IJBM: Pictures with malformed hands

Stormtroper

Thus begins the AI art thread.
Ever since posting those random splotches on the image dump thread, I've been learning how to get stuff that's closer to what I have in mind:




(I can get stuff that's much more similar to the original images, but after a certain thereshold it becomes pointlessly similar (and with weird artifacts) so I like the above more).




(Original character)




A fun toy, but it gets old after a while. I want to learn what I can do to incorporate all this into my 3D work, although I'm not sure I will even if I figure out how. My first experiments didn't turn out very well, but there's lots of things to try out.
(Regardless of how optimistic I might come off as, I'm kinda conflicted about what this AI thing means or could mean for the art world. Nonetheless said world is changing and I better be prepared for it, whatever that means.)

lrdgck

A fun toy, but it gets old after a while. I want to learn what I can do to incorporate all this into my 3D work, although I'm not sure I will even if I figure out how. My first experiments didn't turn out very well, but there's lots of things to try out.
(Regardless of how optimistic I might come off as, I'm kinda conflicted about what this AI thing means or could mean for the art world. Nonetheless said world is changing and I better be prepared for it, whatever that means.)

I'm thinking you can dig up some of your old works, then narrow them down to a bunch of keywords and punch that as a description into the AI. See how close it gets to the original stuff.

Or, frankly, get a verbal description of some famous work of art and do the same.

glennmagusharvey

I'm curious how you got that first picture. Can you enter entire pictures (e.g. my avatar) as prompts?

Stormtroper

^ More or less yeah, it can (attempt to) extract information from one or several pictures (in that case, lineart and depth) and use it to guide itself while generating a picture. Both yours and mine used it heavily. See ControlNet easymodo below.
^^
Something I often try to experiment with is recreating Lunch Atop a Skyscraper. Here's another attempt (no ControlNet, I may have messed up input info, but it can still be seen in the images metadata):

Model: Dreamshaper 7
Positive Prompt: lunch atop a skyscraper, 6+ men, beam, building, construction, worker, sitting, lunchbox, eating, black and white, cityscape, city background, old photo, masterpiece, best quality
Negative Prompt: lowers,(worst quality, low quality:1.4),(bad hands, bad finger, bad fingers:1.2), mutated hands and fingers, poorly drawn face, poorly drawn hands, poorly drawn feet, poorly drawn legs, longneck, big ears, bad legs, bad anatomy, bad art
CFG: 7

Model: Dreamshaper 7
Positive Prompt: lunch atop a skyscraper, 6+ men, beam, building, construction, worker, sitting, lunchbox, eating, black and white, cityscape, city background, old photo, masterpiece, best quality
Negative Prompt: lowers,(worst quality, low quality:1.4),(bad hands, bad finger, bad fingers:1.2), mutated hands and fingers, poorly drawn face, poorly drawn hands, poorly drawn feet, poorly drawn legs, longneck, big ears, bad legs, bad anatomy, bad art
CFG: 7

Model: Dreamshaper 7
Positive Prompt: multiple people, men, beam, building, construction, workers, sitting, lunchbox, eating, lunch, altitude, cityscape, city background,, masterpiece, best quality
Negative Prompt: lowers,(worst quality, low quality:1.4),(bad hands, bad finger, bad fingers:1.2), mutated hands and fingers, poorly drawn face, poorly drawn hands, poorly drawn feet, poorly drawn legs, longneck, big ears, bad legs, bad anatomy, bad art, bad architecture
CFG: 7

Model: Dreamshaper 7
Positive Prompt: multiple people, men, beam, steel beam, construction, building, building under construction, workers, sitting, lunchbox, eating, lunch, altitude, cityscape, city background,, masterpiece, best quality
Negative Prompt: (worst quality, low quality:1.4),(bad hands, bad finger, bad fingers:1.2), mutated hands and fingers, poorly drawn face, poorly drawn hands, poorly drawn feet, poorly drawn legs, longneck, big ears, bad legs, bad anatomy, bad art, bad architecture
CFG: 7

Model: Dreamshaper 7
Positive Prompt: multiple people, men, beam, steel beam, sitting, sitting on beam, construction, building, building under construction, workers, sitting, lunchbox, eating, lunch, altitude, cityscape, city background,, masterpiece, best quality
Negative Prompt:
CFG: 11

