Model "sketches" for free use! Maybe?
Well, inspired by Surt's CC0 Scraps thread, I thought I might as well try my hand at a similar sort of thing. Because, while I can't really rig or animate or even really texture very well, I like to kitbash together models in TinkerCAD, using a combo of the default shapes and CC0 models, and I was thinking they could serve as good bases released under CC0 for somebody who's actually competent at texturing/animating/rigging to turn into resources if they think it's cool!
And, I might as well ask, would that be at all plausible? Linked is an example of such on Dropbox, kinda inspired by the art of a friend Mote-of-Ash, because god knows the best CC- models are inspired by the works of shortstack artists, and I might as well ask, do you think it'd be workable to use as a base? All the assets used are CC0, though I do worry that the natural messiness of elaborately-altered models in TinkerCAD like that might pose an issue, but either way I hope you like it, and I hope it serves as a good first impression for any others to come!
This has a lot of remeshing/resurfacing functions that could possibly take a kit-bashed mess and turn it into something that could possibly be suitable for real-time engine:-
http://www.meshlab.net/
It would also be possible to just automatically unwrap UVs and bake the details onto the model. It's never going to be "great", but with enough time and care something might end up "suitable" to some degree, at least.
Huh. Interesting. I'd used that program before, but only tried it for this one really cheap 3d scanner that kinda sucked. I may have to give it another go.
Looking at the example one I linked, what would you say I'd need to apply there from that remeshing program to get it to be workable? And, how workable would you say it'd be?
:EDIT: Also, here's a secondary version I made with abit of work in Blender and some poly-reduction/re-sculpting in Sculptris, how much easier would that be to work with compared to the original?
Well, here's another one, with a dropbox link you can check out. Now with a good pic too!
She's a fair bit simpler in construction than the other, so I'd presume the mesh'd be less messy. For those who are; again; actually competent at modeling, is she usable?
Also, I made two others, also available for download!
Again, still wondering, how do people who actually know things regarding modeling feel abt this? I know the edges look a bit... ragged with lots of extraneous points, but IDK if that's a problem or not?
The main issues I can tell (without too much examination) - bad topology = inconsistent detail concerning the size of triangles in relation to each other, and intersections and hidden geometry, which not only wastes triangles but potentially causes unnecessary overdraw issues. Neither of these issues are very easy to fix but if so, Meshlab would be the best (free) bet.
Ah. So, basically what you're saying is that the method by which TinkerCAD exports models makes them needlessly messy, and the only way to fix 'em'd be a process well above my skill level; that's probably too much of a pain in the ass for anyone actually competent at 3d modeling to find worth their while to make 'em usable...
Well, poop. Tho, I might as well ask, do you think the actual design concepts are any good regarding the models? Like, my own perversity aside, do the character designs themselves look unique and interesting? And, do you think they could at least be useful as "rough sketch" maquettes to be used as design inspiration for actual competent modelers to make stuff from scratch based on those designs? Like, inspiration for those who want to take those design concepts and make something from scratch that's usable?
Cause, I am more of an artist than an actually-competent-modeler, and I do wonder if these'd work well as "concept art" to inspire. Kinda like Zonked's stuff (Tho Zonked is way cooler and better at it than me)...
First, for removing internal/hidden geometry (which is likely the biggest concern), it can be done with baking a bunch of lights onto the model and then using that information to colourise the vertices and then use that as a selection, deleting everything that isn't affected by the light. Another technique (which is much easier and quicker, although a little less reliable) is this:-
http://meshlabstuff.blogspot.com/2009/04/how-to-remove-internal-faces-wi...
This will shave a bunch of triangles and also help to minimise any overdraw issues when the mesh is being rendered in real-time. Neither approach is 100% reliable, depending on the mesh(es), light can leak through or not be thoroughly caught, resulting in either triangles remaining inside or holes on the outermost geometry. For the latter case, you can try to "close holes".
Then do a bit of cleaning up with "remove duplicate faces, remove duplicate vertex, remove unreferenced vertex, remove zero area faces". Infact this is probably worth doing at different intervals, and definitely at the end of the process.
For merging everything together, you would need a "CSG" operation or a surface reconstruction (Poisson) - one requiring extra time to piece all the different meshes together, and the other requiring a lot of computer power, and resulting in a more high-poly and organic shape. To the former for machines like robotics, and the latter for organic objects like creatures, I'd say.
This will make your model "watertight" and an enclosed contiguous mesh, which is what a GPU wants to chew on, but it would also create a lot of tiny useless triangles and possibly some infinitely sharp ones, so those need to be cleaned up and the entire mesh simplified.
