Midnight's AI Image Peddler Booth

[HokuouTenrou]

.1 ~ .2 if you want almost no changes. .3 ~ .4 if you want some small differences

edit: this is for upscaling correct?

I guess, yeah? I use either hi-res fix or img2img + resize; never simple upscaling

 

[greyblob]

I guess, yeah? I use either hi-res fix or img2img + resize; never simple upscaling

if you don't want simple upscaling. go to img2img and pick SD upscale from the scripts. if you don't have it download if from the extensions. then do 0.15 denoising. and let it run with the script

 

[HokuouTenrou]

if you don't want simple upscaling. go to img2img and pick SD upscale from the scripts. if you don't have it download if from the extensions. then do 0.15 denoising. and let it run with the script

Oh, thanks. I'll try it next time; I don't think I still have quota for the day.

 

[Midnight-Phantom]

i kinda forgot to use vaes bro .. and i was having some issues with it too like it over saturated the images sometime if i use high res fix .. automatic settings do the best in 1111 .. i mostly actively use different vaes when i use comfy ui .. but i am using automatic 1111 now mostly, they launched there new version early and there's some cool stuff like now you can add lora and lycoris in one folder and the extension is not needed .. it makes a whole lot of things easyer.. you guy's should check it out tho ..

upscaleing is quite hard sometimes as that will depend on the available upscales , settings and also on your GPU too .. try different things out and find the thing that's just suited for you .. and as boby said try to use low denoising if default one shows issues ..

i have heard that letest builds with new architecture GPUs meaning rtx 40 series is verry good and shows less issues.. also workstation builds are the best for running anything ai releted... also try to upgrade your graphics drivers if it's outdated it can cause issues with cuda processing graphics..

 

[Midnight-Phantom]

Can you make group of women in the beach, but they're using modest full body female swimming suit instead of bikini ?

I'm rather bored with usual "anime beach episode" where they're using "un-modest" bikini for fanservice. I always use full body swimming suit to increase my mobility and protect my skin in the pool and especially beach.

I would like to see full enthusiastic female divers using full body swimming suit to protect their body during serious swimming.

And also, can you make the same thing above, but in the pool instead of beach ?

Image references :

Spoiler

View attachment 20654

View attachment 20655

View attachment 20656

View attachment 20657

Here you go for Sea :

Spoiler: <->

And here is for Pool:

Spoiler: <->

( i have used the same pose for both of them and did anime style but not the usual school once)

 

[ACertainPassingUser]

Here you go for Sea : -snip-

And here is for Pool: -snip-
( i have used the same pose for both of them and did anime style but not the usual school once)

Thanks for the works.

When compared to the girls size to the surrounding trees and cliffs in the ebach picture, they're basically Gigantic Titans.

The swimming pool looks unorthodox tho. In my head, I'm thinking the story as : the female group already taking group photo in the normal pool, but the facility has beautiful feet bath pool with beautiful aesthetic plant and stones. Therefore they're taking another photo there. And they're taking more photo in the middle of the pool due since nobody was there.

In the beach picture, Multiple anime girl with same expression isn't as creepy as I thought. It's a perfect chance to make twin character. And the Anime style is ignoring facial shape distinction too, so it's understandable. Many swimmer innreal life are using similar hairstyle since they're searching for effectivity.

And the outfits on the beach were almost major success. The girl on the left is using full arm cover and the two twin girl with similar hairstyle and color is covering their legs.

The outfit on the pool tho, its less successful. Although the overall quality and character variances in the pool is better than the beach one.

Both sucess in different directions.

Maybe It's because I forgot to said the keyword "Wetsuit". But whatever, thanks again.

 

[Midnight-Phantom]

 

[MrToothySmiles]

could you make a AI image of a small child staring at his sketchy reflection? a bit erie and desolate? similar to this

Spoiler: Reference

 

[SailusGebel]

Wait a second, your AI can emulate artstyles of different artists?

 

[Midnight-Phantom]

Thanks for the works.

When compared to the girls size to the surrounding trees and cliffs in the ebach picture, they're basically Gigantic Titans.

The swimming pool looks unorthodox tho. In my head, I'm thinking the story as : the female group already taking group photo in the normal pool, but the facility has beautiful feet bath pool with beautiful aesthetic plant and stones. Therefore they're taking another photo there. And they're taking more photo in the middle of the pool due since nobody was there.

In the beach picture, Multiple anime girl with same expression isn't as creepy as I thought. It's a perfect chance to make twin character. And the Anime style is ignoring facial shape distinction too, so it's understandable. Many swimmer innreal life are using similar hairstyle since they're searching for effectivity.

And the outfits on the beach were almost major success. The girl on the left is using full arm cover and the two twin girl with similar hairstyle and color is covering their legs.

The outfit on the pool tho, its less successful. Although the overall quality and character variances in the pool is better than the beach one.

