TheBourbonIsTalking

@ErnestoMantaras @Scow21 @StormslayerDev Every example shown is the same face with modified highlights.

@Scow21 @StormslayerDev @Rev0verDrive It 100% hallucinates and the results are random like all genAI does. The look is determined by the dataset and it works on the final output image of the engine (and motion vectors), only in screen space. There's no control and you can see how no face looks the same across scenes.

My main issue with DLSS5 beyond its uniform, art-sterilizing algorithm is its adding zero value over existing tech. We don’t need to cut pieces out and adjust transparency. It’s an all-or-nothing package: it takes the performance hit and must run at 100% strength. There's no meaningful “in-between.” Traditional upscaling, HDR filtering, and artist-configurable post-processing already deliver flexible, more powerful, performance friendly results that devs and users can precisely tune. DLSS5 have yet to match that control or quality when artists actually get what they want. Judge it the same way as any other tech, turn it to 100%, let it fully transform the image, and see if most people agree it looks great. Right now it fails that test. It can’t be diluted, they need to go back to the drawing board, it needs to be completely redone.

@l2everend @CAMIKAZE78 More high ping garbage. If on PC enable ping on scoreboard

@Rev0verDrive @CAMIKAZE78 But, yet somehow between where we are now with BF6 currently and where we were with the Beta with registration/results quality is completely different. Maybe server quality degradation from Beta to now is the issue more that netcode?

So after another session on BF6 last night, I think the hit reg from a perspective of “me hitting opponents” has gotten better, and more consistent. HOWEVER, getting one-framed and dying around corners is still an issue even in regular, equal ping situations. So… I think the issue now lies with the game’s TTK. Sure, you could improve the hit reg all you want, but a sub 200ms TTK on some guns is always going to feel terrible to go against, and it also will give off the sense of dying around corners because of how the quickly the damage is applied to you. The TTK of the game needs to be looked at. A 1BTK increase at all ranges would go a long way to making the game feel less “1-framey” I think, without making it feel too “bullet spongey” in the process.

@StormslayerDev Are we looking at the same sample? All I see is enhancement via added detail. Take an 8K res photo of someone, then scale it to 2K. Overall to me it looks like they are simply applying inverse tone mapping (SDR -> HDR).

It's probably impossible. It's built on generative AI which doesn't accept inputs, and can only hallucinate a general idea inconsistently. If it functioned like a proper tool where you can insert the face texture and it would add shading and details whenever the character was in view that would be fine, but that would go against how Gen AI functions. The artist never picked these faces, and from what has been shown in footage they cant, it pulls its data from the same uniform place as every other LLM.

@BattleNonSense @VisualFidelity Curious what you think the poly count difference between the two skeletal meshes would be if the devs did go full 8K res textures to get that look? What's the FPS impact comparison then?

@VisualFidelity The only problem with your argument is that what we have seen from DLSS5 does not maintain nor enable the original intent of the game designers. Much of what we have seen in those demos could be achieved without DLSS5 as seen with other games. This is AI slop of the highest order

DLSS 5 off vs DLSS 5 ON illustration DLSS 5 critics complain about retaining ‘artistic intent,’ but seem to confuse art with tech constraints. Lara Croft designers were trying to create the image on the right, but tech limits gave us the one on the left.

@CrusaderTwitch And this is why most games don't bother.

Still can't get over how so many people were BEGGING for Prox Chat in Marathon and it's barely utilized

@steelkrill @ormusai @KakapoCalypse Big Brain problem solver

@ormusai @KakapoCalypse Problem solver right here 🤣

Here is a (theoretical) run-down of what a solo-dev gets from $1.000.000 gross revenue on Steam: Gross: $1000k Chargeback, VAT (~20%): -$200k Steam Cut (30% of 800k): -$240k Publisher (~40% of 560k): -$224k Taxes (Austria, ~45% of 336k): -$151k Result: $185k (18% of gross)

@GameDevMicah @gtsplayground It's a process I forced myself into a long time ago. Refactoring 3-6 months down the road is 5x more time consuming.

