[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip / qa] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


File: 1731890007650844.webm (3.48 MB, 1000x534)
3.48 MB
3.48 MB WEBM
Corner collisions edition

/gedg/ Wiki: https://igwiki.lyci.de/wiki//gedg/_-_Game_and_Engine_Dev_General
IRC: irc.rizon.net #/g/gedg
Progress Day: https://rentry.org/gedg-jams
/gedg/ Compendium: https://rentry.org/gedg
/agdg/: >>>/vg/agdg
Render bugs: https://renderdoc.org/
Previous: >>103214753 #

Requesting Help
-Problem Description: Clearly explain your issue, providing context and relevant background information.
-Relevant Code or Content: If applicable, include relevant code, configuration, or content related to your question. Use code tags.
>>
File: boneless_skeleton.png (107 KB, 921x654)
107 KB
107 KB PNG
>>103215814
>Making my own scripting language: Not a chance. Parsing, lexing, and compiling/executing would take too long
I recommend implementing a forth or a lisp/scheme? Idk about the latter, but if youve got an autistic itch with concatentative languages, the former option will give you a tool you can be quite productive hacking away with.

>and also I'd need to make a JIT if I want scripts in my 3D game to not be slow as fuck.
The advantage with concatenative languages like forth is that even though the average implementation needs to pay the speed tax due to indirect threading, you could bypass it by inlining everything, trivially.

I'm trying to implement one to be used for writing games and small utilities; picrel is essentially the entire execution loop of a typical forth.
>>
>>103256910
I see. Alternatively, I could just use Mono for C# scripting.
>>
>To the guy making the roguelike in the last thread:
Unity is fine. I recommend separating the game logic code from the rendering code, just so if you do make it big you can avoid the Unity fees without too much hassle.
>>
>>103256910
We use forth at the studio I work at professionally and everyone hates it.
We're working to make a gui in-house to replace it
>>
>>103256842
aw man I like being made the OP but not when it's my janky bugs lel

Got it sorted though

>>103254943
I didn't start out using algorithms, I just try shit and then scream and curse at my pc and try other shit.
Eventually I found an algo to find the nearest point on an aabb box and that helped me find the right normal.
Here's my dogshit if you want to see it:
https://github.com/dontpanic5/project-atlas/blob/master/Test/World.cpp#L103
Keep in mind I've been rushing because this has all been for a timed jam.
>>
>>103257904
sorry anon your first bug was so fucking funny even if that wasn't your intention
>>
File: HITBOX COME BACK.webm (2.67 MB, 1243x697)
2.67 MB
2.67 MB WEBM
>>103257991
it's fine, I have a rich history of amusing 4chan with my poorly programmed jank
>>
>>103256842
>(Press Retry to debug the application)
Is this a meme. Are Microsoft devs putting memes in MSVCRT. Just rename the button to "debug". It's a meme.
>>
>>103258557
>He doesn't know about Abort/Retry/Ignore
How young are you
>>
Am I crazy to think of doing shadows in a 2D isometric game by using a shadow map generated in blender?

