post here: Sites worth studying and why they work Tutorials and breakdowns, mostly motion and scroll work Tools, resources, and things I find that are actually good Whatever I'm building at the moment What I'd like from you: post what you're working on, ask questions when you're stuck, and share the work that stops you scrolling. This is more useful when it isn't just me talking. Start by introducing yourself below. What you're building, and one site you keep going back to.