next »

[zacaj]

Im not a pixel artist, but from reading http://www.derekyu.com/?page_id=219 , it looks like it could theoretically be automated.  Render scene without shading, find edges of color, get outline, run some cleanup on outline (maybe optional), fill the outline with flat colors, use lighting information(rendered with scene in separate buffer?) to add bright and dark spots to flat colors, change outline to darker version of color it outlines, done.

[Toeofdoom]

How has no-one mentioned the squads from darwinia yet? Admittedly they basically use a square blur filter thing, but it's related...

Anyway, for the rest your best bet probably lies in some form of edge detection and definitely uses ridiculous amounts of GPU power. Just as Alex May said.

[Chromanoid]

i thought about this for a long time too. but i had more an idea of a (maybe non-realtime) pixelart renderer that can create pixel art from 3d models as a tool for artists or to render (maybe procedurally) customized graphics like game characters on demand. i want a tool that automates outlining, shading, dithering etc. with a 3d model as reference. i would not use shaders/gpu operations but a software renderer that can run on any machine even servers without a gpu to be able to create pixel art on demand i.e. for a MMORPG .

[drChengele]

This, perhaps?

Oh wow.
So am I the only one who's switching to this the second the SDK comes out?

[Chromanoid]

This, perhaps?

Oh wow.
So am I the only one who's switching to this the second the SDK comes out?

i would never switch to something that has such an ugly logo^^. No seriously i think it has its own charm but it is not comparable to pixelart (no outlines, limited palette etc.). i would like to see better art with this unlimited detail technology. why they dont voxelize a human 3d model and animate it to show the capabilities of their technology? until they do that that i have the strong feeling that they keep quiet about a serious problem in their technology. maybe a hybrid approach like you can find in outcast is the best...

[drChengele]

Of course this is just a PR video and there's a ton of stuff they are not telling us. I would fully expect the technology to be full of arcane hacks and shortcuts and work poorly. But the technology is still in its inception, and rendered on CPU power alone to boot. Remember what polygonal 3d was like when it first appeared? Didn't we see a human-oid girl move around though? Of course she is made in 3d software and not an actual human model, but I don't see why that wouldn't be possible. If anything I'm guessing bone animation and skinning should be much easier with voxel-like technologies.

I say give it a few years and the first ATI and nVidia cards will start supporing this... then add in Voxel and Pixel shaders... a few tricks such as "vertex mipmaps" (dynamic level geometry etc.) and I think we're set.

But we digress from pixel art stuff. I don't think you can make "proper" pixel art from 3d models no matter how hard you hack the shader pipeline. The problem is that pixel art is painstakingly drawn - well, pixel by pixel  - and it doesn't always follow the underlying 3d geometry. Pixel art owes much of its appeal to clarity and it achieves this by sticking to fixed angles of lines and shapes.

This is unlike, say, cell-shading, which is perfectly possible because it fits extremely well with 3d geometry and shader stuff.

I would LOVE to be proven wrong on this though. I adore pixel art.

[Chromanoid]

Didn't we see a human-oid girl move around though? Of course she is made in 3d software and not an actual human model, but I don't see why that wouldn't be possible. If anything I'm guessing bone animation and skinning should be much easier with voxel-like technologies.

where? the only girl i see @5:18 is "an ugly example of polygonal rendering".

But we digress from pixel art stuff. I don't think you can make "proper" pixel art from 3d models no matter how hard you hack the shader pipeline. The problem is that pixel art is painstakingly drawn - well, pixel by pixel   - and it doesn't always follow the underlying 3d geometry. Pixel art owes much of its appeal to clarity and it achieves this by sticking to fixed angles of lines and shapes.

This is unlike, say, cell-shading, which is perfectly possible because it fits extremely well with 3d geometry and shader stuff.

I would LOVE to be proven wrong on this though. I adore pixel art.

this is why i would not try to just use some shaders to emulate the pixelation process. I think one has to feed the rendering pipeline with additional information like palettes, edge properties (draw a line only when it is an outline, allow to draw it inline, allow to antialias) etc.

[drChengele]

where? the only girl i see @5:18 is "an ugly example of polygonal rendering".

Oh, I was under the impression she was there to showcase this technology. Well, my bad then. In that case it doesn't seem they have the animation down very well. Either way, I think this would work best if, like you said, they hybridized it. There are some tried and useful techniques, such as geometry instancing, animation sets and stuff, which would just make no sense reinventing from scratch.

this is why i would not try to just use some shaders to emulate the pixelation process. I think one has to feed the rendering pipeline with additional information like palettes, edge properties (draw a line only when it is an outline, allow to draw it inline, allow to antialias) etc.

In a sense most of that CAN be crammed into shaders. For example with cell shading you forward the "gradient" you want to use as a 256x1 texture, essentially a palette.
But yeah, not all of it. And one would first need to quantify and define mathematically all these properties that pixel artists use and frankly I am not even sure if it's possible.

MAYBE it could work with a fixed-angle orthogonal projection camera and carefully textured materials, but in that case, actual PIXEL ART would do just as well. Variable camera angles and full 3d... I don't know, sounds like science fiction to me. Unfortunately.

[Chromanoid]

In a sense most of that CAN be crammed into shaders. For example with cell shading you forward the "gradient" you want to use as a 256x1 texture, essentially a palette.
But yeah, not all of it. And one would first need to quantify and define mathematically all these properties that pixel artists use and frankly I am not even sure if it's possible.

such "palette textures" would be something you would have to provide additional to the model and its textures. this is what i meant with "additional information". you would have to do some extra work to make the 3d models look good... and sharp pixel perfect line rendering is in my opinion nearly impossible with methods used in cel shading graphics...

[Hideous]

I read something somewhere that this infinite detail thing is really only fast when it's static, which is why there's no animation in that thing.

It should, apparently, be used together with regular polygonal stuff.

[Dacke]

I read something somewhere that this infinite detail thing is really only fast when it's static, which is why there's no animation in that thing.

It should, apparently, be used together with regular polygonal stuff.

Makes sense. It works as look-up search, like a search engine. You have lots and lots of dots. Each time you want to render something, you go to your dot-data-bank to find the dot that matches each pixel (from where you are looking). For it to work efficiently the dots have to be compressed and stored in clever ways. Decompression and recompression of compactly stored data is not something you want to do every frame.

[Alex May]

I read something somewhere that this infinite detail thing is really only fast when it's static, which is why there's no animation in that thing.

It should, apparently, be used together with regular polygonal stuff.

As Dacke says the likelihood is that it is based on some non-real-time process of cooking the data in some search-friendly way.

[Dacke]

The youtube-video side-text is my source. It explains it fairly well.

Unlimited Detail is a new technology for making realtime 3D graphics. Unlimited Detail is different from existing 3D graphics systems because it can process unlimited point cloud data in real time, giving the highest level of geometry ever seen.

Unlimited Detail is a fourth system, which is more like a search algorithm than a 3D engine. It is best explained like this: if you had a Word document and you went to the search tool and typed in a word like money the search tool quickly searches for every place that word appeared in the document. Google and Yahoo are also search engines that go looking for things very quickly. Unlimited Detail is basically a point cloud search algorithm. You can build enormous worlds with huge numbers of points, then compress them down to be very small. The Unlimited Detail engine works out which direction the camera is facing and then searches the data to find only the points it needs to put on the screen it doesn't touch any unneeded points, all it wants is 1024*768 (if that is our resolution) points, one for each pixel of the screen. It has a few tricky things to work out, like: what objects are closest to the camera, what objects cover each other, how big should an object be as it gets further back. But all of this is done by a new sort of method that is called "mass connected processing". Mass connected processing is a way of processing masses of data at the same time and then applying the small changes to each part at the end.

The result is a perfect pure bug free 3D engine that gives Unlimited Geometry running super fast, and it's all done in software.

A bit braggy but explains how it works. But "pure bug free" my ass.

[Alex May]

Even the bloody explanation is patronising in tone!

Search for money indeed

[drChengele]

They should really find a way to make it dynamic, even if it means a separate, more inefficient pipeline for the dynamic objects.

Static geometry is useful only to a degree. I was primarily looking forward to this because I do a lot of abstract and particle-based stuff; without moving parts the whole concept kind of falls apart.

[Zaknafein]

About Infinite Detail... it's 2 years old, it hasn't changed since then, and it's very sketchy : http://forum.beyond3d.com/showthread.php?t=47405

[Ivan]

Yeah it's interesting, too bad it looks like shit. There is no fast way of doing the kind of interpolation and antialiasing that would make this look comparable to GPU graphics on the CPU. Their technology doesn't seem all that magical to me, anyway. It's basically a really fast raytracer with a highly optimized voxel space. It's been done before, albeit probably slower. Their whole "You only need to check as many points as there are pixels on the screen" thing is basically what any raytracer is. Their only achievement here is the partitioning of a highly refined voxel space, something that there are plenty of papers on out there for those crazy enough to get into it.

[Ishi]

I was gonna moan about infinite detail but it seems everyone else already did it for me. I have one point to add, which is that they say "this runs on the CPU!" like that's a good thing, but I can't imagine it would run as well on a GPU, so the CPU is gonna be clogged up with graphics. Depends what the data's like though.

Back to the 3D pixelart. I think motion is a big part of the effect too. For example if a windmill is rotating, it should have limited steps - instead of rotating 0.1 degrees per frame, it could rotate 2 degrees every 20 frames. Gives it a more hand-drawn look. (Like in Ghost in the Shell SAC, the Tachikomas are obviously CG because they move way more smoothly than the hand-drawn characters).

[Skofo]

I wonder why everyone's bullshit detectors don't immediately go off as soon as they hear someone offering "unlimited" anything while undermining the power it takes to run Google search and branding old technologies such as voxels, geometry instancing and rendering geometry as pixels on the screen as revolutionary new breakthroughs.

[Mikademus]

My bullshit-o-meter did peak, but still the technology seems very interesting and promising. This video was somewhat more interesting still, I thought





Also, the arrogant little prick gets to rail down not only on their competitors' political deficiencies and the sad state for modellers to have to work within the incredibly archaic confines of polygons, but also enlighten us about his wondrous mastery of English, which I think is a jolly good selling point for the system!