Let's say I have the world in blender and generate a sprite + a depth map from the view of player and then generate a second depth map from the view of the sun.
Have each characters being also rendered sprite with their own depth map too, just no shadow map for them.
And to apply the shadow on the characters would be to turn the fragment position to light space and check the shadow map.
>>
>>103258109
cris-core
>>
File: 344.jpg (428 KB, 1104x519)
428 KB
428 KB JPG
>>103256842
My initial goal was to make a simple housewife manager where you had to balance your business life and your family life, with the risk of getting fired or cuckolded by the pool boy, but somehow it is becoming Depression Quest RPG.
E.g. you start unemployed, you can go job hunting but the odds of finding a job are tied to your college credits, of which you have none.
So, you have to take loans so you can obtain enroll and earn credits.
But even then, the odds of employment are very low.
>>
>>103256842
How do you guys implement menus in SDL? I'm working on a simple VN engine but I'm stuck at the menu part.
>>
one of the reasons I'm actually considering just using an existing engine is UI.
It'll be a PAIN in the ASS to implement, especially if my game uses a mouse-based control scheme.
>>
Should I buy some courses from gamedev.tv? They're all on sale for like $10 each (courses are usually $100+ each)
>>
>>103259998
nuklear, or ultralight, don't make an entire ui engine on your own its so boring
>>
>>103260625
nta but I don't like the idea of using immediate mode GUI libraries (which are meant for debug UIs) as game UI
>>
>>103260471
UIs are absolutely maddening
I’m sure the reason why games have such primitive UIs is simply because programmers hate working on them
>>
Speaking of UIs I remember back when this was a thing but autodesk had to go and kill it
https://en.m.wikipedia.org/wiki/Scaleform_GFx
I’m pretty sure there’s a business opportunity for anyone crazy enough to want to sell a UI middleware
>>
>>103260673
Then don't use it. Everyone uses HTML / JS / CSS for the gui for a reason despite it being heavyweight, because doing UI any other way is usually a pita. I don't like it but it is what it is.
>>
It's not that hard to make your own GUI system
>>
>>103261077
HTML is heavy weight?
>>
>>103261260
I can support this claim.
>>
I want to make a 2D game, would learning GML and using GameMaker or learning lua and using Defold be easier? I'm new to programming.
>>
File: 9ehQ7JT.png (43 KB, 1084x858)
43 KB
43 KB PNG
I'm working on a draggable windowing system for my new text-based RPG that works sort of like a desktop environment. It exposes window creation and management to Lua, with button callbacks, text wrapping, and a toolbar that can make windows permanent or unpermanent. The challenge for me as someone not too proficient in C is making windows and other elements that are instantiated in Lua completely memory safe; in the current version, memory for window contents is de-allocated when the user closes the window, with userdata handles pointing to windows by id rather than the window itself being managed by lua's garbage collection.
The window's contents (paragraphs, buttons, etc.) are serialized in a parsed data format with tags like @p and @b. So I can do things like this:

function window.callbacks:handleClick()
self:setContent("you clicked me")
end

window.create(300, 300, 300, 300, "penis")
window:setContent("@p sjklgjfdgg\n@b handleClick click me")


Soon I will abstract this to :addElement with proper builders. This will all be done on Lua without messing with C data structures. The C side is only responsible for rendering (including parsing the window elements) and animations. I will be writing my entire game in Lua and exposing functionality as I go.
p.s. the bitch on this thread who says you're not making an engine if you use raylib, I am already at 2000 lines of code and I'm not even a quarter of the way done, and haven't written any game logic yet. So fuck you. I haven't changed the font out of spite for you.
>>
>>103260952
>I’m pretty sure there’s a business opportunity for anyone crazy enough to want to sell a UI middleware
You'll be competing with NoesisGUI who have taken over that market, so good luck.
>>
>>103262916
>looks it up, sees "pricing"
Not everyone wants to pay $200 per project.
>>
>>103262936
thats really cheap
>>
>>103262982
not for me
>>
anyone here made something on Pico-8? maybe that would be decent for a noob like me.
>>
>>103262916
I use Noesis professionally and while it's good, it's quite heavy and not simple to use.
The WPF stuff it's based on isn't exactly a great framework, I mean it's a Microsoft GUI framework, it's got jank.
I would recommend small studio making small games do not use it, because of the amount you'll have to learn.
Don't regret choosing to use it in my professional integration though
>>
Lightning rod concept. There would be a separate magic variation for a fantasy theme. Wouldn't need to be a rectangle.
>>
>>103261817
do you understand how slow a HTML renderer is? when something is "blazing fast" in the web world, it means sub 1 second.
>>
How is good looking 3d water made? Especially if you're in it or close to it?
>>
File: 1000fps_comp.webm (2.75 MB, 1022x790)
2.75 MB
2.75 MB WEBM
started working on this again a bit

got variable output resolution working, so here's 1024x768 with 4x pixel scaling (crunchy pixels), running at 500-1000 software rendered frames per second BABY