With ControlNet easymodo:

Positive Prompt: multiple people, men, beam, steel beam, sitting, sitting on beam, construction, building, building under construction, workers, sitting, lunchbox, eating, lunch, altitude, cityscape, city background,, masterpiece, best quality
Negative Prompt: (worst quality, low quality:1.4),(bad hands, bad finger, bad fingers:1.2), mutated hands and fingers, poorly drawn face, poorly drawn hands, poorly drawn feet, poorly drawn legs, longneck, big ears, bad legs, bad anatomy, bad art, bad architecture
ControlNet 1: Depth, Balanced

Positive Prompt: multiple people, men, beam, steel beam, sitting, sitting on beam, construction, building, building under construction, workers, sitting, lunchbox, eating, lunch, altitude, cityscape, city background,, masterpiece, best quality
Negative Prompt: (worst quality, low quality:1.4),(bad hands, bad finger, bad fingers:1.2), mutated hands and fingers, poorly drawn face, poorly drawn hands, poorly drawn feet, poorly drawn legs, longneck, big ears, bad legs, bad anatomy, bad art, bad architecture
ControlNet 1: Depth, Balanced
ControlNet 2: OpenPose, Balanced

Positive Prompt: multiple people, men, beam, steel beam, sitting, sitting on beam, construction, building, building under construction, workers, sitting, lunchbox, eating, lunch, altitude, cityscape, city background,, masterpiece, best quality
Negative Prompt: (worst quality, low quality:1.4),(bad hands, bad finger, bad fingers:1.2), mutated hands and fingers, poorly drawn face, poorly drawn hands, poorly drawn feet, poorly drawn legs, longneck, big ears, bad legs, bad anatomy, bad art, bad architecture
ControlNet 1: Depth, Balanced
ControlNet 2: OpenPose, Balanced
ControlNet 3: Canny, Balanced, actually no because it wasn't enabled

(After stopping to think why there were workers off-camera)

Positive Prompt: multiple people, men, beam, steel beam, sitting, sitting on beam, construction, building, building under construction, workers, sitting, lunchbox, eating, lunch, altitude, cityscape, city background,, masterpiece, best quality
Negative Prompt: (worst quality, low quality:1.4),(bad hands, bad finger, bad fingers:1.2), mutated hands and fingers, poorly drawn face, poorly drawn hands, poorly drawn feet, poorly drawn legs, longneck, big ears, bad legs, bad anatomy, bad art, bad architecture
ControlNet 1: Depth, Balanced
ControlNet 2: OpenPose, Balanced
ControlNet 3: Canny, Balanced, actually no because it wasn't enabled

Positive Prompt: multiple people, men, beam, steel beam, sitting, sitting on beam, construction, building, building under construction, workers, sitting, lunchbox, eating, lunch, altitude, cityscape, city background,, masterpiece, best quality
Negative Prompt: (worst quality, low quality:1.4),(bad hands, bad finger, bad fingers:1.2), mutated hands and fingers, poorly drawn face, poorly drawn hands, poorly drawn feet, poorly drawn legs, longneck, big ears, bad legs, bad anatomy, bad art, bad architecture
ControlNet 1: Depth, Balanced
ControlNet 2: OpenPose, Balanced
ControlNet 3: Canny, Balanced, actually enabled this time

With better prompts for non-drawn stuff, not that it helped much:

Positive Prompt: multiple people, men, beam, steel beam, sitting, sitting on beam, construction, building, building under construction, workers, sitting, lunchbox, eating, lunch, altitude, cityscape, city background,, masterpiece, best quality
Negative Prompt: (worst quality, low quality:1.4),(bad hands, bad finger, bad fingers:1.2), mutated hands and fingers, poorly drawn face, poorly drawn hands, poorly drawn feet, poorly drawn legs, longneck, big ears, bad legs, bad anatomy, bad art, bad architecture
ControlNet 1: Depth, Balanced
ControlNet 2: OpenPose, Balanced
ControlNet 3: Canny, Balanced

What ControlNet "sees" looks like this:




In my old PC I had several pictures of workers having their tea time atop tower-high scrap heaps, but for some reason they didn't show up this time.

glennmagusharvey

It seems to produced some rather cursed details.

Tell it to hide it by using an impressionist style or something.

Stormtroper

Facial integrity not guaranteed.

