modeling hi poly weapons is painful, any tips?
polycount sponsor
Hi, recently I've been using smooth proxy modeling in Maya to create all my high resolution models, this technique works great for organic models like people, after doing two non organic models, both realistic guns, I have found that smooth modeling does not lend itself well to objects with hard edges. The realistic guns were not too bad but now that I have moved on to a futuristic gun the amount of hard edges is extremely painful. I have to create copious amounts of edge loops everywhere to get the hard beveled edge I'm looking for. Just wondering if anyone has any secret tricks for modeling high poly non organic shapes, or do these types of models take the average artist forever to complete? Here is a screenshot of my pain.
Replies
http://home.comcast.net/~dpattenden/images/mp5_wire.jpg
What I ended up figuring out is its easiest to sort of carry your edge loops on rather than try and find awkward places to terminate them. Poop might have some stuff to say ( he's a Max guy, but this is a topology problem really right? ) since I know he has a lot of experience with hi-polygon weapon models. Plus show us your wires and we might be able to better figure out how to help.
I don't want to let this model out of the bag yet, but here is a closeup of the wires on one portion
Its very important to get the microchamfers, and keep your quads for hard surface modeling. If it's for normal maps, you can afford to be messy a little, but I found it was best to just force myself to be neat, and get faster and faster at it, before letting myself cheat all the time. I use orian flames Select limited edge loop and select limited edge ring, as well as his remove all smoothing groups shortcut. I like to have the model all faceted in edit poly, and when I hit the ` key (which I have mapped to show end results) the turbosmooth kicks in and it's all of a sudden smooth. That way I can really see my model in faceted and smooth view. (the picture is in faceted view.) If you use turbosmooth (if you have max 7 it is 10x better than meshsmooth) make sure to check both the isoline and the explicit normals checkboxes. Explicit normals speeds it up even more and looks better. Isoline makes it way easier to read where your lines are.
Limited edge ring and loop allow you to select two edges and then it selects all the ones in between either along a loop or a ring, very useful.
Some tips i've learned. Don't be too anal about the edge loops till the end. I almost always find a spare edge later to connect in and make all quads. If you can mass out your shape early on with all quads, then it makes it much easier to refine. Make sure not to use the same chamfer value every time, as it will feel too similar. I like using .1 and .2, and sometimes using a .6 or a .8 and then putting a .1 on the two resulting edges. It makes a nice bevel where the edges are hard.
Any time you have a non four way intersection of edges, like a triangle or a 5 way intersecton (even if it's all quads, it's naughtier than a 4 way) hide it on a relatively low detail area. I try to have perfect four ways at all major direction changes, as they smooth the most evenly. any time you have details that you can float over the surface, do so, if it's for normal maps. Screws, rivets, and nubblies don't need to be welded in, they can just float over top and the normal rays will grab them if you check "furthest" in your normal map baker of choice. It even works on concave surfaces, like this
See how from the top you can't tell it's not welded? Neither can a normal map baker in this instance. I would chamfer that 90 degree edge if this was a high poly, but you should get my point. I have a whole library of floating bits I can add to a model to increase the detail of the object.
Hope this helps, but like Daz said, it's all about practice, working in your shortcut keys, and wrapping your head around it. You start to know when to do what steps when in the modeling process so that you don't have to rework stuff.
I saw some solid tools and enhanced boolean operations for 3dsmax as well and while I cant afford, still can recommend the stuff. cause although triangulators and file exchange possibilities exist between cad and art modellers the good ones seem to be really expensive, was able to get rather fine results using STL export/import in max from nx3.0 so. not with big models so, but parts.
what I really love is all the associations, like keeping things parallel, aligned on surface... its really cool. anyway if anyone ever gets a chance to try some of the tools, I found them more artist friendly then first expected. although to do big curved surfaces and such its a bit messier than subdiv, cause its more rigid the way the data is processed.
however its always cool to see the really good models like here come out of "normal" boxmodelling, sometimes frustrating seeing how powerful perfect booleans could be so.
As for hipoly topology, there's quite a few ways to approach things, mostly dependant on what your tools allow you to get away with and what you're doing. Some things are done better with SubD, others look as good and are quicker and easier to do in plain polys with beveling.
Like Daz said, there's tools (like modo or LW) that let you weight edges... while it's cool at first and looks like a time and polygon saver, it can be rather frustrating. You end up with weightmaps that are only visible in funky vertex colors when you switch to weight maps view (unshaded horror). Specially when you revisit areas after a while, it's hard to tell which edges exactly had how much weight etc.
