AKA “Restarting these weekly things after a 2 and a half year hiatus to see if it a) sticks and/or b) is worth doing”.

I’ve been pretty inactive here on the old blogstead, aside from keeping up with daily photos¹². One of the things I keep trying to get into is a meaningful reflective practice. That butts up against what I want to reflect on, if/where it might be appropriate to store/share it, and does it even matter. I think it does, and I’m going to try to reflect on what I can here in the /reflections thing. I have no idea how regularly I’ll be doing this. Maybe every week. Maybe occasionally. Maybe it’ll go dormant for another 2 and a half years. Who knows.

After consolidating my Unreal Engine log notes from a separate WordPress site into my main website (this one), I realized that I hadn’t captured any of the final (or “final”, as I basically stopped that work to focus on supporting student-led deliverables that took different directions) “Taylor Institute Game” level.

It’s still empty, but has more textures and materials applied. I didn’t add a roof because I found the environmental lighting useful without having to manually light every nook and cranny…

Inspired by Benj Edwards’ post on Ars Technica this morning, I wanted to try out the latest version of DALL-E that’s integrated into ChatGPT 4 Turbo™, to see what it could create to visualize potential future learning spaces.

I started off with a basic “Create an image of a video game depicting students exploring classrooms and informal learning spaces. Cinematic, 8K, studio lighting.”

🤖 - AI-Generated Content via DALL-E

Interesting. Chalkboards, but sure. Doesn’t look very video-game-y, though. “make a version of that image that’s pixellated as if rendered by an Amiga 1000.”

This post was originally going to be part of a series exploring the topic, but if I take that approach I’ll probably never actually make time to write each post in the series, so here’s an omnibus “I need to think about this stuff and writing a blog post is the best way to formalize my thinking.”

We’ve all been trying to figure out generative AI since ChatGPT was launched almost 1 year ago. (yes, it’s been less than a year since ChatGPT was originally launched on Nov. 30, 2022. yes, it feels so, so much longer than that.) Much of my work over the last year has involved conversations with team members and university leadership about the nature and implications of generative AI tools, most prominently ChatGPT.

So much for the pristine fresh start… I’ve added a few of Unreal Engine’s free asset collections to try out some ideas:

• VisAI Modern AI Framework (to try adding some bots)
• Basic Pickups VFX Set (to try adding some game elements)
• Advanced Cel Shader Lite (to try non-photorealistic stylistic rendering)
• 4K Materials: Wood Flooring (I mean. The TI has so much wood, it’d be a shame to just use “wood” as the material…)

And then Unreal Engine had an update, from 5.3 to 5.3¹, apparently? Regardless. Always apply updates, so hopefully the project works now that I’ve added a couple GB of materials and some complexity.

I spun up a fresh project, re-imported the FBX models of the TI, and set up a first/third-person player with keyboard toggle. There are no textures, no lights, and lots of meshes to fix. But I let go of over 8GB of kruft that had accumulated from messing around in the previous project. There’s probably a better way to clean out old gunk, but a full restart of a project should work.

I’m trying to practice modelling/staging/lighting the Taylor Institute learning spaces in order to give me a chance to learn how to build stuff in Unreal Engine 5. I have some .FBX exports that a staff member had created back in 2016 from the original Revit .RVT files for the first and second floors of the TI, and have been trying to process/fix/import those. I don’t have the original .RVT files, to these .FBX exports are as close as I’ll get. They look like they get some things right – individual steps and railings on staircases are modelled – and some things quite wrong. It looks like all “ways” were just extruded from the floorplan at an estimated full-storey-height. There are no windows, and no doors. No floors, no ceilings. But, it’s a good start to get a feel for the layout of the space.

I had a few minutes to play around with textures a bit, and added some MediaPlate objects for each of the “displays” in the space.

I messed around with the scale of my model as imported from Blender. In UE5, I just selected the folder and all descendants and bumped the Scale value to 2x2x2. It applied uniformly to all objects within the model. I had to adjust the position of the lights, but otherwise the scale of the room “feels” right.

The textures are a hot mess, but that’s easily changeable. Also, I don’t think “shiny metallic cyber femmebot” is the look I’m going for. stock models…

I spent a few more minutes trying to futz around with getting my Blender model of a learning space (loosely modelled on TI 110) into Unreal Engine. Following a godawful YouTube video that I’m not going to link to because holy crap what an annoying mess but I did glean one bit if useful info from sitting through it.

In Blender, I exported an .FBX file using these settings:

I’ve been trying to model a space (starting with a learning space in the TI) in UE5 and have been stuck by the editing tools. So, I figured why not try building a model in Blender and export that for use in a UE5 project?

I was able to get the dimensions right (based on the architectural files I have), and cut out windows and doors using boolean subtraction. It looks like it works, but I don’t have time to import it into a new UE5 project right now, but that’s the next step.