>>9752 The >>>/aichan/ BO would probably let you start a thread of this and let you do whatever. In case the current BO decides that your AI isn't /loli/ enough. >>9786 Normally, I don't like to point things out, because I feel like its criticizing you, and I have no business criticizing you, but ... Gwen has a tiny foot.

>>9789 There are things disallowed on /loli/ that I'd mostly-to-fully agree with & those disallowed things are in abundance on /aichan/. >>9786 >1990s PC >Boris Vallejo Some of this sounds familiar so you've probably said it before & I forgot so sorry for asking again if that's the case. >use the old Pony Realism model for the hair. Not always I've changed LoRA weights while inpainting a few times specifically on faces, but I don't recall switching model. I've toyed with Pony Realism since it had neat LoRAs for different styles of films, but anatomy issues are more prevalent. I'd think you'd use an Illustrious Realism model to get finer hair details since you'd still be on IL with benefits that come with that. Personally, Illustrious Realism models I tried were underwhelming due to lacking the nice film style LoRAs that Pony Realism has, so I usually ended up deleting them. Anyways, the painting styles I've tried are not quite the same as what you're mixing with, but I'm always curious to experiment when I have the time.

>>9789 >Normally, I don't like to point things out, because I feel like its criticizing you I'm pretty sure he posts here specifically to get criticism and help spotting errors.

>>9789 >Gwen has a tiny foot. Oh man, thanks for pointing that out. It's pretty embarrassing that I missed such a glaring issue. Looking at the pic again, it's actually surprisingly tricky to judge how big the foot should look from that angle, but I think this should be more like it. >>9791 Yeah I tried the IL Realism models but they specifically can't seem to replicate the nice looking hair I get from Pony Realism, so I've stuck with that old model for the hair. For me, I do a lot of changing styles while making a pic to get a blend of things I find interesting. I like to pick and choose styles and models for particular parts of the pic, and it's the most fun to me to blend different styles together to get something you don't usually see. >>9794 Well, these threads are really my main thing and I started uploading to ATFbooru at the suggestion of anons, having errors pointed out here has been a nice bonus. I do miss a lot of errors and I get pretty annoyed by how much I miss despite my best efforts.

Tried mixing Boris Valejjo, Jan Matejko & cryinggsilence: https://civitai.com/models/1208185 since it is a "painted" kind of style with anime faces, but I overdid the painting influences overall since I get a canvas texture that, like film grain, is annoying to inpaint as it creates uneven textures. I later had to reintroduce canvas & grain in photoshop to even it out. Correct about inpainting with Pony Realism when you want finer hair detail. It just werks. But I think it'd be easier if a LoRA for Illustrious could do it since you wouldn't have to wait to swap to an entirely different model. I also noticed these art styles want to do full body shots from a distance. I had to use controlnet to get closer. I'll have to experiment more to avoid the canvas texture since I'd rather have a smooth output, then choose whether to add those textures later in photoshop. Currently I think I just went too high on the weight of everything. <lora:Boris_Valejjo_Ill_V2:0.9>, v4le3jjo, oil painting \(medium\), <lora:JanMatejkoStyle-IL.V1.0:0.4>, janmatejkostyle-il.v1.0, classic media, <lora:cryinggsilence:0.7>, crying_illu, <lora:darkness_XL:0.3>, I really like getting darker scenes with the darkness LoRA, but it can fuck up inpainting by trying to make things darker and darker. Especially messes up inpainted skin tones which have to be fixed later in photoshop. Edit: Seems code tags are disabled on this board for some reason. Also cannot repost the same image after deleting/revising a post because R9K junk is also enabled. I'll just put the code into italics & bump contrast on the image since it needed more anyway.

>>9798 Oh that looks neat. As another tip, using the "thick thighs" tag under 1 weight (something like 0.8 or 0.6 depending on your taste) can give you lolis with a bit more meat to them without actually making them chubby. The Beebz pic here >>9562 and the body paint loli here >>9015 make use of that. It also tends to give you nicer butts when viewed from behind.

>>9799 I see. I hadn't used "thick thighs" but there were many times I used "chibi" or "toddlercon" at a lower weight to get a larger head or flatter chest (which can be necessary with certain styles).

