|
spottedfeces posted:New scene with no textures save the sky and the trees, and only a half-assed lighting set-up, but this is essentially what it's going to look like. Comments and crits would be well appreciated. I would remove the lamps altogether. They don't add anything to the scene, they're not even casting light in that example (Although maybe they cast interesting lights when finished? Who knows), and their architectural style is completely dissociated with the rest of the scene. If you have them in there to break up empty spots in your composition, I'd replace them with some shrubs or something. Try lots of small short versions of your trees.
|
#
¿
Jul 16, 2008 08:45
|
|
|
|
| # ¿ May 7, 2024 05:37 |
|
|
Raytracing is the new (old) overhyped buzzword nowadays. Being able to raytrace games will correct some problems, and introduce others. It's not going to revolutionize anything apart from letting you do a few neat tricks. Just look at the movie effects industry to see this - films are still done as a hybrid of rasterizing and raytracing. There are many cases where it's a hell of a lot easier to leverage your computing power towards brute forcing triangles than raytracing stuff. The best CG films are still done with the majority of the work as "throw a huge shitload of polygons at Renderman to rasterize" (or in the case of Avatar, throw a HUGE HUGE HUGE shitload of polies to be rasterized in the form of jungle foliage). Many raytracing techniques don't scale very well due to complexity issues. Goreld fucked around with this message at 23:00 on Sep 16, 2009 |
|
#
¿
Sep 16, 2009 22:55
|
|
|
Fearian posted:Ah, sorry I guess I was being unclear. I was baking it down in Max for a game prop, something I know how to do, when I realised I know nothing about procedural materials or rendering in Max/Maya. Which is what I was interested in learning about. Vague I know, that's because I don't have a good starting point. v For proper rendering of metallic surfaces you'll want to try a Cook-Torrance shader. Or at the very least, something using Beckman distribution for the highlight (Cook-Torrance shaders use Beckman distribution for the highlights; I think it's the 'g' term in the shader, but don't quote me on that) I don't remember if Mental Ray has an explicit Cook-Torrance shader though, or if they called it a BRDF shader or something else. A Cook-Torrance shader is generally how a metal shader is implemented (although a lot of people just use 3 samples for red, green, and blue instead of integrating the whole spectrum for proper brdf, but this won't matter unless you're working on super-photorealism). So just look for something using the words Cook-Torrance, or BRDF on it. The characteristics of a metal shader are that they have very bright highlights, and a Fresnel effect across the surface. Sorry if that's a lot of words, but I'm just throwing out a bunch of terms so you'll know what to look for when playing with shaders. And you *will* have to play with them - no shader program is perfect, and often times stuff may be mislabeled or use slightly incorrect formulas. (I've seen shaders that have added when they were supposed to multiply and result in odd ring artifacts - I think the Max Blinn shader does this)
|
|
#
¿
Nov 30, 2009 23:13
|
|
v
|
Odddzy posted:Sorry, I wasn't clear enough when I said the road, what I meant as that the scene with the wireframe in the reel was entirely done by me not just the road, I agree it's really nothing important all of itself if it was just that. I'll give some comments on the reel. Hopefully constructive. Many modelers you'll go up against are obsessive compulsively insane perfectionists. They're typically the people who post the super detailed stuff on CGtalk. You'd need a lot, a hell of a lot more modeling detail in those scenes to compete against them if you come at them from the angle that your demo reel seems to show (ie. turntables of full scenes). What I would do is focus more on proper modeling techniques. For example, have a model made purely of 4|4 polygons - quadrilaterals and 4 valence vertices. A model made of nothing but quads and 4-valence vertices will smooth perfectly for certain mathematical reasons I won't go into here, so if you can show you can model some weird object (sea slug, something with a lot of holes in it, etc.) but keep this property perfectly or near perfectly on a model you'll impress someone, even if you don't have some 8 million poly model. If you're going for the 'impress with detail' route you'd better load up Zbrush and chug away, though. The other angle you could come from is creativity. It seems like you're kind of trying to express that in the demo reel, but I would go full bore with it. Make stuff a lot more cartoonish, bend things around, exaggerate stuff. Actually, to be honest creativity can't ever really hurt. You either want your reel to be "wow, that looks realistic" (if you don't have an obsessive personality