Notgames Forum

I love realtime 3D. Merely being able to navigate a virtual environment is magical to me. And in realtime 3D, the feeling of being present in that space is very visceral.

But creating 3D spaces is a pain.
The biggest problem is that everything needs to be made. You can't just artistically suggest the presence of a feature in some way, you actually have to build it. Even if you're going for a vague, impressionistic style, everything needs to be built first, modeled and textured. So you end up engineering things, instead of sketching and painting.

3D is so damn clean.
I love how in 2D, you can easily dirty up things. Even a certain sloppiness can be charming, and improve the aesthetics. In 3D, everything is clean. And sloppiness only leads to models that are rendered completely wrong, texture seams, bad performance, things going haywire. If you want something to be dirty, you need to build the dirt, which takes all the spontaneity out of it.

3D takes a lot of time.
Everything you do in 3D is a lot of work. You can buy prefab objects but they are never quite the right style and they often don't go together, so you have to tweak so much that you might as well have made it yourself. By the time you're ready to judge your idea, and get a feel for the space you've created, you've had too much time to think things over.

This is not mean to just be a just a complaint. As I said, I love 3D and want to use it in my work. But I'd like to have ways to make things quicker, easier and dirtier. Any tips are welcome!

Voxels or similar? What makes much of the 3d requirement so strict is because the polygonal base with requires very strict math.

For examples, have 3D built on basic atomic shapes makes it a lot less prone to the error. An early example is Ballz: http://www.mobygames.com/game/snes/ballz-3d-fighting-at-its-ballziest
That way you can go dirtier and less exact without everything breaking up.

Another related issue is texturing. If you skip textures and uv-mapping you can be a lot less strict with rules of model building.

When you go with voxels, objects become more like sculptures as they have an inside, and is not the fragile paper-thin representation that polygon meshes are.

All of the above have other problems though of course, but as always there is no perfect way.

Look at Radio Silence (http://distractionware.com/blog/2010/04/radio-silence/). This game has simple 3d models, but it looks really good.
It is possible to create simple meshes and make them look awesome without a textures by using some methods such, in example, post effects - blur, depth of field, noise, glow - it will make game looks interesting.
Also look at Hide (http://www.superfriendshipclub.com/forum/viewtopic.php?f=12&t=117), this game has pixelated 3d view, and it looks awesome

This was striking me as I was working on teeth, recently, and felt claustrophobic literally having to place the camera inside the character's head to even see what I was doing...

3D has its advantages, however. In terms of modelling I never feel there is much you can trust the 3d software with (I think of the 'smooth' or 'relax' tools as the 'drain characteristics' tools), but in terms of lighting there is so much you can leave it to do. There is a certain 'realness' which properly calculated (linear-space, HDR, &c) lighting can do which I very rarely see done by hand. Not photo-realness, but 'spaciousness', like Vermeer.

You have to engineer it, but then when it is there it can work automatically, often. For instance, my facial system is built to simulate a light tension in the flesh and certain features of the face moving slower than others as they are 'pulled' by lip movement. Once this is there, it can do anything. Although arguably this might as well work in 2d. Cloth physics is another that, once it is set-up, it does its thing. (Although I still have to make this, and may change my opinion on it accordingly.)

Something which endlessly frustrates me is things like having no 'touchable' cameras. I have to make changes to wobbly animation code and such, which is very annoying. I do sometimes get a 'locked out' syndrome feeling about my game world.

But in the endless bag of metaphors, I see myself closer to a set-designer or sculptor than draughtsman, and that helps me cope.

The challenge is that we have been investing heavily in the polygon pipeline for over two decades. Hardware is optimized for it, engines build around it, tools made on top of it, formats specified for it .. and artists made careers out of it.

Hunter ( 1991 )
(http://www.dazeland.com/images/Hunter-2.png)

Skyrim ( 2011 )
(http://cdn.gamerant.com/wp-content/uploads/Skyrim-TF2-Mode-Hoarders-Theme-Performed.jpg)

So, it's not simply a matter of questioning whether polygons have surpassed their expiration date. It takes significant effort to veer from the standard model.