Impressionism:

Model: impressionismOil_sd15
Positive Prompt: , painting, multiple people, men, faceless, steel beam, sitting on beam, construction, building, workers, lunchbox, eating, lunch, altitude, cityscape, city background,, masterpiece, best quality
Negative Prompt: (worst quality, low quality:1.4),(bad hands, bad finger, bad fingers:1.2), mutated hands and fingers, bad face, bad hands, bad feet, longneck, big ears, bad legs, bad anatomy, bad art, bad architecture
CFG: 7
ControlNet 1: Depth, My prompt is more important
ControlNet 2: OpenPose, My prompt is more important
ControlNet 3: Canny, My prompt is more important

Art Nouveau:

Model: dreamshaper_7
Positive Prompt: , , multiple people, men, faceless, steel beam, sitting on beam, construction, building, workers, lunchbox, eating, lunch, altitude, cityscape, city background,, masterpiece, best quality
Negative Prompt: (worst quality, low quality:1.4),(bad hands, bad finger, bad fingers:1.2), mutated hands and fingers, bad face, bad hands, bad feet, longneck, big ears, bad legs, bad anatomy, bad art, bad architecture
CFG: 7
ControlNet 1: Depth, My prompt is more important
ControlNet 2: OpenPose, My prompt is more important
ControlNet 3: Canny, My prompt is more important

Art Nouveau but without the flat illustration prompt:

Model: dreamshaper_7
Positive Prompt: , , multiple people, men, faceless, steel beam, sitting on beam, construction, building, workers, lunchbox, eating, lunch, altitude, cityscape, city background,, masterpiece, best quality
Negative Prompt: (worst quality, low quality:1.4),(bad hands, bad finger, bad fingers:1.2), mutated hands and fingers, bad face, bad hands, bad feet, longneck, big ears, bad legs, bad anatomy, bad art, bad architecture
CFG: 7
ControlNet 1: Depth, My prompt is more important
ControlNet 2: OpenPose, My prompt is more important
ControlNet 3: Canny, My prompt is more important

Art-Deco:

Model: yabalCartoonMix_v1
Positive Prompt: , multiple people, men, faceless, stylized, poster, steel beam, sitting on beam, construction, building, workers, lunchbox, eating, lunch, altitude, cityscape, city background,, masterpiece, best quality
Negative Prompt: (worst quality, low quality:1.4),(bad hands, bad finger, bad fingers:1.2), mutated hands and fingers, bad face, bad hands, bad feet, longneck, big ears, bad legs, bad anatomy, bad art, bad architecture
CFG: 7
ControlNet 1: Depth, My prompt is more important
ControlNet 2: OpenPose, My prompt is more important
ControlNet 3: Canny, My prompt is more important

fourteenwings

I didn't know this thread had actually kicked off.

Honestly I like gacek's avatar more than mine.

Great that we started right after I deleted my NovelAI and Holara archives.

I'll post some of my work when I'm not on mobile.

Stormtroper

I only just now realized the generated gck avatar has two noses. I blame my monitor.

Great that we started right after I deleted my NovelAI and Holara archives.

Oops.


So, experimenting.
Initial 3D base (not AI):




Forgot to disable ControlNet preprocessing (i.e. it tries to figure out depth/normal data from the depth/normal maps instead of using those maps' data directly).

Model: YabalCartoonMix_v1
Positive Prompt: Cave, tunnel, lamp lights, railway, rock walls, masterpiece, environment, dark, best quality
Negative Prompt: bad quality, worst quality, bad art,
CFG Scale: 7
ControlNet 1: Reference, My prompt is more important
ControlNet 2: Depth, ControlNet is more important
ControlNet3: Normal, Balanced

No ControlNet preprocessing, otherwise same setup
:

Same but with dreamshaper_7 model:


Same but all ControlNets set as "My prompt is more important"


Same but all ControlNets set as "ControlNet is more important"


Bunch of different settings:

Positive Prompt: Cave, tunnel, lamp lights, railway, wooden planks, rock walls, masterpiece, good scenery, ultra detailed, dark, cave background, best quality
Negative Prompt: bad quality, worst quality, bad art, bad lighting,
CFG Scale: 7
ControlNet1: Reference, My prompt is more important
ControlNet2: Depth, My prompt is more important
ControlNet3: Normal, My prompt is more important

