Lora trained on 43336 images of all Genshin Impact Characters.
This does not include Male characters, Paimon, Kirara, or NPCs and will not ever.
This took me over 17 hours to make since I only have a 3070ti 8gb.
I release all my models for free, if you feel like supporting me and my work you can support me here and buy me a drink.
Please note that some characters may not come out perfectly due to not individually being able to select the epochs since this is a combined training and not a merge.
This is dim32 so it is only ~40mb instead of having to download all 42 of the loras separately so I hope some people can appreciate it.
Default Outfit Prompts:
kukishinobudef, luminedef, doridef, fischldef, keqingdef, lisadef, raidenshogundef, yaemikodef, kujousaradef, beidoudef, dionadef, ganyudef, kamisatoayakadef, shenhedef, euladef, rosariadef, qiqidef, layladef, niloudef, kokomidef, yelandef, monadef, barbaradef, candacedef, colleidef, yaoyaodef, nahidadef, faruzandef, jeanfavonian, sucrosedef, sayudef, xianglingdef, dehyadef, yoimiyadef, kleedef, hutaodef, xinyandef, amber5star, yanfeidef, noelledef, yunjindef, ningguangdef
Alternative Outfit Prompts:
fischlein, keqingopulent, lisasobriquet, kamisatoayakaspring, rosariafreespirit, monastarpact, barbarasum, jeanseabreeze, jeangunnhildr, amberpizza, noellekfc, ningguangorc,
Random Prompts:
luminernd, dorirnd, fischlrnd, keqingrnd, lisarnd, raidenshogunrnd, yaemikornd, kujousararnd, beidournd, kukishinoburnd, dionarnd, ganyurnd, kamisatoayakarnd, shenhernd, eularnd, rosariarnd, qiqirnd, laylarnd, nilournd, kokomirnd, yelanrnd, monarnd, barbararnd, candacernd, colleirnd, yaoyaornd, nahidarnd, faruzanrnd, jeanrnd, sucrosernd, sayurnd, xianglingrnd, dehyarnd, yoimiyarnd, kleernd, hutaornd, xinyanrnd, amberrnd, yanfeirnd, noellernd, yunjinrnd, ningguangrnd,
Trained on NAI, gen'd on AOM2 hard
Description
FAQ
Comments (76)
Can I change the trigger word by myself? Some of the name is too long in daily use, for example, change the "kamisatoayakadef" into a simplier "ayakadef"?
Ngl this is probably one of the most impressive LoRa I've ever seen
Thx
Oh my lord! This is just mind blowing. Thank you for your hard work!
Also, I was hoping if I could request for a character?
You can always ask.
@Numeratic I wondering if you could make Fiona and her wedding dress from Tower Of Fantasy. It would be great if can change different outfits for her too! :)
@Tsukilicious I honestly have no idea where to get pictures for her at lol
@Numeratic Lmao no worries! XD
How about High School DxD girls? Lots of pictures of the girls from that series and they are really popular!
@Tsukilicious its just that most of the danbooru source are cards...
@Numeratic Oh, would it be hard then? Nobody had done good lora on them...
If High school DxD isn't possible, Oshi No Ko characters would be great. And again, thank you for the genshin batch file. Really saved a lot of space for me.
Great! Now I can try generating Genshin group images without worrying about different LoRAs polluting each other
I wish you luck I have not tried it myself cuz idk how lmao.
@Numeratic Wasn't thinking about huge groups, just want to generate small groups e.g. Loli all-stars or even simpler, cute couples. ControlNet + LatentCouple should work, will post results here if I succeed
@53rdturtle Keep us updated lol, kinda wanna see if multi-character generation is easier with these character pack style LoRas. I'm too busy to try to do it myself.
Amazing work!
you are my hero!!!! you have done my dream work
I don't know enough of the how or why behind this stuff to properly explain this, but while this works fine with simple prompts there seems to be huge problems with anything too complex. If I take a prompt made while using the individual models and swap this in I'll end up getting a chimera of god knows how many characters instead. If I prompt for Lisa then it spits out a girl with short black hair and an amalgam of every hat in the game. Switch back to the individual model for Lisa and the same prompt spits out Lisa again. If I strip the prompt down to bare bones I can force it to give the character I want and I've tried to use that to narrow down what conflicts in the prompt might be causing the error, but it seems to change with each different character/trigger. I fix the prompt to work for Lisa and if I switch to Klee it spits out someone that sort of looks like Klee with dark brown hair and a white dress.
can you share what prompt you are trying?
Im actually interested in knowing what did you do to quack it up. Like I'm very curious you know.