Though to be fair, for the coming era of engines it comes down to shaders, more than polygons. Ever since Doom III shaders have suggested more geometry than there was, and with geometry shaders becoming part of 'the standard', more triangles exist inside the shader pipeline than 3d software could handle. It sort-of comes down to textures again, at that point. I am not sure whether that will be good for sketching or just make the whole process even more meticulous.

I am aware of stylization and I think such games often look pretty. But I also like to appeal to the sense of touch of the player by simulating textures. Above all, I want to create a sense of space, not any space, but a very specific space with a very specific environment. I do want to simulate, not merely represent.

I love playing with shaders. That's definitely a plus. But you still need objects and textures to shade with. Normal maps do not reduce the amount of modeling since you need to model the complex model before you can make the bump map. And a focus on 3D textures does not allow for the same tricks and shortcuts as working in 3D.

Hear hear. I wish there was some easy way of adding dirt, dust, scratches and imperfections procedurally. Particles perhaps?

Maybe a shader that messes up the normal maps of all objects to make them look uneven and a bit crooked and that adds variation between multiple instances of the same model/material. And perhaps add dirt like we would add a lightmap (with a smart algorithm that knows that corners and bottoms are dirtier).

Maybe we can use prefab 3D models (bought on TurboSquid and the like) and then use post-processing effects to mess things up and unify the look.

Side not: all this is why pixel art is so popular :)

http://www.spot3d.com/vray/help/150SP1/examples_vraydirt.htm (http://www.spot3d.com/vray/help/150SP1/examples_vraydirt.htm) I think something similar could be done in a realtime environment as well.

I think procedural graphics could be used more. The problem is, the mentality about how they are used. They are not seen as an artistic tool, rather they are viewed as a shortcut or something that is either used for everything or nothing.

Interesting you should mention this, I had a phase where my character's model was too smooth and I just applied noise to her face and tried to restore her. The model has never been symmetrical since, too.

The dirt lightmap idea can be baked easily in 3d software, too, with the VRay example (or Mental Ray). I was actually planning to use it for my marble areas and have the dirt properly change the surface appearance.

Michael, I would like to return to your initial comment, that started this thread.
 Maybe the difficulties you feel, working with current 3D-world modelling, has to do with what Lev Manovich says, in ”The language of new media”, I am currently studying:

Virtual spaces are most often not true spaces but collections of separate objects.

What is missing from computer space is space in the sense of medium – an environment in which objects are embedded, and the effect of these objects on each other.

The concept of  ”space-medium” is something mainstream computer graphics still has to discover.

Although 3-D computer generated virtual worlds are usually rendered in linear perspective, they are really collections of separate objects, unrelated to each other.

The ontology of virtual space as defined by software itself is fundamentally aggregate, a set of objects without a unifying point of view.

This is interesting. It seems to me that what he means is that an ensemble of objects should define its own space by their presence and mutual interactions.  (As in physics: Newton believed in an absolute space, whereas in general relativity the mere presence of material objects influences the metrics of space, e.g. what is meant by travelling in a straight line.) 

I have been into this in connection with ”colour space”.  Given a number of colour samples, you are asked to pairwise visually estimate their degree of simularity.  (For instance an orange sample has an evident similarity with red and yellow, but much less with green, and none at all with blue. )  Now, if the degree of likeness is represented as a distance, you get a kind of space – a colour space – defined by the given samples.  By statistical methods you can investigate the dimensionality of this space. It turns out that 2 dimensions is not appropriate, but 3 dimensions fits quite well, 4 dimensions would not make a better fit.  So we conclude: colour space is 3-dimensional (for a normal observer).
So the 3D space of colour is not postulated a priori, but follows from the interrelations of the ”inhabitants” of the space.

Could we accomplish something like that with computer-game spaces??
Pehr

This thing about color space being objectively 3 dimensional is interesting.  Can you post a link or something that describes this more in detail?  I don't want to derail the existing thread with that discussion, but I'd like to hear more about it.

Maybe what we need is a way to record 3D reality. So we don't have to start with this empty "space-less" world.

Or maybe I shouldn't deny the cleanliness of 3D.
After all, on a super-clean surface, the slightest speck of dust will disturb. It's easy to dirty/warm up a clean/cold place.

