Some feature requests

I already made a few posts like this one but I probably wasn’t clear enough, so here are a few features and ideas for the Unreal Engine. They probably won’t even read this, but just in case.

  1. More Hammer editor-like BSP editing
    The current geometry editing tools are trying to be simple, but they SIMPLY AREN’T.
    BSP geometry is NOT a mesh, and shouldn’t be treated like one.
    Simply do it the excact same way GTKRadiant and Hammer did it.
    This video explains it beautifully:
    Level Editing In UE4 Kinda Needs To Catch Up To Quake 1 - YouTube

  2. More optical camera
    (Lens flare types)


    More types of lens flares, light glow, more user-friendly DOF…
    These are just a few things that can be added to the camera to achieve a little more lifelike gameplay experience.

  3. Actual water
    The Source engine and even the Quake engine have it.
    I’m talking about making swimmable water easily.
    Unreal Engine 4 requires you to add a mesh, apply a water material to it, disable the collision, add a physics volume and then add a post-processing volume. Why?
    I’m not asking for fluid simulations or caustics (although caustics would be nice), just simple water creation. A water volume with the material on top and the physics and the post processing?
    A new blueprint type, like the character blueprint but for water?

  4. Lighting compilations
    I have 8 GB RAM and an Intel Core i5. Not the ideal Unreal Engine 4 setup, but it will do. So, when I build the lighting in a big level, the computer freezes. I will upgrade my computer soon, but I’m still sure that there is a way to decrease compilation times. I don’t know how the Swarm engine works, never really dug into it’s source code, but there must be a way of building lights faster.

  5. Dynamic reflections without these planar reflection entities
    I find these planar reflection entities bulky and just unnecessary. Look at the Darkplaces engine: it is capable of randering real time reflections for multiple surfaces, it does not show other reflective surfaces in a reflection and it works wonders. Also, making reflections like Rockstar did with GTA V, where there is a camera inside the player that creates a cubemap in real time, can work.

I’m sorry for comparing UE4 to other engines, and I don’t mean to offend UE4 users or Epic. I love the engine and want it to improve and become the best.
Hopefully an Epic worker will find this and do something with it.

Thanks for reading!

As far as building lighting goes–there’s nothing that can improve that, for the render to work it has to load a lot into memory. That’s why a lot of big games use either a render farm for lighting or they drop baked lighting altogether and use a dynamic solution.

Planar reflections where added because many users asked it. If you think it’s unnecessary then just don’t use it. Screenspace reflections can work most of the time and static reflection probes are good fallback. There just isn’t any magic silver bullet that would give dynamic reflections everywhere without very high cost. Just updating single dynamic planar reflection cost hefty amount of performance and you are asking dynamic cubemap. 6 times the cost for non accurate reflections everywhere.

Screenspace reflections are distracting and there is no need to create an object for planar reflections. A checkbox in the material’s properties can work.

Of course there is. Have a look at OctaneRender. It’s a fast, physically correct renderer and light simulator. The secret? It uses the GPU instead of the CPU.

And that’s even more limited because it has to load everything into the GPU memory which typically is much lower than your system memory. The best use is Redshift, which can use both but is still limited by memory.

Swarm engine can calculate and divide the world into parts, each part woud take up the amount of VRAM that you have or less.
500 cores (about 1Ghz each) is much better than about 4-12 cores (about 3-4Ghz each) at some tasks, you just need to know how to use them.
I’m sorry if my solution sounds dumb and unprofessional, I never “explored” the Swarm engine section of Unreal’s source code.

It can’t divide the content into pieces, you don’t know what might affect something else which is why the full scene has to be loaded. The way a GPU renderer works is different, a CPU core is not the same as a GPU core. Both types of renderers have the memory limitation.

No, not dividing the content but splitting the level into smaller levels. And it’s just an idea.

If object is shaped like sphere. How that would work? It’s really crucial to have the control for where and when plane reflections does happen.