@Rev0verDrive @gtsplayground I wish I ran it that way

We just cleared 27,000 Wishlists across the Epic Game Store and Steam! Thank you!! Along with our Patreon Community and over 5,000 Early Access Supporters on EGS 🙏🙏🙏🙏💪💪💪indiegamedev wins! Unreal Engine 5 .@gtsplayground 🔥🔥🔥You keep us going😇

@GameDevMicah @gtsplayground Understanding quality and performance isn't a race to completion. Task, breath, review, test, tweak, breath.

@Rev0verDrive @gtsplayground Small teams with big tech dreams, I don’t know how I get a second to breathe!

@psousa @CobraCodeDev Similar system would be used for modular character customization. Abilities would be actor component based.

@CobraCodeDev That seems much better for reusable npcs

All characters here are using PaperZD's new skin system. This means they all share the same Animation Source, Animation Blueprint and Character Blueprint. The only thing that is different is the Skin asset, which makes updating things or adding new characters a breeze!

@doomednightmare @CodeByNZ @nicola_sosio Centered and responsive?

@CodeByNZ @nicola_sosio Nonsense. We have good enough browser support to ship the basics in every browser that counts.

Do backend devs really find joy in doing this ?

@CodeByNZ @nicola_sosio Then chunking your css and js so each page only loads what it needs. Then making it 100% dynamic on the backend.

@nicola_sosio Frontend joy is solving it in 2 minutes… then spending 2 hours making it work in every browser 😭

@CodeByNZ @AbhishekDadwa16 Like generating a lrg relational json array and caching it... Say a complex main menu that's db generated, user data that's needed every page load, shopping cart etc.

@AbhishekDadwa16 That’s real engineering joy… when performance starts feeling like a superpower.

@l2everend @CAMIKAZE78 But that's not netcode. I can say this because I write netcode. 99% of the issues are latency based. Then it's client and server processing times. Server needs to get the data before it can send you the results.

@Rev0verDrive @CAMIKAZE78 Basically just leads to the same core issue which is the netcode itself and its still there after being "fixed"

@BattleNonSense @TidyWire The 2 5090s aren't required. It's just what the build workstation was running at the time in an unfinished game.

@TidyWire And we get that for the low price of adding a 2. 5090 to achieve that with DLSS5..... 🤣 None of what we have seen has *ANY* impact on the actual gameplay, unlike features such as PhysX - but instead nvidia forces "AI prompt slop" visuals down our throats. They lost the plot.

I like that since everyone called attention to how shitty the faces looked so now they have to turn to environments. And all thats different is they upped the exposure and contrast a bit lmfao. Jesus christ these dipshits.

@Gbdawg331 @PunkyRaccoon @Normalssbbbx Not all of them play ranked hos. A lot of them are weak beta males.

@PunkyRaccoon @Normalssbbbx True competitive players WANT SBMM. Theres a reason they play ranked, to play the game at the skill ceiling.

There’s a very weird subset of people (casual players that don’t want to improve but also can’t handle losing) who wish that all competitive games were just rock paper scissors.

@MightyBlueFish @Normalssbbbx On average 80% of a shooter player base sits at 0.85 KDR. A 1.1 KDR puts you in the top 20%. AAA studios tend to design around the largest consumer base. $$$ keeps them in business.

@Normalssbbbx good players are a very tiny portion of a game's playerbase. Would you prefer to get demolished by a sweat every few games or to have a match where the algorithm guarantees you lose ~45% of the time with no way to improve your odds unless you no-life the game anyway?

@Normalssbbbx The UI would identify the mode clearly, so it shows in clips. Stats and any public facing profile would also note your skill based on the mode you play. e.g. 3.67, avg opponent skill 0.85

@Normalssbbbx Studios could easily walk the dog on this one by having 2 play modes based on competitiveness and skill. Simply naming and identifying them would solve a lot of the issue. Casual mode. Skill mode.