I think this is a better approach. You shouldn't fight the medium; you'll always be unhappy.

Agreed. And to those longing for more power .. there's light at the end of the tunnel.

http://www.newelectronics.co.uk/electronics-news/graphene-electronics-breakthrough-turns-previous-research-on-head/40034/

In fact, Manovich mentions such a project, "Aspen Movie Map" (1978) and remarks:
The idea of constructing a large-scale virtual space from photographs or a video of a real space was never systematically attempted again, despite the fact that it opens up unique aesthetic possibilities not available with 3-D computer graphics.

Sounds a bit like Google Street View. Which I actually find interesting. I just wish we could all record our own worlds instead of browsing the one Google presents. Are there any cameras available that can record precise 3D position and rotation together with the image, so you can reconstruct the scene in a virtual environment later? (This sounds like Patrick Juchli's Stilla.)

My happiness is irrelevant. In my experience, struggle often brings good results.

http://www-video.eecs.berkeley.edu/research/indoor/

A found this
http://www.easypano.com
Better than I had expected. Yet, well .. hm ... don't know. Pehaps worthwhile to look a bit closer into.
Pehr

And now I found Autodesks 123D Catch
http://www.123dapp.com/catch

which looks really promising ..
Pehr

This whole topic reminds me too much of QTVR. It always seemed really clunky and slow. Photosynth is something along the same lines though I think is an interesting improvement of the same concept but it's still clunky and slow. I like the way Trauma had a sort of Photosynth-like look. Idk, I guess I just never like the static image pseudo-virtual-reality thing. To me it will always feel like a weak, stopgap for true full-motion, explorable 3d spaces.

I agree. I'm only curious about this as a basis to build upon or as a detail within a real 3D simulation. I'm looking for texture, to compensate for the sterility of 3D -without needing to hire the team that built Uncharted 3.

Probably 123D Catch is more useful for modelling objects than for environments?

By the way, I downloaded and looked into  ”The Hunter” (a hunting simulator). It has such an astonishingly dense, wild and variable forest to walk in. I cannot imagine that all that could have been ”hand-crafted”. On the other hand it doesn’t look as if made to pattern. It is very photo-realistic. It must have been based on photos, or films, in some way. Not to say that photo-realism is what we are looking for, but I just wonder how they did it.

The technology I linked to generates texture-mapped 3D environments from the point-cloud data that is collected by the device.


It's part procedural-generation, part smart editor / tools that let you "paint" the environment rapidly.

None the less you are looking at a 2D-projection of your perfect virtual 3D-world!
And it is a camera-imitating projection.  This faitfulness to the cinematic tradition from the last century is no necessity, when you are using a computer. You do not have to think about the camera as a physical object among the other objects on the stage.

See for instance the computer animated films by the ingenious artist Tamás Waliczky, www.waliczky.com  especially The Garden(1992), The Forest(1993), The Way(1994) and Landscape(1997) – each one of which utilizes a particular perspectival system. 

-Pehr

I don't see what that has to do with my statement. I wasn't talking about cameras or cinema or imitating cinema.


Sorry, this stuff doesn't strike me as particularly ingenious. The forest one might be nice without the flight-sim platform.

Sorry, my thought took a leap there.  Let me try to express it in other words, and also elaborate it a bit more.
It seemed to me, from what you said,  that you are against the use of static 2D images.  Then the following reflection suggested itself to me.

You (and anyone using standard software for the purpose) put a lot of effort into constructing a digital representation of a 3D world in order to be able to produce a series of 2D images to be displayed in rapid sequence, giving the player the illusion of a 3D space with moving objects.
So in the end the experience of a virtual three-dimensional world is based on static framed two-dimensional images, after all.  That is what I call ”the cinematic paradigm”.

Now, to think one step further, it seems that two virtual 3D worlds are involved here:
-   the one given by a digitally encoded description in the computer
-   the one constructed by your perceptual system, from the impression of a flow of 2D images.

They are of course not independent but the connection is far from strictly one-to-one.
The perceptual one, present in your mind, is the one you directly experience. 
The encoded one, present in the computer memory,  the one you hope to find pleasure in exploring.

