Trained on NAI.
Best weight: around 0.8
Prompt activator: shannon
Major prompt word (maid outfit): thigh tattoo
Minor prompt word (only if thigh tattoo is not enough, maid outfit): ushiromiya tattooOutfit prompts:
Maid outfit: maid outfit, apron, bonnet, brooch, kneehighs, necktie
Casual outfit: shannon, casual outfit, brown dress, brown clothes, purple bow, short sleeves, hatIf you have any issues with this LoRA or if you have any suggestions, don't hesitate to communicate them.
There are some variations with the tattoo and sometimes with the outfit, but the character is respected.
Description
FAQ
Comments (19)
I thought about some memes of Shannon, but no matter which one I bring out, it seems to be a spoiler.🤔
Yes, it's quite challenging not to spoil. I've included 3 (relatively) similar images to events from the visual novel, but luckily, it's not very obvious.
Otherwise, I struggled a bit with this LoRA; I hope it turns out alright.
@UnknownNo3 Hi! he issue is that I was using a Google Colab Notebook to train my Loras, but it's no longer working. I need to find another method. On Civitai, it's not really possible, 500 buzz points is a lot, especially if I have to start over.
@SOSDANÂ Don't know whether Civitai's training is good and how much buzz it costs each time. I tipped some models and left only 300 buzz, I will give them to you.
@UnknownNo3Â Thanks! I did a test by training a LoRA on Civitai, and it seems ok. The issue is that usually I make like 2-3 attempts to adjust some details that I notice, so it's not a good solution, especially considering it takes 500 buzz to train a LoRA. I'll let you know if I resolve my issue on Colab or if I find another solution
@SOSDAN Looking forward to your next work!❤️ Looks like upgrading hardware is still the final solution better than any online service right now.🤦‍♂️
This is gorgeous, thank you. Any chance you'll be able to do one for spoilers, the true identity of the culprit according to their design in Confessions with those appropriate proportions? I tried using fan-art, but am too new at this so I didn't get any good results but their design has been my main goal for a while.
Hello! Thx, I've thought about it because I really like her design as well, but I'm hesitating a bit because if someone types "umineko" in the search bar, sees a character they don't know, and decides to look it up on the internet, everything will be spoiled...
@SOSDANÂ Yeah.... I get it, haha. I think, at some point, people gotta take responsibility themselves. If you're on Umineko, you should know not to search about the series, and we can't police them and deprive actual fans of whatever content we'd be hindering. Anyway, up to you, of course. I think it'd be really cool, cause I love her design so much and don't see enough art where she's in the dress (and i feel like nobody will really make a model for her lol) but with her correct flat proportions and darker hair and haircut and everything. But whatever you decide, of course. Just letting you know there's interest. The work you create is super cool regardless.
@Tharizdun03Â I looked around a bit, and indeed, I hardly see any fan art of Yasuda... Have you tried seeing if it's possible to start from Shannon's LoRA, change a few prompt words to get a decent image? I can try creating a LoRA with something like 20 similar images of Yasuda, in the same outfit, but I don't see enough of them.
@SOSDANÂ There isn't a ton a ton, but i found a bit I scooped up stuff from yande.re, danbooru, and and gelbooru. If I may, I tried doing it before as my first try on making a lora. If you wanna check out the dataset, I used. Maybe it could help as a base? (https://drive.google.com/drive/folders/13IviFUKdwH8TnSE9UdXLlNo_X8pettn3?usp=drive_link) Altho I'm sure you could get a better result regardless.
I did try with just Shannon, but Yasu does have a slightly different face in most drawings, and her hair is different. You can probably get something close if you really want to using the Shannon LORA, but if it can be done, there are definitely enough changes that you can make an entirely separate one, you know?
@Tharizdun03Â You need to request access in order to access the drive :/
@SOSDANÂ Sorry! That's me being silly, haha. Try now.
@Tharizdun03Â I will try to see before the end of the month. First, I will attempt to generate a LoRA with Natsuhi, and then I will run some tests to see if it works.
Thank you again, and have a great day!
@Tharizdun03 https://zupimages.net/up/23/39/slpy.png
I created a LoRa with the provided images; I have this as a test image, but the difference is quite small.
@SOSDAN Hmm, I suppose. It's up to you if you wanna publish it! I'd be in favour of it. I suppose a flat chest is the only thing missing, but that's something you can do by prompt too.
@Tharizdun03Â The maid outfit is trash (https://www.zupimages.net/viewer.php?id=23/39/kvk7.png) , I am going to try with some images of shannon. You are right, flat is not justice here but I can't do anything about it; the data images are like that.
https://zupimages.net/viewer.php?id=23/39/8lhe.png
I will probably release it tomorrow.
@SOSDANÂ I blame fan-artists not doing her actual proportions justice smh.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.