• Sign up or login, and you'll have full access to opportunities of forum.

Art by Riodoro

Go to CruxDreams.com
Riodoro said:
While I'm also working on my other stuff, I still have a lot of fun playing around with Fooocus. So here's another new AI pic.
View attachment 1527963
It shows the point of view of a scrivener during a witch trial (I've always found the fantasy of being a passive observer recording everything that happens during a torture session enticing). The suspected witch has failed to answer the Witchfinder's questions to his satisfaction so he has ordered the torturer to undress her. Next she will be put on the dreaded Witch's Throne...
Click to expand...
Great post. There’s so much to take in viewing that pic.❤️
 
Riodoro said:
While I'm also working on my other stuff, I still have a lot of fun playing around with Fooocus. So here's another new AI pic.
View attachment 1527963
It shows the point of view of a scrivener during a witch trial (I've always found the fantasy of being a passive observer recording everything that happens during a torture session enticing). The suspected witch has failed to answer the Witchfinder's questions to his satisfaction so he has ordered the torturer to undress her. Next she will be put on the dreaded Witch's Throne...
Click to expand...
Very nicely done, and I love the chair. :devil2:
Thanks for sharing.
 
Riodoro said:
While I'm also working on my other stuff, I still have a lot of fun playing around with Fooocus. So here's another new AI pic.
View attachment 1527963
It shows the point of view of a scrivener during a witch trial (I've always found the fantasy of being a passive observer recording everything that happens during a torture session enticing). The suspected witch has failed to answer the Witchfinder's questions to his satisfaction so he has ordered the torturer to undress her. Next she will be put on the dreaded Witch's Throne...
Click to expand...
A nicely crisp image, shackles seems to be something that challenges AI but at least the wrist cuffs are connected.

She has a good facial expression

The witchfinder generals hat has an oddness in its peak

Overall I quite like it, Miss loinclothslave would take her place
 
supplice said:
The girl is cute, the "throne" is interesting, but where did that Moore-ish executioner come from?
Click to expand...
Well, it's a good question ;) The promt was something like "medieval executioner with a leather hood" without specifying ethnicity so the AI gave me a wide variation. Most (I'd say more than 90%) were obvious rubbish like these:
botched torturer.jpg
From the rest I chose the "Moore-ish" one because I thought he was the best. With the style of the pic I was aiming for a "Nineteenth Century Historical Painting" look, a bit like Bessonov did with his pics:
kol03.jpg
"orentalist" elements are actually not at all out of place there and are a good fit for this style, I would say. Also, I have no problems fitting him into the backstory of the pic - my personal headcanon, if you like: He's a former Turkish slave who came to Western Europe during the first siege of Vienna. Now he is in the employ of a famous Witch Hunter who hired him due to his extensive knowledge of exotic torture techniques... Something like that. So, that's where he came from.

Loinclothslave said:
A nicely crisp image, shackles seems to be something that challenges AI but at least the wrist cuffs are connected.

She has a good facial expression

The witchfinder generals hat has an oddness in its peak

Overall I quite like it, Miss loinclothslave would take her place
Click to expand...
I'm glad you like it. Shackles are a problem, so are hands and feet and a lot of other stuff. You may notice that I tried my best to hide most of the girl's feet :D It's just one of the limitations of AI picture generation. I find that I can tolerate the little errors and inconsistencies much better when the pic is not photorealistic which is another reason I prefer it looking more like an artistic painting.
 
Riodoro said:
To make only one of those in Poser is 30 minutes or so of work and it would not even look half as good. On the other hand, some things are a real pain. Getting the hands and feet right took me forever and involved a lot of Photoshop and the fists still don’t look good. The torturer looks gorgeous, but he just stands there like an NPC in idle mode which gives the pic a slightly static feeling. I would have loved to be able to pose him like in a 3d app, let him and the victim interact, maybe let her look at him and plead while he shows her a branding iron or a torch, but this one pose is all I have from this specific character and for some reason the AI seems to always prefer characters looking at the viewer instead of each others. I think some of those problems are due to my inexperience and some are due to the inherent limitations of the technology, but anyway, sometimes I found myself pining for good old Poser where I can move the Brazier to the right with just one click.
Click to expand...
Probably this won't be needed with better prompt-following models like Flux, but you can achieve most of what you mentioned by providing a sufficient context while inpainting.

For example, if you just render a girl's face, the chances are high that she'll look at the viewer. To make her look at another character, you can select the faces of both of the characters and change the prompt to something like "a girl looking at the man's face in horror", and you'll have a much better chance at getting what you intended. If you don't want to change the other character's face, just selecting a tiny pixel to expand the bounding box can do the trick.

The same goes for depicting hands or other things. If you just select the hand area when inpainting, the AI may have trouble understanding if it's the right or left side, for example. Including other parts of the body to give it a better context usually helps mitigate the problem.

Also, you can manipulate the control net images for finer control, like posing a character in a specific manner. You can even do that without control net images, which I do often when I feel lazy, like drawing a very crude image of a branding iron by hand and inpaint over the area with low denoising value to make it look real.

Hope this may make it less stressful for you. :)
 
fallenmystic said:
...
Hope this may make it less stressful for you. :)
Click to expand...
Thanks so much, @fallenmystic your advice is greatly appreciated! I'm still learning how to use Image Prompt and Inpaint properly but what I have figured out so far has already vastly improved my experience working with fooocus. I have some free time next week so I'm planning on learning some more, especially about LORAS, maybe about how to use Pony models. Hopefully you all will be able to see the results here soon...
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom