So, I actually like generative AI (disclaimer I feel I have to include every time: local open models only), and my main problem with that image is how genericized the new face is. If you’ve seen a lot of AI images, it’s immediately recognizable as the default mixed Asian/Caucasian face you get when not prompting something more specific than “woman” due to the datasets dominating the training data. It heavily implies all faces will be similarly genericized.
I don’t think this tech will be viable unless creators can give the AI a reference image of what a character should look like when photorealistic, and that’s just going to increase the workload of running this in realtime.
Qwel@sopuli.xyz 4 days ago
It just isn’t the same person. Makeup was added, hair coloring, veins in one eye but not the other, lip and nose texture were changed to be rougher, lighting was put on the back of the jaw to make it look square, even the skin tint was changed. It’s not a bad image, it’s just not the design that the author was going for, at all.
NocturnalMorning@lemmy.world 4 days ago
It’s just typical gamers being hyper picky. We used to rant and rave about 16-bit graphics. N64 graphics were cutting edge when the system came out. And now we’re not happy unless the game has hyper realistic graphics, open world, and an amazing riveting story. It’s annoying.
The goal posts for a good game are always moving. Seems impossible to satisfy people today. And that’s nothing to do with AI.
Qwel@sopuli.xyz 4 days ago
The issue here is that the filter is taking a not-realistic graphism and forcing it to be realistic (and, along the way, completely changing unrelated artistic intentions). The desired outcome would be to leave thing as unrealistic as they were, as intended initially. It’s not related to being picky about realism.
If an AI company takes Ocarina of Time, turns it into a realistic game while changing everyone’s hair color and skin tint, and claim it is obviously superior, gamers are not going to be happy. The inside of that tree is not going to inspire the same feeling when using fully realistic lighting.
The goal posts did move a few years ago, as a lot of people got bored of realism and more into stylized 3D. All the realistic games kind of end up looking the same and the Sifu-Lorn’s Lure range of polygon count is much more exciting.
NocturnalMorning@lemmy.world 3 days ago
Anything to win an internet argument…
ericwdhs@discuss.online 4 days ago
Mostly agreed. For me the actual biggest problem here is Nvidia presenting this as the assumed default experience everyone obviously wants and using a heavily genericized face as a win. The tech needs to be much more energy efficient and configurable on both the developer and end-user side before I’ll give it any serious attention.
Regarding future versions of this tech, I think “death of the author” still applies to video games, so changing artistic intent isn’t always bad, especially for games that get frequently replayed. I certainly don’t play stock Skyrim or Minecraft anymore. To use your example, yes, a photorealistic (attempt of) Ocarina of Time would probably be too off-putting, but give me style options like BotW, Spiderverse, Pixar, anime, etc.? I’d be down to try those.