This Summer, I had occasion to dive into the Unreal 4 Engine, and I thought I would share my experiences, along with a starting link for modelers and designers so we can a better idea of what is coming. I am looking at this as a 3D modeler, a software developer, and a huge fat TS20xx railfan. Of course by “had occasion to,” we mean, “still looking for a job.”
I'll cut to the chase: I think Unreal 4 is a wise choice.
A Starting Place For Modelers: https://docs.unrealengine.com/latest/IN ... index.html
Now comes the Fine Print, what we all want to know, the age-old question which has tormented us since Sid Meier was a child, “and this affects me HOW, again?” What of my models? My routes? My careful attention to detail while painting ground textures? My revenue streams from DLC?
I was able to export, from Blender, using the AutoDesk FBX format, to Unreal 4 in less than 30 seconds. It preserved all my materials and textures. There was no intermediate “blueprint” step as there is in the Kuju railsim pipeline. It was a fairly complex model with many materials. FBX supports many meshes per file, so you export all your LODs in place, if you take the time to make them. Unreal 4, in its native state, can be made to create them for you, with varied results depending on the model's complexity.
The Basic Deal
A great deal depends on what DTG does once they get their hands on the source code. For $19 per month, me, you, anyone can download the complete C++ source code, stick a fork in it, and go on your merry way with your OWN way of doing things, up to and including modifying or completely replacing the Unreal editor, and the very engine itself. Given the complexity of editor replacement, such a choice would add a year to the development cycle, in my estimation. That may be a bit hopeful. And misguided, I think, at least from what I see.
Yes, I said COMPLETE source code. We're in a more civilized era. They also want 5% of your gross should you market what you create, but only if it is distributed as an executable. You could use this software to make movies and those are not subject to royalties. There are also exceptions for presentations, and for educators. DTG will engage in a custom contract with Unreal 4 to establish rights and distribution and all that good legal stuff so the price we pay to DTG has been realized.
Notice C++, not LUA, the scripting language of the existing product. This is likely the real rub as far as any direct conversion goes from where we are to where DTG will be taking Unreal 4. LUA underpins so many crucial railsim functions, it simply cannot be “done without.” In an ideal world, the sooner we saw some blueprint from DTG on their approach to this, the better.
I think it would be a better course to extend the existing Unreal 4 editor for the rails, if only for consistency's sake with other Unreal projects, which are going to grow exponentially. The complexity of these simulated worlds will demand a convergence of tool sets. Unreal, Unity, Cry, Steam... one of those is going to set the standard. Unreal has the momentum just now. $19/5% is stark, simple, and quite a blow to the other three.
Because my personal aim is much more realism than can be achieved with DirectX 9, I am having trouble justifying the investment of time on models for it. Railsimming is not going anywhere for sure. I can afford to wait. Unreal 4 supports DirectX 11, and the difference is just stunning. Like, when you got HD for the first time, for those who remember the days without it.
As I do not ever intend to wade into PayWare as it relates to railsimming, it's difficult for me to judge what kind of “deal” DTG will be offering payware authors. That 5% of the gross has to be accounted for somewhere along the line. This is another area where DTG has some serious deciding to do.
Peeling It Back
Now let me geek out for a moment on the details of conversion, because I do see a pathway for preserving some aspects of existing routes, such as terrain elevation data, track and signal layout, and even model placement. The C++ to LUA problem can be addressed with shims and stubs after a one-time pass through an LUA “source adjuster,” if you will. A subsequent release of the product would then have he ability to easily clean those up.
As for the Kuju blueprint system, I think it's history. To get the performance benefits from Unreal 4, there really can't be any stand-alone asset database clogging up the works. They will have to conform to the way assets are currently handled. DTG will almost certainly extend the Unreal 4 code library to include railsim-specific constructs, and these new properties will at least resemble the ones we set in the Kuju blueprints today.
What the existing Unreal 4 Editor feels like to me is a cross between Blender and the editor for Second Life. On import, you are able to set most of the options you currently make both in the Kuju blueprints and from within your modeling software, such as LOD, behaviors of materials, assignment to scripts and so on. What's more, unlike the railsim, which only lets you rotate and change elevation, Unreal 4 lets you apply practically any transformation at any time. Think: “everything I have ever made can now be made to pulse and blink in a horrid shade of green.”
