Face restoration is optional, may provide better result if not activated.
Use CM as trigger.
Important:
This Lora model is a virtual representation of a natural person and it is important to emphasize that its use must not be illegal or degrading. I want to remind you that any infringement in this regard is the responsibility of the user and not that of the designer of this model.
It is crucial to respect laws and ethical standards regarding the use of this technology. Any abusive or illegal use of this model is strictly prohibited and could result in legal consequences for the user.
As the designer of this model, I take the responsibility of ensuring that it is used in an ethical and responsible manner very seriously. I rely on each user to do the same and to act in compliance with the laws and ethical standards in force.
The tags were made using the WD14, so it'll use a lot of anime pics tags (ie. 1girl, 1boy, bald man, etc...)
Keywords to use this guy: (blkd-cowgirl, testicles, interracial, sex, penis, erection, 1boy, lying, on back, ass, pussy, straddling, girl on top, hetero, cowgirl position, looking back, Latina woman, spread ass)
set the lora weight to 0.6:
<lora:blkd-cowgirl:0.6>
I'll probably prune the keywords in the future, but for now, it's fine.
Works great on:
512x512
512x768
768x512
Be sure to use hires.fix, and restore faces to get high quality images.
Тэги: poseswomansexblackedmature
Токены: blkd-cowgirltesticlescowgirl positionstraddlinggirl on top
Jules Winfield, played by Samuel L Jackson, is the overall deuteragonist of Pulp Fiction Initially he is a hitman working alongside Vincent Vega but after revelation, or as he refers to it "a moment of clarity" he decides to leave to "Walk the Earth." During the film he is stated to be from Inglewood, California.
This model is trained with SFW sources only. If you get NSFW content, well, it's on you.
Please provide constructive criticism.
You can support me here to be able to render better models (https://ko-fi.com/djo226). Supporters may request model, so I will never gonna give you up.
Prompt : SLJ, a man, serious, photography, trending on artstation, sharp focus, studio photo, intricate details, highly detailed, by greg rutkowski <lora:SLJ:0.9>
Face restoration is optional, may provide better result if not activated.
Use SLJ as trigger.
Important:
This Lora model is a virtual representation of a natural person and it is important to emphasize that its use must not be illegal or degrading. I want to remind you that any infringement in this regard is the responsibility of the user and not that of the designer of this model.
It is crucial to respect laws and ethical standards regarding the use of this technology. Any abusive or illegal use of this model is strictly prohibited and could result in legal consequences for the user.
As the designer of this model, I take the responsibility of ensuring that it is used in an ethical and responsible manner very seriously. I rely on each user to do the same and to act in compliance with the laws and ethical standards in force.
The famous Renault FT (also known as FT17) is the first tank in history to have a mobile turret, a design still used today.
Recommended parameters
Weighted around 1 for best results but will vary according to the checkpoint and other settings. You can go as low as 0.5 (it will look more like a generic tank) or you can go high like 1.5 for experimental results.
Trained on SD 1.5 base. Tested on all top 20 checkpoints. Epoch 7 on a total of 10. Images from real tanks and toy tanks.
This is a model that focuses on anime style, it won't work well in realistic style . I recommend using the "SuikaMix" main model . Manual:
LoRA weight: 0.3 ~ 0.4 Trigger words weight: 0.6 ~ 1.2 This model will make cameltoe plump and beautiful . Please obey the laws of your country, don't use this model to make illegal picture . V3 is trained from 768*768 pixels pictures, so V3 has better performance in generating high-resolution pictures than V2 . When using this model, please reset the "Height" of the picture to 768 pixels or more, or using the Hires. fix . If you like this model, please give me a good review .
Another one of the most requested JP_VA model. please try this one as well.
(As Requested) Trained on 54 photos with ChillOutMix, work best on DPM++ SDE Karras sampling steps 20-30 CFG Scale 6. weight 0.8-1.0 is recommended for better results.
lightweight LoRA, works on any generator (easydiffusion,stablediffusion etc). share your results.
My Ko-Fi Shop is now open. Get more Epochs of your favorite model, new unpublished and premium models, as well as exclusive image sets here : https://ko-fi.com/meantweetanthony/shop
Thankyou for downloading. Please leave a rating and comment if you like my models. also leave a comment on what model would you like to see next. Please consider to leave a tip if you like my work 😌☕️ enjoy
Disclaimer : Usage of my LoRA models is 100% under your own responsibilities. I bear no responsibilities on however the models are used and it is beyond by jurisdiction.
LoRA trained to generate infirmary backgrounds, together with characters in different positions and poses, or just the background itself.
Please rate and leave a like if you enjoyed it! And be sure to follow to check out the upcoming models.
Trigger: infirmary
Optional useful tags: curtains, hospital bed, etc. Be creative with your prompts and post them here!
Additional info
I make these models for fun and for the challenge and release them for free for everyone to enjoy, for non-commercial use (see license). The one thing I ask of you is to provide rating and feedback! That helps me understand where I can improve and helps me stay motivated to produce more content!
I accept suggestions or requests here or on Civitai discord server.
