Hello all!
I’m currently starting my journey through udk4 and I wish to start doing architecture visualizations.
I’ve already learned a lot about the engine itself, but I’m a bit confused when it comes to arch models and techniques.
First of all: where do you get your models? I know there are a lot of packs (like the ones from evermotion) but they aren’t usually game-ready, are VERY high poly and often don’t come UV-unwrapped. I know there are one or two packs on the marketplace, but I might need a lot more variety than that. Due to the lack of models available for me, I tried to model each part and furniture on my own, but it takes a massive amount of time…
In this matter, what’s an acceptable production time for, let’s say, a 90m² apartment scene? (using existing models vs. making them manually)
Also, for the average computer running an interior scene, how many polygons make for a nice polybudget? I can’t help but wonder that if one uses traditional architectural models, even the smallest scene would blow up the highest budget.
I use a combination of 3DSMax and time.
Just as you said :
… See it this way: The modeling si already done and “only” the unwrapping job is left to do.
Its not the sheer polzcount that affects performance. The Kite demo is the best example for that.
I could easily build a level with only a cube in a room and put an obscenely expensive material on it and performance will be less than the same room with a 60000+ poly mesh which just has a simple diffuse color material.
If zou have to make tradeoofs, for a wood cabinet for example, it would depend where your focus is, detailed geometry or a really nice material.
A nice case study for you would probably be the Berlin-flat example from the Marketplace/Learning tab.
Compare that to the Kite demo
Kite demo : Moderate polycount but texture maps, big like Godzilla. -> Low performance. Needs much resources.
Berlin-flat : Excessive polycount (500.000 to 1.000.000+ per asset). Smaller texture maps and simpler materials -> Speedy performance, no lags, no skips.
If you are able to break up these models into smaller pieces and use UE4’s automatic UV unwrap when you import into unreal, the end results looks pretty decent sometimes without having to do it manually yourself. That should speed up your workflow.
I also get my models on Evermotion, Design Connected and Dimensiva. The low poly assets aren’t enough for architectural visualisation imo. Just take em, unwrap them. I never had problems with high poly meshes. I have a 2,9 million polygons scene right now and it’s running well. (with a gtx 980 tho :-). The only thing that sometimes sucks is when the models are not well modeled…it can be a pain to unwrap, but when you are modelling it yourself it’s easy to think about the unwraping job and do clean modelling right off the bat to minimize the hassle later on.
I also break my objects in many parts. Even a simple chair, I’m going to break it in 6+ parts and manually (but could use auto from ue4) unwrap each one. Import them as separate meshes and group them (CTRL+G) in ue4 after I drop them in the scene.
Thank you all for your inputs.
It’s nice to know that the engine can handle models that high… also, the Berlin-flat scene is crazy. Lots of polys and things to learn from
@heartlessphil, that’s an interesting approach. How do you manage to keep things organized in the content viewer that way?
I just create a new folder for each object when they contain more than 5-6 meshes. I set up some filters to easily pick the geometry out of all the stuff I put in each folders (materials, textures, etc). When you select all meshes at once and drop them, they keep their relative position to each others. You don’t even have to manually place them one by one. Once in the scene I group them if I want to move the object around. Every parts have been flatten mapped in 3ds max. This way I don’t get overlapping uv’s and the islands are big enough so it looks decent.
I do this because I’m very bad at uv unwrapping, I don’t know how to make seams and stitches and I’m lazy lol It’s probably not an optimized way to do it but I’m not doing games and my scenes are relatively small so I don’t think it’s much of an issue.