Mixed Boris Valejjo & Jan Matejko again, but this time I only added artist kanikama at low weight to make the eyes bigger & face a little more 2D. <lora:Boris_Valejjo_Ill_V2:0.9>, v4le3jjo, oil painting \(medium\), traditional media, <lora:JanMatejkoStyle-IL.V1.0:0.5>, janmatejkostyle-il.v1.0, (no lineart), (artist:kanikama:0.1), So I got the soft glossy shading on the body with a flatter more anime styled face. I didn't do any Pony Realism inpainting on the hair this time as I think it's fine without it.

>>9798 That's pretty good. The only thing that bothers me about the pic is her line of sight. She's reaching out to the cluster of fireflies/fairies(?) in front of her hand, but appears to be blankly staring at open space above them. >Seems code tags are disabled on this board for some reason Didn't realize it was off. Code and latex tags are now enabled. >Also cannot repost the same image after deleting/revising a post because R9K junk is also enabled Changing a single pixel would be enough. Sorry I have threadwise R9K enabled to prevent the same things being posted over and over. Most threads here are slow don't need a ton of duplicate images. As an experiment, I'm going to unlink the file from the deleted post and see if it can then be reposted in this one.

>>9804 >Code and latex tags are now enabled Thanks. >her line of sight I imagined she was looking at one that's farther out-of-frame, but I photoshopped another firefly (or fairy) higher up anyway. So the object of her focus is in-frame now.

>>9805 Nice.

Forgot a few images during my earlier image dump.

>>9787 hina-chan! :3 cute

anyone else here doing animations?

>>9812 I approve of this hina posting :3

Just finished this!~🍊💘💦

Style Prompt: (artist:alkemanubis:0.4), (artist:iesupa:0.8), <lora:Beautiful_Artstyle:1>, beautystyle, semirealistic, detailed, <lora:Oil_Painting_Extra_Large:0.6>, oil painting, painting, Oil Painting LoRA: https://civitai.com/models/1405337 Beautiful Artstyle LoRA (DELETED, NO ARCHIVE): https://civitaiarchive.com/models/1285597?modelVersionId=1450483 Now only available on tensorart & download is (((paywalled))): https://tensor.art/models/833328053464684953 Reminds me of how Epic ruined Unreal Tournament 4.

>>9825 :3

Darkness LoRA: https://civitai.com/models/902671 Velvet's Mythic Fantasy Styles Darklines LoRA: https://civitai.com/models/599757?modelVersionId=2110071 v2 of this LoRA went up a few days ago. I used v1. Haven't tried v2 yet. Style Prompt: <lora:darkness_XL:0.2>, (dark, night), <lora:iLLMythD4rkL1nes:0.8>, D4rkL1nes, anime screenshot, flat color,

>>9829 So beautiful :3 I wish I could save it <3

More Hina-chan! Taking selfies with her favorite toys

Finished!~👅🍊💜💦

Finished!~ 💘🍊💦

Trying V2 of this now. >>9831 Penny (Inspector Gadget) LoRA: https://civitai.com/models/984429 Velvet's Mythic Fantasy Styles Darklines v2 LoRA: https://civitai.com/models/599757?modelVersionId=2291129 Mitsudomoe LoRA: https://civitai.com/models/1246190 Mixed with Mitsudomoe style to keep a round face. The pose was difficult to make.

Finished!~💜🍊💘💦

Orenjipiiru style: https://civitai.com/models/111792

>>9872 Sugoi. How do you match his font? Does the AI have the ability to transform text you type in into certain styles?

Guys, i'm too stupid to use the coding and the stable atuff (also my pc is shitty and crashes), so i was using civitai to make my stuff, when they started censoring everything i migrated to seaart, bt they are censoring stuff too... theres any other easy option for idiots to do hentai stuff easily? (of course grok, gpt and similars are impossible to, they censor even kisses)

>>9873 I recreated the text using pen tool in photoshop. Only AI model I have that can generate working text in an image is Flux, but it doesn't use booru tags which makes it more obtuse to use. Of what I have used so far, Illustrious & NAI models are best for illustrations. >>9874 As always, the only way to fully evade censors is local generation. It is what it is. To be brutally honest, I've worked with stable diffusion for years & the web UIs are python jank that easily breaks whether you're a beginner or novice. Just yesterday I had to reinstall it after something broke it.

>>9873 >>9875 thats the problem, anything that needs more than two steps to work, i just can't do 😭 my brain is too tiny for this world...

Today civitai announced that its image generator does no longer allow nsfw petitions unless it is from its yellow currency ( aka premium currency - you have to pay for it)