don't go this way) or "wow, how in the HELL did he do that?" Try to do something that you haven't ever seen someone else do before. Example - if you REALLY want to show off texturing, and feel free to use this idea, then texture one of these insane birds. I'm being perfectly serious - birds of paradise not only have some really really weird shapes, there's some extremely subtle stuff going on with their texturing and coloring, such as the green sheen on the bird at 1:40 or the infamously strange bird at 2:43. You can't just color them with the texture either. You have to look at the way they're reflecting light, try to figure out how exactly the 2:43 bird is doing that crazy optical illusion with the intensely saturated blue and incredibly strong black color. If you just throw a Phong shader with ambient occlusion you'll never properly capture the bird's appearance. If you could model and/or texture something that plausibly looks like *any* of those birds, I guarantee you your demo reel would get instant attention. edit: Also, one more thing - take some traditional art classes. Getting a strong foundation in figure drawing, shading, etc. through classes will far outweigh any little software tricks you learn. All of the great artists at effects companies have strong backgrounds or at least strong familiarity in traditional art. Figure drawing, ie. drawing the naked human body from a purely objective standpoint, will teach you all sorts of weird poo poo that you didn't even realize about shape and form. You can't really replicate the experience any other way, which is why figure drawing is always included in art curricula. Goreld fucked around with this message at 10:11 on Jan 19, 2010 |
|
#
¿
Jan 19, 2010 10:05
|
|
|
FLX posted:This is a cool idea. I tried it just now, but it's difficult to get a perfectly round shape, and the top/bottom surface of the lens is a little wobbly. I did play around with it though until I reached the following shape: In layman's terms - if the mesh is not 4/4 (4 valence vertices, 4-sided faces), then portions of the mesh will not be C2 continuous. C2 continuous means you have smooth reflections. In other words, this is what creates hosed up specular highlights all over a model if you try to smooth a triangulated model with Catmull-Clark (for triangular models you use Loop subdivision, which has its own set of problems) The explanation behind it is a bunch of topological math involving subdivision kernels and other wacky poo poo. But more or less, if you make a model that's made of nothing but quads and 4-valence vertices, it's guaranteed to look very, very good when smoothed. If you can't get it down to that, it's best to hide the extraordinary ie. non-4-valence vertices in a part of the model you won't see very often.
|
|
#
¿
Apr 13, 2010 00:24
|
|
?
|
Mazdrol posted:
My advice - find reference footage. Use it to figure out timing. Professionals do this all the time, but it's especially important for students. In that footage, the guy's neck is way too limp. The human spine is very flexible and springy, and a person's head bounces back much more readily. There's not really any reason not to use reference footage nowadays, because youtube makes finding reference of just about anything a cinch. Example: https://www.youtube.com/watch?v=04W5CZgtItE In this video (which conveniently even has a slow motion part), notice exactly how her head moves. The head is hit by the ball, and moves slightly but immediately transfers force to the shoulders, so you end up with this compound motion: Head moves shoulders begin moving and the head is already rebounding backwards The head slightly flexes back forward in a spring pattern again, and her shoulder muscles begin clenching as she begins reacting to the ball. In other words, the head is actually reverberating back and forth due to the springiness of the spine. Take your animation and put it side-by-side with that footage and it'll help out a lot. The best teacher is observation. Oh and also, your guy's head moves so much that it looks like his spine collapsed. If you're trying to go for cartoony exaggerated motion, you still need to respect the springiness of the spine. In other words, if he's hit so hard that his head flies down like that, his shoulders would be following and he'd be doing a full cartwheel in the air. Goreld fucked around with this message at 06:45 on Nov 9, 2010 |
|
#
¿
Nov 9, 2010 06:41
|
|
|
Mazdrol posted:
Looks good, although I'd use a FFD lattice to squeeze the body down a little bit. Pineapples aren't ovoids, they're rounded cylinders (which is why it's easy to make pineapple rings!) Something more like this (I did this for a class assignment a long time ago - displacements + specular + diffuse maps from photographs): 
|
|
#
¿
Feb 21, 2011 20:57
|
|
|
