Non-PhotoRealistic Quake (2002) (research.cs.wisc.edu)

There are couple hidden in subpages on the submission (look for "sketchy/blueprint/brushshroke NPR links" in the sidebar), + some in the paper here: http://research.cs.wisc.edu/graphics/Gallery/HijackGL/.

Here's a video of the "sketchy" shader. Couldn't find the others. https://www.youtube.com/watch?v=jrr3dt01jNc

I don't know why everyone's saying there are no screenshots. There's one here: https://research.cs.wisc.edu/graphics/Gallery/NPRQuake/sketc...

Is that really 3D models in the video though? It looks more like regular neural-style-transfer done on something like series of images from google street view.

One of the suggestions on that video was this analog approach which looks very interesting:

https://www.youtube.com/watch?v=uMHHfMTJTDk

Need's A-ha as a soundtrack :-)

It really depends on what style you want to achieve but it can get quite hairy pretty fast.

Non-photorealistic rendering is actually a whole (fascinating) academic field in itself.

Since you have conflicting goals (temporal coherence, spatial coherence, "flatness", relatively constant stroke/feature size, ...) it's full of interesting tradeoffs and techniques.

Even simple things like beautiful real-time stylized outlining can be really hard if you are picky about the result.

For anime style you'll need a little more than just a few shaders. The models themselves need to be distortable and perspective-dependent.

That might be what Unreal Engine refers to it as, but in academia it's referred to as NPR Rendering - Non-Photo-Realistic Rendering.

https://en.wikipedia.org/wiki/Non-photorealistic_rendering

or some 3d storyboarding, but then you don't need quake, it could be a shader for blender for example

> still doing interesting stuff with Quake

I like it too, but NPRQuake was a spring 2000 class project and was widely available in 2001. I remember making a version which intercepted calls with GL_TRIANGLE to instead call glutSolidTeapot. The framerate was… degraded.

A related, interesting rabbit hole I wandered through the other day (prompted by a thread on Mastodon), was that the early Sierra games included stronger art than the systems they were developed for supported. The thread started by wondering if there were smart ways to upscale some of the games because Sierra often used a lot of vector art, though the conclusion there is the vector/raster mix is too brittle in most of their games, because it is almost always a pixel-specific mix in the end result.

That lead to the interesting bit that SCI0 games from Sierra in the EGA era had artwork drawn at higher color palettes than EGA supported at the time (or more specifically, the artwork individually follows EGA restrictions of 16 color palettes, but a larger mix of different colors palettes for different artwork mixed in the same scenes/games than EGA can quickly cycle through and than other EGA games would have done in the same time frame), and included that detail in the games with the SCI0 engine dithering the results at runtime, rather than standardizing palettes at dev time. Sierra at the time was quite proud of their dithering tech because it gave the artists greater freedom and looked good enough at the time. Arguably, too they knew the technology would get less restrictive eventually and they could support that when that happened, though that was before the era of easy game patches so who knows if they ever thought to act on that for "easy" VGA upgrades.

The interesting part of that to me is that ScummVM reimplemented the dithering, then eventually turned it off by default. If you play a SCI0 game in ScummVM today you get the wider color mix of the original artwork. You can turn the dithering back on if you wish, but the artwork really does look better with modern color support rather than emulated dithering.

The PlayStation upscaling image is related to an observation I made upon playing the FF7 and FF9 re-releases on PS4: the polygonal models look pretty good! But the static backgrounds don't, and the juxtaposition is sometimes jarring. In their original form, the polygonal models looked well integrated with the static background, but not any longer.

I find it interesting how, for these games, upscaling helps only part of the graphics. I understand why, of course, but I was surprised how well the polygonal models have aged.

On the topic of changing look and feel, there is SUPERQOT [0], i.e. Quake in the style of SUPERHOT [1], meaning that time flows in proportion to how fast you move.

[0] https://www.youtube.com/watch?v=ciBgg-Fuf0o

[1] https://www.youtube.com/watch?v=dW-RJyI8LGs

> A similar trick is used in Crash Team Racing and the Homeworld PC game (or was it Homeworld 2? I don't remember).

It was Homeworld 2, apparently:

https://simonschreibt.de/gat/homeworld-2-backgrounds/

Here I was messing with running Colin McRae Rally (v2 I think) on Linux through Wine, oops: https://i.imgur.com/wJZ9cFG.png

I had a ton of fun with Valkyria Chronicles back in the day, and I’m not sure anything’s pulled of that sketched style in quite the same way. The brushstroke renderer also reminded me a lot of Okami[1], one of my favourite games from the PS2 era. It had a very distinctive style based on Japanese ink brush art.

[1]: https://youtu.be/G6NPoX7_ehc

Also appropriate, since Quake has become synonymous with wide angle field of view.

Perfect! Next we need a Bakshi-style rotoscoped Quake.

The jitteryness is an artistic choice (no tech limitation).

Madworld? https://en.wikipedia.org/wiki/MadWorld

While no-one would argue that Quake was truly photorealistic, it was as realistic as the hardware of the day would allow at a videogame framerate (30+fps).