Hi Sketchfab! My name’s Noah Summers. I’m a hobbyist illustrator and 3D modeler. I discovered Blender (and 3D artistry) in 2007, and learned most of what I know during high school (before getting a job, getting married, and becoming a parent). I’ve settled firmly into the modeling specialization, but I try to keep up with industry developments and broaden my skillset whenever I can find the time. I used to shy away from hard-surface modeling or modern game engine assets, for example, but now those are some of my favorite subjects!
The Blender Foundation is also to be credited with my obsession for Open Source Software. I’ve been exclusively using Linux for two years now (no more dual-booting), and I find there isn’t much I can’t do anymore with Blender or Krita (the incredible Substance domain excepted). I haven’t yet joined the development side of my favorite open-source projects, but I did finally take the plunge into programming last year. As of this month, I’m living in Dallas, Texas, and looking for a job in web development (call me!).
While I no longer daydream of a career as an artist, I have happened to work on some commercial 3D-printing projects that include film props, fine jewelry, and (featured here) board game figurines. Turtley Democracy is a physical game currently in development by a former coworker of mine. When he found out about my ability as a digital sculptor through a mutual friend of ours, he asked if I would be willing to create the player tokens.
Originally, I was handed the model sheets for three of the characters, and I went straight to work without much planning.
Since the characters consisted mostly of round, tubular forms, I created my base meshes from simple wireframes and a Skin modifier (subdivided a few times), and started sculpting from there.
Before I got much further, I noticed that I had strayed too far off model in the case of the small turtle: arms were too long and bulky, and the head was too small relative to the body.
I had another model sheet showing the different varieties of turtle that would feature in the game. Without knowing in advance which ones I would need, I went ahead and created new base meshes for each unique type, this time much closer to model.
I used a similar process: starting with skinned wireframes, I sculpted some smoothed off suggestions of joints and and musculature. I modeled some common elements like eyes, and the suggestion of mouths.
With these basic details already in place, I could pose new characters fairly quickly, allowing myself more time to worry about expressions, clothing, and finer details, rather than foundations like bodily proportions.
After all that, I retopologized the ones I knew I would need, to make them more armature-friendly. This may have been an unnecessary step, but the clean topology made my life easier down the road: I was able to quickly create form-fitted clothing items (after posing the characters) by duplicating and scaling sections of the existing body topology, rather than manually sculpting them out later. By the way, CG Cookie’s RetopoFlow addon is a godsend, even for simple models like these.
As for the process of sculpting itself, I really don’t use any remarkable techniques (not even on my much more intricate Slender Man piece). I almost exclusively use the crease and clay tools, mostly with default settings, and I always leave Dynamic Topology on because I don’t see why not.
I originally intended to sculpt fully expressive faces on the characters, but it was requested that I remove these and restrict sculpted details to the eyes, as the mouths would be painted onto the physical figurines.
After finishing the cast of characters using the new process, I restyled the initial two models to be consistent with the rest. The blonde character was pretty close already, but the hair and skirt were very soft, where the newer models made use of chunkier, sharper forms.
The smaller turtle needed to be recreated from scratch, but I’m glad that I did. The improvement is self-explanatory, I think.
Revising for 3D Printing
Given that these were to be 3D-printed, one problem I did need to solve was interior geometry, especially self-intersections near joints and creases. For most of the characters, it wasn’t too much trouble to zoom to an interior perspective to manually select and delete offending geometry.
For this character, however, I made such heavy use of the crease and inflate tools that self-intersections were just everywhere. I believe Meshlab has a superior set of tools for dealing with this exact problem, but at the time, I figured out a quick-and-extremely-dirty method of my own using Blender’s built-in “3D Printing Toolbox” addon.
First, I would click to check for intersections. Then, click to select the identified geometry…
After deleting, make sure to deselect everything, then hover and press “L” to select the main mesh.
Press “Ctrl+I” to invert selection (which should leave you with all of the remaining interior geometry highlighted) and then delete.
Lastly, click the “Make Manifold” button back in the 3D Printing Toolbox, which (most of the time) will fill the hole, which can then be quickly and carefully touched up using sculpt tools. The key is to work on one area at a time, as having too many holes in the mesh will cause the “Make Manifold” function to fail.
As for creating the watertight meshes for printing, that’s pretty much it.
Months later, I decided it would be cool to start using Sketchfab to show off some of the 3D projects sitting on my hard drive (screenshots and renders just aren’t cool anymore). The lazy side of me wanted to upload the plain sculptures and call it a day, but I’ve learned that the average viewer tends to be underwhelmed, if not confused, by untextured models.
Sketchfab’s user base doesn’t fall into that category, but I actually intended these to be displayed alongside my programming job applications. If possible, I want my portfolio to say, “I’m not just your average entry-level programmer. Check out the other cool tech stuff I’ve taught myself to do!” And if a recruiter is glancing through at my website, I’d like to use that half a second to elicit memories of a fun video game or animated movie, rather than remind them of a chess set.
Given the intended scale of the figurines, I went for a Substance graph that would roughly emulate the painting style common to tabletop wargaming miniatures. The result wasn’t perfect, by any means, but I think the point comes across.
First, I needed to unwrap the meshes. I decimated them as far as they would go without creating smooth-shading errors, and then unwrapped using Smart Projection.
I tried to strike a balance between efficient use of UV space and an unholy mess of texture seams. In the end, this battle was lost before it began, but I didn’t have time to retopologize the models a second time, so I tweaked the settings as well as I could.
Generally, tabletop miniatures are painted in layers, from shadows, to solid colors, to highlights. Because of the small scale, colors are applied lightly with a technique called “dry-brushing”: repeatedly brushing across the surface, away from the direction of the light source, allowing the lighter-colored paint to catch any protruding surfaces along the way.
I might have been able to mimic this effect more accurately with manual techniques, but I needed something quick and procedural.
Other than the base color map, the substance used a few bakes as input: ambient occlusion (to darken the cracks), top-down hemi lighting (for general large-scale light and shadow), and curvature (to be used for small-scale “dry-brushed” highlights).
It was a pretty simple graph: just a bunch of levels adjustments, blur filters, and blending nodes. The darker shades were made cooler, and highlights were made warmer. On top of all that went a softened perlin noise normal map, to simulate a mottled plastic surface.
After uploading the final scene to Sketchfab, I noticed it had trouble loading, let alone rendering, on my low-end mobile phone (a Samsung Galaxy J5), which I wasn’t happy with. Ideally, I wanted loading time to be reasonable on an average office computer.
At first, I thought I would need to:
Decimate the models further
Re-bake all of the input textures in addition to normal maps
Re-project all of my paint strokes onto the new meshes in Substance Painter
While testing this process, however, I learned something super cool: Blender’s Decimate modifier also decimates the model’s UVs and preserves the original layout! So all I needed to do was decimate the models and bake normal maps. All the existing texture maps would just work!
After baking the normal maps, the new models looked identical to the originals with roughly one fourth of the original polygon count (and without a lot of extra work, to boot). Despite seven characters and texture sets, the new scene loaded and displayed quite nicely even on my low-end smartphone and ten-year-old ThinkPad.
At this point, I decided to call it a day. One of my long-hidden portfolio pieces was finally up for display, looking better than ever, and without requiring a gaming computer to view it.
If you found this interesting at all, let me know in the comments, on the article or the model, or email me! I love to hear from people!
Lastly, many thanks to Sketchfab for the opportunity to showcase this piece. I can’t wait to upload more (and better yet, create new stuff)!