99* Days of Blender

[joseph (J)]

Day 35 (Part 1?) - Modeling Practice

(Light test/matcap)

Based on this work by Marcel Hampel:

I do plan on doing the other two lanterns, possibly today, and I do plan on texturing this, hopefully today, right now it feels good just to have it modeled. And check out this nice clean quad topology (the icosphere doesn’t count):

I’m going to try and texture this in the painterly style you saw yesterday on my building. My inability to paint backgrounds is annoying, but real (see yesterday’s entry in my other sketchbook), so it will remain unbackgrounded. Shut up Grammarly, that is a word!

Why did I pick this reference image?

Marcel Hampel is a master of shape language- using combinations of simple shapes to create remarkable results. Knowing his process of starting with extremely simple shapes, I did the same thing here- started with the most basic possible shapes, and then refined by layers. This isn’t a new process for me, but it’s always good to practice.

What did I learn today?

Recently, inspired by @piranha4D , I’ve been using the bevel modifier more. I’m proud to say that there’s not one single subdiv modifier anywhere on here - it’s all bevels. It’s a different workflow for sure, but I like it.

Music

I’m currently on my third playthrough of the FFVIIR OST in three days. I’m once again listening to One-Winged Angel (Rebirth) while writing this post, and if I had a nickel for each time that’s happened recently, I’d have two nickels. That’s not a lot, but it’s still interesting.

[joseph (J)]

Day 35, Part 2 - Lantern Done

What did I learn?

That it’s really hard to get the right amount of Bloom in Eevee, the normal Bloom settings don’t even kind of cut it. I used compositing to fix this- I cryptomatted out the lantern, applied a Simple Fog Glare just to that, blurred that twice, and added it to the main image twice:

For context, it looks like this without any compositing:

I ended up going in a different direction with shading than I was planning, but I really how it turned out.

I made the rings texture myself- by hand, I copied-and-pasted a bunch of (black outline, transparent fill) concentric circles, transforming more or less randomly, and then grouped those randomly and filled (just the transparency) those groups with white so I have the layering effect you see. I then applied a simple Paint with Brush filter:

As always, these textures are free for anyone to use for any purpose (except NFTs, if you use anything I’ve made for an NFT I will hunt you down and break your heat/AC)

[Minamookevlar (Minami)]

Nice rendition! Grammarly is scanning me right now as well, never thought they’ll be here too…

[piranha4D (Pir)]

@joseph

That it’s really hard to get the right amount of Bloom in Eevee, the normal Bloom settings don’t even kind of cut it.

Yeah. I haven’t even touched rendering/lighting beyond the very basics, but I honestly don’t get why the default values are so ridiculous. Same goes for AO too; it’s absolutely useless with the default values. Before I caught on, I blithely assumed those default values were set for a reason, and stuck with them for way too long. Oh, samples in Cycles too – what beginner needs 4096 for a basic render? I didn’t use Cycles at all at the start because it took SO VERY LONG. Well, it doesn’t take that long with 100 samples which is usually enough for my purposes.

Yay, inspiration! Indeed, you totally don’t need subdiv – especially useful when you want to/have to keep your polycount down. Always good to have options.

This lantern looks great! Nice job translating it to 3D.

[joseph (J)]

The worst part about Bloom is that it’s only white… so it’s not strong enough, even cranked all the way up, and it’s only white even the glare node isn’t enough- it only goes from 6-9, which is an absurd range, and 9 does barely anything. It shouldn’t take two full add layers and two blur layers to get this level of Bloom.

I mean, sure, I could add volumetrics, but that seems unnecessary when Bloom is a screen space effect, there’s no reason why it’s so weak and why the max values are so low. It’s very odd.

Also, where’s your status? All the cool kids have one

[Minamookevlar (Minami)]

Yeah, Eevee takes time just to set up… I thought it was a bad renderer and didn’t touch it
And Cycles had their samples bumped up 4x for some reason, that, I was surprised

[piranha4D (Pir)]

Volumetrics seems kinda overkill, yeah.

