haar

@Syddy___ @WaleyMark @Pirat_Nation Byproduct? That was the whole point. Upscaling from 1080p to 4K means generating lots of new pixels, so it is genAI. DLSS5 doesn’t generate new pixels, it just changes the colors of existing ones. Let’s leave it here. We disagree, and in a few years we’ll see who was right :D

@Varzelhaar @WaleyMark @Pirat_Nation supersampling creating new pixels was a byproduct, DLSS5 editing shit is literally what it was meant to do

NVIDIA has confirmed that DLSS 5 its like a “filter” and works by taking two main inputs: - A single rendered 2D frame (the normal color image the game produces) - Motion vectors (data showing how objects move between frames) It does not get access to deeper game data like 3D geometry, depth buffers, or material properties. The AI looks only at the flat image and motion information, then guesses details about the scene, such as skin, hair, fabric, characters, and lighting.

@Syddy___ @WaleyMark @Pirat_Nation Supersampling also caused a lot of ghosting, blurring, and artifacts in DLSS 2.0. Both supersampling and DLSS5 produce details that were not present in the original input. Supersampling generates new pixels, while DLSS 5 alters colors in the color buffer

@Varzelhaar @WaleyMark @Pirat_Nation Because dlls5 does more than just sharpening, it literally adds stuff that wasn’t there earlier. It put make up on grace and gave her dark roots.

@NBSaphierMD Uhm, I don’t see other offers from private sector players, so that sounds like Elon at his best

Private sector at its best 🇺🇸

@Syddy___ @WaleyMark @Pirat_Nation Sharpening the image is also a form of enhancement or beautfying. Neural rendering just focuses more on lighting and shading to make frames look more photorealistic. My point is that both work on 2D frames, so I don’t get why people act like only one of them can ultimately work.

@Varzelhaar @WaleyMark @Pirat_Nation supersampling doesn't aim to "beautify" anything, its goal is to just upscale.

@Syddy___ @WaleyMark @Pirat_Nation tbh I don’t get this argument at all. By that logic, DLSS supersampling is also just a filter. So calling it a filter over and over again doesn’t prove anything. If DLSS supersampling has gotten this good, then that’s not an argument against DLSS 5

@Varzelhaar @WaleyMark @Pirat_Nation They did all that „work“ to end up with a glorified filter

@dedicatedmicro1 @opinali @Oratuomi I think you underestimate AI ability to learn patterns. The lighting and shading are already in the color buffer. That’s enough information. That’s also why Nvidia said it works better with RT/PT: the more accurate the base frame, the better the result. Time will tell who’s right

@Varzelhaar @opinali @Oratuomi Because dlss 5’s advertised job is to add realistic lighting. If it doesn’t know the positioning of lighting in a scene, it cannot do a good job. It’ll just be higher performance cost for worse visuals.

Daniel got important clarifications from NVIDIA. TLDR: the "DLSS5 skeptical" were right about *everything*. 1⃣ It's a 2D AI Filter. Input is only color buffer & motion vectors. The model doesn't see geometry, lights, PBR properties, normals, anything🧵 youtu.be/D0EM1vKt36s

@stuffybag @ptlemic @Mallchad @opinali seems like you think 2D can’t be enough. DLSS supersampling works with 2D input

@ptlemic @Mallchad @opinali seems like you have difficulty understanding that a 3D scene has vastly more information stored and calculated than a 2D image

@HUMINT_GO @ptlemic @opinali Is DLSS supersampling a scam? It doesn’t have 3D scene understanding either, yet it works very well.

@ptlemic @opinali if it doesn't have full context of the 3D scene then it's a scam that will never reach its full potential.

@opinali @Oratuomi So what’s the point of all this outrage? If supersampling can be really good while being “just a filter,” why does everyone conclude the new model can’t be?

@Oratuomi Yes, most of those technologies can be described as filters: upscaling, FG, and now DLSS5. Only denoisers including DLSS/RR and FSR/RR, also NRC, I believe have deeper integration to the RT pipeline (but don't quote me on this, I'm not a game engine dev)

@Syddy___ @WaleyMark @Pirat_Nation And this sentence still didn’t make you wonder why it’s so computationally expensive? Hmm, maybe it’s not a filter. Maybe it’s a huge model trained on millions of data, learning patterns of lighting and shading of photorealism. Show me a filter like that.

