Introducing the ultimate anime model that will redefine your artistic horizons. Dive into a world where anime transcends the ordinary, with my state-of-the-art model designed to bring your wildest imaginations to life. This isn't just any model; it's a groundbreaking blend of vibrant soft colorings and nuanced shading that elevates characters from mere 2D to a realm just shy of 2.5D. This, combined with thicker outlines make a wonderful blend for a unique anime experience offered by very few others. All with the convenience of a baked-in VAE
HOW TO USE
Prompt what you want. It should come out in a similar style as the sample images.
If it does NOT produce anime style, please add "anime" to the positive prompts.
If you're feeling fancy, add "80s anime" or "90s anime", it gives amazing results.
Produces beautiful, eyepopping CURSED NSFW images!
Tested on many schedulers and samplers, but I personally use Euler a Karras
After some experimentation DPMPP_3M_SDE works really well too for internal details
Please enjoy.
It is a combination of Unstable Illustrator, AAM, GhostXL, & Osorubeshi XL.
Please give them a look as well, they are all excellent models in their own right. That being said, the model I present to you brings out what I think are some of the best elements each have to offer in an easy to use package.
I HIGHLY RECOMMEND Negative Embedding for SDXL Based Anime Models - v0.5 | Stable Diffusion Embedding | Civitai when using this model, it will help with hands on all anime sdxl models.
It also works excellently with the famous HARRLOGOS lora, which I highly recommend if you are seeking text in your creations.
Below are supported dimensions to get intended results
Dimensions Aspect Ratio
1024 x 1024 1:1 Square
1152 x 896 9:7
896 x 1152 7:9
1216 x 832 19:13
832 x 1216 13:19
1344 x 768 7:4 Horizontal
768 x 1344 4:7 Vertical
1536 x 640 12:5 Horizontal
640 x 1536 5:12 VerticalSpecial shout out to Cheshire, FreckledVixon, Hellrunner, RCNZ, Joemcey, MajMorse and others for their guidance through this process and their playtesting.
Description
Hi everyone, it's been an incredible journey and for those along for the ride, I deeply thank you for downloading Black Magic XL. I hope that it has been good to you and you like it. If you do... I have some excellent news!
I am pleased to present to you: BLACK MAGIC XL v1.45, the best version yet.
This time I merged a percentage of Raemu 3.0 with MBW weight adjustments. The goal this time was simple:
Maintain the specific clear, crisp style of Black Magic XL with strong, clean outlines.
Improve the diversity of poses, the dynamics and facial expressions With Raemu.
Improve the general quality of eyes regardless of distance from the camera.
I spent days on this one and I am extremely pleased by the results.
I admit it's not perfect, it cannot do specific characters without a lora. It can however produce "Walmart" versions of characters. (E.g Vegeta will have a widow's peak and the same hair, with a similar style to DBZ, but there will be clear differences in his specific clothing and possibly body shape).
Prompting is very simple as always with Black Magic XL:
Subject (required) - (e.g. 1girl)
Medium - Will almost always produce anime. Adding something like "photorealistic" will produce awkward results as this is an ANIME model. Pastels, line art and other illustrative mediums work best.
Style - 90's Anime, 80's Anime, (By default Black Magic XL is largely considered modern anime)
Artist - Spin the wheel and try your luck!
Website - Artstation, deviantart, tumblr, whatever works.
Resolution - 4k, 8k, "Best Quality" etc. are added to the end. They are not necessary, but they are helpful if you get some peculiar results which happens with some prompts.
Additional details - Custom.
Color - Give it some color hues or what have you
Lighting - Things like: Soft Lighting, glow, neon, fluorescent, volumetric all tend to work well.
THERE IS NO NEED TO UTILIZE score_9, score_8_up, score_7_up, source_anime, AS THIS IS NOT A PONY MODEL. Though, I have seen Pony XL loras work, at least a few of them, but the weight may need to be cranked high to produce the results you're trying to get, but it's possible it won't work at all. Most SDXL models cannot boast this after some testing. Must be some kind of black magic or something... 😉
Same rules apply as other versions: No sharing to other websites, please give me credit if creating a merge.
If you can make some money off the pictures generated with my model, by all means, get that paper!
In any case, I sincerely hope you enjoy this model as much as I do. It was a lot of fun making it and I adore the results, even if the changes are small.
Thank you for downloading.
If you have any questions, comments or concerns send me a chat or ASK FOR RAMTHRUST IN THE CIVITAI DISCORD'S GENERAL CHAT. I'm always ready to help people.Thank you,
RAMTHRUST
FAQ
Comments (21)
Yay! The promised release has arrived
Haha hopefully it's a good one!
It wasn't. Black Magic 1.45 4 lyfe
@RAMTHRUST Time to make a version for Flux!
@badassdragoon2 Lol, Lora in progress
@badassdragoon2 Hey, can I use some of your gens in the training set for my V2 lora? It would help a lot and I think bring some cool results.
why can't choose on the website?
Unfortunately not all models can be hosted on the online generator; it just costs too much money for Civit. Though I gave them permission to host it if that ever changes. Bummer I know.
Hiding your prompt is not cool bro..! But the checkpoint is cool though!
Ah, that's a comfy issue where it doesn't transfer the prompt automagically through the flow. I'm working with Civitai to get it resolved but it's slow going.
Thanks for the compliment though!
@RAMTHRUST Just passing by to confirm that does happen and it's incredibly annoying. Nice to know I'm not the only one with that issue. lol
Even if you decide to manually add the prompts, it still won't let you properly tag what loras, embeddings, and checkpoints you used.
It makes me think "hope no one thinks I'm hiding the prompt on purpose..." every time I post a picture.
@_Kapot_ Yeah it's super frustrating, I wish they'd change it but I can see why they don't. ⚡⚡⚡
@RAMTHRUSTThey cant "change" it since it's a side effect of some images being uploaded. I presume you have zero programming experience to understand this. To make it simple, it's something both the model uploder AND the website that provides you FREE models cannot do anything about it. Use your eyes, describe the picture, this should be relatively easy. But if you're expecting a prompt you copy from an image that was provided to provide you an identical image - this won't EVER happen. It's due to Pytorch version discrepancies including a number of other variables. Use your brain, make your own prompts. ROFLMAO
@SOLTIS Laughable. They have a metadata save node now which fixes the problem. You've been a real peach so let me explain something to you this time: grow up and get a life.
@RAMTHRUST Cry about it. clearly you know nothing. There are workspaces out that do not get saved in the meta data and only show [custom node] in the actual body of the node where the prompt would be. How bout you be less emotionally shallow and accept that the world clearly isn't what you assume it is. Remember, this is a public forum. Just because someone doesn't agree with you doesn't mean you should be a child about it and rant and rave and THINK you are proving someone wrong. You've wasted your pathetic breath because it's null and void. There is NO FIX for it, not even in Comfyui v1 desktop. Now sit down kiddo.
@ZUSIMO k
@ZUSIMO The fellow read a two lines long comment and immediately started writing a small wall of text and hyperventilating. Amazing.
@_Kapot_ now he's going through all my gens and laughing everything I posted over the past year. This was already sad but now it's going full tragic.
@RAMTHRUST Wow. In all seriousness, that's not exactly normal. He may actually be mentally challenged. It's probably best to avoid any other interactions and just block him, or else he may hurt himself irl.
@ZUSIMO I know you're still doing it and it's hilarious.
Fuck I love this checkpoint!