ThereIsNoJustice's Sketchbook

[Leo Ki]

Coding jobs are all going to be taken by the AIs anyway, whereas arts will remain... oh... wait.... No, seriously, the fight is worth fighting. Are you considering ramping up expressivity and stylization, even within booba commissions? (I didn't know there was such a market.)

[dimensional-knight]

That sucks. From anecdotes I know there's still money to be made from coding, but it's more specialized work thus requiring investing more into it, and it takes luck to be in the right place when opportunity crops up.

Nice work, you seem to be humming along in the anatomy studies fine. But you seem to have hit a snag with that foreshortened pose you drew four times, right? It's likely the waist that's throwing you off, you're making it narrower and longer than in the other studies. The bottom edge of the ribcage too forward for the pelvis angle, and instead of taking the waist to the iliac crest you're placing a lot of emphasis on the buttocks, so it's flattening it a bit and creating a wasp-like figure. It absolutely can be a stylistic choice, but it doesn't seem to be since you put more of the crest into the other poses.




Here's a reference. Her stomach is absolutely flat due the strain of balancing on her hands, but it's a 3d scan you can turn around. Check out her pelvis bone and where the ribcage ends, how it still aligns to the pelvis. Depending on height, on women the crest can be a couple of fingers away from the bottom of the ribs when simply standing.

[ThereIsNoJustice]

Darktiste, good points about studying from other artists. I'll throw that on the pile of stuff to study.

Coding jobs are all going to be taken by the AIs anyway, whereas arts will remain... oh... wait.... No, seriously, the fight is worth fighting. Are you considering ramping up expressivity and stylization, even within booba commissions? (I didn't know there was such a market.)

I think the key to the booba market is finding a more idealized figure and doing variations on it. Speaking of AI, that's definitely something I've been learning from it. When perverts people are iterating over the stylized figures quickly and without much limitation, it's a lot easier to see what is appealing. Needless to say, I have been rather tame.

Also, yeah, I would put a lot of well known artists out there in the booba/nsfw market. Reiq, sakimichan, those kinda folks.

But you seem to have hit a snag with that foreshortened pose you drew four times, right? It's likely the waist that's throwing you off, you're making it narrower and longer than in the other studies. The bottom edge of the ribcage too forward for the pelvis angle, and instead of taking the waist to the iliac crest you're placing a lot of emphasis on the buttocks, so it's flattening it a bit and creating a wasp-like figure. It absolutely can be a stylistic choice, but it doesn't seem to be since you put more of the crest into the other poses.

That's me trying to push for more dynamic poses. For the most part, I have avoided poses with twisting torsos and such. I felt like the last figure on the right came out mostly ok, but looking at it now, it feels like it's missing detail around the hip area. Probably it is just pushed too far.

I think I used this same 3D skeleton for that past pose. Anyway, I'm going to continue w/ trying to find a balance in too much/too little in dynamic poses, so expect plenty more strange looking figures in that regard. :D

[Leo Ki]

Glad you found a way to use the AI galore as a marketing survey!
So, I looked up the artists you mentioned. It looks pretty codified but there might be ways to shake up the code and break through?

(Knight, the link you posted to the contorsionist's 3d model is both fascinating and painful to watch.)

[dimensional-knight]

Ah, I see that. You nailed the 3D reference! But the bad news is that this one isn't a very good ref.

Sometimes 3d artists get things wrong. It's dangerous to use 3d as reference without double-checking if it's correct. I don't think the thoracic spine should be able to bend like this. You have a sort of hinge where it meets the lumbar spine, then it's gets limited in the forward/backwards axis, otherwise you'd yeet your ribcage out like in this ref. And another important issue is that this is a male skeleton.

This ribcage is large and wide at the bottom; It narrows in women, with the ^ at the bottom creating a smaller angle. The pelvis is a man's narrow pelvis, it's not only narrower in width relative to their bodies, but check the bottom ^. Women's pelvis look more like a round butterfly. These two structures combined are what creates a waist, or the lack of one. That's why your figure's waist ended up being off; the ref is doing you no favors.

