News. (Patreon)

Some Info. (read it or stay confused)

If you don't want to read it, let google read it for you. Install, select all text, and click the megaphone icon: https://chromewebstore.google.com/detail/read-aloud-a-text-to-spee/hdhinadidafjejdhmfkjgnolgimiaplp?hl=en

So let's assume you know nothing about Ai-generating.

There are many ways how you can generate sfw and nsfw images.

Whether it's online (usually paid) like pixai.art or offline (free, but requires RTX GPU or an infinite amount of time) like Stable Diffusion.

If you want to generate anime characters that exist (Sakura from Naruto, Nami from One Piece,...) you need to use LoRA (ai-character model) All models are available here: civitai.com .

I use/used Stabble diffusion 1.5. It's the most supported version of SD (Stable diffusion) almost every anime character has its ai-model created for SD1.5. The problem for SD1.5 is that it's very unresponsive.

This goes to the second problem about 'poses' ... simply... when I type "jumping girl, Sakura" in SD1.5 in 90% of results she will just stand and stare at you, and the other 10% will be full of glitches. Even for stupid 'jump pose' you need LoRA (a specific model that knows how to project a jumping character).

Sometimes even a simple image needs up to 6+ different LoRA models.

LIKE THIS: <lora:Hinata[boruto]V2:0.77>, <lora:more_details:0.4> , <lora:weight_slider_v2:-0.2>, <lora:breasts out:0.5>,<lora:nicetits:0.2>,<lora:InstantPhotoX3:0.3>

Sometimes searching for all needed LoRA takes hours at then you still need to test them and find the fine balance between them.

in terms of posing, responses, and glitches SD1.5 isn't good at all.

In terms of community support, variations of art styles/characters, and simplicity to create a completely original style/character SD1.5 has no competition, it's just best.

I'm trying to create imgs in 2.5D half realistic half anime. For that, SD1.5 is the best. But when I go for NSFW hardcore SD1.5 struggles hard, every 15+ picture ends decent and every 50 generation is perfect.

Why am I saying it? well, you've probably noticed that in the last 2 months (especially on Pixiv) everybody's picture dramatically improved, in poses and angles from where the pose is taken. BUT all these images share the same art-style, I won't lie I really dislike that art style.

That thing is 'Stable Difusion Pony XL' . Pony has outstanding responses for posing. You can type 'sakura, jumps, eating spaghetti' and you'll get it, but only in that typical Pony-art-style.

SD-XL is here for a very long time like 7 months but only recently Pony appeared. Originally I planned to wait for SD3.0 which was already announced and Alfa version was released a few weeks ago. But then I realized it can take another 6+ months for SD3.0 to become relevant for Anime content.

So I will keep my standard-looking galleries as I do in SD1.5.

But I really want to find a way to do comics and manga with Ai and for that reason I will jump in SD XL Pony. I will try to find the most original-looking art-style/character that Pony allows. And outside of standard galleries you can expect some very experimental NSFW comic/manga posts + tutorials in PONY XL (if I discover something cool).