Much depends on how much of the editor DTG allows to stand. It is EXCEEDINGLY powerful. For me, this will be the make-or-break decision for them.
Blueprints v Blueprints
This is going to drive us railsimmers a bit nuts at first. There are blueprints in Unreal 4, but they do not provide the same function as in Kuju. In Unreal 4, you are insulated from having to write C++ code by a whole visual scripting layer. The Blender Node Editor is one example of how it generally works, where each node is a computing device, and you visually link inputs and outputs like wires. In Unreal 4, a “blueprint” is a collection of these nodes and wires. As you may suspect, they are heavily re-used, so modularity is a huge key to mastering this tool.
And if the blueprints aren't enough, you can wade into C++.
Visuals
We're going to love Unreal 4. By the time a DTG product arrives, they will have moved up to DirectX 12. The realism many of us crave will soon be at hand. But what has been painted on the ground before Unreal 4 won't make it through any conversion, I don't think. Personally, it has held me back because even getting fair results requires way too much time. Unreal 4's system is more intuitive, and lets you paint in direct colors is that is your fancy. You paint with textures and materials as well.
For animations, so long as they cross the FBX pipeline intact, they should work. And then, of course, you get a whole boatload of options AFTER import to perfect it.I see spending less time in Blender as a result, using it to beef up the mesh, do some texturing and then get into Unreal 4 for the final polish. Assuming DTG lets us do that. You can even join meshes in Unreal 4's editor, vertex to vertex.
Hints From Helouise
Wish I had more of these, frankly. But what is available is instructive. For example, to get the best out of Unreal's lighting system, there are specific model texturing guidelines to achieve the look you want.
So there may be a Rewrapping Era on your models during your pipeline process in which you adjust your texturing to meet the unreality. It seems some of this work could be started now, if only an inventory of those models which will certainly need tweaks. The Unreal Engine of today can serve as a stage for that effort.
A note should be made regarding documentation. Another burden relieved from DTG is a lot of detail documentation that already exists for Unreal 4, details that have never been sufficiently compiled by the makers of the TS20xx software, IMHO. I have consumed virtually all the Unreal 4 application manual, and am presently knee-deep in examples. They are even so nice as to tell you, “this aspect of the engine is new and we're playing with it, and we'll update this as we have results.” So you get an idea at least of what is coming down the pike.
Prophesy
What I predict will happen is DTG will not make any great effort to convert any data. They will bank on the long-standing community to keep writing content for the exiting railsim, which cannot possibly “go” anywhere anytime soon. The market will determine what makes it to the new world. A lot of OK models are going to look just crappy, I hate to say it, so there will be some of that.
It will be left to an existing railsim coding master, such as those that brought us RWTools and ReDEM, and the Blender IGS/IGA exporter, to imagine and create these tools, should they even be possible. Again, much depends on the choices DTG makes. It leaves me to predict any LUA to C++ pipeline is a pipe dream. Please feel free prove me wrong! Access to C++ also means texture maps of existing terrain could be processed into visual elements on the ground, opening the door to very realistic routes to come alive quickly.
I think it wise to hold that any conversions are going to be unsupported piecemeal affairs.
What is clear is we are headed into a much more open environment. Unreal 4 has no DRM. Since everyone can get every line of C++ code, DTG won't be able to hide what it's doing NEARLY as well. I regularly read fantastic ideas here for the railsim that aren't possible today simply because there is no feasible way to know how things actually work. It turns away the vast majority.
Unreal 4's statement to the world is that engine won't start. And DTG could earn a lot of instant loyalty by releasing design details ahead of the launch. It would be most civilized, and there really isn't any excuse not to. We know too much. And, revenue streams are a serious issue that needs talking about now.
I look forward to what y'all have to say about it!
(Disclaimer: I am not affiliated with Unreal 4 or DTG in any capacity other than a user of their products. See “had occasion to,” above.)