@WaleyMark @Pirat_Nation An instagram filter that needs a whole ass 5090 to run

@WaleyMark @Pirat_Nation For fuck’s sake, has everyone gone mad? Yeah, sure, it’s an Instagram filter. Show me any filter that takes a color buffer and motion vectors as input and outputs a more photorealistic image. Why would Instagram even have that? Reality is already photorealistic.

@Pirat_Nation So in other words: It's a fancy instagram filter

@coop92968 @MisterOptical @Pirat_Nation Lmao, how can you call others retards when you write a sentence like that? Because if your sentence were correct, DLSS supersampling wouldn’t be able to work at all. But we all know it does.

@MisterOptical @Pirat_Nation >how is generating 3d objects off a 2d image of a 3d object bad I cant tell if you're being retarded or not

@muxmutt @MisterOptical @Pirat_Nation Does DLSS supersampling understand the content being rendered? No. Do its “hallucinations” (the errors it makes) make it worthless? No. What might this imply for the new model?

@MisterOptical @Pirat_Nation No deep understanding of the content being rendered = it's just going to hallucinate like any other AI model because that's all it can do

@gigib82 @SawyerMerritt @Tesla Hmm, useful context, thanks, tho I can’t find the original post in Dutch, only this: rdw.nl/en/news/2026/r… Nevertheless, given their previous news about FSD and the approval dates, that seems to be the case. I wonder how Tesla came up with this date if RDW didn’t tell them

@Varzelhaar @SawyerMerritt @Tesla That’s quite clear if you read the original message in Dutch. They are irritated that Tesla posted “again” an expected approval date.

NEWS: Dutch regulators (RDW), the key authority that would clear the path for Tesla FSD approval across much of Europe, has issued a response to @Tesla’s 𝕏 post today. "We know that Tesla's application has the interest of many people. The RDW gets a lot of (media) questions about this. Given this great interest and the many speculations, we would like to give a short response to Tesla's message and request. Normally, we never make statements about requests from manufacturers because of the market-sensitive information. Assessment process: In the message, Tesla states that they are in the final phase of the assessment process. That's right, Tesla and the RDW are currently going through the final stages of the assessment process. About 18 months ago, the joint intensive testing program began. During the term and in this final phase, the RDW thoroughly looks at the test results and analyzes the data. During this final phase, our inspectors will review all data and test results and, after completion of this process, a decision will be made on the approval of the driver's assistance system FSD Supervised. For the RDW, (traffic) safety is paramount."

@digitalfoundry Wtf? Why would it have been prudent to wait and see the audience’s reaction? Are you here to protect people’s feelings? Don’t bend the knee to the mob. You did nothing wrong. The video was about sharing your hands-on experience with DLSS5, and that’s exactly what you did.

The big DLSS 5 machine learning debate and why we should have waited before posting our first round of coverage - today's video: youtu.be/5dTTfjBAFzc

@TheMG3D Yeah, I’m sure you know it’ll be bad, even though it hasn’t even been released yet. Artists are just afraid they’ll become irrelevant because of AI

You are lying to yourself if you actually think DLSS 5 is good.

@kauraii @SlySinatraa it's not an AI filter :D or define what you mean by an AI filter. it's an AI model that takes motion vectors + colors as input, and outputs the image for photorealistic lightning and shadows. and it's not released, not even in beta, and it will improve. also, nvidia said this:

@Varzelhaar @SlySinatraa yeah applying an ai filter to your games definitely isn't destroying it's art direction... lol

That DLSS 5 shit is genuinely disgusting and I have no idea how anyone can support it.

@kauraii @SlySinatraa It’s fine if you don’t understand the topic, but then don’t pretend you do. GenAI is, by definition, ML (AI → ML → DL → GenAI). DLSS supersampling generates NEW pixels. How else do you think going from 1080p to 4K works? Also, it’s not destroying artistic intent

@Varzelhaar @SlySinatraa dlss supersampling isn't gen ai, it's machine learning, it's not actually generating anything new, dlss5 is basically a shitty filter that destroys artistic intention

@szmrek27 Áttörés 3d grafikában, ezzel közelebb kerülve a fotorealisztikus játékokhoz

Mi a fasz ez?