If you're working with plain polygons (no SubD) then a bevel/rounding tool is your best friend. I'm guessing all apps have some kind of edge rounding tool nowadays so that you can have edges a bit rounded for that fancy hipoly look.
For SubD you usually slice rows of polygons to create hard edges. It can be messy but doesnt has to be if yuor app supports Ngons. You can just finish your slicing whereever you want and let it flow into an Ngon. The Ngon has to be positioned in a rather simple area, as there usually some tension on it from all the edges flowing into it. Therefore, it's always good to have 'relaxing' cages around details that normalize the topology a bit again.
Well, i dunno if that helps, hipoly models do need more polys, but you should terminate edge loops wherever you can rather than letting them flow around the whole model. AT some point while adding details, all those edge loops will come back and bit you in the butt.
Personally, i mix both modeling methods as i see fit, i don't bother much keeping quads (just dont have tris if you can avoid them) and have as much or little topology as i need in the specific area. In organic models it's more difficult to find spots to 'dump' your edgeloops into Ngons, but in mechanical model, you can go pretty wild usually. I often ignore the topology completely, welding all possible polygons together to have large clear surfaces, do whatever i have to do in that area detail wise, and then just connect the vertices needed to keep the SubD topology working by hand.
If you want to skin the hipoly models, then you'll be thankfull if you can have decent UV seams where you actually need them, rather than something that looks like a car accident, otherwise, if it looks clean and looks good, ship it.
Some plain poly wireframes, the upper model is a good example of how a model shouldn't look with all the slices left there (was my second or so hipoly model 2 years ago).
http://www.strangefate.com/webby/modeling-plain2.jpg
http://www.strangefate.com/webby/modeling-plain.jpg
Rounded edges are way more expensive than with SubD (or meshsmooth and Co.) but you save yourself some thinking trouble when cutting into rounded surfaces and you almost don't have to bother about any topology at all.
Some SubD...
http://www.strangefate.com/webby/modeling-subd.jpg
And some of the rendered out objects
http://www.strangefate.com/portfolio/Mercury-rad.jpg
http://www.strangefate.com/portfolio/beebugbot-rad.jpg
http://www.strangefate.com/portfolio/freighter-rad.jpg
Edgeweighting can help you in some areas, but as Daz pointed it you can't make a full model with just them I guess. However in Max you can use a combination of both to achieve interesting results.
Edge weighting in Max basically adds smoothing groups to you model, even at the lowest value. A small crease is created, and the highest value you put, the closer your angle will look like the one on your proxy model.
Vertex weighting is something else. It's similar to Zbrush's pinch : the verts of the final highrez mesh will be dragged close to the vert you weight if you give it a high value.
On the pic below you can clearly see why its popular amongst car modeling enthousiasts :
One thing to note is that you might need an extra row of edges/verts to counterbalance the 'force' the weighted verts have. Still have to experiment myself ; its a weird approach but can prove out to be handy.
One final thing I find useful : A neat way to model is to model the whole shape first without worrying about the hightrez problem, and then add the needed bevels for subdivision. This way you focus one 'large scale' shapes before jumping to the 'small scale' details.
Now show us your progress!
[StrangeFate] Rounded edges are way more expensive than with SubD (or meshsmooth and Co.) but you save yourself some thinking trouble when cutting into rounded surfaces and you almost don't have to bother about any topology at all.
[/ QUOTE ]
I'd like to hear more about how you guys deal with cutting details into curved Sub-D surfaces. This has proven to be difficult in the past, keeping a smooth curved surface when cutting bevels into the middle of it. Spherical or domed surfaces seem to tbe the toughest to do this with. You get little pinches or soft creases where you don't want them. It just seems to be a pain to deal with.
Cool models there, btw.
How about cutting across the tip of a hemisphere? If I use a longitudinal sphere as the base, I get a pole of triangles. Cutting a beven through that is hell.
If I use a geosphere I get rid of the pole but I get tons of tris that need to be converted into quads, and then terminated.
I guess I could manually edit out the pole, but then where to terminate the loops? I'd love to see some wires of how people conquer this.
hmm Don't know if this helps, for cutting into rounded surfaces.
http://www.strangefate.com/webby/modeling-rings.jpg
Edge or vertex weighting can be usefull and easier there sometimes, although i really don't like using them, you really forget where you used them in large models and while edge loops transfer to pretty much any 3d package, weight maps are often not compatible from one app to another.
