Should we still use World Composition ?


World composition has always been “complicated” to use, plus a lot of issues can happen (:confused: oops, some information was lost, lets restart !) and it’s buggy (to say the least) with the great but sadly useless Water plugin. :frowning:

Will we be able to, at least, convert a project using it for UE5 or should we start creating maps with a big (not tiled) landscape ? Does anyone have an answer ?


I want to know as well. If we can’t convert, that would be useless of course and leave many out of the loop, and I can’t imagine them doing that :slight_smile:

Hey folks,

We’ve spoken with the team to confirm- you’ll be able to convert your projects to UE5, including those that use World Composition. If you are starting a project or just delving into World Composition, the team recommends using it with only one layer fixed to a medium distance and then let the engine handle the rest. We hope this helps, and good luck with your projects! :slight_smile:

They said in a video they’re working on a new “World Grid” system, so that will probably replace world composition

Could you provide a link to that video? Thanks in advance! :slight_smile:

Niemas Problemas . Linkos… mmm… el pronto! :slight_smile:

Is there any docs for world composition ? like how to package a world comp map? Or how the heck do you place a player start on a world comp map? If i place player start on the persistent it works , but when i package i add all the heightmap levels to the package list , then in packaged game player just falls through map. And if i choose game starting map as one of the heightmap tiles , then level is just all black in packaged game.

If the machine takes time to load the streaming level, the character will fall. Maybe you should do as I do : try to trace for the ground and wait if it’s not here yet.
But yes, documentation est very thin. Like Procedural Foliage in World Composition : the options are not explained and nothing is said about how to optimize instances, how to limit usage and maybe not have every tree of the same type in the same instance, making it’s destruction at runtime impossible (or HUGE lag). Or how to radial collide with foliage damage… but It’s off-topic.

I know you are supposed to be the one wearing the tin-foil, Sam, but Iet me snatch it for a post or 2…

It sound a bit like “the team” has never actually tried to publish anything after Kite using world composition…

To be able to have a project run with world composition there’s a TON of things you have to literally micromanage.
You definitely need more than one layer - at least one of which set for manual override - if you actually plan for the world to be able to load interiors.

So saying that “it’s OK just keep only 1 layer” is more than a red flag.

Are you seriously listing a video from 2019 as a source for something which might happen which hasn’t yet been released in 2021

All he was saying is pretty much pertinent to world composition.
and used the same images as

from 2017.

It’s probably about as factual as those folks out there that believe the earth is flat…

Relax a bit :slight_smile:

I agree that World Composition is left unfinished. It works well in my case, but I had to some editor code for it. So it does automatically load levels with bounds encapsulating the camera when working in the viewport. Otherwise, everyone would need to load level manually…

Open-world streaming presented there probably is coming with UE5. That a proper solution while World Composition is just a workaround, so you might be happy… in few months :wink:

You assume the new engine will just work.
Unfortunatley, the track record so far is that every release they break the engine just a little more. To the point you have to go and build from source after fixing things to get a working version…
And let’s not even get started on how Chaos performance went down 400% between the first released demo and the (at least prior to last patch) official release of .26. I suppose “it works”, even if a lot slower and more expensive than physx with the same settings. But that’s been exactly the same trend for anything they released the past 2 years.
including RTV, left unfinished. Procedural foliage, left unfinished. hair groom, labeled finished and crashing the editor. Etc. Could go on for days like that.

^ This ^…

It doesn’t seem like Epic is doing anything particularly interesting with landscapes anymore anyway. For example, there’s far more interesting developments coming from the Community here. And that’s after all of the fortnite billions. Consider this… The best Epic could come up with recently to show off the new sky / atmosphere system was a 100% fake planet (using a material trick to fake spherical terrain). And that was from Hourences, not even Epic. So when it comes to landscapes? YAWN… Happiness comes from avoiding them altogether. :stuck_out_tongue:

Obviously, a lot comes down to urban environments versus rural plus genre too. Sci-Fi offers flexibility, whereas medieval RPG / RTS not so much. But overall the reason ‘Epic are swamped’, is that its not that easy to scale up from hundreds of workers to thousands and still stay nimble. There’s probably a relevant Sun Tzu quote for this. But that’s why the engine has been massively more unstable from 4.19 - 4.26 versus 4.10 - 4.18. It may not always be so simple for Epic to stay at the forefront of game dev going forward. They may even struggle. :wink:

Its nice to see some of the new features and options in 4.26 , but i think they should focus on optimizing the engine and polishing the features as well
, as this is important to developers , especially with open worlds , i mean "ray tracing and volumetric clouds and water shaders " are cool and all , but if it drags down your system , then its not so great anymore . Heck you can’t even get a better video card hardly these days , with the shortage and crypto miners.

TBTH, I think they went Down from 100 to like 10 employees.

Otherwise stuff should eventually be getting completed, not abandoned halfway.

The only other reason would be upper management being at fault with idiotic stuff of type “you, this thing you made is absolute trash I don’t understand. You now have to do this instead”.
across the whole trello board…

Also, ray tracing is another BIG ball drop. 1 year of reported issues, most of which are so BAD that nvidia is releasing engine fixes for them… (and guys. If you use ray tracing those engine fixes are actually good. Can vouch).

There’s probably some truth in that. But its probably about more basic things. Business schools and academics study this subject all the time. Basically, at a certain point, the bigger the team, the more complex communication becomes and the more difficult it is to get anything done. Here’s an example from Harvard:

Epic isn’t doomed, its just that this stuff is hard to get right. Sweeney’s push for acquisitions right now may not be helping either. As there’s always a sweet spot that tech firms need to reach to be at their most productive. If that’s not bang on (being too small / too large / expanding too quickly / growing too slowly), then you tend to get more chaos than productivity.:wink:

In my opinion it was a shift in priorities. At the beginning of UE4 whole team was mostly focused only on the engine itself. Their game development efforts were minimal, so they could focus on the tools and think what would be best for (at that time paying) engine users and the engine itself. When the engine was mature enough they started developing their own games and many new engine features (certianly not all of them) are driven by their game dev needs what makes sense, especially when games like Fortnite are so successful. Tools were developed to the minimum to meet the needs for their projects and then sometimes left in that state. With UE5 on the horizon it seems engine developers now have enough time to work again on long term goals and from what I heard they are already revisting some unfinished or outdated systems. This is surely exciting time for Unreal Engine and let’s see what they come up with :slight_smile: (Disclaimer: this is mostly just speculation).