This is my first attempt at creating a good mix model. It's a mix of a lot of other models, both anime and realistic. Good at rendering Anime as well as drawn and semi-realistic images. You can achieve a high level of realism if you want.
Easily reproduces both SFW and NSFW images.
If you enjoyed this mix, please leave a review or comment. It motivates me to create new versions.
Change log:
V.2.01:
- Several additional models have been added.
- More emphasis on Anime.
- More flexibility.
- Works much better with NSFW
Additional info:
Recommended params:
CFG scale: 5 - 8.5
Steps: 30
Resolution: 896x1252 (Works fine with other resolutions too, that's my personal preference.)
Sampler: Euler - better for Anime, dpmpp_sde - better for realism.
Example promt:
realistic, hyperrealism, masterpiece, photography, 1girl, sfw, cute dress, on a beach, long hair, vivid colors, detailed face, high quality, best quality
Negative:
disfigured, kitsch, ugly, oversaturated, grain, low-res, Deformed, blurry, bad anatomy, disfigured, poorly drawn face, mutation, mutated, extra limb, ugly, poorly drawn hands, missing limb, blurry, floating limbs, disconnected limbs, malformed hands, blur, out of focus, long neck, long body, ugly, disgusting, poorly drawn, mutilated, mangled
The model was tested in Swarm UI.
Description
- Several additional models have been added.
- More emphasis on Anime.
- More flexibility.
FAQ
Comments (21)
Are you able to add some more info? Things like generation info for sample images, what VAE is used or baked in, good CFG / Sampler / Clip info etc... and does it change depending on the level of anime or realism trying to be achieved?
Yeah, I hadn't thought of that. I'll be sure to add all the information soon. Thanks.
are you really liz omg i'm your biggest fan
just to let you know:
my "dumb" way of thinking is working on all your checkpoints too....
I don't see it working. I see works that are completely different in style and centerpiece. If I take a Lora model and include a strong influence I can get the same images from different models. So not convinced.
you just don't understand conceptual art. all you do is your programming stuff and not even trying your brain in a correct way. all i do is copying styles and stuff and you just keep on barking. and not understanding what is easy to copy and what not.
you just ignore and downvote the fact that i can recreate any style or model in general and you keep on being that ignorant "know-it-all-nerd" that never has tried anything else...
@themancalledmojo999 Okay, let's do it this way. I'm going to find Lora with styles that I don't think I can recreate on my Boltning 1.0 model (because it's not flexible at all in terms of styles). If you can do that, I will admit that your statement is correct. But you can't, because you are talking about some styles and trying to claim that a programmer knows less than you about what he programs 😂 Let's start with this style: https://civitai.com/models/183354/sdxl-ms-paint-portraits?modelVersionId=205793 I want to get exactly the same pictures as the author.
@georgebanjog i'm on it, but something else i found out: your apogee can generate very nice comics. even with multiple panels without having to prompt it.
@georgebanjog https://images.woh.to/image/yqS
@themancalledmojo999 "https://images.woh.to/image/yqS" This picture was made on the model - "pornmasterAnime_v5"... Nothing to do with my models. By trying to not play fair, you're only proving me right. I guess you couldn't get the results you wanted with my model without using Lora and decided to try a model that could handle it.
@georgebanjog ??? you never said that i have to use your models....
@georgebanjog https://images.woh.to/image/yqT
not the final result, but https://images.woh.to/image/yqX
maybe you know that meme. it just crosed my mind during tests.
that checkpoint can generate very nice pixelart without much prompting. not every checkpoint is able to generate that on the spot.
still not the result i want, but this looks nice on your checkpoint Image posted by themancalledmojo999 (civitai.com)
getting some awesome pixelart styles during tests Image posted by themancalledmojo999 (civitai.com)
will continue this tomorrow. have to go to work now....
@themancalledmojo999 The pictures aren't bad, but as expected, not even close to the style you can get with MSPaint Lora. The thing is that when training models, certain Data Sets are used, and if they don't contain the style you need, the model won't be able to reproduce it, because the weights of neurons are not calibrated to reproduce these styles. I don't argue that you are good at making prompts, but you won't be able to replace Lora.
@georgebanjog i fucking found what i was searching for!!!!!!!!!!!!!!
sorry for that harsh words, but i am just tired right now and wanted to inform you on my finding. no images from your checkpoints, but use these prompts:
(((vector-drawing-style, doodle-like))) 1girl, outside
as basic and change the thickness of those lines until you are satisfied....
add colors or put in negative "grey-in-grey"....
another one and i'm off now: (((vector-drawing-style, simplistic-minimalistic-drawing-style))) 1female
@Stapfschuh02 I like those styles... But it's not even close to MS-Paint Lora results. So I think your claim about useless of Lora - debunked. Good day sir 🙂