You're doing it wrong.
mod
So, i thought i would try something here, and if people think this is a terrible idea, just ignore it and do not post(or post to tell me how i'm an idiot, in the true spirit of the thread).
So its a pretty simple concept, people like to argue on the internet, right? Right, so how about we try and focus some of that energy into a positive use. Now, where this gets a little fun is here:
I would like to encourage people to post in-depth descriptions of their work-flow, this could be any number of things. From how you bake a normal map, to how you start concepting a character to how you only use 16 color animated gifs for your normals. Various people will then tell you why what you're doing is terribly wrong, how they would do it, or tell you to never even think about doing it again.
Ideally i see this thread as a place where people will not get offended when they are called names, told to fuck off, etc. Because it is inevitable. But can actually learn from the collective differences in opinion from our wide range of users. And even if you do get offended, because you inevitably will, that there is enough of a purpose here to keep you posting, keeping the dialog open, and truely offer today's children of tomorrow, a wonderful world, free of outdated misconceptions, where they will be free to frolic in the wild, with the bearsk.
Now, for this to work at all we just need one person to start, so please by all means, if anyone finds this at all worthwhile, tell me how you make a model, or texture a goat, or how you painted a totally sweet portrait of your favorite character from saved by the bell.
So its a pretty simple concept, people like to argue on the internet, right? Right, so how about we try and focus some of that energy into a positive use. Now, where this gets a little fun is here:
I would like to encourage people to post in-depth descriptions of their work-flow, this could be any number of things. From how you bake a normal map, to how you start concepting a character to how you only use 16 color animated gifs for your normals. Various people will then tell you why what you're doing is terribly wrong, how they would do it, or tell you to never even think about doing it again.
Ideally i see this thread as a place where people will not get offended when they are called names, told to fuck off, etc. Because it is inevitable. But can actually learn from the collective differences in opinion from our wide range of users. And even if you do get offended, because you inevitably will, that there is enough of a purpose here to keep you posting, keeping the dialog open, and truely offer today's children of tomorrow, a wonderful world, free of outdated misconceptions, where they will be free to frolic in the wild, with the bearsk.
Now, for this to work at all we just need one person to start, so please by all means, if anyone finds this at all worthwhile, tell me how you make a model, or texture a goat, or how you painted a totally sweet portrait of your favorite character from saved by the bell.
Replies
- Theme
- collect reference images
- draw out a map
- Look over the map and add a focal structure that the player can easily relate to where there are.
- Grey map the level in the editor
- Game play pass (lead or have others play it)
- Add meshes
- Final visual pass
not sure if this is the correct way, but its the one I've learned from a few lead designers.
Seriously, one of the things that i've always seem a little fishy about that workflow is this:
Don't you run into some major problems when you get to the meshing phase, where as you're adding visual detail, you end up changing gameplay specific elements?
What sort of things do you to do combat this, if any?
Sorry i havent dont much LD work so i cant give a very detailed response here.
1) make lowpoly out of various primitives, cutting and merging, keeping it clean, then unwrap the UVs;
2) duplicate the low and either reconfigure the topology to subdivide and retain shape, or subdivide it directly, split model up...
3) sculpt the highres versions, or model out the detail for functional normals;
4) bake the normal map in sections within Silo and/or Mudbox, composited in Photoshop, previewing it along the way;
5) bake an AO from the sculpt or sub-d;
6) gradient maps on the AO to get material colour started, and carry on into texturing the object with spec and shit in a way that I haven't gotten down quite to a routine yet. (Am still anxious to experiment more thoroughly within this stage.)
For mostly organic things, I start out by sculpting volumes roughly but as accurately as I can and do edge modeling of a better sculpt base to that surface, which I can later further compress to effective game-res geometry. I also like blocking out areas and the location of specific details if doing a very detailed diffuse texture using the 3D paint tools of whichever program, before exporting and refining within Photoshop.
Where I feel the most wrong at any given moment is how haphazard and fucked up my method for generating the high resolution detail has been. I mean, everything is so split up and pasted together, it feels like I'm just tearing the mesh up, doing whatever works to get an adequate result. But it really varies from asset to asset how smoothly this goes; for example, baking the normals to a strap element that wraps around the rest of the lowpoly without actually being attached - how the fuck am I supposed to get that to retain shape and bake properly to the lowpoly? Talking about it here makes me want to go try something though, so brb.
gradient map on an AO? how does this look like? how does it help?
IMHO, makes a decent start, if doing your textures in Photoshop. Unless using AO as a base for diffuse is one of those "you're doing it wrong" things, in which case, woohoo, intended use of EQ's thread fulfilled!
