What Is a 3D Website, Actually? A Plain-English Guide
October 18, 2026 · 4 min read
The short version
"3D website" describes two genuinely different things right now, and most people use the term without knowing which one they mean. One is a real 3D scene rendered live in your browser - you could, in principle, rotate it, walk a camera through it, click into it. The other is a flat video with camera movement baked in, playing back in sync with your scroll position. Both can look identical in a screenshot. They aren't the same thing.
Type 1: Real-time 3D (WebGL / Three.js)
An actual 3D scene, built from real geometry, materials, and lighting, running live in your browser via WebGL (usually through a library like Three.js). The "camera" moving as you scroll is a real camera in a real 3D space, not a pre-recorded shot.
- Genuinely interactive in principle - rotating an object, clicking into detail, changing a variant
- No large video file to download, since nothing is pre-rendered
- More technically demanding to build well
Type 2: AI video sliced into scroll-triggered frames
An AI video model generates a short clip with camera movement already in it. The site then extracts individual frames from that video and swaps between them as you scroll, so scroll position maps to a point in the video's timeline.
- Feels 3D because the source video had real camera motion, but there's no actual 3D scene in the browser
- Nothing to rotate or click into - it's closer to a very well-synced video than an interactive scene
- Can depict literally anything a video model can render (a flyover of a fictional planet, real-feeling materials, photoreal lighting) that real-time rendering can't easily fake
We wrote up the full technical comparison, including which real products use which approach, in How AI "Cinematic Scroll" Websites Actually Work.
How to tell which one you're looking at
Honestly, you usually can't tell from scrolling alone - that's the whole point of the video-frame approach. A few tells: try right-clicking the backdrop (a video often shows video-player context options); check if the page ships a large video or image-sequence asset in your browser's network tab; or just ask the product directly, since any honest one will tell you.
Which one should you want?
Depends what you're optimizing for. Real-time 3D is the more extensible foundation if you eventually want genuine interactivity. The AI-video approach can currently achieve a more photoreal, cinematic look for less engineering effort, and - depending on the product's cost model - can be dramatically cheaper to offer for free, since matching a prompt to an existing scene costs nothing close to what generating a fresh video per site does.
Parallux uses the AI-video approach specifically because of that cost structure: a shared scene library means free, unlimited generation and refinement, with cost only entering the picture once you publish. See real examples by industry to judge the output for yourself.