After realizing I had the incorrect normal map set, otherwise same setup as earlier:


Same but made with human love:

Made by an actual breathing, living being after several hours of painstaking human work, with lots of love and sovl:

Same but AI and using the previous scene as ControlNet inputs:

Model: Feiwu_v10SlightFlat
Positive Prompt: Cave, tunnel, lamp lights, railway, wooden planks, rock walls, masterpiece, good scenery, ultra detailed, dark, cave background, best quality
Negative Prompt: bad quality, worst quality, bad art, bad lighting,
CFG Scale: 7
ControlNet1: Reference, My prompt is more important
ControlNet2: Normal, My prompt is more important
ControlNet3: Depth, My prompt is more important

Same but with the right model:

Model: dreamshaper_7
Positive Prompt: Cave, tunnel, lamp lights, railway, wooden planks, rock walls, masterpiece, good scenery, ultra detailed, dark, cave background, best quality
Negative Prompt: bad quality, worst quality, bad art, bad lighting,
CFG Scale: 7
ControlNet1: Reference, My prompt is more important
ControlNet2: Normal, My prompt is more important
ControlNet3: Depth, My prompt is more important

Same but icy:

Positive Prompt: ice cave, ice, cave, frozen, frost, glaciers, cold, masterpiece, good scenery, ultra detailed, best quality
Negative Prompt: bad quality, worst quality, bad art, bad lighting,
CFG Scale: 7
ControlNet1: Reference, My prompt is more important
ControlNet2: Normal, My prompt is more important
ControlNet3: Depth, My prompt is more important

Same but fiery:

Model: dreamshaper_7
Positive Prompt: fire cave, fire, cave, lava, magma, smoke, burning, red hot, masterpiece, good scenery, ultra detailed, best quality
Negative Prompt: bad quality, worst quality, bad art, bad lighting,
CFG Scale: 7
ControlNet1: Reference, My prompt is more important
ControlNet2: Normal, My prompt is more important
ControlNet3: Depth, My prompt is more important

Regular cave but ControlNet is more important:

Model: dreamshaper_7
Positive Prompt: Cave, tunnel, lamp lights, railway, wooden planks, rock walls, masterpiece, good scenery, ultra detailed, dark, cave background, best quality
Negative Prompt: bad quality, worst quality, bad art, bad lighting,
CFG Scale: 7
ControlNet1: Reference, ControlNet is more important
ControlNet2: Normal, ControlNet is more important
ControlNet3: Depth, ControlNet is more important

I was hoping I wouldn't gain much from passing it through SD, but that's a lot of detail I wouldn't get in a timely manner otherwise, let alone the 2D-drawing stuff. Oh well, I still like mine more.

fourteenwings

Pictures with malformed hands

Well, if you want thematic.



This came out so well except specifically for all appendages. Seriously annoying.

A while ago I thought about making AI Album Cover Art to go with various Prima J songs (they were just on my mind) so we got:

Song Prompt: Tame




Overboard with those sunglasses there, Jessica.



Song Prompt: Corazon




Song Prompt: Flip the Script





Random other impressionist ideas for various Prima J songs





I had ideas for other, non Prima-J related songs too.

Song Prompt: Never Loved (Like This Before)



Song Prompt: Show Me What I'm Looking For

This one was hard to get right, and in the end I just went with what looked best.



Song Prompt: World Famous



You've probably noted that none of these songs are Japanese, but this was just what I felt like at the time.

My AI "art" is not usually this eclectic and unfocused, but I'm afraid when we get to the focused stuff it'll become extremely samey and make me seem kind of obsessed.

Stormtroper

I really like that semi-flat shading it uses.
Also I wonder how it figured out each lens of glasses have to reflect a similar picture.
Maybe those hands can be fixed with inpainting?

Made by an actual breathing, living being after several hours of painstaking human work, with lots of love and sovl:

Wouldn't it be poignant if something like "made with love" became a commonly used prompt?
IJBM: Despite all those advances in automation, there's still no one-click solution to the tedium of retopologizing.

fourteenwings

Also I wonder how it figured out each lens of glasses have to reflect a similar picture.

I'm sure that at first AI was very bad at that, but this was one of the things that was fixed long before I could get into it.

Maybe those hands can be fixed with inpainting?

I feel like inpainting makes things like this worse, but maybe it's because I haven't really tried hard enough to make it work.