Mazdrol posted:An undead clown trying to say the word "happy" without any lips: Assuming he's trying to say "happy" without lips, you need to hold the last phoneme? or whatever a lot longer. At least for me, saying the "peeee" sound at the end causes cheek contractions which is designed to pull back the lips, but if I say it with my lips open (acting like they're missing) the cheek flexes a lot more and longer. I say hold the cheek contraction for the entire length of time he's making noise at the end. Oxygencult posted:crossposting dis,the hand position looks a bit funky, i'll fix it tomorrow. The hand is shifted too far forward from the arm. It makes it look like his wrist is broken. It's probably a side effect from the bone binding - the pivot in your wrist is probably slightly off. Wrists are difficult to get right because they're sliding joints, so you get weird poo poo like that if you just have a single joint. The same goes for head motion - proper head motion mimics the first few vertebrae, ie. instead of having a spherical joint you have a twisting joint connected to a turning joint closeby. Goreld fucked around with this message at 08:04 on Mar 10, 2011 |
|
#
¿
Mar 10, 2011 08:01
|
|
|
BonoMan posted:Most excellent. Although either the bump is SLIGHTTTTTLY too much or the pore sizes are SLIGHTTTTLY too big, but something is kinda standing out about that. Other than that that's amazing. That was my initial impression, but I think what's happening there is just that there's no hair. That is, in a real person you'll have some hair, even if it's extremely tiny wispy hair or stubble (both of which would be on a character like that). Without the tiny fuzz, it makes the skin feel a bit plastic and less 'living'. Especially for males - even if you have a perfect razor, you're going to have *some* stubble. I'm not sure if that applies to chemo patients, but otherwise both men and women have hair all over their skin, even on their face. This is why 'wolfman' genetic mutations have such a crazy effect - the mutation just causes these thin, wispy hairs to grow like normal hair. I think the only parts of your face that might not have hair would be immediately around the eyelids (apart from the eyelashes), the lips/inside of mouth, and well, that's about it. If you look closely, people have hair everywhere, and this subtly affects light interaction with skin. Also importantly, it affects the silhouette of your face, giving you a slightly glowing halo rather than a hard edge. I'd try rendering out a light pass and doing a (VERY) slight blur on it to see if that diminishes the shiny/plastic feel somewhat. That should also add a slight halo for free, instead of having to deal with some crazyass BRDF for the skin.
|
|
#
¿
Apr 13, 2011 23:52
|
|
|
Surprise T Rex posted:I'm just finishing a game development course and after two years I still feel like I'm UV mapping 'wrong' and that I'm basically somehow managing to fluke my way through it. If you want a good way to practice, try a combination of a) things that are hard to UV map, especially high genus surfaces. If you're insane you can try something really high genus, like handle shapes made in Topmod. There's mathematical reasons why high genus shapes are difficult to UV map, but mapping them can teach you good ways to hide seams. b) organic asymmetric shapes like ears. Sometimes they can be easy when using a relaxation/angle-based-flattening, but it's a good way to experience distortion on maps when you can't just lay out nice rectangularly oriented regions and c) using textures that don't play well with distortion, such as chain link fences or any repeating texture. Or with a low rez texture where any distortion creates aliasing artifacts.
|
|
#
¿
Jun 21, 2011 23:46
|
|
|
EoinCannon posted:Something I'm working on at the moment in my spare time Try putting cuticles on the horns.
|
|
#
¿
Aug 29, 2011 05:49
|
|
|
|
| # ¿ May 7, 2024 05:37 |
|
|
ExtraNoise posted:Any game/realtime artists care to give a newbie some advice on terrain? That's an extremely complex question which is dependent on many factors. What's the possible camera angles? What's on the terrain? Sparse houses? A city? A single house? What will your illumination be? Ambient gray sky light? Direct sunlight? Do you want to use shadows? Will there be water on the terrain? What is the vegetation? Grassland? Forest? Dense or sparse forest? Will you walk between trees? If so, will you handle shadowing from the canopy? How do you plan on generating the terrain? Heightmap? Procedurally generated shapes and/or heightmap? Isosurface? Hit detection? Path finding? Deformable terrain? etc. I guess what I'm trying to say is, the raw poly budget for terrain isn't important as a standalone metric - you have to consider a whole shitload of factors all combined. As for advice, it would be best to be able to split apart the terrain into sections, preferably of a size you can change. That way you can do a lot of wedgetests (ie. try sizes in a range from x-y) to determine which level of complexity works best within your limits.
|
|
#
¿
Apr 4, 2013 03:23
|
|