I’m currently working on a (non-archviz) scene for UE. The big problem is: it’s so much more work to do it right (or even acceptable) … You must make sure everything has good non-overlapping UVs for baking and as soon as something moves (especially interior) everything gets a lot more complicated. You can’t simply take the models you use for your normal renderings, it won’t work that way. Screen-space reflections and reflection spheres have both pros and cons but it’s nothing compared to raytraced reflections - you will always have to fight against ugly side effects.
Another thing is that some features are just not really ready yet. It’s also very easy to run into limitations (max texture size, limited amount of dynamic lights, distance field shadows and the geometry you use etc…)
I’m not a game artist (and I probably will never be ) but I doubt that this will revolutionize the Archviz industry anywhere soon.
I have Unreal4 for 6 months by now I think and CE3 for far longer (I used CE in school before I knew 3dsMax, I imported by Sketchup models and created environment from Crysis Assets) and I knew it’s useless for regular Archviz, the work amount is simply not worth it unless budget and time don’t matter anymore. And that’s the practicality of the output..well.
But I am just having fun, creating small game assets for personal pleasure and watching others daydream of the “revolution” around corner…
That’s a very healthy way to see it
Unfortunately I had some very long talks with clients why those archviz UE4 demos are not just “realtime” versions of what people normally do in max with vray or corona.
“It looks good, is realtime and you can even change things on the fly in the editor…” yeah.. no…
C’mon… i’m starting to feel curious about. And since Corona has make my life less complicated i think maybe i have the time to do some unreal architecture levels to demolish interactively.
Corona render to texture (Cof!.. cof!)—
You would have to “hard-bake” it as overlay for each object’s diffuse texture (and create unique variation of the same shader thus) because at the moment Unreal doesn’t support importing of lightmaps.
Exactly!.. thats the only way i can think of using a full render solution from corona. That way i only use Unreal engine as a realtime walktrough generator.
Sadly the amounts of unwrapping to make a level is insane, it remembers my time doing modeling and rendering for games. Very fun times.
AFAIK Corona can output diffuse, reflection, refraction, translucency and emission channels using shading components both direct and indirect and I guess is fairly enough for now for a game engine (also AO pass) that way we can use unreal’s own ligth system, but for really stand out needs a pbr material like output.
But almost every renderer can. So the main idea of using Corona as a map generator is useless.
Maybe is time to let aside my obsession with animated dispersive caustics and start doing some unwrapping.
So… corona is about materials… unreal is about unwraping.
I will try to find a interior scene with minimum unwrapping required… is there a problem if i try with sample scene?— lets hands on.
The thing with Unreal 4 is that it still is a game engine. Before start it pays off having some basic knowledge of how game engines work. The things that DeadClown mentioned are part of the process to create games and in this case to create arch viz.
Because automatic sucks. Most of the time it will create way too many UV-shells which then will produce ugly seams somewhere clearly visible in your mesh. It’s better to have a continuous unwrap with specific seams somewhere hidden (corners, on the bottom, etc).
Another example: Since the lightmaps are usually very small in resolution (e.g 64x64 ) it’s is highly probably that little grooves (the gap between a car door and the rest of the car) which normally would be darker only get hit by a couple of samples, so you get ugly “jaggy” dark pixels/lines in your baking. It’s better to plan your unwrapping accordingly.
Don’t get me wrong, I would be extremely happy if there was an automatic solution
we are exploring Unreal aswell, and untill now we came to the same conclusions as Deadclown and Juraj.
But we are not giving up [I do not mean that Deadclown and Juraj are doing this]
Yeah, I have seen that blog post about X-Ray Unwrap update. It is quite cheap and obviously it does the work done pretty good (considering the reviews)… but I simply haven’t had enough time to give it a try.. yet
Because it won’t look good. Even with best automatic unwrap you will be stitching isles together to avoid seams, divisions and other artifacts.
Sometimes the amount of work to clean-up is bigger than to simply unwrap it manually from start.
The GI is photon mapping + final gather. Every artifacts shows heavily and can break the Archviz look.
Edit: Oh…Deadclown answered, I had no idea there was 3rd page, I don’t know why it opens on previous for me always.
Best is an automatic by hand method using some tools (like the one from bertrands forum) best of both worlds.
But you must first start with a good mesh and known some basics about topology .
I have seen some assets came out like a bunch of triangles. Its a pain in the ass to unwrap those unless you do some retopo.
Not even talk about starting with those mesh generated by photogrametry, those are the worst.
So workflow by now: mesh->retopo->create new UV and unwrapp->reproyection->optimize texture size->asset (more or less).
Lots of work…
Coming from a game art background this thread is awesome to read as I had pretty much the reverse experience when I learned visualisation. Imagine always having to unwrap everything by hand thinking about seam placement, overlapping, padding, wasted uv space etc. and then find out that a simple box unwrap with some small edits will be enough in 90% of the cases.
I’m currently working on a product visualisation which I’m going to do in UE4, it’s a lot of fun to create PBR materials using the Quixel suite. It’s not possible to get completely photorealistic results in real time yet but if you can accept a bit of a stylished look you can get some really nice results.
If anyone ever needs some help with “game art” related issues feel free to send me a message.
I am also interested in Unreal, just signed up and tried the demo scene from Xoio, had some trouble for some reason at first to open the scene (it was stuck at %29 and then at %74 loading for like 20 min, is it typical?).
I am curious about this, because I if you can combine this experience with Oculus, then clients could go wild for something like that and it already looks like some are willing to pay top dollar for this experience:
I wonder how they did this, did they bake it in vray or corona and then compiled it in unreal or unity as .exe file? Baking requires so much work that I have given up that practice couple of years ago when I tried one project with that technique, it just too much work when some element changes in the scene… Is Unreal 100% capable of producing its own GI and baking? So you can import just the model and bring it to photoreal level?
I quickly tug the Oculus Rift off my head, ending my virtual trip into a luxury apartment—a trip that costs $95,000.
What the.. I am moving to Miami to make this shit right now.
Yes and No. Yes you can import just geometry with textures, because Unreal does in fact, not allow bringing custom shadow maps (unlike Unity), it can only generate it by itself through:
1)Baking- Using Lightmass (which is Final Gathering + Photon Maps, basically simplified and ugly MentalRay). Photoreal ? Perhaps…if great care is taken. Lots of work.
2)Dynamic GI (Currently only Light Propagation Volumes)- Easy, looks bad. Not suitable for architecture (you can only choose big…or small detail, but whatever you choose, rest is ignored).
) but I doubt that this will revolutionize the Archviz industry anywhere soon.
[I do not mean that Deadclown and Juraj are doing this]
