Dark Magician Girl LoRA
Turn on visibility to see all the examples
Making models can be expensive. Do you like what I do? Consider supporting me on Patreon 🅿️ or feel free to buy me a coffee ☕
Great character design from a great franchise. Should work on most anime models at 1 weight (0.6-0.7 for the old one). Triggers with dark magician girl plus her outfit description so she can be customized like you want. For the first pic I also used dpep2 LoRA to give it some style.
Thanks to @plsdontultme for some of the examples.
How to use LoRA's in auto1111:
Update webui (use
git pulllike here or redownload it)Copy the file to
stable-diffusion-webui/models/loraSelect your LoRA like in this video
Make sure to change the weight (by default it's
:1which is usually too high)
Description
FAQ
Comments (29)
Was waiting for this one, thanks!
Hoping for a Mikasa LoRA at some point (fingers crossed)
theses Lora's are amazing do you take requests?
I take suggestions :)
Ey Lykon, I have a question, do you train your Loras at more than 1 epoch? gradient accumulation more than 1?
@lykon ok how about a fire emblem lora of Yusuke Kozaki style (awakening/fates/heroes) or the TCG Cipher, also do you have a discord or anything because I wanted to ask you some questions?
@stygianwizard42 noted, thanks
you made the first one with which basic model and VAE, I can't reproduce it to test the model
i love all your Lora!
Do you have a discord or something for suggestions?
I would like to offer you to do Chel - The Road to El Dorado if you like her, she is quite a popular character
I'll think about it.
The first one however is done with AnyPastel I think. Can't check the hash right now, but I'm relatively sure. The vae is anything4 vae.
@Lykon can you double check? The hash says AOM2_nsfw, but I also can't reproduce it with that model
@credipl08 uhm I probably used that by mistake for img2img upscale, let me get the original low res version.
@credipl08 lowres pnginfo says
parameters
dark magician girl, masterpiece, best quality, 1girl, blonde hair, blue footwear, blue headwear, breasts, duel monster, hat, hexagram, large breasts, long hair, looking at viewer, nail polish, open mouth, pentacle, pentagram, solo, staff, wand, wizard hat,
yu-gi-oh!,
<lora:dark_magician_girl:0.7> <lora:dpep2768:0.25>
Negative prompt: lowres, ((bad anatomy)), ((bad hands)), text, missing finger, extra digits, fewer digits, blurry, ((mutated hands and fingers)), (poorly drawn face), ((mutation)), ((deformed face)), (ugly), ((bad proportions)), ((extra limbs)), extra face, (double head), (extra head), ((extra feet)), monster, logo, cropped, worst quality, low quality, normal quality, jpeg, humpbacked, long body, long neck, ((jpeg artifacts))
Steps: 20, Sampler: DPM++ 2M Karras, CFG scale: 9, Seed: 161252005, Size: 512x768, Model hash: 0873291ac5, Clip skip: 2, ENSD: 31337
@Lykon That worked, thanks!
what is the lora:dpep(2)? Where can i find it?
It's a lora by closertpdeath. You can find it on Huggingface
Is this Lora 3.9GB?
Hi,
Can you provide the settings on how you did your training and your training data please.
Thanks
Default kohya settings. They're really not important. Data is 98% of what matters
@Lykon can you provide you training data?
@goodfun danbooru
@Lykon Can you be more precise. You probably filtered and cropped the images, added text descriptions. Did you do it all automatically? What were the steps? Interested in knowing how this all works in detail.
Thanks
@goodfun no need to crop, use bucketing
@Lykon can you provide more details on the steps please?
Details
Files
dark_magician_girl_offset.safetensors
Mirrors
dark_magician_girl_offset.safetensors
dark_magician_girl_offset.safetensors
714_dark_magician_girl_offset.safetensors
dark_magician_girl_offset.safetensors
dark_magician_girl_offset.safetensors
dark_magician_girl_offset.safetensors
dark_magician_girl_offset.safetensors
dark_magician_girl_offset.safetensors
dark_magician_girl_offset.safetensors
dark_magician_girl_offset.safetensors
dark_magician_girl_offset.safetensors
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.