Profile for taylor
About taylor
Fields
- Git Forge
- https://forge.axfive.net/Taylor
Bio
I am Taylor. I am a programmer who mostly uses Rust, C++, Ruby, Python, Lua, and JavaScript at work, and work intermittently in roughly a dozen other languages.
I'm a hobbyist artist and musician, trying (poorly) to develop some 3D art skills. I mostly work in free software like Blender.
- Joined
- Posts
- 407
- Followed by
- 115
- Following
- 471
Stats
PSA: Modular power supply SATA cables are not interchangeable. I needed an extra set of SATA power cables and grabbed one from my PSU box, and didn't notice that it was actually a leftover from a different power supply, and simultaneously fried two 4TB hard drives, both members of a RAID1 array.
At least they should still be under warranty, but still painful.
A "bigger on the inside" proof of concept in Blender.
This was pretty challenging. The basic effect is pretty easy, with compositing multiple view layers and using a holdout to layer them, but I need to have an object that can extend from the inside to the outside.
I ended up effectively splitting the noodle object in half with geometry nodes and an invisible inside bounding box. The noodle is fully rendered on the inside layer, but only the portion outside the bounding box is rendered on the outside layer.
This is still not fully sufficient to do what I want, though, because when I place a light under the box, there is a seam between the inside and outside layers. I still haven't figured out exactly how I'm going to address this yet.
edit: I figured it out and fixed the broken lighting. I have to have raytracing off to make it work, but this did the job. There is no hope in hell of this effectively working in Cycles as-is, though, unfortunately. I could manage it if I could get depth tests to work with anti-aliasing. I can probably manage it with super-sampling, if I want my renders to get incredibly more expensive in terms of every computing resource I have.
#Blender #GeometryNodes #Compositing #Compositor #BiggerOnTheInside #3DArt #Blender3D #WIP #ProofOfConcept
Been learning Japanese and kanji for the past month or so via Mango and WaniKani. Started off the right way, learning all of hiragana and katakana via Anki before I did anything else.
Japanese is kind of hard, but kanji are hard. It's incredible to me that there's an entire country and culture where basic literacy requires you to learn thousands of characters, each of which have multiple pronunciations and often multiple meanings.
#Japanese #LearningJapanese #Mango #MangoApp #WaniKani #Anki
I liked this piece by @sako , so I recreated it in Blender with Geometry nodes.
I spent a while mostly doing the Instance Grid node group. I got a simple version working pretty easily at first, but wanted something more robust, with functionality for edge margins, different margin types, margin collapsing, etc. I failed to just hack at that for a while before whiteboarding it out, which I probably should have done in the first place.
#Blender #GeometryNodes #MoGraph #MotionGraphics #Looping #Loop #AbstractArt #3DArt
Can somebody who sees this like and/or reply to this post?
I broke my gotosocial instance by restarting it while running migrations, and I ended up having to restore it from a backup that I'm not certain was finished (and therefore might be corrupted). I'd appreciate some reassurance that things aren't broken.
Yes, I done goofed. From now on, I'm not trying to do live backups or migrations. Shut down instance, backup, run migrations, start new version. Learned my lesson.
I spent a couple days reimplementing much of this piece by @Goldstarknight .
The principles were relatively easy, but the details took a ton of fiddling, especially getting it to loop properly. At first, I tried a BSDF for the material, but that proved too challenging to get the two-tone look I wanted, so I ended up just doing a simple color mix with a cross product. The background cubes use a simulation node to spawn on a timer with a random timeout in an interval.
Took probably about 5 or 6 hours of work altogether. I might see if I can get a dithered look without any anti-aliasing, and see if I can turn it into a compact APNG. I know I could do it via GIMP scripting, but I'd rather do it all in Blender if I can manage (I probably can't).