[ThereIsNoJustice]

Knight, good points about the skeleton. I'm always wary of 3D ref and in this case you're right it would need to be adapted a lot to make it look appropriately feminine.

Anyway, continuing to study the neck, torso anatomy, and simplification lately. I'm going to start doing nicer lines and some more value studies. I feel like doing quick sketches is great but ultimately it's important to put the time in refining the sketches as well.

[ThereIsNoJustice]

Some more general anatomy stuff, eyes, feet... Currently in a weird spot where I feel like I should grind studies on anatomy to at least refresh my memory on all the various parts, before I go on to try to do any more finished works.

Also, aiming toward DnD and MMO type commissions would widen the net I'm casting, so I've got a thought to start studying all those fantasy type clothing after the anatomy pile.

I've got a pretty great grayscale to color workflow process figured out, and I need to write a little tutorial on it soon. I've never seen anyone else do it, but it's the only grayscale to color process I've figured out that means I don't have to paint the colors and values twice. Anyway, if you're reading this, you'll see when I get around to posting it.

[Leo Ki]

The RPG scene has a great many varied settings that can power the imagination, good decision I think.
I've never been a fan of value to color, I find it too analytical, but I'm interested in your new method if it is more intuitive.

[dimensional-knight]

Your style would fit MMOs and Applibot-style commissions very well! For MMOs it's a mixed bag, but for cards you'll need to eventually focus on color, composition and a clean rendering style because image need some quick readability to have narrative power. Your lines would work well with these elements.

So, for the anatomy studies, your torsos are awesome. There's a minor detail about the women's, something that flies under most male artists radars — let's talk about breasts.

You're depicting them very well. They react to gravity in a natural way, they point in the right direction, everything is good. It's just that for natural breasts as opposed to implants, most of the time their outer edges won't follow the pectoralis up to the armpit, it'll go in a bit of a horizontal line under armpits instead, fading into the ribcage.


(source)

They'll get flattened when raising the arms, then having more the appearance of climbing up to armpits with the muscle but still more to the side of it.


(source)

Like I said, it's a minor detail. Stylistically you can ignore it if you want, but getting it right will set you apart from 90% of male artists. xD

For the heads, they're almost there. You're making the farther eye in the 3/4 a tad too long, ears and the back of the skull are a tad too small (but look better with hair). How do you feel about the construction method?

The feet are also nice, though you're making them too tubular at times. They'd be more triangular with a carved out arch on the underside instead. Because it's not articulated at the top the surface only changes slightly when standing on the toes. The dorso can't curve, changes happing due flexing and contracting tendons and the lean muscles there, and it's the change of direction in relationship to the lower legs and toes which makes it look more or less curved.

[ThereIsNoJustice]

Well, I bounced off the tutorial idea. It's a lot of work for something that most people will probably skim over once and then forget. I'll try to explain it in text, though.

The process works like the following. The 'trick' of it is creating a folder and essentially making my own blend modes. Do that for different lights and shadows, and that will take a painting 90% of the way to finished. Then paint over the top.

• Linework
• Grayscale flats below linework (pick flats in the range of 30-70 value so there's room for going lighter and darker). It's even better to separate the different materials to different flat layers, as that will come in handy later for selections.
• Create folder w/ passthrough and mask, put brightness/contrast adjustment layer inside.
• Paint on the mask of the folder to create light or shadow.
• Create new folders for new lights and shadows.
• Repeat until grayscale is complete.

Now to add color, modify the grayscale flats to color, and then add color balance adjustment layers to the folders.
• Change the grayscale flats layer to colors.
• Add color variation on a layer above the flats, if you'd like.
• Add a color balance adjustment layer in each folder to change the color of each light/shadow.
• Once happy with all the adjustment layers, create a normal layer at the top and paint over the top until done.

Sometimes you need to make a certain surface look different. For example, more shiny/contrasty. So in that case make a new folder and put a brightness/contrast adjustment in there, and use the mask to paint in the boosted contrast.