Somehow every time I redirect some topology I wonder how other artists go at it. So if you have a better technique, please tell!
This one method works well for me but maybe there is something faster?
1. Paper and pencil, dozens of drawings, views, angles, design forms, versions, etc. Too much time for a solid design, and idea.
2. In Modo, i start from a low low poly model for subdivs, having all loops needed for a good mapping (edge flow), rigging and smoothing (i test deforms previously), and then, i subdivide x1. I avoid the need to correct a topology not wanted.
3. I refine details with the sculpting tools in modo (pinch, smooth, etc.), i make the uv mapping, and then i texturize in modo aswell, previewing the results like in real time (all the maps, spec, glossiness, etc.). With some models (those that need a fine detail pass), i use Zbrush to make some textures using polypainting.
4. Once i have all the texture work done and shaders, i export to max and i pose the model. Render and that's all.
All my workflow is focused in Subdiv modelling and for texturing, i bake a lot but i prefer to paint.
having missed the 70's as an adult, I think I'm going to have
to start incorporating that into my vernacular! :poly142:
Yeah you say you think you're doing this a bit backwards and i would tend to agree. It is possible to get some decent results from doing the lowpoly first, but i really wouldn't recommend it. Really the only reason i've ever seen that made any sense to do this, is some people needed to have their lowpoly finalized, uved, rigged, etc before they start on the highpoly. While i can sort of understand why, i really think it is just a bad idea, and much too limiting when it comes to changing forms, and getting an overall quality result in the end.
Being locked into your lowpoly forms can be a really bad move, because to get a good result with your normals, you need to have a lowpoly that is an approximation of your highres. You want edges flowing to match certain elements, all those sort of things. You may find out that when you get to the highres stage, some of your elements were not thought out as well as they could have been, but what can you do? If you change it now you're totally reworking, or redoing your lowpoly, basically doing your lowpoly model twice. Or stuck to a lame shape that was based off of a lowpoly model, because lets be honest, its just a totally different game, working with forms on a simple, lowpoly mesh than it is working with a highly detailed source mesh.
Now what i would suggest here, is making a blockout/cage mesh to start with, try to get all of your proportions nailed down here, do not worry about optimization or any of that stuff yet, just get all of your forms roughed out. When you have everything rouged in, save a copy, because you can come back this, and use it as a base for your lowpoly. You will still have to do a lot of work to match it up correctly with the high, but it should save some time as aposed to redoing the entire lowpoly after the highpoly is done.
Alternatively, you can start from your finished cage mesh(lowest level sculpt, or cage mesh for sub-d) and optimize down from there. I tend to do this a lot, with sub-d stuff it can often be just a matter off killing off some extra edges and doing a little clean up. In some cases, where i have a really complicated cage, that i'm going to project onto a pretty simple lowpoly, i will just tottally remodel it, because its easier. This is done on pretty much a case by case basis.
In the end, the better your lowpoly matches your highpoly, the better bake you're going to have, and the better overall result you'll get. Remember, the whole point of using normals is to emulate the Highpoly model, not to try and conform the highpoly to the low.
As far as the split up part. I'm not entire sure what you're saying, but one thing i can tell you is this:
Always use as many different meshes as possible in your highres, there is no glory in modeling everything as a solid mesh(and its a little backwards a lot of the time, things in life are made up of may differe components). If it makes it easier, split it up, dont even think twice about it, intersecting objects are perfectly fine, and AA will help to bring them together in the final bake.
For this sort of stuff i tend to just use the edge slice tool in modo, slice out the topology i want, and then go back in and remove the excess edges that are bound to be there.
What i like about this method, is that i could make a bunch of cuts all over a model in one pass, and then in another pass remove all the extra edges. Either way it seems like a similar amount of steps, and you could probabbly do the same as far as doing it in "passes" as well.
Also, fuck you, FRENCHIE
if you work on this on a single case basis, one could go so far and automatically position the new vertex with the intersection of original mesh and mouse cursor when running the script.
BUT, instead of posting exact work flow which can be a personal preference thing, i am going to post some work ethics which i follow daily:
*don't be lazy
*don't "always" look for shortcuts and try to make scripts for everything fucking task
*don't try to over analyze art
*when in doubt look up reference
*visualize end result
*work in chunks
*ask for critiques
*details matter, but they mean shit without good form and structure
*contrast and variety matters
*beauty lies in some imperfections (nothing is 100% perfect, neither should it be)
*work hard
*don't be lazy, seriously.