@joseph

Also, where’s your status? All the cool kids have one

The conclusion is obvious: I’m just not cool. ; )

I did have a status pretty much from the moment I saw we could do it, but I let it expire. I’m way too tempted to put obnoxious things there because it feels a little twittery: “Sitting on the loo, working on a big one”. Ok, just for you, something more appropriate…

[joseph (J)]

Day 36 - What the heck is wrong with cryptomatte today?

I did the second lantern- this cold, which was almost gone yesterday, has completely wiped me out today, so I've been low on energy, creativity, and motivation. I still got this done, although it took quite a bit longer than yesterday (because of the cold).

If your immediate reaction is “that looks drab”, yes, it does, and I’m deeply unhappy about it- because the solution is straightforward (my compositor setup from yesterday), except for some reason it doesn’t work today- specifically, cryptomatte will not let me isolate the lantern itself no matter what I do.
See:

Object will not show up in Cryptomatte

[2022-11-04 15_56_11-Blender Render] [2022-11-04 15_56_04-Blender Render] The orange diamond lantern is not showing up in cryptomatte- I have no way to Pick it, and I very much need to. I've tried all the Crypto object layers, as well as the Crypto material layers. I've duplicated the object, applied all transforms, etc, no dice. I *know* this works fine in theory, because I did it yesterday in the same file, and I can do it again today: [2022-11-04 15_59_51-Blender Render] [2022-11-04 …

If anyone knows how to fix this, please help

I remeshed/decimated the middle loop thing, since that’s a cheap trick to get a welded look,. hence the bad topology. Everything else has perfect quad topology

Also, sidenote:

No, Grammarly, I’m not heartbroken about cryptomatte not working, I’m deeply unhappy. I will not choose a different word, thank you very much. Sometimes I swear Grammarly is trying to prank me. It’s cryptomatte, man, I’m not getting divorced

What did I learn today?

Cryptomatte is so cool when it works and so frustrating when it doesn’t. Also, I can still whip polygons into shape despite feeling like garbage. I might even have a bit of a fever right now, but that will not stop me! Suck on that, polygons!

Music

Kabalevsky, Piano Concertos - performed by Kathryn Stott. I don’t listen to them on Spotify, I use MusicBee and physical copies, but they are on Spotify if you want some excellent concertos in your life:

Kabalevsky: Colas Breugnon: Overture / Piano Concerto Nos. 2 and 3 / The...

Kabalevsky: Piano Concertos, Vol. 2

[stray]

“Heartbroken” sounds much more dramatic, as is proper for an artist =D If you want to be a real pro, you should spend a day or too staring dejectedly at the ceiling: the universe has turned on you - act like it!

*ahem *
The main lantern has a curious topo as well. A subdivided Icosphere?
I love these models, they look so nice and clean

[joseph (J)]

You have a point. I would go lie outside in the dirt and cry for a few hours like Hans Christian Andersen, but it’s cold, dark, and wet outside, and I’m very comfy in my blanket at the moment

I do have to wonder who at Grammarly HQ decided that “deeply unhappy” is less appropriate than “heartbroken”, and what their criteria are for “choose a different word”. Why should I choose a different word? It’s so ominous

Yeah, the main lantern is a subdivided icosphere. You can see some subdivision on the upper… paddle things? as well. I’m quite proud of how I was able to add the circular bolts while maintaining quad topology, it’s a simple trick and it works great.

@stray

I love these models, they look so nice and clean

Thanks, they’re fun to make. I really recommend using Marcel Hampel’s work for concept, he’s super cool about 3D artists making renditions of his stuff and he has a wide range of rocks, organics, mechanical constructs, etc.

[Minamookevlar (Minami)]

I feel like they forget that there is a yellow color in their library. It’s not wrong, so please tell me so!
Oh, premium? Uh… Uh-m… no thanks

[piranha4D (Pir)]

Sorry, I have zero knowledge about cryptomattes, so I’ll just go on about much less relevant stuff. From your description something simply has got to be different about this lantern, but your guess is gonna be better than mine.

