AI [AI Art] - Show Us Your AI Skill *NO TEENS*

[Mr-Fox]

That's because you are getting too demanding.

Well that's what makes it fun for me. To push the envelop so to speak. There is no "sport" without the effort. The big part of the job were all the many images I have generated with tweaking of prompt and settings, to get the image that I then fixed up. If you look closely on the eyes they are still not right. But they are much better now compared to the original image. Then there is this thing when you stare at the same image for way too long time, everything looks weird and off. With new eyes tomorrow I will probably appreciate it more.

 

[fr34ky]

Well that's what makes it fun for me. To push the envelop so to speak. There is no "sport" without the effort. The big part of the job were all the many images I have generated with tweaking of prompt and settings, to get the image that I then fixed up. If you look closely on the eyes they are still not right. But they are much better now compared to the original image. Then there is this thing when you stare at the same image for way too long time, everything looks weird and off. With new eyes tomorrow I will probably appreciate it more.

Or hate it more...

There is a certain point where you have to admit that's all you got at the moment and use the experience for the next picture/project.

I apply that philosophy a lot in different things, don't get me wrong I'm very obsessive when it comes to art and learning or creating things, but I learned to let go to avoid paralyze.

 

[Mr-Fox]

Great advice and I agree. I have to confess that letting go is something I'm struggling with. I was banging my head against the wall quite a bit yesterday. The solution was to not give up on the project but the checkpoint. I was trying to make

You must be registered to see the links

work with my concept but in the end it simply wouldn't. I switched to

You must be registered to see the links

wich in my experience is much superior and voilá.

 

[baloneysammich]

View attachment 2394446

This is phenomenal, I don't suppose you remember the prompt? Or at least the model?

 

[atheran]

And now for something completely different.

I liked it so much I have like 500 renders of that style. I just posted 4 at random.

 

[sharlotte]

And..... back to the grind! (see how she's partly floating on water, she's a Jesus chick )

 

[Mr-Fox]

Pinup Betty

(Edited heavy with inpaint and photoshop).

 

[tdree]

This is phenomenal, I don't suppose you remember the prompt? Or at least the model?

Thank you. The original image was from a Girl Genius webcomic. I practiced with img2img and inpainting (using a lot of prompts for different parts of image) to make it more realistic.
The model was Deliberate v1.

Spoiler: Original image

You don't have permission to view the spoiler content. Log in or register now.

 

[baloneysammich]

Thank you. The original image was from a Girl Genius webcomic. I practiced with img2img and inpainting (using a lot of prompts for different parts of image) to make it more realistic.
The model was Deliberate v1.

Spoiler: Original image

You don't have permission to view the spoiler content. Log in or register now.

Wow, I never tried img2img, no idea it could do stuff like that. Thanks for the info.

It's funny though, I assumed the abundance of watches was due to the AI going overboard.

 

[fr34ky]

Great advice and I agree. I have to confess that letting go is something I'm struggling with. I was banging my head against the wall quite a bit yesterday. The solution was to not give up on the project but the checkpoint. I was trying to make

You must be registered to see the links

work with my concept but in the end it simply wouldn't. I switched to Clarity wich in my experience is much superior and voilá.

That's a good one, I find myself changing checkpoint when for some reason the concept is not working, Clarity is one of my favorites currently

 

[MoeMoeGames]

Tried different styles for the characters of my game.

! These are not in the Game !

 

[fr34ky]

Tried different styles for the characters of my game.

View attachment 2429256
View attachment 2429253
View attachment 2429276

The models Samdoessexy and babes have a lot of future for game characters.

 

[MoeMoeGames]

The models Samdoessexy and babes have a lot of future for game characters.

Could be in the future, but right now you can't replicate characters for different scenes/positions/angles as far as I know.

Oh and here another rose I just generated.

 

[fr34ky]

Could be in the future, but right now you can't replicate characters for different scenes/positions/angles as far as I know.

Oh and here another rose I just generated.

View attachment 2429332

I have some testings to do about creating consistent characters, I have some ideas of certain techniques that can be used.

At the moment I see that most tutorials are based on the most superficial part of AI art, the stuff that can be done fast and has high impact, so achieving any 'niche oriented' like result will need a lot personal experimentation.

 

[atheran]

I have some testings to do about creating consistent characters, I have some ideas of certain techniques that can be used.

At the moment I see that most tutorials are based on the most superficial part of AI art, the stuff that can be done fast and has high impact, so achieving any 'niche oriented' like result will need a lot personal experimentation.

If you figure it out, please share. What I'm thinking is to generate a bunch of images of a specific character and train a Lora model for it, but I lack the vram to train Loras. Maybe I will with google collab at some point.

 

[Mr-Fox]

Could be in the future, but right now you can't replicate characters for different scenes/positions/angles as far as I know.

Oh and here another rose I just generated.

View attachment 2429332

You can with the new

You must be registered to see the links

.

You must be registered to see the links

(ControlNet YT Tutorial)

You must be registered to see the links

(Live Pose YT Tutorial)



I have not messed with it myself yet but I have skimmed through some videos about it.

 

[Elefy]

but right now you can't replicate characters

You can!
Firs and lame route is to use face of some known "star" or mix of two like ,[Person One:Woman Two:0.5],
Pro route is to choice your favorite generated character and train LORA for.

 

[KingBel]

I haven't tested this out myself yet, but if you can train a checkpoint (or Lora as Elefy said) of your custom character (created in DAZ for example), you could then create consistent gens with that character. This Youtube video explains the process:

You must be registered to see the links

.

Seems like it would work. I am gonna do a test with one of the game characters that I normally do fanart for (possibly Ann from Big Brother) and see what the results look like.

 

[sharlotte]

If you figure it out, please share. What I'm thinking is to generate a bunch of images of a specific character and train a LORA model for it, but I lack the vram to train Loras. Maybe I will with google collab at some point.

You should watch this one:

You must be registered to see the links

on how to create a Lora using Kohya gui. I have an 'old' 1070ti with 8Gb VRAM and have been able to generate a few to practice until I get a card with more vram. It's well explained. If you want more details, go and watch

You must be registered to see the links

and

You must be registered to see the links

which are the two parts on explaining how to create the LORA.

 

[Elefy]

custom character (created in DAZ

If you render DAZ, why you need SD then?
DAZ characters is so ugly and with wrong proportions can guess G8 character/face from first look.