Takamura Yui/篁 唯依 from Muv Luv Dimensions/マブラヴ：ディメンションズ SDXL for Animagine XL 3.1. V1 is trained with 86 images, 5 epoch, 10 repeat and 4300 steps. Contains multiple style and outfit. Activate using "takamura yui"
V2 is trained with 175 images, redesigned the whole lora and training setting, 10 epoch, 5250 steps, the aim is to increase the accuracy, versatility, quality, and merging the whole art style into one style. V2.3 is small update to v2 to increase accuracy. Pony version is trained using 2.3 dataset, but. it doesn't seem to be as good as v2.3.
THE "VERSION NAMED" LORA IS DESIGNED FOR ANIMAGINE XL 3.1 AND IT'S VARIANT, USING THE "VERSION NAMED" LORA ON PONY MODEL DOES NOT GUARANTEE IT WILL WORK. PLEASE USE PONY VERSION INSTEAD IF YOU WANT TO USE WITH PONY. THE SAME GOES FOR ILLUSTRIOUS VERSION.
Check out the doujin I made based on this LORA!
Takamura Yui Enkou R18 Doujin | Civitai
It's recommended to run Illustrious version with @advokat's Illustrious-XL SmoothFT - smoothFT-SOLID | Illustrious Checkpoint | Civitai. Recommended setting is 28 step, 5- 5.5 CFG, euler a.
This version NOW follows stylized version of Animagine XL 3.1, so you can literally design any style as long as it's available on Animagine XL base model.
IF YOU CAN'T DETECT THE LORA using A1111 or FORGEUI, follow this step;
1 — Go to the "Settings" menu.
2 — Click on the sub-menu "Extra Networks".
3 — Scroll down and click on the option "Always show all networks on the Lora page".
4 — Click on the "Apply Settings" button (at the top of the page).
5 — Go to your Lora page, click the "Refresh" button,
Feel free to support me at Buymecoffee!
Credits to
CagliostroLab Creator Profile | Civitai
Description
V2 has so many adjustments made based on how I improved my LORA making skills. I also added gekka outfit. You can learn from example on how to design the outfit. One of the aims for V2 is to HUGELY increases versatility, in which this model is capable of.
FAQ
Comments (14)
It's great, and I would love to have a pony version.
there is already pony version by someone else. So, I won't be touching pony model for it.
Thank you very much, I will try to get the LORA to include the original costumes.That one did not yet support the original Muv Luv Dimensions outfit, so I wanted a LORA that included it.I will ask the author over there.
@mumumu1295 sure. in case if the owner doesnt want to after few weeks, ill take over it over.
@Madafada1991 Oh, thank you very much.
The V2 Fortified Suit is worse/less accurate than the V1. The legs are completely different.
i'll take note of that. its probably merged altogether because lack good quality images for her fortified suit, and what about the rest?
@Madafada1991 You can find more images of her fortified suit from the artbooks, this is a good one:
https://e-hentai.org/g/1543806/f1c639c52a/
Only other model I've tried is the kimono and that's missing the rope that's tied around her ankles. It may be difficult to find more images of the art but there are lots of photos of the figure based on that art:
https://myfigurecollection.net/item/78535
If you want/need more reference materials or suggestions please PM me.
Another small quip is that the hands/feet aren't as stable on this model as they were on V1, especially getting 5 fingers/5 toes is a challenge.
All that said, V2 is way better for making porn than V1 so really good job on that.
@zip_ee803 thanks for your input. After all thats why the model is in early access first.
@Madafada1991 Happy to help. Do you know if/when you plan to make changes?
@zip_ee803 ill have to redesign it first. Especially for those fingers part. This are the hardest to get it correct it. I cant say for sure. But it will take some time because Animagine XL 3.1 has its own finger model, so Im slightly contemplating to whether redesign the fingers from scratch or let the model do the job. Leaving to the model makes it easier but it will weaken non close up generation. But redesigning from scratch will take even longer time .
@Madafada1991 The fingers/toes worked well on V1 so I don't think you need to remake from scratch. Maybe it's just a hiccup with the compatibility with the V2 model and Animagine.
Been playing around with the 2.3 update and while the outfits appear more accurate, the model is a lot more susceptible to mutations with any terms given weights above 1. It feels more unstable than before.
@zip_ee803 as I said in the description, this is the most balanced between them. Let's say if I adjusted it, there will be finger problem, if I adjusted there, there will be outfit problem, its likely Animagine XL 3.1 limitation. I did like 30 different variant of training setting, and this is the most balanced between all of them. Literally the seven whole days doing this last week. Rendering me incapable of doing other LORAs at same time frame. It's definitely not the same as Pony models as I could just slap in picture, and it will work.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.