@joseph

Sometimes I swear Grammarly is trying to prank me.

See, this is the sort of thing where I’d like to see advances in AI long before I’d like to see them in creating art. Grammarly annoyed me to no end when I gave it a try, so I don’t use it, I just haul out a thesaurus (digitally, these days) when I feel I didn’t quite get le mot juste. Natural language is legitimately a hard subject, even for humans themselves.

Good on you for doing this despite the cold. I really like these pieces; I can imagine a whole environment into which they’d fit. Oh, a question – do you ask 2D concept artists explicitly for permission before using their art, even if it’s just for practice? I need to come up with a policy for that now that I am publishing my practice models.

[joseph (J)]

@piranha4D

From your description something simply has got to be different about this lantern, but your guess is gonna be better than mine.

You’re not going to believe this, I don’t even believe it myself, but this lantern is in fact a duplicate of the one that works that I shrink wrapped I thought this might solve the problem - and it did, until I shrinkwrapped it and applied that modifier, and then it suddenly didn’t again. Yeah, beats me. Honestly I’m probably just going to make a mask by hand in Affinity (Fill tool goes hard ) and cheat. I’d love a solution eventually, though, because this is for sure not a sustainable workaround.

@piranha4D

Natural language is legitimately a hard subject, even for humans themselves.

I’ve been thinking all evening about this- and I can’t think of many use cases in which “heartbroken” would actually be better. I can be deeply unhappy that something isn’t working, I can be deeply unhappy with someone’s job performance, I can be deeply unhappy that I have a cold… but in all those cases, “annoyed” would be more accurate. I desperately want to know how the AI ended up with “heartbroken”

@piranha4D

do you ask 2D concept artists explicitly for permission before using their art, even if it’s just for practice?

Generally speaking, yes. If I see them engaging with people that have made 3D versions of their work, then I figure they’re cool with it, but if I don’t see that, I like to ask.

With that said, one caveat- most of my reference images are meticulously ripped (haha, I can say it) from a collection on Instagram by a script I wrote (and it was extraordinarily difficult, they do not want you to scrape images off there!) but I neglected to plan a way to collect metadata like “who made this” in that process. So… I don’t know who most of my concept artists are. If I do ever share anything I’ve made from one of those cases on social media, I like to reverse image search the reference first and see if can find the author.

In this specific case, I’ve been making 3D versions of Marcel’s work for quite some time now, couple years at least, and he’s always been super cool with it (he usually retweets it, which is especially nice). That’s another reason I like to recommend him as a concept artist

[piranha4D (Pir)]

@joseph

I’ve been thinking all evening about this- and I can’t think of many use cases in which “heartbroken” would actually be better.

Me neither. Some dictionaries list it as a synonym for plain “unhappy”, but I think that’s insufficiently granular; unhappy covers a vastly greater amount of ground, and “deeply” narrows that only a bit.

I don’t know how Grammarly are training their AI, and whether there’s any Deep Learning involved – maybe not, which could account for hyperbolic results. They certainly weren’t doing that when I tried it out. But that was long ago. Like the early days of Google Translate.

[asking permission]

@joseph

Generally speaking, yes. If I see them engaging with people that have made 3D versions of their work, then I figure they’re cool with it, but if I don’t see that, I like to ask.

I’ve been mostly wondering about the latter case, unless there is a general statement to that effect. Since I don’t know whether those other people asked. I think I’ll just ask.

[r-i-i-i-i-i-p]
You can say it but I wonder whether I can even quote it. Tease. ; )

Yeah, I have a huge image collection on disk, and for many of them I don’t have creator metadata – that was before I did anything with it other than look and enjoy, and images often came without name of the artists attached, and there was no reverse image search. If I used any of that to directly model from, I’d try to track down the artist. But I admit to still just taking largely generic clipart as reference, if I just use it as inspiration. Probably not the right thing to do.

:ugh: another long power outage. Don’t know what I’d do without my ebook reader and its lovely light.