>>103257736
forth is one of those languages ive played around with a bunch, written a couple half assed interpreters, etc, always felt more interesting to build one than actually build with one. very interesting that it's used at a game studio, some people are huge fans, everyone else hates it lol
>>
>>103265148
reminds me of teardown
>>
>>103265155
voxlap is prob the biggest inspiration
>>
File: 1717196888267634.webm (1.74 MB, 1280x720)
1.74 MB
1.74 MB WEBM
>>103259998
Making the UI for a VN is very simple once you've learned the SDL library. It's 90% just creating SDL_Rects, drawing stuff on top of said rectangles, and getting the mouse state. Use an int or enum to track which menu you're on. Pic related, all those buttons are just default rectangles. Parsing all the dialogue is the real pain and is the reason I hesitate to make a visual novel.
>>
SDLMixer or MiniAudio? Just looking to do simple ambient looping wind noises and other slightly pitch-shifted sound effects, maybe music later on.
>>
>>103262816
>I want to make a 2D game, would learning GML and using GameMaker or learning lua and using Defold be easier? I'm new to programming.
>>103263195
>anyone here made something on Pico-8? maybe that would be decent for a noob like me.

What you want is an easy engine with a big community so there would be tutorials/examples/open source projects and games. Godot is that but can be a bit overwhelming as a first engine.
Gamemaker used to be that 10 years ago but YoYo killed it. They killed Sandbox with 43k members, Community Forum with 286k members, GMS1 with Steam Workshop and most recently Marketplace. Everyone left. Good luck finding any working projects or tutorials. GM is dead now. They had a topic to celebrate 25 year anniversary a couple days ago and from all those thousands just 20 people showed up... It's over.
https://forum.gamemaker.io/index.php?threads/celebrating-25-years-of-gamemaker.116513/
PICO-8 is decent as a tricycle. Since every project in it is basically open-source you can learn basic things as well as complex game concepts. Also scene guys seem to love it so you can pick up on some neat optimization techniques. You won't be able to make a commercial game in it though.
Defold would be hardest to get started in since it has no community but it is Lua based as is PICO-8. So you could start on PICO-8 and then move to Defold to make larger projects.
There is another Lua based engine like Defold - Love2D. It used to be quite popular and I think it has a larger community but it somewhat fell off the radar in the past few years.
>>
>>103265073
shaders, lots of shaders and sometimes even a combination of rendering techniques where things that are further away are "just" a tessellated mesh but nearby it's ray marching SDF's
>>
>>103265353
I actually found parsing easier than UI. I'm still not done with the parser but it'll be done soon. C++ has some insane functions in the standard library for parsing. I went for a scripting style that has a command in the first part and then parameters for later. For example it'll be something like:
label Main {
show Lily happy
speak Lily "Welcome to the game!"
move Lily 100, 0, 0.5
}


