Some bugs I found-
Clicking on a complex StaticMesh actor I have added, a car, crashes the editor. I can drag materials onto it but I can’t click it.
I cannot select subtractive brushes from the viewport sometimes, when they’re at the same distance from other objects
I just checked again and one StaticMesh, when I “replace with composited blueprint”, just gets deleted for no apparent reason, so it has effectively deleted my work. For another StaticMesh earlier today I had this same feature save a textured version of a plain StaticMesh, so that’s great. This time though I was trying to save the transform, but/now I respect the decision to draw a clear distinction between content editor(/Maya/) and level editor. However for this issue, it’s a silly stipulation. (It really helps me to flip my model 90 degrees and not ex/import it again, because of time, and I will leave actors all over a demo scene just for this problem, probably. The t3d will help)
When I highlight all the assets in a folder and move them, one always seems to get left behind.
Rotating-then-scaling cube brushes makes them skewed- I’m not good at math but AFAIK the only order I’m supposed to do it in in 3d math is: Scale -> Rotate -> Translate. When I clicked on a rotated cube brush the editor crashed, so I associate it with this but I could be wrong. All the crashes are from either my direct action or FBX importing, though.
When I try to increase the sides of a cylinder brush after positioning/transforming it, it resets to the initial position- this is most unwelcome.
Most of my FBX StaticMeshes crashed the editor, maybe because I’m not using the right FBX version. I tried some large meshes, but I know that doesn’t narrow it down.
Suggestions-
3dsMax& Cryptic AR (GPL’d which is fine for tools)
A complete sample of this workflow would be positive—that is, if the FBX exporter is compatible. Possibly you can download the newer FBX exporter for 2010; all I know is I haven’t had a problem yet with static meshes.
How about a “hatchback car mesh” and similar for convex collision generation? Taking only 1000 points at random might be enough to suggest a shape of a very complex model in some cases. I could tick a “suggest shape” checkbox for working with this particular mesh asset, and then it might prompt me if it finds one of many shape profiles suitable.
For the rapid prototyper- a project setting to flag fbx’s as “use complex collision info as simple”.
Things like setting specular to 0 is something I want to be able to automate. If I am importing hundreds of meshes, what do I want most? Maybe a Blueprint for handling every new FBX StaticMesh and so on (I haven’t studied Blueprints yet).
The rotation gizmo/tool is even worse than 3ds to use! Please make it intuitive ASAP
What would really help with materials, I think, is if I could see where the “choke points” of the shader compilation are happening so I can avoid those parameters.
I’m using Light Propagation Volumes for everything now (no lighting set up(!!)), so I really want to know how to speed up the Eye Adaptation; I hear Blueprint support for shader params is coming soon. I’m hoping it will be possible to stop the dark rooms “going” to full brightness because of no windows open. But I understand this stuff is complex to implement.
For the impatient like myself, a high quality car paint material would really add more value. I don’t know why there aren’t already hundreds of materials and samples being offered… because, an investment in Unreal 4 is a wise investment in its brand/perception, and yet with all the existing UDK games/content having been made, surely there are some assets here and there lying around that would be compatible. But anyway, outsourcing lots of samples/materials is what I would do if I was managing Unreal 4 and wanting to appeal to such a wide user base in the cheapest way.
To me it seems that the memory usage is far more than with CryEngine. With CryEngine’s editor, last year, I loaded really huge terrainsbeyond the engine GUI creation limit, by modifying the files, but still it didn’t top 4gb of RAM usage and I am very sure about this as my motherboard only maps/supports 4gb. The racing/rally demo in Unreal 4 was too much to even load, with no other apps, on Windows 8 with 4gb RAM. Kind of took me by surprise, since I never noticed I had only 4gb of RAM with CryEngine! All of its demo maps, and some more complex maps, and all of the FarCry 3 game loaded fine with 4gb RAM as well as on my laptop, but my Windows PC choked on this racing sample. Interesting, for me. (I will probably develop side by side with both engines anyway, since the price is right.)
Lack of DDS and TIF(F) support is a huge annoyance; luckily for me I know how to get around problems like this with even hundreds of assets at a time. But really, excepting licensing issues I don’t see any point in excluding popular formats, unless they’re hard to develop for. An unsuccessful conversion with some MIT-style library would at least be closer to an automatic import, and it really makes all the difference for me. Maybe I’m just lazy, but that means all of its significance for my resulting impression of the Unreal brand/service, and a good product designer knows that pleasing the lazy people is effective when contrasted against the competition’s effort(s).
If I am an indie developer throwing together a basic model collection and c++ code, things like wood detail enhancement materials are the things I’d pay for artists to put in the marketplace. Things I wouldn’t be able to do with a median level of knowledge: these things (like Light Propagation Volumes) add immense and fundamental value to the Products and thus the Unreal brand, and at LITTLE cost to Epic Games/whoever (in the materials case). What I would like to see with Unreal 4 is the rational prioritizing of how value can be brought out of these Products. Breakpoints in Blueprints was a very intelligent idea. Maybe animation is the Next Big Thing that really needs a bit of a push-start, because I took one look at Havok’s solution and said no. Havok’s Anarchy editor/campaign/thing has physics going for it, but then so do all game engines now. Things like city generation and terrain generation are laughably slow to evolve out of Unity’s content emporium. These things, like Light Propagation Volumes and the Detour library, can and should be packaged-off for the indie tier of developers who(m) are only troubled by the increased complexity. What’s needed to develop Unreal 4 into the definitive best Cold War game engine, though, is a much deeper knowledge of all the problem area of not just modern game development but development hell. Full-functioning samples of the more complex aspects of game development, like say Runevision’s “Semi-Procedural Animation for Character Locomotion”](?v=0ea-CXxBEYg), ease the mind and lift the anxieties about that engine’s capabilities, which is powerful subtext. Fundamentally, the game industry has so many gold tech nuggets that ought to be democratized by now, if the development of games in the next few decades is to genuinely improve on more than a presentational level. Blueprints and similar technologies have potentials behind them, but like their forerunners at each iteration, that potential is often left latent when there is no incentive to fully bring it out. Fundamentally, then, you are also talking about the act of engaging more directly with licensees in more than lip service, to genuinely figure out the meat of what their problems are, by which means some real genuine progress of the overall quality of video games is surely possible. We don’t want to look back in another 10 years and see mission-based games with no depth and all the marketing ploys. Well, I don’t. And to that end, sharing=evolution is a must and not a luxury, because the technology barriers aren’t the only thing I’m interested in tackling.