Anyway, I've battle tested this process, and I've tried to use all the other grayscale to color methods and they all tend to fail when things get complicated. This one just gets the job done with less fuss, in my experience. (I'd show a couple paintings with it but they're nsfw, so I figure I'll keep that kinda thing off the forum.)


Recent study stuff. I was trying different hair methods through all these, doing more anime style faces (redrew them about 10 times each, the proportions always feel wrong somehow) and interpreting reference. Also a boob painting as dimensional knight suggested paying closer attention to that connection to the torso.

DK, you asked about the construction method. It's great but it's also easy to draw something with form, but the proportions are off, etc. A lot of that for me is drawing from imagination too early. With 3D models and AI now, there's not much excuse... I should always be using ref of some kind.

[cgmythology]

Really strong figure works with your latest drawings, the forms are very well done. Keep it up!

[Leo Ki]

Thank you for your process for value to color, I thought I was understanding it until I bumped into the last step: "Paint over the top." With all the adjustments underneath, do you just pick the resulting color and paint in it, or is there more magic going on? The last layer is a normal layer, so it's not like painting just the texture and have it get colored by what's beneath, right?

BTW I think NSFW is OK here, or you can post just fragments of the images that illustrate your technique.

I'm in a drastic layer-reduction phase right now and I might end up going back to paper to be honest, but I'm always on the lookout for layer wizardry.

Hair is hell to get the right amount of detail for the style you're after. On the first page, anime-like, you might have too much detail. If you ink them simpler then add some detail with the coloring, it might be more like it? The long hair on the second page is awesome, but now with the mention of AI on it I'm wondering if it's from the machine, lol. I'm not sure what the "AI development" arrows mean.

[ThereIsNoJustice]

CGM: Thanks!

Leo: The 'paint over the top' for me is mostly color picking and fixing any areas that look weird still. Or adding details where it's just easier and faster to color pick and paint here and there. If you try it once or twice, just do whichever is faster and easier; you can try to get the adjustment layer folders' masks really close and do very little paintover at the end, or you can be a little sloppy and spend more time in the paintover stage. The reason to try to do it as much as possible with the folder masks is that you can change your mind at any point, of course, and change the colors.

Having a big layer stack is annoying to work with, but for me it's balanced out by giving me confidence in the workflow. Your mileage may vary. I know a lot of people out there use very few layers or blend modes and do some amazing stuff.

The hair stuff... The first page was trying different ideas and most of them weren't what I was looking for, but the second page things started to make more sense. The drawing to the right of the painting is a study I did off an AI/Stable Diffusion image. But that's all my drawing, no actual AI in it.

The two drawings with arrows pointing at them are Stable Diffusion interpretations of the drawing above.

I will be sure to mark the AI stuff more clearly as I definitely intend to use it more. I know people are rightfully unhappy about AI and I've certainly been part of a bunch of discussions along those lines. However, if everyone else in the world who's not an artist is going to take advantage of this tech, then you can bet I will too. I won't be taking credit for the stuff the AI generates, but I will be using it to create reference and variations of my own works.

I actually had a conversation about this earlier today. I was trying to convince someone that AI was useful as a reference and learning tool. With his permission, I took one of his quick sketches and Img2Img it into eight more refined drawings, picked the best out of that bunch, turned it into a 'controlnet' image, and then turned that into eight rendered illustrations. Somehow, in his opinion, that wasn't all that useful. To go from a five minute sketch to having eight possible finished looking paintings in a few more minutes. Seems useful to me.

Here's what I mean. Here are a couple of my old quick sketches that I used controlnet on. And keep in mind, you can generate a lot of variations. These are the ones I liked the best.

If you don't want to use any AI stuff because you hate it too much, I can honestly respect that. But thinking this isn't useful? C'mon.

[Berebobe2]

Where are you getting your commissions from? I see there are artists that advertise on reddit but I dont know if they are actually making money there or if there's a better site?

[Leo Ki]

Ha, got it about the last layer, thanks! I also understand the advantage of the many layers, it's remotely similar to being able to change the lighting and other aspects of a 3D scene without having to touch the whole model.

