3D Photogrammetry Scans of Universal’s Hogwarts

One of the ideas that struck me fairly early in this project is that it would be cool to compare the theme park versions of the castle to the “real” versions in the films. If you’ve ever walked through the Wizarding World of Harry Potter at Universal Studios in Orlando, Hollywood, or Japan, you know just how overwhelming it is to approach the towering facade of Hogwarts. Very effective theme park design.

Even so, much of the castle has been removed, the remaining sections have been modified, and the scale has been significantly reduced. (The filmmakers built their original Hogwarts miniature with a very specific real-world scale in mind.) The theme park version is highly evocative of the castle seen on the silver screen – specifically in Half-Blood Prince – but it’s still its own thing. (Three things, actually, since there are even slight differences between the versions at the three resorts.)

So…what would it look like to plop the Universal version down next to the cinematic version? That’s a question I intend to answer as the project progresses – hopefully with renders of them side by side! – but for now, you’re going to have to settle for these incomplete and very rough 3D scans of the Hollywood version.

These images are the fruit of photogrammetry, the science of extracting 3D measurements from photos. On a recent visit to Universal Studios Hollywood, I took about 400 photos of the castle. After heavily color-correcting these shots in Photoshop to bring out shadow detail, I threw them all into Meshroom, which dutifully set out to create a textured 3D model from the photos without any need for human assistance.

The first attempt failed. Unfortunate, but not surprising. There were angles I couldn’t capture due to the geography of the park and my own time constraints. (Not to mention the fact that we were trying to, you know, enjoy a day in the park!) Meshroom got confused and couldn’t connect all the disparate parts into one unified 3D model.

Not a problem. When I split the photos into six different sets that focused on smaller sections of the castle, Meshroom handled each set just fine on its own. Then I brought them together in Blender and manually lined them up with each other on top of an aerial photo. An HDRI sky added a bit of visual interest to the environment.

The scans needed a lot of manual cleanup, and even so, you can still see a lot of yucky fringes and artifacts, especially around the edges of each scan. That’s totally fine for my purposes…this is all just a rough guide for when I create my own model from scratch.

Still, interesting to look at, huh?

•