PlayStation

No it is. We’re talking different types of quality. There’s subjective quality, which would be appearance, art style, direction, cohesion. You can make something high quality with fairly low detail meshes and textures.

Then there’s quality in the sense of fidelity. You could decimate a mesh and reduce the amount of surfaces on it by an order of magnitude, make it much smaller, faster to render, etc. and have it almost imperceptibly different from the original mesh. Same thing with textures and audio.

Then there are other optimisations, like cutting a mesh you’ll only ever see from one side in half, so instead of rendering an entire high-rise building, you’re basically rendering a cutout.

Doing this takes precision and time though, but it’s worth it because it makes the game run much better, and the asset smaller, at no cost to the visual fidelity, assuming the player doesn’t go out of bounds and views the asset from a side it wasn’t meant to be viewed.

Modern hardware and rendering techniques have gotten so good we can basically forego this though. Deep Learning Super Sampling was initially suggested as a way for lower performance hardware to run games better, but what we’ve ended up with is developers taking shortcuts, not optimising their games, and rendering them on lower resolutions while having DLSS take care of up scaling and improving the image quality.

You can have really high fidelity textures, meshes, sound, and VFX, that takes up a tonne of space, while still looking/sounding/feeling rubbish.