#Blender #EeveeRender #EeveeRenderer #GeometryNodes #SimulationNodes #3DArt #Art #Cube #Cubes #AbstractArt #AbstractArtwork #MotionGraphics #MoGraph
First render in a while. It's a simulation nodes gravity simulation. I went through many phases of overcomplicating it to varying degrees of success before going the absolutely simplest approach of Newtonian gravity with a force filter that drops very close particles to 0 influence (otherwise particles get flung into the Twilight Zone). I cheated a bit and optimized calculation time by using a Merge by Distance in my simulation zone and only calculating the influence from those. It does cause closely-clustered points to have only as much influence as a single point. In another study, I had found a way to make the merged points have the mass of all the points in the merge, but the way that I did it had worse performance than the raw N^2 N-Body calculation.
The color of each particle is simply the absolute value of the velocity, mapping XYZ to RGB.
I should work with Geometry nodes more. They are very stimulating. My real addiction is programming, and I often neglect my art, and geo nodes give me the opportunity to scratch both itches at once.
I wish I could get this one done in Cycles, but it's just too damn slow. Low frame time and denoising alone would make a 30 second render take over 5 hours, and it doesn't really look that much better than Eevee.
#Blender #BlenderEevee #EeveeRenderer #EeveeRender #Video #MotionGraphics #MoGraph #Gravity #GeometryNodes #SimulationNodes #Suzanne #NBody #NBodySimulation
Getting a bit sick of fielding LLM PRs. I understand why many projects have a no-LLM policy. I'm generally fine getting good PRs, LLM-aided or not, but the poor ones are just complete wastes of time. They all read the same, good or not, so I have to spend quite a bit of time sussing through it to determine which ones are useless or not. They're all "bug fixes" or "security improvements".
No, it does not make sense for an CLI argument parsing library to scan through argv to make sure it doesn't contain null pointers. That's ridiculous. No, it does not make sense for a CLI argument parsing library to defend against an input argument size overflowing size_t. When the fuck would that ever happen?
The PRs that actually fix a missed fallthrough causing undefined behavior look exactly the same as the ones that defend against somebody passing a 4GiB string as a CLI argument. They all have descriptions full of nothing but fluff, they all add much more code than they remove, they all bloat the code with more abstractions and indirections, and they all take way longer to review than the equivalent human-written code.
I'm even not against getting LLM contributions, but I'm getting sick of the fucking spam, and I'm losing patience with figuring out the difference between the two.
#LLM #LLMs #LLMSlop #AI #Spam #Code #Programming #PR #PullRequests #PullRequest
Felt like texturing something I didn't make for practice. Basically digital miniature painting.
The model is this mimic by Yasashii Kyojin Studio on Thingiverse.
The brick floor is this one from polyhaven.
This took me probably about 4 or 5 hours of work. A surprising amount of work went into getting the boundaries sharp and in the right place. I almost feel like it might have been less work to model and sculpt it from scratch and keep the materials separated by object. I might have to do a mimic of my own soon.
The turntable video is rendered in #Eevee and the still is rendered in #Cycles. Unfortuntately, I forgot to turn on screen-space reflections for Eevee.
#DungeonsAndDragons #Blender #Blender3D #3DArt #DigitalArt #Mimic #Monster #Turntable #Animation #texturing #EeveeRender #CyclesRender
@btp Finally started playing Axiom Verge during my lunch break today. Pretty good for the first couple dozen minutes. I'm looking forward to the rest; thanks for the recommendation. I've got a list of like 10 games in the same vein that I plan on playing through this year; Axiom Verge is just the first.
Refrenched my quest tonight at Pathfinder.
Art Advent 10: Baubles
I put a lot of work into the shaders for this one, and I'm pretty proud of them. The glitter shader uses the voronoi color output, projects it onto the normal, and perturbs the normal by that amount in an adjustable way. That took a bit of work, because I initially tried to output an RGB normal map and feed it into a Normal Map node, but I kept having artifacts on UV seams. Skipping UV and tangent space entirely bypassed that.
