Shrunken women board

[size=150][quote=MrWitness post_id=33554 time=1694860139 user_id=7367]
[i][color=#BF0000]Some time later, with new experience and knowledge, I come back with what I am currently doing using Stable Diffusion. It's still not perfect, but these are some concrete results that look pretty good.[/color][/i][/quote]
The stuff you’ve done with Stable Diffusion looks pretty damn good.[/size]

Some time later, with new experience and knowledge, I come back with what I am currently doing using Stable Diffusion. It's still not perfect, but these are some concrete results that look pretty good.

I'm open to suggestions and advice. If you have ideas on how to improve something or if you want to use my knowledge - just let me know. I see huge potential in this. The graphics below are generated using SD 1.5, which means that with just 4 GB of graphics memory you can reproduce them or get similar ones. I know that some people achieve better results using SDXL, but in this case at least 8 GB of graphics memory is necessary.


[url=https://postimg.cc/5j225HzG][img]https://i.postimg.cc/Dfq4TLw2/00003-2934684304.png[/img][/url]

[url=https://postimg.cc/3yZ3m7xY][img]https://i.postimg.cc/jjpCVSVw/00007-548393239.png[/img][/url]

[url=https://postimg.cc/47LZ25c0][img]https://i.postimg.cc/yxBddp87/00070-1692770032.png[/img][/url]

[url=https://postimg.cc/TKgd2n8y][img]https://i.postimg.cc/9QNrpBJB/00115-1497105356.png[/img][/url]

[url=https://postimg.cc/1n5sKT86][img]https://i.postimg.cc/HxbnT1HS/00130-1820694140.png[/img][/url]

[url=https://postimg.cc/TyxRKhGt][img]https://i.postimg.cc/0NMrHJ41/00132-2569106702.png[/img][/url]

[quote="Prof Sai" post_id=30317 time=1685229323 user_id=63]
I think there comes a point where it is easier to have the AI do everything in bits and assemble them in a paint program. Tell it to put your figures in front of a green screen, for example. It is also sometimes better to fix details like those faces that way.
[/quote]

I think Master Mini does something to that effect

https://www.deviantart.com/notifications/watch/deviations/MasterMini

Or you can do what I do blend "real" models with my render program for hands, props ect.

[url=https://postimg.cc/mc3tm9pX][img]https://i.postimg.cc/pd4zXYfH/1.png[/img][/url]

[url=https://postimg.cc/MfHHZF6N][img]https://i.postimg.cc/cCMnmpVx/2.png[/img][/url]

I think there comes a point where it is easier to have the AI do everything in bits and assemble them in a paint program. Tell it to put your figures in front of a green screen, for example. It is also sometimes better to fix details like those faces that way.

Well, I feel the same way. I'm experimenting a lot with Midjourney, but I don't achieve much. Personally, I'm not interested in such a large shrinkage, I'm most often interested in something around 40-60% of normal size. I managed to generate some images where a woman about 60-70% of the man's height is standing next to him. Whatever I try to generate below this size, Midjourney sees these characters as children or dolls. If anyone has any thoughts or managed to do something in this regard, please share

And here are some of my works with Midjourney "shrunken" women:

Here are some problems witf faces...

[url=https://postimg.cc/nM2QXB9q][img]https://i.postimg.cc/ncbKw1HS/1.png[/img][/url]

This is the graphic where I reached the smallest size, but as you can see the face remains deformed as well and the size of the man is unnaturally large. Plus, something strange happened with his clothes...

[url=https://postimg.cc/Mc1nqQ8M][img]https://i.postimg.cc/XvzdKfH8/2.png[/img][/url]

[url=https://postimg.cc/BXZHpMfh][img]https://i.postimg.cc/BQHpqVYG/3.png[/img][/url]

[url=https://postimg.cc/HrRMpmcK][img]https://i.postimg.cc/7LYnnYxx/4.png[/img][/url]

[url=https://postimg.cc/RqLww2y6][img]https://i.postimg.cc/PfBKnjM4/5.png[/img][/url]

[url=https://postimg.cc/GHS8QRQ2][img]https://i.postimg.cc/JzRNR4Hb/6.png[/img][/url]

[url=https://postimg.cc/k68BMt3R][img]https://i.postimg.cc/TwNm9rt0/7.png[/img][/url]

Has anyone played around enough with Midjourney or any other comparable AI program to generate good, clear, cohesive images of, say, different women in a giant hand, or King Kong's hand? My little experiments have spat out pretty much garbage, but maybe I'm not being specific enough, or just misusing it for what I want. I imagine we can basically cast any hot actress into the Fay Wray role without having to deal with photo-manips. But I haven't seen anyone do that yet. Just curious.