Both sucess in different directions.

Maybe It's because I forgot to said the keyword "Wetsuit". But whatever, thanks again.

It's doable, though. I have to spend a lot of time finding a model that can do that kind of background in an anime style or similar loras. After finalizing the model, I have to work on the characters. The more characters in an image, the lower the quality will be. I also have to work on their hands, faces, and other details. After doing all of that, I put the resulting image in a control net and make the renders based on that. If an image with a lot of aliasing comes up, I have to upscale it and hope it doesn't break the image. Yeah, it sounds like a pain.

Wait a second, your AI can emulate artstyles of different artists?

I think you're asking the wrong question, bro. All models can do all kinds of art styles if they're trained on that kind of art, or if you download a lora that's only trained on the artist's art and nothing else. You can add that lora to any model, and the resulting artwork will be based on the layering of both. The AI interfaces that people mostly use online, don't include those features, or if they do, they cost money.

Since I run everything on my own device, I don't have those restrictions. I can just download models and loras and create whatever I want, but it takes up a lot of space. I also have issues because I have more than 2,000 model, lora, ti, and hipernetwork files combined. To accommodate that, I run the whole thing on an hard drive because it takes up a little less than 2TB of space. Models load slower from an hard drive, so it would be a lot easier if I had a 4TB M.2.

 

[greyblob]

Wait a second, your AI can emulate artstyles of different artists?

It's doable, though. I have to spend a lot of time finding a model that can do that kind of background in an anime style or similar loras. After finalizing the model, I have to work on the characters. The more characters in an image, the lower the quality will be. I also have to work on their hands, faces, and other details. After doing all of that, I put the resulting image in a control net and make the renders based on that. If an image with a lot of aliasing comes up, I have to upscale it and hope it doesn't break the image. Yeah, it sounds like a pain.


I think you're asking the wrong question, bro. All models can do all kinds of art styles if they're trained on that kind of art, or if you download a lora that's only trained on the artist's art and nothing else. You can add that lora to any model, and the resulting artwork will be based on the layering of both. The AI interfaces that people mostly use online, don't include those features, or if they do, they cost money.

Since I run everything on my own device, I don't have those restrictions. I can just download models and loras and create whatever I want, but it takes up a lot of space. I also have issues because I have more than 2,000 model, lora, ti, and hipernetwork files combined. To accommodate that, I run the whole thing on an hard drive because it takes up a little less than 2TB of space. Models load slower from an hard drive, so it would be a lot easier if I had a 4TB M.2.

t2i adapter style is supposed to emulate styles from an input image. meaning you can take small or obscure artists and emulate their style without having a model or lora trained on their images

 

[SailusGebel]

t2i adapter style is supposed to emulate styles from an input image

Oh, that's because of an input image. Thank you for the enlightment.

 

[greyblob]

Oh, that's because of an input image. Thank you for the enlightment.

you can also emulate styles without an input image but that requires the model to 'understand' the style. meaning it needs to be trained on it (or have a lora trained on it).

there are like 500+ artists' styles in almost all models. there are also a ton of loras trained by random users. though the majority of those are for hentai

 

[Midnight-Phantom]

t2i adapter style is supposed to emulate styles from an input image. meaning you can take small or obscure artists and emulate their style without having a model or lora trained on their images

Now I have to add a new thing to the list for me to try out ..

 

[Midnight-Phantom]

could you make a AI image of a small child staring at his sketchy reflection? a bit erie and desolate? similar to this

Spoiler: Reference

View attachment 20721

I tried bro but cant create something like that at all ..

these are the best types i have gotten ...

Spoiler: <-->

 

[MrToothySmiles]

I tried bro but cant create something like that at all ..

these are the best types i have gotten ...

Spoiler: <-->

Woah! You've got skill!

 

[Midnight-Phantom]

Woah! You've got skill!

Thanks, bro,

but I really can't get the one like you wanted, you know, like the reflection on the wall. I've tried everything, downloaded various images, put them through the control net, and always— I mean always— ended up with either a version of the character with an age difference or something like a mirror or a portal, but not the shadow. This is the case for all the models; I've literally used probably more than 30 different anime-type models, pixel, slight realism and 2D, 3D types, and it just won't work, even when I use the exact prompt you give me as a sentence!

 

[MrToothySmiles]

Thanks, bro,

but I really can't get the one like you wanted, you know, like the reflection on the wall. I've tried everything, downloaded various images, put them through the control net, and always— I mean always— ended up with either a version of the character with an age difference or something like a mirror or a portal, but not the shadow. This is the case for all the models; I've literally used probably more than 30 different anime-type models, pixel, slight realism and 2D, 3D types, and it just won't work, even when I use the exact prompt you give me as a sentence!

lmao the difference between AI and Human is strong!

 

[Midnight-Phantom]

lmao the difference between AI and Human is strong!

that's true .. ( at least for the next 5 years )