For the love of the light, this cross-gen period must end soon, please.
I'm not a game dev but my understanding of the way textures have looked last gen is all down to the speed of the physical HDD... So many PBO photo textures needed to be artistically 'modified' in order to make the file size smaller. RAW photo textures can be very large and there is no way they're getting a whole scene worth of them in whatever little bandwidth they had available in physical memory because asset streaming was impossible so entire loaded zones needed to fit into 8-12GB of RAM. Again, I'm not a developer but this is where I understand the bottleneck to be. Are you saying it's not and devs have been deliberately making their games look worse for the sake of pride or 'artistic' choice? You're saying someone like Naughty Dog deliberately held back the resolution and quality of the textures in The Last of Us 2 (as an example of where I feel a games realism level is affected greatly by the quality of the textures specifically)? I just don't think I understand what you're saying. So you said you worked on Halo in the past... Halo Infinite just came out and uses a different texture for the ground rubble on the X1/X1X compared to the XSX/S. Can you please explain more precisely what the devs are doing specifically? Also to me personally, the most incredible thing about the Matrix Awakens demo is the lighting and physics, the textures are still not perfect when you get close enough but the lighting is essentially ray-traced and that's definitely not possible on previous-gen without baking light in like TLOU2 and don't get me started on the previous-gen CPU trying to do those physics...
I've done some 3D sculpting and while I'm not fast, I know and have seen some people throw together a model sculpt in the time you're saying so I believe that. But my understanding of the months time to work on this Troll is the complete model for its use in this sequence, all the rigging, muscles, animation everything was prepared using ML. This is what took the month, they built the base model which yes probably took a day, then using ML they trained the animation system in a month. That's pretty outstanding considering I believe it would take longer than a month to fully ground up manually keyframe animate this whole sequence without the ML assistance. Any further instancing of this Troll in the game should have its dev time much reduced because of the time invested to train the ML animation system. I believe this kind of ML efficiency insertion into game development is absolutely critical to bring down the ever ballooning dev time for modern games.
Still, I do not properly understand what you're saying. Not having a go, but I assume English is not your first language? Using so many capitalised words affects your credibility because you're saying you're a skilled 3D artist yet your language skills do not seem to match what you're saying so I automatically assume that's the case but I'd still like to confirm as it will impact the credibility I give to your statements. It's always interesting to get perspectives from developers but I'm on the fence about the truth of your statements. I shouldn't need convincing but this is the internet of games and it wouldn't be the first time someone has artificially claimed to be an insider to try and ramp the importance of their statements. I hope you understand.
On Your Question of me and My Origin, I'm American. My First Console Was An Amiga, Then A Commodore 64, My First Next Gen Console - A Nintendo N.E.S.
SSD As A Requirement due to Texture Size? No, SSD's aren't required to stream in Large Photo Reference Textures, besides that, most Textures can be sufficiently reduced in size and applied to assets
using highly ML optimized methods that fully today mitigates file size. I'm Saying, It has been very obvious for now over a decade in actuality, across many case examples (see below) - that most asset's rely on artistic expression opposed to photo reference and that is Tragic as it has held the Industry Back at the sake of what, finally delivering one Amazing "The Matrix Awakens" Demo?
We Have seen Photoreal Assets Demonstrated on Unreal Engine 4 on the GTX 680, released in 2011. Inferior to Last Gen Consoles.
And The Above Method Could Easily Be Expanded On.
You Could Build Larger scenes, without much foliage, and still achieve largely the same photoreal effect on the same type of Hardware. As Stated, this is running on a GTX 680, and a Demo like this could easily be expanded into to work for a Larger City setting. Now I'm not saying Last Gen Hardware Could Fully Conquer the Matrix Awakens Tech Demo, I'm saying Many elements of what is on Display - Could Easily had been accomplished Last Gen, merely by using genuine Photo Reference Materials.
So I Am 100% Implying, Yes the industry Has been Underselling Fully - What These Machines This Gen are capable of As Far As Plain/Vanilla Low Geometry Assets are Concerned, Artist's are not building Assets to scale, where you could have easily imposed photorealism by Maintaining and Building Assets at a 1-1 Scale mirroring The Reference Material.
Last Gen With many Plain Assets, Artists instead as standard all of Last Gen - Built Assets from the Ground Up at a less than 1-1 Scale Likeness Ratio For Most Assets... and Gamers/Consumers Have Lapt it up as Gospel - It's actually Extremely Sad From My Point Of Reference. It Proves How Helpless Consumers are when measured against the Standards Narrated By an Entire Industry. My Point, Consumers/Gamers - Start Demanding More if Graphics are Important to You.
Because the Industry as A Whole has been yanking your chain. And If my Example of A GTX 680, Achieving the Same level of Photorealism arguably, isn't enough - then you are in complete denial I must inform you.
Photoreal textures is a thing that is so last gen, so why were they hardly ever used last gen? When I Could Build A Comparable Smaller City, and Run it on a 680 with Photoreal Textures.
Because Unfortunately the People that really care - get spat upon by the Community while the Industry At Large, tells you what you should be impressed by.
And I Don't Sculpt Assets to use As Assets, but can Sculpt Too, I Build Models Vertex By Vertex, Tile by Tile, the Proper Way so that it is Already Heavily Optomized on Completion - And can still manage To Build such a Model as Long as the Reference Material Is Nailed Down in 3 1/2 Hours.
If I have to go off a description, and come up with it myself, based on many descriptions - increase the time spent on such an asset 3-10x depending on the scrutiny provided.