The orange selected edges are what i call 'salvage rings', they are located around a detail, encircling it. Their job is to normalize the tension or 'force' between the outher shape and the detail, so you don't have any weird distorsion from the cuts carried over to the main object.
I then connected some of the vertex from that cage to the outher shape by hand to keep the whole thing from imploding or creating small problem zones.
In the second screenshot you can see that i welded all the incoming loops from the sides together into one. Keep the mesh as clean as possible. Keep quads where you can bother, they're good to make edge loops and tension free, but other than that, Ngons are your best friends. The more complex your cuts and slices, the better Ngons perform over quads, and are a lot cleaner.
Detailing and cutting things into SubD shapes makes only limited sense. If you were modeling an airplane in SubD and wanted to cut in all the panels (rather than use bumpmaps) and still keep the surface perfectly smooth you'd be in for some work. It can be done, it's just an insane amount of balancing 'the force' (i like that, thx Pior
Another way i've seen to deal with exXxtreme SubD slicing is to build the wanted shape roughly in SubD, then use it as guide and rebuild it again panel by panel (or detail by detail) as separate objects. I guess this would come close to face modeling (like Per has in his P&P thread), and lets you deal with each part of the model separately, without having it's properties and shape break the tension and flow of other parts.
I've never tried it out myself tho, saw a guy do it here and after 3 seconds i fell like watching something else.
My patience < patience required.
I'd keep the guns all in SubD, most of the shapes are rounded which do better in SubD and they're overall a good practice.
The most important thing i feel like repeating endlessly is to keep your mesh clean of unnecessary loops and slices. If you have an empty flat surface with unneded slices, weld it all together. If you later need to add details there, you'll have a clean Ngon and no wasted loops screwing up your new details (booleans or whatever you do).
**edit: what i forgot to say is, that complex or rounded SubD surfaces do sorta limit where the details go and how you have to add them in order to not break stuff. It's all a matter of knowing a bit beforehand where you want the stuff and improvising the rest.
If you look at other SubD hipoly models you'll see that the artists don't always bend the shapes and topology to fit the details they want, but instead, they take advantage of the existing topology as best as they can and add whatever looks cool within the restrains of the topology. As a watcher you'll never see that unless you see the wireframes.
Normal non SubD hipoly modeling on the other hand (while admittedly not as cool and appreciated as SubD) is like riding a bulldozer over a garden. You can do whatever you want with the geometry, giving you all the freedom you want, great for machinery.
I agree about how vert weights can be a pain, at least in max. The biggest beef I have is that they don't survive a collapse of the stack... the results can be baked in, but they're no longer editable. If I use extra edges instead, I can collapse whenever I like, which is pretty often. I guess that's a max thing though.
Do you guys ever really transfer the cage from one program to another? I think this would make sense if you're keeping a model morgue. I wonder if anyone organizes their morgue into universal formats, like say OBJ.
Even if i did everything in 1 app, i'd still want to keep my models correctly readable in other apps, just in case.
Personal paranohia i guess.
I see Poop is maintaining a library of widgets. We have a library here, but mostly textural. Anyone being good and anal about stripping out the good assets at the end of each project? Or maybe instead just tossing them in a folder as they go?
I used to keep a library with details, but by now it takes me less time to model them than to open the library file and grab the detail.
What i do look for is reusability of complex or neat parts/chunks of a model. Things that stand well on their own and can be integrated in other models.
I found this to be informative about cutting holes.
There is probably a super boolean alternative in Maya. This technique is absolutely fabulous when it comes to fillets, holes and all those other pesky modeling details.
...yet despite this topic's vintage, there's still a substantial validity in terms of underlying fundementals irrespective of technique shared by ppl here. Although yes, tech has obviously progressed since the 'swinging noughtie's' however basic modeling principles remain I think largely unchanged.
Now with that said, in the current era we're basically spoilt for choice, whether NURBS, CAD, Curves/Splines, Pro/Boolean, Procedual, Polygonal Parametric...etcetera, either via explicit adaptive software and/or inspired tools/plugin alternatives graciously made available by equally innovative devs for all too enjoy **shoutout** Boxcutter-Carver, HardOps and the like....which enables modeling stuff that happens too tickle ones particular fancy.
Anyhow I'm an extremely appreciative all things mechanical hard surface bloke for the continued opportunity to do so.