I'm just calling you a fucker because EQ basically told us to be offensive in this thread, so I'm starting things off right. It's nothing personal, you lazy fucker.
On a more serious note, I like the idea, and next time I find myself doing something that strikes me as being needlessly complicated or slow, I'll post that in here.
am i doing it right?
well my work flow is simple and nothing out of the ordinary i think, but it tends to be something like this in short (for character art):
1. take in art direction and look at any concept/reference if available
2. plan out what needs to be normal mapped, high poly modeled, detailed etc(not everything needs to be high poly sculpted or modeled)
3. create clean base meshes, recycle if possible but dont be lazy to create something from scratch.
4. before sculpting, check surface normals of all base meshes.
5. after high poly is done i export a semi-high poly as a ref for creating the low poly
6. UV layout low poly and bake normals and AO
7. create textures and materials
8. fix texture seams
9. skinning/rigging, QA etc.
i usually do this:
i just cut one of the tri sides in the middle and curve snap that vertex to center to maintain surface normal.
I'm gonna go ahead and say you're doing it wrong - sure there are things that it doesn't make any sense to try and make a script for, but I have seen a huge amount of times when artists have been using painfully slow methods to do repetitive actions and just keep on doing it because they are used to it and refuse to consider any alternatives.
So yeah - always look for shortcuts.
well you are WRONG
i would not "always" look for shortcuts
i need to edit the ethic to:
"dont always look for shortcuts"
it is a judgment call.
i would script something that would help in the long term and not just something i might need right now.
things like the topology change above does not require a script.
relying too much on scripts makes an artist too handicapped or lazy in my opinion.
i tend to try it the hardcore way and i seen my speed increase tremendously because of that.
usually i see people wasting time looking for shortcuts, trying to cut corners, and at the end they have to go back anyways to fix some shit they totally bypassed earlier because they just wanted to take the easy route.
most of the times not looking for shortcuts is the shortcut to getting the work done fast.
Why is making a cut, deleting an edge, and eyeballing the center of the face better than a script that more precisely does the same thing in one step?
for the simple case above, the less you deal with scripts the faster/better you will get. at least in my experience i gotten lot better at eyeballing/accuracy and speed.
i use scripts a lot, but not for simple things like that. i actually have fun doing simple things like that and dont want to script everything.
sure the end result is what matters, but i also love the process of getting to that end result.
lol, i really dont care about it mathemetically being is perfect center.
i usually get it looking good eyeballing whether it is hard surface or organic and thats all that matters.
you say my method is less freedom, how so ? it was a simple topology edit, not much to it really.
now, if i was doing something scientific with real word application like prototyping i would do all sorts of shit like constraints and what not and blow peoples mind away.
you are implying organic modeling is less challenging ? only if you consider artistic anatomy to be mindless.
hard surface modeling is more technical, tedious and somewhat mindless repetitive tasks. at a broader scale it involves some aesthetics but not as much as artistic anatomy.
looking at GOW or UT3 its hard surface stuff is more about creating non-functional details and rivets and staples everywhere. sure it involves more accuracy and structure, but not more challenging. it may take more time but i dont consider it more challenging.
All ways commented to solve what pior posted are valid, and i must recognize i would have done the same as pior, but thank to god, i don't need to solve those kind of problems due to my modelling technique hehehe.
When i model, i mainly use bevel tool, only one shortcut, you have extrude+bevel+inset all in one operation. Scripts are time saving, and macros... ufff more of the same but better!. Know your tools and work in the fast and easy way possible.
That's something very subjetive Perna, not all ppl have the same eye to catch human forms/proportions and making good 3d humans. In contrast, for me, mechanic objects with subidvs are very easy, the forms are very simple, is all reduced to technique and someone could do it without problems. Once you learn how subdivs works, it's easy, slow, but easy. If you are going to model a human with poligons, subidvs, and not a Zbrush model from a shitty base mesh.. the thing is quite different, and more if you haven't got the enough experience.
A person, a world, a way. The important thing is the result, not the way.
well i have three edges to curve snap to, which seems to me more than enough freedom to stay in same surface normal while getting to the mid-point.
well i guess you define the word "easy" differently than me.
and, what you consider as "human" might be a lumpy pile of shit with tons of anatomical and aesthetic flaws to me.
modeling a car or a weapon from a blue-print or a concept is pretty straight forward technical modeling, and does not really challenge my thinking ability for the most part. that is why i say its more tedious but more easy. something taking more time does not mean its more difficult.
As an example : I use that technique alot when trying to find the best loops for a deformable face model.
