INTRODUCTION
maxxxx_Pony_Mix is an experiment to create a Pony model which has a wide variety of use and has that Western style anime look. While creating the merge the goal was to create something which is different enough from its origins and brings a new look-and-feel to the Pony world. I hope you will find this true - I think it is : -)
Tag-based or Natural-language prompts
Based on my experience the checkpoint likes natural language prompts too, sometimes they even give better results.
Useful:
these words add extra to the pictures: bright colors, high contrast, dark background, vivid lighting, realistic, freckles, large lips, beautiful eyes
if you want to fo go for the more traditional 2D look, just add 3d to the negatives
you might use these for your generation: 'source_pony', 'source_furry', 'source_anime', 'rating_safe', 'rating_questionable' and 'rating_explicit'.
Porn
based on my exerimentation, this checkpoint excels at creating NSFW images probably the most
although sometimes it will add dicks to places where they make no sense (which is kinda general for pony checkpoints) it creates scenes and poses with ease which I have never seen before with other PONY checkpoints
Themes without a human
This is a general purpose checkpoint and can create a large variety of image types but it does have the want to put some person (or sometimes pony xD) in the picture. You will be able to create pictures in a lot of different areas but sometimes a person will appear in scene where you would not expect it (e.g. it is pretty hard to create a bunny with a carrot, the checkpoint naturally will replace the animal with a playboy girl). Yet, as you can see it from the samples, it can deliver great results even wihout humans.
Pony numbers
They are not always necessary but they can help you a lot of times to get really specific about what you want - or don't want. My proposal is to use them :-)
Wish you a happy exploration and please share you creations and thoughts, as always!
The credit goes to the creators of the basis models which I used to create this mix:
DucHaiten-Pony-XL (no-score) v6.0:
https://civarchive.com/models/376450?modelVersionId=817261
Pony FaeTality v1.1:
https://civarchive.com/models/331116?modelVersionId=464939
Incursio's Meme Diffusion (SDXL, Pony) v2.7DPXL:
https://civarchive.com/models/314404/incursios-meme-diffusion-sdxl-pony
Description
FAQ
Comments (17)
Western style sure interests me, but I don't see it as an option when generating on Civitai? Is it possible you didn't enable that option?
hello, glad to hear:)
Unfortunately Civitai allows only a few amount of checkpoints to be used on-site - they say it would take too much resources... they allow only the big ones to be used on-site. So, you cannot use it on Civitai's own creator platform :/
@maxxxx_civitai Oh, that's too bad. Maybe some day, then. =) Thanks for the swift reply!
hello, CIVITAI has just turned ON all checkpoints, so now you can try it here online :-)
@maxxxx_civitai
I read the message and instantly went to try your model! =)
I didn't get any results I like, though... I was going for an image of a woman on a beach. Very, very nice backgrounds with your model, but the woman always seemed... off. Like... it was originally a realistic photo of a woman, but then with some basic "cartoon" filter applied to it. It's hard to explain, so I left it for now and will give it a second go later.
I do have a quite specific result in mind that I'm still chasing, and perhaps I was pushing your model in a direction it's not really meant for. We'll see, I'll keep you posted.
@Otto_Mattiq hey, thanks for trying it:) I am not sure what you are looking for but if it is not this style, you can try the followings: use 3d in the negatives --> this will create a more anime-like look. You can also use 'realistic' as a positive prompt which will enhance the details but not making it 3d-ish yet. Hope these might help you :)
@maxxxx_civitai
Thanks for the tips. =) Frankly, I'm not sure exactly what I'm looking for, except that I know it when I see it. I grew up in Belgium, during the golden age of Franco-Belgian comics. A lot of it is a visual style that is in many ways distinctly different from the Asian anime and US superhero comics looks. I respect the importance and value of those styles, but it's not what I grew up with, and I feel like "my" styles are severely underrepresented internationally.
Here's an example of what I really love myself:
https://civitai.com/images/58424861
If you checked my latest posts, I did a series of portraits using (Italian) Paolo Serpieri style, intentionally very soft lines and muted colours. I think they're also a good example of me trying to go with visuals that are quite different from the vast comic/cartoon images here.
@Otto_Mattiq hey, yes, now I see it. Nice style:) Very specific indeed. But if you have loras for that, is that now enough? Does that work for you? Or the loras are not either? Or you wanted to have a checkpoint which produces this?
@maxxxx_civitai Ideally, I'd like the checkpoint to be as unbiased as possible, leaving me open maximal options. One example: "very dark skintone" is rarely giving me a result that I like. And "colored skintones" are definitely fighting me as well. My theory has been that, in anime, a major datapoint source for these checkpoints, fair skintones are most common. In tests with Pony Realism, I do easily get the skintones that I'm trying to achieve. So: I definitely prefer checkpoints that at most push color saturations and contrasts to my liking.
For anything else I prefer LoRA's, and there I try to limit the number. I've been trying to create common character's by prompting alone and not using dedicated character LoRA's. Much of it as an exercise to learn and improve my prompting.
To be honest, I'm kinda going all over the place, testing things out. Dany's (Belgian artist) visual style is phenomenal in my opinion, but iconic characters like Disney's very quickly become too different and I have to dial it down a bit. There's only a few constants: I generally try to move away and distinguish from anime and US superhero comics visuals, I know what I like when I see it, and I'm like a kid in a toy store, excitedly running from one thing to another and going, "Ooo! Ooo! What does this do?!??"
@Otto_Mattiq cool :-)) keep on with the good stuff and thanks for the thoughts:)
@maxxxx_civitai
As for LoRA's... it's hard to predict how they'll respond to the different Checkpoints. It's a lot of trial and error, and some produce amazing results in draft, but way too contrasty and sharp in higher settings.
Ex, this is draft!
https://orchestration.civitai.com/v2/consumer/blobs/5YYF5AF81RMQ7R9HV7YZ4VSSR0.jpeg
@maxxxx_civitai Here's the result I get, exact same prompt and draft quality, only swapped out to your checkpoint. It's a lot darker, and following the prompt less accurately. (The LoRA has a strong bias to wavy hair, so I needed prompts to force sleek hair. Your checkpoint enforces it less.) There's definitely potential, though!!
https://orchestration.civitai.com/v2/consumer/blobs/QNPYG0R5N3KGKFT5QKBVAFRDK0.jpeg
https://orchestration.civitai.com/v2/consumer/blobs/ME3KE0P7PKATKHR3KB3CSJ6QN0.jpeg
https://orchestration.civitai.com/v2/consumer/blobs/Y7KP1WR69S6B23SCKPQ3DG1EE0.jpeg
@Otto_Mattiq :-)))
@maxxxx_civitai
It took a while, but I found time to give your checkpoint a real run. It sure can produce some very nice detail, colours and contrast!
@Otto_Mattiq glad to hear you had fun and liked it:)) thanks for trying it:)
a really cool and versatile checkpoint, nice stuff
thanks, glad you liked it:))