#3497 - suggestive, ai assisted, ai generated, automatically imported, derpibooru import, editor:paracompact, generator:purplesmart.ai, generator:stable diffusion, oc, oc only, earth pony, pony, autumn, butt, clothes, female, hoodie, leaves, looking at you, looking back, looking back at you, mountain, mountain range, plot, random pony, rear view, scenery, sitting, solo, solo female, tree

Image imported from Derpibooru.
Original: https://derpibooru.org/images/3005971
Metadata at the time of import:

Field	Data
Favorites	53
Upvotes	90
Downvotes	7
Score	83
Comments	3
Uploader	Paracompact

Original description below this line

If the tail made sense, you would see less of the butt, and who wants that?

Made with a fine-tuned model of Stable Diffusion (continued training from Astralite’s pony model v1) and some hours of inpainting and outpainting, with some manual touching up in between and at the end. Ask me if you want the checkpoint, or any prompts/settings used along the way.

Syntax quick reference: **bold** *italic* ||hide text|| `code` __underline__ ~~strike~~ ^{^sup^} _~sub~

Detailed syntax guide

https://derpibooru.org/images/3005971

Comments

Background Pony #146D

@Background Pony #146D

I made something of an attempt to fix the tail issue early on. This is one result of mockup + inpaint: one result of mockup + inpaint

Interesting, my gut instinct is I’d expect their tail to swoop up and out much higher/cartoonishly to the up-right before S-bending down and settling to the ground further to the right.

it has a sometimes crude understanding of emphasis in general.

That was certainly my first experience with it. Ex. I’d ask for pink shirt and blue pants and it’d swap the colors between the two more often than not. I expect if/when they figure out emphasis properly, they’ll also simultaneously solve the ability to render more than one character at a time.

Anyway… I guess I’ll just see this like I see the “eyes clipping through hair” trope in anime :p

More and more I’m starting to get accustomed to the routine faults in the stable diffusion output, so I feel I’m less sensitive to it, almost like an otherworldly artistic signature by miscounting the number of fingers on a hand, arms per body, cutie marks per butt, etc

It’s about half due to my fine-tuned model (just a Dreambooth on some favorite pony art of mine), half due to generic prompt juicing. I have some of my favorite modifiers here, and I find that the PA Works + Photoshoot in particular works wonders. Simply adding “3d” to the negative prompt (and positive prompting for e.g. “2d graphic novel illustration” for good measure) makes it crazy smart and talented with illustrated stuff.

Oh, “graphic novel”, I’ll have to try that one, thanks

_{Imported from Derpibooru - Posted by dualreason}

Posted 2 years ago Report

Link Quote Reply

Background Pony #146D

@Background Pony #146D

Stable diffusion made that? Really love the colors. It’s so spot on…

Thanks. I did some color work of my own (in between generations, to alleviate the inevitable color drift from too much inpainting, and at the end, to my own personal preference), but if there’s one thing AI art is already superhuman at it’s color theory, IMO.

If the tail made sense, you would see less of the butt
Dang it, of coarse. Wouldn’t be AI without a really subtle catastrophic failure, lol

I made something of an attempt to fix the tail issue early on. This is one result of mockup + inpaint:

but in the end I couldn’t be bothered to go through with it. When you prompt for too much butt stuff, it will inevitably try to erase anything that gets in the way of that—it has a sometimes crude understanding of emphasis in general.

Anyway… I guess I’ll just see this like I see the “eyes clipping through hair” trope in anime :p

or any prompts/settings used along the way.
Did you do any specific prompting to get this art style? The shading on the jacket looks really slick.

It’s about half due to my fine-tuned model (just a Dreambooth on some favorite pony art of mine), half due to generic prompt juicing. I have some of my favorite modifiers here, and I find that the PA Works + Photoshoot in particular works wonders. Simply adding “3d” to the negative prompt (and positive prompting for e.g. “2d graphic novel illustration” for good measure) makes it crazy smart and talented with illustrated stuff.

_{Imported from Derpibooru - Posted by Paracompact}

Posted 2 years ago Report

Link Quote Reply

Background Pony #146D

Stable diffusion made that? Really love the colors. It’s so spot on…

If the tail made sense, you would see less of the butt
Dang it, of coarse. Wouldn’t be AI without a really subtle catastrophic failure, lol

or any prompts/settings used along the way.
Did you do any specific prompting to get this art style? The shading on the jacket looks really slick. Love the boarder, that’s a nice touch

_{Imported from Derpibooru - Posted by dualreason}

Posted 2 years ago Report

Link Quote Reply

Tantabus

BBCode

Comments