[joseph (J)]

Day 37, Part 1- Yep, I did the mask by hand

Sooo… I tried the suggestion in my problem thread to not use an Emission shader. This, which worked for the person commenting, did not work for me, so I threw up my hands in disgust (figuratively, don’t go getting ideas, Grammarly!) and made a quick and dirty mask with the Magic Wand and the Fill tool in Affinity Photo:

I’m quite happy with how this one turned out, I think I like it even more than the first one. Just for the sake of completeness, here’s both:

I would actually really love to 3D print these as miniatures, get some chains and metal paint from a hardware store, and make the lanterns from tissue paper (stiffened with white glue) and LEDs. I don’t have a 3D printer, which puts a bit of a wrench in that plan, but maybe someday

What did I learn?

Cryptomatte is great when it works, and there doesn’t seem to be much you can do about it when it doesn’t.

What’s next?

I’m not sure yet- something for sure, this doesn’t really count since I did most of the work already yesterday. I’ve been mostly working on this:
https://blenderartists.org/t/adding-custom-links-custom-categories-to-the-new-sidebar/1415240/
so far today. Stay tuned for more, there’s a lot of time left in my day

[joseph (J)]

Day 38 - Don’t import Mixamo rigs!

I watched a very good YouTube video today:

10 Tips to make Your 3D Render more Cinematic

That leads me to try out some of the tips in a very simple scene using Mixamo rigs. This, I thought, would be a good idea. It was, except for the fact that apparently, Mixamo rigs are poisonous to Blender- every time I added a new light, all the other lights would stop working until I closed the file and opened it again. When I clicked Render, the poses reset themselves completely and the lights stopped working; luckily, I had an autosave, so I opened that and just did a viewport screenshot.

I have no idea what’s going on, I saw this behavior in both 3.3 and 3.4. But hey, at least I got a screenshot

Anyway, this is my result:

And this is my result pre-post-work:

And lastly, this was my reference:

I don’t really know what this picture is, I found it by chance in an old Instagram-rip folder, and honestly, given the somewhat intimate nature, I think I’m ok without more details. It really just spoke to me as a simple, character-focused scene. And it’s very relatable, my wife also steals all the blankets at night I don’t think that’s the impression I’m supposed to be getting from this picture, but oh well.

You’ll notice I tightened up the shot quite a bit in post, the composition of this scene was bothering me the whole time. The reference is perfectly centered, and there’s a lot of empty space distracting from the focal points. Also, I wanted to add more lights to mine (it needs more!) but I physically cannot, without my Mixamo friends resetting themselves and all the other lights failing, so this is the best I can do

What did I learn or improve?

I fought off the temptation to do any texturing- there’s a point in the linked YouTube video where he says something about “relying on shaders and textures” and I immediately felt called out- with good reason, those are my strong points, and I lean on them far too much. So I used no shaders at all, and focused on lighting, camera, and composition exclusively. This was really good for me, as an exercise, and I’ll be doing it again for sure. Just maybe not with Mixamo rigs

This was also really good posing practice for me. These two poses are more complicated than they look- they’re relaxed, which is a lot harder to pose than action-esque poses. Most of my posing experience is with action-oriented poses, and it was really good for me to practice this more relaxed, loose posing. It’s not perfect, of course, but I’m happy with it

Music

Piano Concertos- Bortkiewicz and Scriabin

[anon12677551]

I like to see a narrative element. Love the colours too.

[piranha4D (Pir)]

The illustrator is a May Lee (https://linktr.ee/MayLeeillust), and this seems to be part of a greater story; she draws these two people a lot. It’s called “Half Light”.
https://www.inprnt.com/gallery/maylee/half-light/

Your posing looks really excellent.

[joseph (J)]

I salute you for doing what I should have done and looked up my references,I’ll try to be less lazy so you don’t have to again

[piranha4D (Pir)]

Don’t worry about it, I do this stuff for fun (practices my search-fu), not because I feel I have to. I am nowhere near that anal about giving credit.