The bulby node tree (I think of it as a pumpkin) is done in UV space via two sine calculations, one offset from the other by half a period, and taking the max of it. It's a neat effect, and I've enabled a "Y Influence" input for twisting it along the V coordinate and a "Band Number" output to allow alternating colors and other effects. Getting the band number playing nicely with the Y influence took a bit of trial and error, but it's turned out well, I think.
#Art #AdventOfArt #ArtAdvent #ArtAdventCalendar #3DArt #Blender #Blender3D #CyclesRender #Christmas #Ornaments #Ornament #ChristmasOrnaments #ChristmasOrnament #Bauble #Baubles
Art advent day 9: eggnog
Spent a couple days working on this one. Couldn't get the cinnamon (both the powder and the stick) looking just the way I wanted, but it's not terrible. I spent forever trying to deform the cinnamon stick with a curve deform modifier, but couldn't crack it, so I used a lattice deform instead.
The foam probably took me 5 hours alone, and I gave up on everything I was trying to do. I wanted bubbles in it, but couldn't figure that out. I tried geometry nodes with a hand-crafted bubble, but it never looked quite right. I also tried using volumetric rendering, creating the porous appearance with a high density volume, but the part that was supposed to be diffuse never looked correct. I should probably just buy a fancy foam material at some point, but I want to know how to do this myself.
#Blender #Blender3D #3DArt #ArtAdvent #Advent #Christmas #3DRendering #CyclesRender #Cycles #Art #ArtAdvent #ArtAdventCalendar
Art advent day 8: Stocking
This was a pretty easy one. Just a couple tubes, some cloth deformation in sculpt mode, a quick fur object, and borrowing the candy cane from a previous day. I was going to put it over a fireplace or something, but "Fireplace" is a later day's, so I'll do it then.
#Blender #Blender3D #3DArt #ArtAdvent #Advent #Christmas #3DRendering #CyclesRender #Cycles #Art #ArtAdvent #ArtAdventCalendar
Art advent day 7: Underground
I don't even know what I'm doing, man.
#Blender #Blender3D #3DArt #Santa #ArtAdvent #Advent #Christmas #Dirt #Underground #3DRendering #CyclesRender #Cycles #Art #ArtAdvent #ARtAdventCalendar
Advent of Code 2025 day 12 solution
I really didn't like this one. Basically have to do some bounds checking "it obviously always works" or "it obviously will never work" on it to get it to work on the actual input, but it still NEVER FINISHES on the example. I'm slightly bothered that the problem is more or less unsolvable as written, but I'm really bothered by the actual solution for the input not working on the example. That's not a fun puzzle, that feels like I'm being tricked, like the puzzle was a prank on me. Leaves a real sour taste, especially being the last puzzle of this year's AoC.
#AdventOfCode #AdventOfCode2025 #AdventOfCode2025Day12 #Day12 #Rust #RustLang #Programming #CodingChallenges #AoC #AoC2025 #AoC2025Day12
Art advent day 6: Snowman
Still 5 days behind. Planning to catch up over the weekend. This one was an excuse to sculpt a decent-enough male basemesh, then destroy it with my horrible rigging and posing skills and wobbly displacement in the snow shader. I'm still pretty happy with it.
#Blender #Blender3D #3DArt #Snowman #ArtAdvent #Advent #Christmas #Snow #Snowball #3DRendering #CyclesRender #Cycles #Art #ArtAdvent #ARtAdventCalendar
Advent of Code 2025 Day 11 solution
Memoization actually worked for this one. My first instinct was to just crawl the graph with a memo, and it fortunately did the right thing. It was effectively an exhaustive depth-first search. Really not much to discuss.
My code is not nice or clean, but after yesterday's, I can't even be assed to clean it up or properly comment it. I barely gathered the gumption to pass errors up, and when I started writing it, I was just throwing unwrap() everywhere.
#AdventOfCode #AdventOfCode2025 #AdventOfCode2025Day11 #Day11 #Rust #RustLang #Programming #CodingChallenges