I'm incredibly impressed. Within a week, I've mastered the training methods for the LoRa model through the clues you provided in your comments. You always manage to do such astonishing work in such a short period of time, and I really admire that. You've become a mentor to me, even though we've never met. I noticed that you haven't added the Kariri character yet, and if you need it, I'd like to offer my labeled Kariri data as a token of my gratitude. I only have over 100 images, but I hope it can be helpful to you.
I also have a small question. When using the Grabber tool, I download over 1,000 images each time, but the precision isn't high. I still haven't finished cleaning up my Zhongli data. I wonder if there's a better way to filter the data, such as removing low-quality or complex background data, or do I just have to delete them manually?
Thanks for the kind words.
I did make an article on how to train vs having to read my old comments.
In regards to grabber I just manually delete. I did show the search filters I used in grabber in the article I made which should help remove lower quality sources.
In regards to the missing characters. Ill probably just make individual loras first before updating this full one. There's kariri and that blue haired girl idk the name of. Maybe ill add paimon but idk if I want to sift through that dataset.
Wow, I haven't even tried this yet but super impressive effort to put together.
You're doing a god's work, please keep it free. I will buy you a coffee
Hello,
well done for this nice job, it's amazing !
By the way, does your message about Kirara means you will never ever do it ? Or do you consider making it, outside of the pack ?
Thanks !
Sweet Jesus... I'm giving this a like simply on account of you managing to shove 42 characters into it and it not being a horrible mess..
Finally nude genshin characters...
大佬,怎么让多个原神角色同时出现在一个图上。还是说只能两个一起出现。
Regional prompter https://github.com/hako-mikan/sd-webui-regional-prompter
@Numeratic 谢谢大佬!
话说大佬有制作水神芙卡洛斯吗?
jeanrnd is broken
Thanks for all the Loras. These work best for me. Can you also do this with the Honkai Star Rail Girls? Pleaassseeeee
我对你的敬仰犹如滔滔江水,延绵不绝!
求教方法,怎么把这么多个角色融合到一个lora,而不相互干扰的。。。??
How i install this?
For A1111 webui, you have to move downloaded genshinfull1-000006.safetensors file in folder stable-diffusion-webui\models\Lora
Then use it in the prompt like this: lisadef <lora:genshinfull1:1>
No boys???
m*les lol
lol im kind of joking but im not fond of scrolling through datasets and seeing male genitalia but im kind of memeing at the same time so idk, might make NumeraticMale account lel
@Numeratic Nether am I, you shouldn't need to have nudes. I hope you do make them at some point.
Amazing!! Added to the big list https://civitai.com/articles/262 Thanks!!
whats the difference between the random and deafult prompt?
Outfit is default, random is random images stored so that it can probably change clothes easier
@Numeratic not like im using any clothes anyways
@neonek1232363 xD , so random would probably do better at it then. I also think you can just use the name without def or rnd attached and it should still work.
Can you add Male characters to the women's collection? so we can use the Lora to ship both male characters and females at the same time, in one Lora, please?
there are many that have downvoted don't know why, I feel If add males to the Lora then you can Prompt Ex. Traveler Aether and Jean in any kind of scenario you want or Hutao and Zhongli ect.
@RanceHikary >This does not include Male characters, Paimon, Kirara, or NPCs and will not ever.
@RanceHikary That's exactly why you are downvoted, the Genshin playerbase prefers hilichurls/ugly bastard partners over male characters in any kind of nsfw escenario. Just look bellow lol
what model adjusts this lora well?
你做的好啊,做的好啊
really cool,but how can i get the old version v1.4
There has only been 1 version of this
AMBER5STAR
lmao
this is the goat of genshin-lora
Amazing work.. <3
Calling it "All in One" is a lil bit overreated when your not adding the male characters as well, dont you think?
Nonetheless.. thank you for your hard work.
Sorry I only had motivation to do the female chars, I would not have the patience to do it if I had to do the males (not attracted).
@Numeratic its okay, really. just sad bout it! Still great work! ❤️
why are you gay
@day411 what kind of strange question is that? 😆
Is it necessary to be gay to want to create male characters? 😯
Fortunately, the world isn't just black and white. 😉
will you add fucalors and new charactors, my lord
😭
Which Checkpoint you are using for this? When i download it, it doesnt work for me, even with the keywords
Hello this uses a Novel AI base model so anything made with SD1.5 would work
I just posted two images using Seaside Mix and Dark Sushi, so I can confirm it works pretty good with those, AbyssOrangeMix3 (AOM3) gives mixed results but also tends to work from time to time.
Details
Files
genshinfull1-000006.safetensors
Mirrors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
genshin_chars.safetensors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
genshinfull1-000006.safetensors
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.

