Stylized Swords
interpolator
I made a fantasy stylized sword to show off some smart materials I was working on, and wound up having so much fun with it I've decided to make a full set of them.
Here are those I've completed so far;
Here are those I've completed so far;
Replies
With that in mind, I'm working on a cleaver next.
I didn't even end up adding the hanging cloth strap, because I decided the physics would be too engine-dependent if I turned it into an asset.
-Edit- Updated.
I might need to move on to the next sword now, and just come back to the rust later.
I'd also like to work on a smart material which adds polish to the edges of the blades, which I foresee being CRAAAAZY rough to set up.
Something along the lines of combining a thickness with a curvature mask, and using noise to refine the inner sides. Maybe a directional blur.
With that in mind, here's the new lowpoly.
I've realized that I should be able to work much more effectively by just jumping straight into making the highpoly in ZBrush, then retopologizing it.
On top of that though, I think I may be able to just make a tool for Houdini which would handle the retopology and UV-mapping automatically, due to the simple shapes and lack of animation. If it works, between that and the smart materials, I could pump out a stupid number of weapons easily.
Here's my test result;
It has an acceptable polycount for a modern game (I think this one is sitting at ~1,800?), and I can get the final result to look perfectly fine.
Though if you saw either the wireframe or the UV map, you would probably vomit.
-Update- I think I've gotten the UV issues down to a negligible level. I'm actually extremely impressed. The topology is still kinda gross, though.
I may need to dial back the automation a bit on that front, and add in a weight painting step to control the geometry density across various areas. It would still be stupidly fast and easy, though.
Here's the new vs old;
The packing algorithm is actually miles better than Modo's. It created a lot more islands than when it was done by hand, but that actually allowed it to make far more effective use of the UV space.
This is also in spite of the bad topology, which severely exacerbates problems in the auto UV mapping.
As you can see from the wireframe, the topology is quite the mess and needs some improvement. Despite that, it's only a difference of like ~600 total polys. Though that is before accounting for all the extra UV splits.
-Update- I've been able to optimize the workflow further by adjusting how I handle the highpoly with automation in mind.
I removed unnecessary higher detail bits that could easily be handled in Painter instead, like the runes, and this prevents the auto-retopo from trying to match that geometry.
Even without changing the setup in Houdini, doing this already made the result much cleaner and dropped the polycount to 1,500. And there's still more I could remove.
I think I can start on the next sword now with this workflow in mind. I'll sculpt out a sort of "mid-poly" which has the shape, but none of the higher details I don't want in the lowpoly. I'll GoZ that over and generate the lowpoly and UVs from it. Then I'll continue sculpting the higher details onto the mid-poly over in ZBrush to get the highpoly.
I figured out the perfect way to test this workflow. I took the lowpoly I'd already made, then sub-divided it to get a super dense mesh equivalent to the mid-poly I would make. I then ran that through the automatic retopology / UV mapping process, and compared it to the original done by hand. Here's the result.
So the UV mapping is debate-ably better than when I did it by hand, at least in this particular case. The only issue is the increased number of islands, though that still falls within what I would consider perfectly acceptable. This mostly comes down to Houdini's packing algorithm, because I believe I got the best result by packing the original UVs inside Houdini.
The topology looks messy, but it shades perfectly, bakes perfectly, and actually has a polycount of slightly more than 100 less than when I made it by hand. Which.... makes it better? Maybe? That probably calls my modeling skills into question, if anything...
Below we have the final results for each.
The two are virtually identical, which is incredibly impressive to me.
Doing this, I went from highpoly to finished in literally minutes. That includes retopology, UV mapping, baking, and texturing.
Even without the auto-lowpoly stuff, making it in ZBrush from the start is feeling way better and faster for me.
I still need to work out a few kinks in the auto-lowpoly before it'll be production-ready. Right now I need to;
1) Make it automatically delete occluded areas of the mesh so it isn't wasting space on stuff you never see. I may be close on this one.
2) Figure out how to get it to split long UV islands. This is a prevalent issue in stuff like swords, which have long, unbroken planes. Normally I'd just split them in half manually, but obviously that won't work here.
I'm also looking into using non-square textures, because I think that might also fix the issue with long islands.
3) And this is a tough one; I need to get the UV mapping to create fewer small islands. The Houdini tools have a feature for this built in, but it's buggy as hell right now and doesn't work right.
Seeing it complete, I'm happy with the whole design now.
I'll definitely have to redo the first two at some point, because they really don't stand up anymore. Besides which, they were made with a completely different workflow.
As I was trying to work on my next sword, I realized my design skills were severely lacking. Not only was I struggling to come up with a new design, but the more I looked at my existing swords, the more poorly designed I realized they were.
With that in mind, I decided to take some time to practice my concepting and design skills over the past month.
I'm going to redo all the swords, and I'm planning six in total.
With that in mind, I'm reworking the first sword right now. Here's the basic form of the highpoly. I haven't added the finer details and wear yet.