As a 3d modeler I can confidently say 20 people have never worked on 1 chair, ever, even at the dawn of CG. I can model a simple wooden chair in about 30-60 seconds, or a few hours if it's very ornate with cushions that need to have perfect seams when viewed super close, etc., but that's almost never needed. I can make it photoreal in Blender in about 10 mins to a few hours depending on the complexity of the materials if I'm making the materials from scratch, which also is not really how many studios do things; they will have a catalog of materials or just buy them somewhere unless they're a huge studio or overly ambitious. So yes the tools are monumentally better than they've been in years past, but we're not that close.
Photogrammetry (scanning objects/surroundings and having a program weave it all together) is not as straightforward as you make it seem. There are many artifacts that need to be edited out manually. It's a vast skill all on its own. It is getting better all the time, though. If you have an iPhone X or greater you have access to an even better and simpler kind of photogrammetry.
Then, artistic vision comes into play. Not every game wants to be photoreal. How many games use cartoony characters with wild proportions? Many. If your idea of games begins and ends at call of duty-likes and assassin's creed-likes, then yes those will become easier and easier to make when it comes to producing assets such as characters, environments, and even weapons if the tech becomes so easy that it's quicker than modelling the weapon from scratch (currently modelling weapons and materials that look photoreal is really not difficult; it falls under "hard-surface" modelling, which, in general, is far easier to trick the human eye into believing is real than organic modelling, which additionally needs equally-perfect animations to look real).
So let's say you don't actually mean "photoreal." You mean: "really fucking good; good resolution, high polycount models, infinite detail, and great lighting." Well that is certainly becoming easier. I'm sure you've seen the Unreal Engine 5 demos showcasing basically infinite-detail environments. An easy thing to forget is they had to scan these objects at some point, manually, and edit them (you've heard them call these "megascans"). Someone had to actually do that, for all this stuff, for these demos. This is drawn from an established library curated by Quixel. So there's a significant price tag on these objects (https://quixel.com/pricing). You will need that $199/month plan to get any kind of practical usage (I hope your game doesn't flop!). And then, again, what if you don't want realistic? What if you want stylized? You're shit out of luck on the megascans department. That just leaves you with Lumen lighting and nothin but a lot of asset generation ahead of you.
So, all that is for today, or the next few years. What about 2030-2040? I don't see it changing a huge amount more. Keep in mind that the more photoreal something looks, the more photoreal it needs to behave, too. You still need to pair your character rig perfectly to the animation, which is a literal job all its own, otherwise you end up with uncanny valley faces or that "guy in mascot uniform" body movement. There are efforts to automate the more difficult parts of this, but it will take longer than 20 years IMO to get actual real use-case-ready tools such as auto-retopology, auto-rigging, and auto-weight-painting. All these things exist today, but they are bad and only serve as starting points which you then must spend hours tweaking.
This became longer than I intended.
tl;dr I didn't answer the question. I kinda just said it'll be longer than 20 years. To answer the question? I'll be too distracted trying to keep from becoming homeless in deteriorating America to buy and play more than a few games per year, so I'll just pick my favorite from those based on the execution of the art team.
Log in to comment