From an aesthetic point of view I would say the perceptually created one is the most important. It is dynamic, holistically integrated, adaptive, sensible for influences from mood, attitude, preconceptions, memories etc.

Pehr

I agree that the aesthetics of the 2D plane of the computer screen are important. But my best experiences in videogames have been those where I felt I was in a place, I was there, in this three-dimensional environment. When that happens, the videogame ceases to appear as cinematic, but becomes architectural perhaps. Videogames are only cinematic in a very technical sense. Even the camera in 3D is fake. Computer screens feel much more like windows to me, while cinema screens feel like moving pictures.

Meanwhile, I found this:

http://player.vimeo.com/video/36414371

The space is scanned with lasers by ScanLAB (http://www.scanlabprojects.co.uk)

It's beautiful!
I love the artifacts and the transparency. It definitely gets away from the cleanliness of 3D space.

Well, essentially, I am. I'm with Michael on this. The illusion is important.

"Ceci n'est pas une pipe" So what. I don't care that the image is paint on a canvas, I care what the image represents.

Radiohead used raw point-cloud data for their House of Cards videoclip a while back. Part of the data can be downloaded here ..

http://code.google.com/creative/radiohead/

It's certainly possible to do a game using point-cloud visuals. You just need to generate a polygon / octree / primitives data-set to solve collisions with .. or go the other way around, generate point-clouds from a polygon source :)

Wow, I really like the electronic, Matrix-y grittiness of this virtual 3D space. The "vibrating" artifact that sometimes happens is pretty disturbing to me on a visceral level, but if that were a lot more infrequent I would really enjoy trying this approach.

Are there any point-cloud rendering engines or content-creation tools available that could be used for games?

OpenGL & Direct3D have build-in support for point-clouds. Concerning content-creation, you usually want to use a scanner .. but you could generate the data from polygons / NURBS in most 3D tools as well ( by ray-casting your scene ).

I'm getting a lot out of using largely automated processes in modelling, it's slightly (And in some ways, significantly) less optimised, but doubling the modelling speed and freeing me up to make large changes at any point is very worth it. Plus, I never even have to think about UVs. Here's a breakdown if it's of any use:

(http://payload80.cargocollective.com/1/8/272905/3911984/CINEMA-4D-64-Bit-2014-08-20-19-19-50-81.png)

- First the model is created parametrically with splines in Cinema 4D using largely standard processes. I personally prefer spline-based modelling as it allows extensive tweaking without worrying about having to redo large sections of the model.

(http://payload80.cargocollective.com/1/8/272905/3911984/GateOuter_1-Gloss.png)

- The model then has it's UVs and relevant texture maps baked out automatically inside Cinema 4D (Auto generated UVs are far, far less optimised then hand made UVs, however since I'm doing this project alone I can't factor in the time needed to make these) and is sent to Quixel Studio's excellent DDO, a procedural texture generator.

(http://payload80.cargocollective.com/1/8/272905/3911984/UE4Editor-2014-06-29-11-43-49-45.png)

-The maps generated by DDO are then imported into Unreal Engine 4 and used as masks to blend between multiple world space aligned materials. More info on this process, http://www.polycount.com/forum/showthread.php?p=2090428#post2090428 (http://www.polycount.com/forum/showthread.php?p=2090428#post2090428).

(http://payload80.cargocollective.com/1/8/272905/3911984/UE4Editor-2014-08-17-19-00-55-99_52.png)

-The model is then placed in the world, and assembled along with other meshes built with the same process, into the structure seen here.

Thanks so much for sharing, this is really helpful

This thread strikes me as selfish. The obvious solution is to make a big commons of artifacts in different styles all meticulously cultivated. Call it a cyberspace if you want, but not a facelift for the WWW, just a catalog of all digital art that might have practical future applications.

No TurboSquid dumping grounds. A concerted and open effort to make tools and cohesive elements available to use. If you need a character, making a Frakenstein monster is always suboptimal. What you want to do is cast all of the characters together, in order to optimize onscreen chemistry. Scout for locations in the cyberspace before resorting to desperate measures. Reduce, reuse, recycle.