Once I parse that, I'll store them in a vector and have a global index variable to get the current element. Every click, it'll increase the index by one and the game will execute the next line. Of course it'll auto increment after stuff like show or move. Parsing seems really hard at first but if you go step by step, it's actually pretty easy and fun.
>>
>>103266033
Sounds complicated. I'll have to do a different type of game I guess.
>>
>>103265148
Kino. How will it peform at 1080p? (with 180p or 360p render res obviously)
>>
>>103265148
How will characters look in this (assuming you have them)? I want to do a similar style, but my game ideas require characters in them and I fear making dynamic animated characters fit in this style in a way that doesn't look like minecraft is going to be difficult.
Of course, one option would be "3D voxel versions" of 2D SNES-esque sprites, but that isn't really dynamic
>>
>>103266232
To me it's not the parser itself that's a pain, it's manually inserting all the variables for every single line. Even if you have a really good parser it only reduces the amount of work you have to do by 10%. For example, I made a parser that fetches dialogue and variables from a text file using fstream but I haven't even implemented it yet since I figured it's neither significantly faster nor less confusing than just push_backing every vector directly in the game's code. That being said, my parser could be a lot more efficient (for example, it doesn't identify individual characters like yours but gets a string for the name and a texture instead) and if you have a ton of lines the increase in efficiency will probably matter.
>>
>>103267815
My main objective is not to make a game but rather learn C++ and SDL for future. I don't have much experience with both of them so I'm trying to create a modular VN engine just to learn about stuff like event loops, file reading and parsing, rendering, program structures, etc. I believe it'll be a good learning experience once I finish it(and maybe something I can put on my resume).
>>
should i learn Odin as a programmer noob?
>>
>>103266033
Why not SDF all the way?
>>
>>103268462
Meh, probably better to just learn C. Odin is nice but there's no info online and nobody uses it
>>
>>103266818
1080p with standard 2x pixel doubling is 140-160fps, full 1080 around 60-80. Depends on your cpu of course. I haven't bothered with anything more than 4x sized pixels, it's already low res enough.

>>103266851
Haven't thought about it, don't have any characters, I just found maps online in voxlap format. I'd probably do a 3d sprite style to start, maybe something like voxlap characters (higher voxel density)
>>
>>103269800
duplicated maps up to 1024x1024 grid (maps are 512x512)
pretty minimal performance decrease
>>
File: duplicated_maps.png (651 KB, 1923x1085)
651 KB
651 KB PNG
>>103270722
forgot image
>>
>>103265148
>running at 500-1000 software rendered
Damn, nice.
>>
>>103265148
>software rendered
based
did you do any kind of simd optimisation? multithreading?
do you scale with hardware acceleration or you're outputting to native resolution with cpu rendering too?
I always though that you can achieve pretty much any low res game with software rendering then you just upscale with hardware acceleration. not much of a difference.
>>
>>103265796
sunvox runtime
>>
File: IvyMen_pOwUDamIS8.webm (2.5 MB, 1008x724)
2.5 MB
2.5 MB WEBM
>>
>>103271332
Is SunVox good for SFX?
>>
>>103271944
Yes.
>>
>>103265796
OpenAL is pretty cool but the docs suck. Does a whole 3D DSP
>>
>>103271944
It's a DAW with all the blows and whistles, very fast too, you can play sunvox files directly without rendering them to wav/ogg/whatever
this is great because you can do dynamic soundtrack better than even balatro (which just had multiple tracks that switch between them)
it makes me think of banjo kazoi amazing soundtrack that used multiple channels.
you can also overlay and use any and all audio effects in realtime. or even generate procedural notes.
>>
Say I want to make it so that there's a Sims-like with moddability. Should I make my own game or just make an open source reimplementation of the sims 1/2 or something
>>
Which Anti-Aliasing solution should I use? FXAA, MSAA or TAA?
>>
>>103274838
If you're making a pre-modern game you should use SSAA, because it's the best, simplest method and the performance impact will be negligable
otherwise TAA, FXAA looks like shit
>>
>>103256842
i'm making a space invaders clone with SDL, trying to figure out the best way to handle input. if i simply do something like this:
while (true) {
while (SDL_PollEvent(&event.type)) {
switch(event.type) {
case SDL_EVENT_KEY_DOWN:
switch (event.key.key) {
case SDLK_W:
player.y -= speed * delta_time;
break;
...
}
}
}
}

where `player` is an SDL_FRect, there's gonna be a short delay when you hold down W as if you're typing.
i ended up making a struct like this:
typedef struct player {
const float speed;
SDL_FRect sprite;
bool m_right;
bool m_left;
bool m_up;
bool m_down;
} player;

and then the code becomes
player p = { ... };

