bsky.app/profile/dekk... yeah, guilty :P Ironically, it's my most popular video/image post, by far. Folks like jiggly bits...

Wouldn't be the first time that a series' canon got expanded without the author being the first to know. Fanon is a powerful thing, sometimes...

Still requires a decent understanding of anatomy and stuff, but the difference between this and my usual stuff is huge. It's a lot of fun, and doesn't take much time. Should allow me to make lots of 3D models quickly, that I can pose to serve as reference for drawings, or use directly in 3D.

So, the workflow thing: I really enjoy 3D sculpting but find it hard to really define shapes properly. So I repeatedly set up a camera, rendered an image, ran it through img2img in SD (with a model trained on my 2D art style), overlaid that on the camera view, and used that to guide my sculpting.

Hah, yeah! And it's easier to learn lighting that way, when learning digital painting. That's my main reason for dabbling in it, right now. I often went right into full colour stuff and got confused about why it just... wasn't "working".

Hah, a friend often asks me to check her art for mistakes and stuff to improve. Kinda got into the habit.

Also... 88k bsky followers and nobody accused you of using AI yet? Damn. Wild.

And yeah, grayscale with a splash of colour is nice. Makes a statement. Also called "selective colour", I think.

The two ends of the pencil don't line up. You ain't fooling anyone! This was made with AI.

Joking, of course. You've got a really nice art style, and my best guess is just that a lot of it is grayscale. AI users rarely gen anything in grayscale. Most don't think about style stuff at all.

That way, you won't see any unloaded chunks before you hit a new chunk and do your load/unload check again. But you might see jitter: you might want to get fancy and load stuff in advance based on the direction the player's moving, and load them in the background. But that can be implemented later.

This approach does mean you need to add a bit of a "safe area" of chunks that are loaded despite not being in view yet, but that's good practice anyway, depending on your chunk loading mechanics. I'd expand the "needs to be loaded" range by 1 chunk further than the view reaches...

Make a new list of chunks that should be loaded based on the current view position, check your list of loaded chunks, and remove chunks from this new list as you see them in the list of loaded chunks. Any that remain on the list are chunks that you need to load now...

If your current chunk doesn't match the stored one, you've moved to a different chunk (though you've already got something like this, so use whatever works). When that happens, check your list of loaded chunks for any that are too far from you...

So... my advise, if you're still struggling with this: You're really not entering/exiting chunks often, so it doesn't have to be super optimized. Keep a list of chunks that you've already loaded, and a vector2i to indicate which chunk you're in (divide the position by the chunk size)...

One looks like a dork and the other looks like an all-powerful demigod, and the dork is cuter and more relatable.

Legit though: a change in body language is also part of the character, I think? And that might make a difference too.

Well, I clicked the link in your post again and got the same error. Anyway, it was a pretty good vid. I watched the entirety of it. It's a nice update on what SlimeVR's been up to. I've got my own similar project that I recently picked back up, so it's kinda fun to see.

YouTube says the video's been removed by the uploader, though. You may have broken the link by reuploading?

It seems that www.youtube.com/watch?v=_a12... is the correct link.

Man, I might have to do the same. On the other hand, I found out yesterday that I've been waiting since 2010 for them to finally add ptex, so I'd no longer have to deal with UV maps.

And if it didn't, someone would have weirdly-thorough house rules for it instead. Or, even if it did, really.

Knowing the current state of 3D generative AI, seeing an AI error end up in a printable design is both funny and kinda impressive. I'd expect most AI output to require quite a bit of manual fixing to be printable.

I had to look up what the clarity effect does. Looks like the clarity trick you're describing is a bit like a bloom effect?