For cleaning and optimising the model, use "Quadratic edge collapse decimation", or (in combination with) some other simplification methods. Every model is different and you'll just need to play with parameters until you get something that brings down the polycount, helps get a consistent topology and doesn't make your mesh too ugly in the process. Depending on if it's something mechanical or organic, different methods and parameters would be used.
At this point, you would really need to weigh 2 considerations - if the extra amount of triangles (even after cleaning up) is a worthwhile benefit over the possible overdraw penalties from how it originally was. It should be, but not always, so keep your eye on the polycount when you're doing this. Each case would be different and there's no general advice to give. To generalise, a modest amount of a higher polycount is worth it against the overdraw and batches that a "kit-basher" will face in real-time rendering.
Finally, you can do things like take the original version, using subdivision, texture-painting, displacements, adding extra geometric details, etc. and unwrapping your clean low-poly version, and baking everything onto that. Textures, baked lighting/AO, normalmaps, etc. Auto-unwrapped UVs are never going to be as good as a UV layout that has been carefully unwrapped and orientated manually, but you might be lucky come out with minimal artifacts and decent texel scales. For the high-poly version you're going to bake from, there are basically no rules, just do anything to make it how you want it to look and bake it all out.
If you want to animate, that's a different story, but it may just require some cutting up the mesh to create a couple of new loops for joints, with a little bit of cleaning. Again, it's too case-specific to go into, but sometimes a model can be easy to set up for animators with a few little cuts here and there.
Ultimately, at the end of all this, the real question is - is it worth doing all of this or just making a real-time 3D model correctly - from the beginning?
Ah. Well, as you can see, again that's way beyond my skill level. And, while I do try to work with from-scratch, I consider the kitbashes different artistically, with the weirdness created by the mingling of elements being distinct from the sort of designs I'd make from scratch. Sorta like the difference between painting and collage art if you get me.
I may just use 'em as "maquettes" for model sheets to put up as concept art, ala again the one Grapple Girl was built from or the way Zonked uses his 3d sculpts, so people can use 'em more properly as guides.
Tho, from what I can tell, Tinkercad-kitbashed meshes don't really have that issue of internal geometry and duplicate faces, aside from non-manifold verticies I have to manually fix in Blender before doing stuff in Sculptris with 'em.
I'm presuming you're coming from a place of kitbashing in Blender n such, given that TinkerCAD handles it differently. Namely it handles it more like solid objects rather than networks of verticies; which I'd presume is where the weird higher-poly edges come from when exporting. Does that change anything regarding your advice?
The first paragraph of my previous post, the link to that webpage I posted, is probably the most important thing. Sometimes you might halve the polycount of your model, or even more. These are just a couple of easy steps in Meshlab so I doubt it's beyond anyone's abilities. For that example, something built out of virtual LEGO blocks, the reduction in polycount must be pretty huge there. In my experience, it usually leaves more internal leftovers than any gaps in the outer geometry.
Well, I did try, but on the Al Menoir one at least; there were no internal faces detected and even using your tips I found Meshlab difficult to work with at best.
Again, I must ask, are you familiar with TinkerCAD? Because, I mean, I'd presume it has a lot of quirks compared to other modeling programs in terms of how it exports, and I have noticed it can cause these sorts of internal-hidden-geometry issues, but not as much as probably straight-up kitbashing in Blender does. Since it seems to take the external mesh (Or what it interprets as such) and export just that directly.
But, I did do some stuff with two of the models above, specifically using Decimate in Blender on 'em both to bring down the polycount, and detecting non-manifold verticies to find what's wrong with the geometry so I can fix it. From looking at the new versions, do they look better/more usable for game stuff?
Ah, it seems this TinkerCAD already performs CSG merging on export, so the screenshots you posted before really threw me off. Still, the decimation in Meshlab is far superior to the Blender modifier, to bring down the polycount even more and get them ready for baking the high-poly details back onto. If you need any specific help in any other area, feel free to ask me. Maybe all this stuff I posted before will be of use to anyone else looking at kit-bashing their way into content creation.
AH! Well, it's been a while since I posted in this topic. But that's because I was preparing the attached file!
It has all the models I made for this and then some! I was thinking about posting it to the site as a sketch-thingy for people to use, under OGA_BY with the mention that I'd love for any full-on actually rigged/textured characters made with this to be under a similar (Non-SA) license because I'd love for there to be more in the Creative Commons!
But, the question is, do you think the stuff is useful? I'm fully aware it's not really "generic" enough to use as generic bases, but that wasn't quite my intention as much as providing inspiration to make all new characters!