while (true) {
while (SDL_PollEvent(&event.type)) {
switch(event.type) {
case SDL_EVENT_KEY_DOWN:
switch (event.key.key) {
case SDLK_W:
p.m_up = true;
break;
...
}
case SDL_EVENT_KEY_UP:
switch(event.key.key) {
case SDLK_W:
p.m_up = false;
break;
...
}
}
}


if (p.m_up)
p.sprite.y -= p.speed * delta_time;
...
}

is there a better way to do this? i feel like this really sucks but i don't know how else to write it.
>>
>>103274856
I see. What about MSAA?
>>
>>103275173
MSAA is downgraded SSAA, and it doesnt work with deferred rendering
>>
>>103265148
>>103270741
Looks cool
>>
>>103275147
That's pretty much it, you could pack the bools into a byte and use bitwise flags if you wanna be fancy. And use if statements rather than a switch if you want smooth diagonal movement at any point so that you can process multiple key-presses in one go of the event processing loop.
>>
>>103271106
Lotta data structure/cache optimization, it's a 2d grid of run length encoded voxel spans, with the spans and colors stored in separate arrays for cache purposes.

all cpu rendered, two passes.
first pass is multithreaded raycasting (no simd, algorithm isnt suited for it, trying to figure out a vectorizable version), which is the bottleneck.
second pass is a deferred lighting pass, also multithreaded, and using avx2.

No hardware acceleration is used for scaling, although that could be a minor performance win.
>>
File: ygg-112324.webm (2.51 MB, 1366x768)
2.51 MB
2.51 MB WEBM
>>
>>103275916
wow, didn't expect that. thanks!
>>
>>103276267
that's very cool, looking forward more progress.

>No hardware acceleration is used for scaling, although that could be a minor performance win.
It could be major performance win, on my old computer, there's a significant performance hit when you get to 720p, 1080p is even more slower, it was still within 30 FPS for me which is fine, but still.
Have you looked into Lattice optimization?
https://www.youtube.com/watch?v=5hmKH2pohVk
https://www.youtube.com/watch?v=4xs66m1Of4A
I imagine you can do an additional pass to smooth out the distant voxels with this technique.
>>
>>103256842
Just learned that floats are non-deterministic and not platform independant. Damn. https://gafferongames.com/post/floating_point_determinism/
So does this means there's no way to use glm for collisions, hitboxes, raycasts, and physics with determinism since glm uses floating point precision / math?
>>
Been playing some StS recently. It's so demoralizing knowing I wouldn't be able to make anything half as good even if I had all the time in the world.
>>
>>103278769
No, collisions use
>, <, >=, <=,
so it's not a real problem.
>>
>>103258912
Is there a suicide game over?
>>
>>103280509
yes, if your Euphoria runs to 0, you will commit suicide.
If your libido reaches 100% you will rape someone and go to prison for life
>>
I just drank my coffee and milk, my progress can't be stopped now
>>
>>103278769
It's possible even across different microarchitectures with aggressive rounding, throwing away some precision in exchange for determinism. Doing this in a way that's both general and efficient is a hard engineering problem though.
>>
>>103278769
>https://github.com/TeamHypersomnia/Hypersomnia?tab=readme-ov-file#tech-highlights
iirc you can force determinism by using certain compiler flags, some custom math function, and just accepting that you will use one compiler with possibly the same version restriction, which isn't that much to ask imo.
The advantage is that you get deterministic behaviour for even your physics engine which is huge for multiplayer game, you can now only send player inputs and simulate instead of having to do some retarded solutions to make physics synced at all, it's almost impossible because even "static" objects can be moving at some very small fraction.
>>
i'm working on a character editor, some people said they don't like blocky legs, so... what can be better?
>>
>>103278769
reject floats, return to integers
how do you think retro platforms without FPU units handled physics?
>>
>>103281664
nice minecraft skin
>>
>>103281681
thanks, the blocky legs are a problem?
>>
File: ygg-112324-2.webm (828 KB, 1920x1200)
828 KB
828 KB WEBM
Fishing to do list:
[x] made concept fishing rod sprite
[ ] make fishing line wobble on movement for aesthetic
[ ] animations for limbs for running with rod, and fishing related animations
[ ] catch a fish
>>
>>103281686
yes, the head is too detailed, the torso and hips are somewhat fine because that's how their shape is.
the arms and feet tho? looks like you got too lazy with them, and it looks like they animate like garbage too. add 2 extra loops to make those rounder, a loop on the front and a loop on the side.
>>
>>103281754
wasn't because i was lazy, it was because keeping blocks make it easier to texture (like minecraft), but is not a use case anymore because my game clothes are meshes now.
Gonna improve legs and come back.
The game originally was a minecraft like, now is something different.
>>
>>103281664
Looks decent apart from the blocky legs like that one anon said
How are you modeling this btw? I've been wanting to make a similar art style
>>
>>103281982
in blender, check the low poly threads on /3/
>>
>>103281754
>>103281982
better legs?
>>
>>103278220
>>103278220
I was thinking minor performance win relative to optimizing the raycasting, because the scaling is done in the lighting/postprocess pass, which only takes a couple ms, but if it wins even 1ms (need to benchmark, it's all consecutive writes) then that's fairly significant.