Wouldn't it be poignant if something like "made with love" became a commonly used prompt?

I'm going to start using it and see what it does. Probably something annoying.

A while ago I thought to myself "What if Frozen starred two brothers instead?" which led to what ended up looking like a modern Chinese blockbuster film instead (to be fair, I did change the races).



As tended to happen with early versions of Stable, it literally created a whole character I didn't ask for, but maybe that's what would make the movie intriguing. Actually, Stable still does this (I think it might have to do with "and" and plural "s") but it rarely changes the sex of the characters anymore (early versions were very female-biased anyways so it was mostly that).

EDIT: It seems the original version ran afoul of imgur's rules but I bet myself that if I stuck a similar prompt into an emulated Stable 1.5 (since I no longer have that version installed) it would spit out something similar.

It did! But it was kind of ehh, so I stuck that result in XL0.9 which gave me:





This version captured a lot of cool stuff but completely missed the "Ice Princes" part.

The next version is vaguely NSFW; this one at least captured the fact that the main characters of Frozen have different hair colors. I like the brunet's outfit a lot.

I decided to remake the prompt (as far as I could remember it) with Stable XL.





Sadly, their outfits became more samey, whereas earlier versions were more free. In addition, it's hard to get the glassy CG look with your characters now unless you use the right LoRA (and I can't use LoRAs much at all).

The backgrounds came out much more in line with what I initially wanted though.

Whilst doing my remake, I also got this, which I thought was extremely pretty for Stable 1.5.



Ran it through XL obviously and it ended up looking much nicer, but less thematic.



I think the guy is way too pretty.

fourteenwings

I was trying to create a very, very specific image, which led me to some CG and cutesy models.

I didn't get what I wanted, but I did get myself a new obsession!






In funnier, more "why AI why" stories, XL1.0 is really really bad at "and", and I think it just wasn't as "Stable" as you'd think as tends to happen right after a "Look, it's a new Stable!!!" announcement.



There was a slight update (as tends to happen) so hopefully it's better now.

Stormtroper

From what I understand, training for Stable Diffusion models (dunno about the others) literally do not attempt to make it understand phrases, only tags, thus "X and Y" does not mean anything regarding X or Y except inasmuch as its training dataset may include the phrase as a tag (e.g. if it's tagged as "black and white" or "Tom and Jerry".
So I tried Dream Textures (Blender's Stable Diffusion integration) and, oh my, it's way more than I expected.

Apparently there's no way to generate normal/displacement/roughness/whatever maps in accordance to what you get, which severely limits how useful it is.

I'm not quite sure how to make texture projection work automatically but I can sometimes manually tweak it to make it look like how's intended.

Just as I start to think there isn't much to this Dream Textures thing, I find out it has an entire node system.

One of the demo scenes for Dream Textures (based on this benchmark scene).



(Same prompts as before, although other settings differ.)
For the previous tests I had spent some time looking for a one-click solution to the issue of setting up ControlNet's preprocessed images, to no avail. Turns out this is what I was looking for (and it doesn't seem that this is well known).
More tests to come.

Stormtroper

Despite my expectations, there doesn't seem to be much I could be doing to change how the scene looks in a controlled way. Well, after inputting this I did get something:

Now with a different scene. Base (not AI):





I can't seem to get rid of that super artificial looking "AI clay" shading, though I did like when it decided to give me a different view of the town altogether (and I can't seem to replicate it).

glennmagusharvey

Did you tell it to fill the outsides with relevant stuff? As opposed to just blank space, or grass, or whatever.

Stormtroper

Nope, it did that on its own. (Note: the base image there was not the final image, the first AI image used a base image with a solid color background, hence why one has a sky instead.)

fourteenwings

Rainbow Rave Tunnel

Pride Month is over man.

I like the water trees.

fourteenwings

Abercrombie & Yuichi






I wanted a whole second set with rooftop pictures but XL was not in the mood, even this last one has a few face issues.



Obligatory Malformed Hands



In a whole other direction, CG Beach!

This first one is my favorite from the set;


The rest.



CG tank tops are very hard to get right;


This guy came out much younger than the average, and also don't look at his fingers.

