Kornel Kisielewicz

Fast-paced and turn-based - the return of a roguelike legend! *Jupiter Hell Classic* packs procedural levels, explosive weapons, and brutal tactical combat. Early Access starts NOW! (link below)

@nimlot26 Nowhere in the replies it is suggested that DLSS 5 has access to anything else than the color buffer and motion vectors. A single frame is analyzed, all "context" information is inferred. Which makes sense as it's supposed to work ~"out of the box" with existing engines.

People keep sending me Daniel Owen's video which basically states that DLSS5 is an instagram filter because it uses just one frame and a motion vector. While his questions were valid and he actually got the replies he needed, his lack of basic technical understanding took him to the wrong conclusion. A lot of you have a very hard time understanding the basics of how a frame is rendered in your game and what kind of information is available before postprocessing and straight up went to the idea is that a rendered frame is like the screenshot you saved on your disk. @ado_tan and his friends have some amazing breakdowns about how different games render stuff. I suggest you to dive on his blog and check all of them out. This way maybe you will understand a bit more about how games are rendered. Here you have an article he wrote about how GTA5 renders a frame. I'm sure it will be eye opening for a lot of you. Till you do your research...I'll just go and play Crimson Desert.

@pivovarit That's not an opinion, that's a fact :P

Unpopular opinion: coding in natural language is way harder than in a programming language when you need deterministic results

Have you tried Jupiter Hell Classic, 0.9 Melee update?

@macson_g That one is level pro :P

Keeping your emergency password list on a floppy disk is way safer than on a piece of paper, just sayin' :P

@SebAaltonen Do it! I need a bookmark to link people to :P

Considering writing a thread about DLSS and frame gen. My feelings are both positive and negative... My thread would likely make everybody angry as these two technologies are such a heated topic nowadays :D

@KrolMichal Ten filmik poniżej to nie DLSS 5, to fanowski remake AIem z youtube'a.

Jeden gracz powie: chcę DLSS 5 Drugi gracz powie: nie chcę DLSS 5 A wszystko zweryfikujemy w momencie, gdy wyjdzie pierwsza gra to wykorzystująca. Jak się przyjmie i dobrze sprzeda, to będą takie robić.

@MCH2024 @SebAaltonen @TvanHelsdingen @jumaron03 @kimmonismus How does it deterministically handle the amount of makeup? In some scenes the girl has more, in others less despite it being the same model in the reference (not hating, genuinely interested).

@SebAaltonen @TvanHelsdingen @epyoncf @jumaron03 @kimmonismus This is not Genie3. Think more RTX Remix Toolkit but automated/engine integrated. It’s deterministic and customizable. Artists advocated for it. It’s an overpowered tool to make photorealistic graphics. RE9 / EA FC are early customized implementations, rest is base model.

I seriously dont get the hate for DLSS5. Even though DLSS5 does enhance the appearance to some extent, the result is still a much more natural look. I'm well aware of the argument that this alters the developers' style, but 1) everyone is free to disable DLSS, and 2) I hardly believe that visible improvements fundamentally change the atmosphere. So far, every image has been fine by me.

@BRobo82 Also, she has time to do redos in the middle of the action!

@BRobo82 Dunno about the game but the person on the right takes time in her day for a decent makeup, and pays for a hair stylist. Also looks 5-10 years older, but that can be the makeup. Dunno what the artistic intention on the right was, but that's what it reads.

So, do you seriously think the image on the left looks technically better? Or is it just a case of 'this isn't the creators' artistic vision, and besides, AI is awful,' because hating on AI is super trendy and drives clicks?

@RadekDubisz Ich już dawno nie obchodzi reakcja środowiska graczy, przecież nie są już ich głównym klientem. Ich obchodzi reakcja tech-bro AI prasy i co za tym idzie tech-bro AI inwestorów :P