As for the lattice stuff, I had watched a bit of that video but not all of it, it might not apply because I don't mesh my voxels, unless it can somehow be applied to LODs/mipmapping which I plan on implementing.
>>
>>103282246
That ass
>>
File: file.png (365 KB, 1264x705)
365 KB
365 KB PNG
>>103282326
Watch towards the end of the greedy meshing video, the author was going through greedy meshing optimisation step by step, by the end, the ultimate greedy meshing (for this 2D face) is to reduce the whole thing to a single quad/texture, in 3D, you just have a grid of those intersecting quads. I recommend actually finishing the video to get the whole idea.
But like I said, the quirk of those lattices is visual artifacts towards the horizon, which could be smoothed.

Are you going to publish a demo or code? it sounds interesting regardless.
>>
>>103278769
make sure all relevant maths ops are performed in simd units, write your own trig approximations and use those exclusively, don't enable fast math and that's about it. will run the same on x86_64 and arm
>>
File: chara3.png (231 KB, 839x686)
231 KB
231 KB PNG
>>103281664
I'm also working on 3D characters but not low poly.
In general I would say straight and square shapes don't look organic at all, first thing they tell you when learning to make anything organic is to avoid anything straight. It works for swords, machines, etc. but those legs don't look like legs at all they look like blocks.