The reason why I'm inching toward a one-layer world is mostly because I'm currently going back in time to reconnect to my fundamentals. For the same reason, I'm not interested in generative networks at the moment although from a scientific point of view they are intriguing indeed, and I envision the next phase when the generators manipulate 3D models before rendering.

So for the "behind up" image, did the generator/controlnet create the smoother shading and the highkey-ness of the hair, or did you do a manual pass over it? Pretty convincing result either way!

The examples in your next post are a mix of convincing and creepy especially the more realistic first one which might be difficult to fix. I do see the interest the technique can bring though. Thank you for sharing!

[ThereIsNoJustice]

Berebobe2: I mostly got commissions just from having people find my work, whether that's on a site like DA or posting it on Discord servers. I haven't tried to advertise yet. But I'd recommend checking out The Art Mentor on youtube. As grating as I find the guy in terms of presentation, he seems to have a lot of good advice for working with non-corporate clients.



Leo Ki: The AI images below are all AI, in that I didn't paint over them. Also, just to be clear, I don't really have any intention to ever paint over AI generated stuff and try to fix it. At least, not right now. You can see what I'm trying to do w/ the workflow below.

---

I've noticed that when I do loose sketches and then go to line art, it's harder to see mistakes. Like my eyes will adjust to mistakes or bad proportions. So I've been trying to go straight to finished lines, and if things look wrong, I just erase an area entirely and start over.









And here's something I started as a quick sketch, just a joke about fantasy armor, which I'm turning into a finished painting. Once I got to the flats, I put it through Stable Diffusion Img2Img to generate some references. These are good for getting some color and lighting suggestions. Then down below I just do the workflow I described before. This time I jumped straight to color since I have the AI references.

Not quite at the point where I can start doing the paint over the top but pretty close.

[Leo Ki]

Yes, loose sketches are double-edged: Great for exploration and dynamics, but drowning the final stroke in a soup. In your direct-ink ten minute drawings there is often something slightly off about the heads and how they connect to the neck and trunk. I can't quite pinpoint it right now, I'll try some drawovers when it pops.

It's funny how the AI dragged your input into what I presume are more conventional attitudes it's been conditioned with. I guess you let the "take some liberty" slider far from zero. Can you control the lighting or did it infer it from your input?

[ThereIsNoJustice]

Re: the heads and necks, I can definitely see something weird with the #1 head in the timed drawings. I'll keep an eye out for that.

As for the AI img2img, there is quite a bit of control, because you have a chance to describe the lighting in text and also use the input image. I could have used flat light and shadow colors on the character and it would have generated something to match that lighting scenario.

In this case, I just used the flats with no real lighting information. The text description I've been using lately is only "cinematic lighting" and that seems to get the classic warm light, cool shadow look without even needing to put something more descriptive in the prompt. But you could specify different colored lights, rim light, backlit, etc.

Not really related, but you can also try to take a mediocre reference and see if the AI can make it more interesting. There are a lot of things I see on pinterest where I think, "That's almost an interesting reference, but not quite..." and this is maybe a helpful tool there. Like what are some options for using some reference more creatively -- what else could this reference have looked like? Just change it around a bit. That's pretty cool.

[Bookend]

Wowoww!! Love your sketches and finished works, nice stuff for sure!

For the latest, really, really good, only crit I have is about the middle lady-- now, these do seem purposely exaggerated in places, but her legs below her knees look cartoonishly thin, and I'm not sure if that's what you were going for :D

But beyond that, awesome work, loving the lighting and rendering you've been doing :D Keep it up!

[ThereIsNoJustice]

Bookend: Good catch w/ the shin areas. I buffed them a bit for the final image. I'm trying to find a balance with making the perspective make the legs and feet noticeably smaller. I'll probably leave it as is.

The whole thing took 15hrs at this point, which is probably too much for what is ultimately a shitpost. But fun to do anyway.

My goal for the next painting I do is to have a nice background. I couldn't justify working much longer on this one.

Also I've seen these robot models on pinterest a lot and started saving them. I learned about them from watching Reiq's youtube vids. They're pretty fun to draw, and here I was testing out a new line brush.