psilocervine

but wife city is two words

56k warning


cohost (arknights)
cohost.org/arkmints

joewintergreen
@joewintergreen

On Twitter once I did a big thread where I went through Half-Life Alyx's SDK - so this is Source 2 - and had a big ol' yarn about 'em. That thread's turned into a useful resource, so here it is in Cohost Post Form.


vectorpoem
@vectorpoem

if you did level design before the early 2010s it's weirdly hard to talk about the state of level design tools today. i feel like at some point much of what i valued in tools just went away, and it was like having my hands blown off in an accident or something. sketching spaces in UE4/UE5 for day job is just not fun, it's so slow and cumbersome and painful to iterate on. i'm sure it was a major contributing reason i started focusing more on older engine personal projects, eg Doom, Quake, etc.

anyways yeah, the torch for better level design tools is still being carried if you know where to look, it's just not reflected in the mainstream tools that everyone new is pushed to learn. and i hate that. i know a better world is possible there, i just don't have the energy to put my weight behind it right now.


psilocervine
@psilocervine

it's so bad! it's the fucking worst!! to say "ProBuilder, for instance, is barely better than nothing for level design" is pretty much the perfect way to put it and it's so bad that it's to the point where nobody on the unity forums, even unity staff recommend it for anything more than greyboxing a level out! frankly, I wouldn't even recommend it for that. probuilder is a feature that goes on the back of a box, like most things in unity, but actually sucks shit from an ass to use

aside from being just a glitchy piece of shit, actually using probuilder is a nightmare. the entire workflow is an utter disaster to the point where I sat down and went "okay, you know what? enough is enough" and started making level design tools for use in fucking blender. how absolutely dire does a situation have to be for somebody to go "I would simply rather make a level in blender than try and do it in the actual engine I'm using." I have had to make multiple custom asset postprocessors just so I can not use unity's tooling

hell, I'm even hooking into a script that lets me convert meshes to unity terrains so I don't have to use the terrain tools in unity either because the less said about them the better

actually doing level design in unity is a greater stumbling block to game dev in that engine than any of its numerous technical issues imhop


You must log in to comment.

in reply to @joewintergreen's post:

I, for one, am enjoying the Joe Wintergreen level snark greatest hits.

I threw some thoughts here: https://cohost.org/egotists-club/post/2270257-today-in-sisyphean-b after your latest round with Ryan, BTW. I have tremendous respect for Ryan, who publishes in the same areas I do, has a number of really good papers out, and really knows his shit; but I really want to know what internal data Epic has on tool use patterns that's telling them that people don't want this sort of thing, that it's not cost-effective to develop because nobody wants it, and that what people really want is more enormous art bills for one off or prefab meshes. That just makes no sense. Or maybe it's just another variant on Henry Ford's statement that if he asked what people wanted, everybody would have voted for a faster horse.

As I said, I really wonder who they are polling here. Even amongst AAA licensees, surely somebody would just like to make some nice corridors?

(I had assumed that the usage pattern for Nanite was that you would take a base set of brushes a la the HL:A stuff and just kitbash the everloving toot out of it, but apparently this isn't a thing. I'm even more confused by the claims further in the thread that we can't do this sort of geometry and must rely on instancing for everything is the GPU's fault, and I want to see the data for that as well.)

I think he really got turned around in that discussion. We don't need instancing at all, Nanite guarantees that. He seems to have a real hard time unpicking a bunch of thoughts from each other - You tell him we don't need high poly stuff, so he gets hung up on instancing; you remind him Nanite makes the instancing not matter, and he says it does because otherwise you're storing too many high poly objects; you remind him they're low poly and he says that's fine then, but now that you've both agreed his initial response was bunk in context you can't remember what it was in response to. I really just want to talk to him in person sometime

This is interesting to see as people are gearing to use the newer version of this(?) to make Counterstrike 2 maps, with cool features like live lighting preview (on raytracing-enabled GPUs)

in reply to @vectorpoem's post:

i got it working!

i discovered a new and unique and very edge case bug that would never affect reasonable people

godot is a modern style engine, it just drops you into an infinite empty space with a cube in it, pretty much like unity and unreal, have not made it much farther yet

ive been working on a plugin that adds brush-like geo that can be authored right in-editor, bypassing awkward import pipelines (even Qodot is unsatisfactory to me because of this)
still very early but if that ends up working out then that'll probably become the best option to use

in reply to @psilocervine's post:

Landscape tools are fucking terrible lmao.

Did you know the general shape of a given landmass is determined by erosion patterns building into networks of creeks that feed together into rivers? Forget that shit, how about you paint some blobs on a heightmap. You'll have to roll your own solution to rivers you bumbling moron.

yeah and people will point towards the new terrain tools but honestly they're so lacking in any fine-grain control that they're only good for things like background or broad stroke stuff and at that point I might as well just use World Machine or, again, even Blender

Yeah and the thing about WM-esque procedural programs is that they really just generate noise soup with only vague utility for designed spaces. Like yeah you can make an island or a mountain or some hills but trying to match a map you sketched involves a lot of wonky masking and invites a lot of randomness in. What's frustrating is basically that it isn't a tool for designed spaces as much as it is a background generator.

My current workflow is making broad shapes in blender, running them through a procedural program for erosion/terracing/flow, and using my own spline solution for waters in unreal. It's super bad 😭

Yeah I think they have something going on with landmass but I found it very frustrating in tests for making large terrains. A high-component map with a few paint layers and blueprint layers it would just chew up 40+ gigabytes of ram and crash. The UI has the same problem being squirreled away in a submenu (a 32 pixel button in the sculpt menu) and the blueprints expect a very granular arrangement of little factors for instead of anything resembling an intuitive UI for making prefabs with lots of examples. Documentation sucks.

The performance especially they could definitely iron out (or make it obvious how I'm probably doing something wrong), and like, the approach seems to work for them for Fortnite (but that's still pretty stylized terrain without many natural principles to worry about). Really I'm just bemoaning it because I can feel the bones of an uncommonly legitimate level design tool where I can make prefab landscape blueprints and arrange them or tweak them in editor. Most importantly, I don't think they've updated it at all recently.

Sorry to ramble lol, the frustration with tools is why I love the OP. I think they'd be great in the island cases they've shown off, but larger landscapes make it feel undercooked.