Thread is heating up, NICE!
well then you are talking about knowing how to cut faces, beveling, extruding, ring selecting, loop selecting etc.
these are about knowing the software, which even a tech demo guy from autodesk can know.
when i use the word "modeling", i mean creating something 3d that has the right form and structure and/or details whether it is hard surface or organic.
i guess you define the term "modeler" as someone who knows the modeling tools/software, rather than someone who knows how to make a believable three dimensional object regardless of the technique.
A technical guy can very well know all the tools of an app, but at the same time, not knowing how to use them properly in a modeling scenario.
For instance I can go and ask a technical guy at work where to find the bevel tool of whatever app. He might be able to show it to me, but that does not mean he knows how to place such bevels for a mesh to subdivide properly! (especially since tech guys seem to only test their tools and scripts on cubes and spheres hehe)
Like that redirection .gif thing : There is select edges, connect edges, collapse edges, and move vert, all in less than a second. There is a difference between knowing such tools (from school or from messing around) and knowing how to combine them (from experience).
anyway, I digress! Fuckers!! (well Pea at least)
i'm probably doing that wrong -_-
that is true, but i was taking that as a given for any modelers. anyone who wants to learn modeling should think cognitively and learn to assess the situation.
it is part of problem solving and it is not unique to modeling only, it is done in lighting, rendering, texturing, etc.
anyways, as far as which is easier you just need to take a look at cgtalk and you will find out.
a good car model will show up every week or so, but once in a blue moon do you get to see a good human model.
also, forget about sculpting. creating a good organic human model with muscle shapes and making that look good at all sub-d level requires very good understand of poly modeling and how polygon smoothing works. on top of this add the good topology for good animation deformation and good UV layout. for hard surface you rarely have to think about good deformation.
An artist can be trained all the technical ins and outs of a particular app in a few weeks/months. No place is going to hire a non artist for an artist position no matter how much he/she knows about a given app. There are rare occasions and it doesn't apply to tech artists. They happen by the title artist by their location on the seating chart.
Enough with the philosophical discussion you nerds!
Oh yeah, and fuck you, whoever you are! Fucker.
I'll post my work flow for environment art. Hopefully I'll fuck something up and contribute.
to be fair i would say both humans and cars can be difficult to model depending on where you set your priority and quality level for each. lets just leave it at that.
there is mathematical precision in characters too, its just close to infinite that we tend to stop at a certain level of accuracy for humans and learn to live with it. so it is up to individuals what they consider to be a believable human model.
for example, take a 3d scan of a head as a visual target. now try to create a 3d head by poly modeling and make its anatomy exactly as that. it would be quite difficult.
for mechanical models the mathematical accuracy is much clearer and more finite.
so for a car, if you just give me blue prints of side, front, back, top etc. i could create that exact model easier than a realistic head. obviously there are technical factors to keep in mind like keeping surface planer, proper way to chamfering etc. the over all process may be much more tedious but the visual target is very clear (as opposed to knowing what a certain good human model should look like).
And i repeat that what you consider as difficult, can be the opposite for another person. All sums up to practice, experience and technique.
For me inorganic modelling is a silly task compared with a realistic character. IMHO is more difficult to achive a credible 3d person than a silly mechanical object. Anyone can make a car and the proof is what ppl post on forums, cars and buildings everywhere.
Almost all 3d humans i see are ugly attemps to make an human (uncanny valley, ridiculous attemps of humans). You really need to be a true artist to make credible things. With cars, buildings, etc. (inorganic forms) you don't need to be an artist to do things well, because is a work a child can do previously learned the technique (That's what we call here, a TOOL, a pawn worker, a person who makes what others artists don't want to make). Also, with making inorganic shapes, you don't need to be good at drawing and painting.
All this is subjetive and i think there ins't lack of intrepretation. Each one will consider his work as the best/most difficult and can be more or less objetive.
I see perfectly the point of MM
back to topic, someone have builded a subdiv shapes library? 3d tiles? etc. It makes 3d work very fast an easy.
