DLSS 5 is on track for a Fall 2026 debut and replaces a game’s original textures with AI-inflused versions to make them hyperreal. Or, out of one uncanny valley and into another!
Maybe this is just imo but making images 10000% sharper does not improve quality.
This is more like painting over an existing painting to make it “better” imo
Huh. This kind of seems like expanding beyond the scope of helping framerates.
Is yassifying each frame going to further drive up system requirements?
How else to sell the new 1000w graphics cards right?

It honestly looks good. Christ.
“The party told you to reject the evidence of your eyes and ears. It was their final, most essential command.” — George Orwell
Sums up life on Lemmy some days.
It’s genuinely disturbing that people like you exist. This is so much worse looking. It removes all of the detail and heart to have an ugly, generic ai image.
Shouldn’t it look the same? I thought the point of upscaling was to reduce the hardware demands of outputting something at a given resolution. This isn’t upscaling, it’s a social media filter
it’s a social media filter
that’s a perfect description
They aren’t upscaling a lower-res render in real time, they’re upscaling the textures in memory in real time and then rendering the usual viewpoint with fake new textures.
I don’t think that’s entirely the case. It might be part of what’s going on, but it’s definitely not everything, because actual shapes are changing https://streamable.com/pzap7c
Per the article:
According to Nvidia, unlike Video AI models, which must run offline and can produce unpredictable output, DLSS 5 operates in real time, using a game engine’s motion vectors and source color as inputs for its AI model. Those inputs are then used to combine the aforementioned photoreal lighting and materials, delivering consistent performance frame-to-frame. DLSS 5 has been trained to apply its visual enhancements across various model details, including hair, skin, and even fabric.
Ah, you’re right, this is even worse than I expected. Merely upscaling the textures would have been notably less awful.
It doesn’t even look good, it looks generic.
Their model wiped out all that atmospheric fog in the background and fucked with the contrast so badly that it changed the mood of this scene from genuinely-creepy Resident Evil screenshot into an uncannilly airbrushed tiktok filter selfie. That water/rain, grime, and rust on the walkway behind the character almost got removed entirely, the door on the ice machine on the opposite side got a lot of extra weathering and scratches instead, their hair has even changed color from blonde to badly bleached brunette in need of a touchup, and the eyes have too, from pale green to deep blue.
The fact that Nvidia thought this was a noteworthy improvement over the original is a sign that the inmates are running the asylum. Tech companies should leave the art to actual artists, because clearly just AI-upscaling a game’s textures without an art director involved in every step of the process is only going to bury it deeper in the uncanny valley.
It’s even worse if you watch the video they posted, it looks kinda ok in a still frame but because it’s an ai filter it’s constantly churning