Stormtroper

The hands thing isn't unexpected but I find it odd that it seems to have trouble building their arms. Regardless, those backgrounds are great.
Most nuanced take I've seen on the whole AI art thing:

(The usual "I don't agree with a lot of it" (mainly the parts that ignore that Stable Diffusion is FOSS and most sentences with the word "corporation" or "corporate" in it) but I thought it was good.)
Also, what's with the association between NFTs and AI art, they have basically nothing to do with each other besides being the latest technology thing.

fourteenwings

I will watch that when I have the time, but probably not soon.

Anyways, AberYui dump;












Significant error section;

Stormtroper

That AI has interesting ideas on what should be seen through/reflected by the windows.
I found out I've been making a mistake by not increasing the samples amount up from the default 20. I tried it once and it only got me images that were "better but not smarter" so I didn't bother until recently.
More experimenting with different models and ControlNet settings. Base (frame from a failed attempt at getting an animated sea of clouds that could be rendered in a reasonable amount of time):




Positive prompt: cloud, sky, white cloud, scenery, masterful, best quality
Negative prompt: bad art, bad environment, worst art, bad quality, ugly










It amuses me that there are two incorrect Miyazaki quotes from that documentary, the fake "anime was a mistake" one and the out-of-context "this is an insult to life" one.

Stormtroper

This post is less about experiments and more about just having fun with SD:

ControlNet reference (from this (edit: I mean this), varying whether it's set to "Balanced", "My prompt is more important" and "ControlNet is more important":


Model: Varying between AnythingV5ink_ink and Feiwuv10_Slightflat

Positive prompt: Several combinations of the following;, aqua shirt, red skirt, aqua skirt, doll dress, doll clothes, frills, frilly dress, ribbons, red ribbons, hidden hand, hand behind skirt, extended arm, big eyes, bug eyes, fat face, crooked smile, off-centered smile, small hat, yellow hat, round hat, short hair, pink hair, yellow shoes, holding skirt, tube legs, full body, masterpiece, best quality"
Negative prompt: knees, ankles,(worst quality, low quality:1.4),(sfw:1.05),(artist name:1.1), (censored:1.1), (mosaic:1.1), (bar censor:1.1), (underwear:1.1), (bra:1.1), (greyscale:1.05) tiling, text title signature






Despite what the thread title might imply, getting bad hands without specifically prompting for it turned out to be surprisingly difficult.

lrdgck

"girl with a sausage", "girl with a third leg", damn, my mind is dirty.

Stormtroper

I thought this was interesting:
https://sambleckley.com/writing/dont-fire-your-illustrator.html
This moreso:
https://www.reddit.com/r/StableDiffusion/comments/15q324o/concept_art_timelapse_5_hrs_in_5_min/
It's more or less what I've wanted to achieve so I'm pleased to find it in working form. I wonder why it took so long for someone to put in practice the obvious idea, or why it took me so long to find someone who did.
No images this time, I've been kinda busy with other stuff and will still be for some more time, but hey, gotta keep posting stuff on IJBM somehow.
Also, I'm kinda missing doing normal 3D work.

fourteenwings

Tried a new model!



I was going for creating some sort of "Strawberry Shortcake, but Boys" characters and they did come out okay actually. Except it took forever. Details below.

Anyways I like them a lot, Blueberry Cola, Strawberry Fizz and Lemon Sorbet. Sorbet was super hard to get right. He kept coming out in a dress.



Meanwhile, Strawberry Fizz somehow got an eldrich monster friend at one point.



Cola was easy to deal with, all he got was some sort of candy rocket weapon.



Seriously though; Sorbet gave me the hardest of times.


Two of him (and a dress).



"Please 14w, I don't want to smile..."

Stormtroper

I thought about posting this on the Touhou or videos threads but I haven't posted here in a while so I thought I would (Edit: Seizure warning, maybe):

glennmagusharvey

oh no is this what i think it is

yes it is

lrdgck

I have been tricked! Shame on me.

By the way, this reminds me of an AI-generated music video I have seen a while ago. Here's the link, because it begins vaguely NSFW.

Stormtroper

I have been tricked! Shame on me.

idgi, but I choose to believe I tricked you into watching an anime game thing.

lrdgck

Don't take it the wrong way man, but yeah, it kind of felt like I let my guard down and got anime-rickrolled. Animerolled.

lrdgck

Also, may I have a request? Do a Pope John Paul II by Frank Frazetta, please.