But if you're trying to emulate some specific style like PS1 graphics, then just do exactly what they did and don't worry about what people say. Having said that I doubt that PS1 graphics looked like that
>>
>>103281644
>The advantage is that you get deterministic behaviour for even your physics engine which is huge for multiplayer game, you can now only send player inputs and simulate
This solution has serious downsides, you can't do lag compensation
So it's only really good for RTS games where you have a lot of things to sync but input delay isn't really an issue
It's bad for action games
>>
File: Oni_PC_07.jpg (70 KB, 640x512)
70 KB
70 KB JPG
>>103282246
Honestly dude, I would just copy Oni if I were doing that kind of low poly
>>
>>103282455
I'll watch it, but I don't really do anything with quads or textures, just raycasts into a 2D grid of voxel columns.
>>
>>103282569
>you can't do lag compensation
huh? lol no, why would you think that?
>So it's only really good for RTS games
again, it works for all deterministic games, you see it work in the very link I posted, and that's a 2D multiplayer shooter.
Sending player inputs and timestamps/frame number is no different than sending the whole game state, because the game is deterministic.
You do of course need to resolve priority and conflicts of course, different inputs can still happen on the same frame, you also need to keep old world states to simulate and fix the game state, you can rollback and do whatever.
I think you're mixing this approach with lockstep, this isn't lockstep.
>>
>>103281664
I would look at what they did with classic runescape. I don't think you need to go as high poly as they did, but something between this and their designs will look more appealing.
>>
>>103282641
oh yeah I forgot about rollback
I wonder how it feels to comparison to regular lag compensation
>>
>>103282576
is cartoon ish
>>
>>103282349
Would you tap it?
>>
>>103282246
>>103282349
>>103282576
legs changed, i want to keep the cartoon ish style because i don't want to spend to much time doing realistic low poly graphics.
>>
>>103283204
also some strips socks
>>
>>103283204
try making limbs that aren't square
>>
How many transform buffers are you supposed to have? 1? 2/3? Number of concurrent frames? Do you have to update all of them when you move something? How do you apply a partial list of updates?
>>
>>103283401
What's a transform buffer?
>>
>>103283409
An array of { rotation, translation }
>>
>>103283204
The legs still look completely square, except the thighs are now shaded differently. What exactly did you change?
>>
>>103283439
sideview
>>
>>103283435
That's a question without context, there's lots of ways to deal with transform data, what are you trying to do?
>>
>>103283451
I think if you do those same kind of geometry changes but to the upper leg now it will look better.
>>
>>103283458
I want to send an array with transform updates to a compute shader that updates the projections array, which will be used by the vertex shader. If the projections array can't be updated because it's in use, how are you supposed to update it? If you have 1 per concurrent frame, you'd have to update all of them, if you have 2 or 3, you'd need some weird barrier scheme, etc.
>>
>>103283509
You should do partial updates, send an array of changed transforms every frame
If you have a multi-stage process happening on the GPU then you will need "weird barrier schemes"
>>
Does good antialiasing exist?
MSAA seems like the gold standard for quality/performance, it's not temporal so no ghosting or jittering and it looks kind of decent? Seems like it's OK for VR as well.
SMAA looks good too at least to me, but SMAA TX2 has the problem that all temporal AA methods have.

Then there's just using a high DPI screen without anti aliasing

TAA/FXAA/TSR/Upsampling are obviously out of the question
>>
>>103284960
AI anti-aliasing (DLAA) seems decent from the one game I played with it
>>
>>103284983
disgusting
>>
>>103285021
People are gonna look pretty stupid when they waste their time trying to solve technical problems on their own when AI can do it for them
>>
>>103285057
What the fuck are you even trying to say
Is this post AI too?
>>
>>103285122
It means rejecting what looks like the best anti-aliasing techinque because it's AI is stupid
>>
>>103285131
Sure, but when you don't provide any details about said technique other than "AI" you're going to get replies like that, specially considering how awful the closely-related methods of AI upscaling look
Does it still suffer from any of the problems mentioned? Is it realistic to use it without specialized/high-end hardware?
>>
>>103285330
There's no difference between AI antialiasing and upscaling in practice. It all looks the same. The kind of person who doesn't mind AI AA won't mind TAA either and is the kind of player who doesn't even know what those funny letters even mean
I would argue 50-80% of players fall in that category.
For me though, I can't stand the blurriness, ghosting and shimmering that all those AA methods cause but there's a solid argument in favor of them, that is to say don't care about what a minority of users think
>>
>>103285330
What details can I provide? It's propietary NVidia shit. I'm just saying it's the best I've seen
>The kind of person who doesn't mind AI AA won't mind TAA either
TAA blurs with motion
I haven't seen and DLAA artifacts, but I've only seen it on BG3
>>
>>103268467
ray marching is expensive, so you want to significantly reduce the marching distance but in doing so you only get high detail water near the near clipping plane and that's really only where it matters, the rest can be a tesselated mesh, then sew it all together with post processing, fog, splashing particles, w/e
>>
Since when do people use SDFs for water and not just a moving triangle mesh?



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.