The task of making a realstics human or "cartoon" character of course is an artistic challenge, hence "harder" than redoing a real-life object with a more predefined outcome. But that wasnt subject to discussion.
it depends on which level you approach it, the technical or the "creative". Of course making a mechanical object still involves creativity when it comes to how to construct it efficiently. But yeah this can go on forever depending on which level you approach it, and how you define such levels...
i was never trying to say that mechanical modeling is very very easy compared to human modeling, in fact it was the other way around.
you think that is exclusive to hard surface only ? no, poly modeling an organic model, or a shirt, or a pant, or a vest or anything that is not mechanic can have the same amount of these technical hurdles. sometimes it can even be more, and sometimes less.
also when you say "we" who are you representing? everybody in polycount ?
you see, you are over simplifying the task of modeling a human. the amount of "just tweaking" you talk about can be a LOT of tweaking. and this tweaking requires good understanding of topology, smoothing etc which are all technical.
all those issues you mention could very well exist when modeling a human model. just the face can have lots of unwanted smoothing, tearing, pinching, inaccurate smoothness/sharpness etc.
but obvioulsy zbrush is the magic answer to everything it seems.
also try to understand that i am not being disrespectful in any way, this is just a civil discussion and no one should be scared to discuss their point out loud.
So for this seems like the most optimal of processes currently imagined, ever.
To model in Zbrush from a box, or bad base mesh is easy, of course.. you only need to subdivide in a crazy way and then sculpt, anyone can do it because you really don't need how to model. The rest is our ability as artists to sculpt: good, very good, bad, or very bad (mucus way).
With CAD tools, with Nurbs, you make inorganic forms uber fast, there isn't technical difficulty, all is know how to use the tools and that's all. There isn't creativity in an organic shape when you copy a form, and that happens in the most cases, of course not all (Someone think that making a glock in 3d has creativity? and a car previouly designed by Audi?).
Anyone knows Catia or similar apps? you can make complex forms in minutes, with subdivs is more time and maybe hard and difficult for many... With subdivs inorganic models all is technique, if you don't have enough technique and practice, all will be harder and more difficult, no doubt.
Got enough technique, are you sure? if it's really that no, then, the technical aspect can be something difficult to solve and we tend to say: "ohhh it's difficult!". It's difficult because yoy don't know it. If you know how to do things, how really subdivs works, there aren't technical problems, all should flow as water and you should model fast and smoothly.
Complex forms (but simple) are not difficult if you know what you need to know, are only time consuming and that's all. Technical problems? where? do you know how to model or not?. All is to put in practice our knowledge, and i can say that to model an inorganic shape in subdiv is easier than making a credible character with the same technique.
With characters, is like we needed more brain power. You can have the technique but if you don't see the "proportion", you don't do anything. That's Plastic (vision and perception; understanding of color, volume, perspective, etc.), and you don't need it for inorganic shapes. The lack of plastic is what makes things to appear like something really difficult, and it's because we don't "see" the form, the volume. We can see our level of plasticity in drawings, and there are dozens and dozens of "artists" drawing as a child of 5 years old... and look what a joke.. they call theirselves as artists
Got enough technique? if not, learn practicing more.
Got a great level of plastic sculpting with Zbrush?, try doing the same with polygons and level up
Ending, time consuming is not equal to difficult. There's nothing difficult when you know it, and human proportions and measures are one of the most difficult things to "see" (you know.. uncanny valley)
guys, good art is hard to make, okay, we get it. Per and pior are talking about the technical process of making polygons do their thing, not the APPLICATION that process to make pretty art. MM, go read everyone's posts again and feel silly about yourself, let's move on.
My workflow is to open 3dsmax, feel an overwhelming rush of depression and anxiety, and cry myself to sleep.
anyways, i see a bit of chronyism here so it is pointless for me to say any more.
lol wut
Does this mean that if you model a realistic human using nothing but NURBS, you're somehow a better artist than someone who sculpted it? I really don't understand what you meant by that phrase.
As for workflow speeding up, here's one I like - I did a script where if you have a diffuse map assigned to your material, you can click a button and Maya will scan for textures matching some arbitrary naming convention (if the diffuse map is _d, it'll stick the _s in specular and _local in normal bump), so that you can easily assemble a full scene of materials without having to manually connect up all the maps. Saved me a fair bit of time already.
I also have a Photoshop script that outputs layer groups matching these naming conventions, so from a single PSD i can get a full shading network set up in Maya in 2 clicks.
Also, yes, what Per says is right, it's silly getting stuck up in semantics, clearly the "technical" and "artistic" things are two completely different issues and don't really have any crossover at all.
Writing a script to speed up rendering a character from 8 angles has absolutely nothing to do with how well I can model a realistic human head. Workflow and artistic integrity (whatever that is) should be kept well apart, IMHO
I guess it's the difference between production art and personal/gallery art. People who oil paint just to oil paint and share feelings with the world of art probably don't really have any connection with people who make 3d models of giant space chainsaw guns, yet they're both artists...
Type some code.
Create a teapot.
Run the code.
Type more if necessary.
Sorry if that is useless but I seem to be getting my work done fast enough, so I thought everyone in the entire world could benefit from it.