Na spokojnie jeszcze jedno przemyślenie o nowym DLSS. Niby trochę nie mieści mi się w głowie, że Nvidia mając wszystkie pieniądze świata i współpracę twórców gier, na przykładzie których chce pokazać nową technologię, decyduje się pokazać to, co wczoraj pokazała. Nawet zakładając, że faktycznie twórcy będą mieli pełną kontrolę nad tym, jak DLSS 5 wpływa na ich gry, a obecnie technologia jest na stosunkowo wczesnym etapie rozwoju i będzie udoskonalana, wydaje się niemożliwe, że Nvidia nie przewidziała reakcji środowiska graczy. Ale tu dochodzimy do wniosku - spodziewała się, ale i tak zdecydowała się to pokazać, bo (jak w przypadku griftu) odezwała się wokalna mniejszość, zapewne większości się to podoba lub jest obojętne. I to do tej większości zaliczają się zapewne inwestorzy spółki, którym neuron aktywuje się na "większy numerek, wincyj, lepiej, AI rewolucja" i reszta się nie liczy. A korporacji tylko o to chodzi - zapewnić inwestorów, że spółka się spektakularnie rozwija.

@gr8_rebellion This is the next step after having a world with two engines and 99% of the games using the default shaders of those two vOv.

That's one of the main problems with this. Nobody will even attempt to make games look a certain way anymore, just put in the barest minimum and apply the slopify-filter that needs 2 GPUs to do the work (everything will look shit in exactly the same way).

@TvanHelsdingen @jumaron03 @SebAaltonen @kimmonismus I wonder if they'll produce any of those cards for the gamer market too, though ;)

@epyoncf @jumaron03 @SebAaltonen @kimmonismus Yeah maybe. They have been getting quite good at fast and realtime diffusion though, and what they showed here is a glimpse of the future (they were running the DLSS on it’s own 5090) My guess is that with 6xxx series we’ll see more NPUs to handle this type of stuff.

@kiwitalkz This kind of contempt is what is wrong with the industry nowadays. No, it's not enlightened devs vs backwater gamers. Many devs also see this as a path towards visual mediocrity.

This video shows the massive gap in perception between gamers and game devs regarding AI. Gamers seem to be anti AI and game devs embrace it and see the potential of it.

@TvanHelsdingen @jumaron03 @SebAaltonen @kimmonismus If the tags are just a vec3 or few of control parametes (say intensity, value, reflectiveness) - it will still work sensibly. If the tag is "worn, slightly reddish leather" or "no makeup", you're screwed.

@epyoncf @jumaron03 @SebAaltonen @kimmonismus DLSS is a post effect, it already uses all the various buffers a game generates. And yes it does have some performance overhead of a few milliseconds. I don’t think adding an additional buffer would change the equation that much, but i could be wrong.

@TvanHelsdingen @jumaron03 @SebAaltonen @kimmonismus Genie doesn't do physics. It predicts physics the same way as everything - by predicting what it will look like the next frame. It's predictive, not simulated. It's still damn impressive how far can we go brute force though.

@epyoncf @jumaron03 @SebAaltonen @kimmonismus Genie isn’t just diffusing though, it builds a mental model of the world it generates, including physics and various other things. So it’s a little different from how regular diffusion models work, it is actually going to be smart about interactions and such.

@TvanHelsdingen @jumaron03 @SebAaltonen @kimmonismus When tagging - when does the computation from the desired result happen? If realtime - performance killer. If dev time then this isn't a universal overlay but something needing deep integration into engines.

@epyoncf @jumaron03 @SebAaltonen @kimmonismus Genie world is not screen space, it’s a world model. Anyways, even with a screen space model like this, tagging is not hard. You tag models, they could render as an AOV that the model could ingest. I don’t see the issue here, would not be too hard to implement.

@KKroQ Also, all of this is impressive! It's just that I personally prefer the "before" effect in most (all?) cases, and I'm sure I'm not the only one. This is not RTX, DLSS, framegen, this significantly impacts the color grading, granularity and look and feel of the scene.

@epyoncf IMO that mediocricity comes from most discussed results - which are from games that were not designed for the process, implemented probably during its active development. Those results imo go both ways. IDK how you can look at Zorah and not be impressed: nvidia.com/en-us/geforce/…

For a man that was raised on triangle boobs and billboard based heads - This is crazy.

@KKroQ There aren't many great deep dives on the web in the deluge of hate and hype, but this guy (his content i generally decent BTW) made a nice primer on the mediocrity effect - about LLMs, but it also applies to other nn transformer architectures - youtube.com/watch?v=0PB09f…