You can also check my more polished images and upcoming models on my pixiv profile.
p.s. this is not a version update, this is a set of characters in the future it will be replenished :)
You can use the hints from the pictures :)
Тэги: animecharactersexygirlscharactersgirlpackfemalcharacterpackbeelzebubi’ve been killing slimes for 300 years and maxed out my levelslime taoshite 300-nen, shiranai uchi ni level max ni nattemashitaazusalaikaazusa aizawaaizawa azusaадзусаraikaアズサ / 相沢 梓ライカлайкаベルゼブブвельзевулpecoraпекораペコラ
Checkpoint trained on 102 images of swords of various kinds. It can generate interesting images but I have not tested if those can be inserted in a context (like in the hands of a subject). Try yourself and see.
The trained modifiers are: ICE, FIRE, MAGIC, CHAOS, NATURE.
I also added RAPIER, SCIMITAR, FLAMBERGE as additional ones, but I cannot guarantee if they were understood correctly.
Lora model for the fictional character Sten Berglund from the virtual reality roleplay Neon Divide. Trained on screenshots of the 3D model.
Main models used in the images attached:
abyssorangemix3AOM3_aom3a1b
neverendingDreamNED_bakedVae
The metadata was lost for some of the images. Trained on the model AbyssOrangeMix2_nsfw using Booru tags. The training images includes outfits that are armored, jacket, shirt and barechested which you can all use specific tags for.
This is a LoRA for puffy nipples. Who doesn't love those!? It favors smaller breasts, and it also has rather good "texturing" of breast skin and nipples. Just use the "puffies" keyword to help engage it.
I recommend a strength of approximately 0.7
Got a pretty good bake on it, ~300 images for 16 epochs for about 100k steps.
Did you know that not all my LoRAs are hosted on CivitAI? Check out my Patreon to learn more!
What's this LoRA
This is a LoRA trained on commissioned pictures from a user on DeviantArt. It has a nice western cartoon style with a bias towards smaller body types.
My models are not trained to achieve the exact style of their source, but to change the general vibe of the image. They are also fine-tuned on anime models, so they may not work perfectly with realistic ones.
The images in the examples are created using only this LoRA and different models (mainly my custom):
best quality, Elizabeth Olsen, 1girl, solo, long hair, breasts, looking at viewer, smile, brown hair, black hair, cleavage, upper body, lips, realistic <lora:elizabeth_olsen:0.6>
Created according to the guide provided by konyconi, cheers to them for the awesome work.
The LoRa aims to produce an art style that resembles the aesthetics of the Zerg from Starcraft. Especially good at creating creatures, otherwise it tends to insert zerg-like creatures into the image, even if no creature is specified in the prompt. I'm not entirely satisfied with the result, so I might revisit it later.
Make sure you are aware on the usage instructions of LORA
Sampler: DPM++ SDE Karras (Recommended for best quality, you may try other samplers)
CFG Scale : 5 to 10
LORA weight for txt2img: reccomended 0.5-0.9.
Keys: side_view_perspective
My advice is to use the keyword with 1.3 weight and atleast 0.5 on the model.
As far as i could see it works with every other model, but it might need to inpaint the face after generation.
Seems to work very well with anime.
Enjoy!
If you'd like to support my efforts and have access to early and bleeding edge builds, please think about joining my Patreon: https://www.patreon.com/betterCuteAsians
With this model, you can make the iconic backgrounds of early Hanna Barbera cartoons. It's not trained on character design, in fact no image has any character in it.
I have decided to split the model into nature, as you would find in shows such as Yogi Bear, and civilization, where you would find the indoors, city, house models you'd see in other shows. Because putting them together did NOT work out well.
By the way, I see at least some liked the model, if you generate some awesome things, you can share, I'd love to see :D
Miss fortune video game league of legends work with 0.6 weight enjoy it
tiggers words:
miss fortune
follow me on Instagram @creativehotia
Тэги: animecharacteranime modelvideo game characterartstylehentaianimel-yangyipowergirloriginal characterartworkpatreonwomen on topskinpublicationanime ultra high resolution{1girl}voluptuous womenanime stylewoman anime
If you don't add short hair to the prompt it keeps trying to add a ponytail, twintails or ahoge for some reason, even though none of the dataset had any of those.
If you don't add brown eyes the eyes often turn black or blue.
Feel free to share some of the pictures you create with me and post them here!
Тэги: animecharactermovieanime girlmovie characterpaprikasatoshi kon
Sincerely invite you to try this Lora. I am eager for your feedback and comment. It would be great if you can share the beautiful images generated by this lora. Thanks~ The training expectation of this LORA only contains the dress, not including other elements such as faces and backgrounds. You can mix this LORA with elements (for example : face) of other lora or checkpoint to create a complete picture. If this LORA pollutes other elements, pls let me know by comment. I will try to fix it.
I will keep updated Lolita fashion dress by this post.
Recommend LORA weight : You can try 0.7 first. If you feel that other elements are polluted, please reduce the weight as appropriate. It is not recommended to exceed 0.85.
Make sure you are aware on the usage instructions of LORA
Sampler: DPM++ SDE Karras (Recommended for best quality, you may try other samplers)
CFG Scale : 5 to 10
LORA weight for txt2img: reccomended 0.5-0.9.
Keys: top_view_perspective
My advice is to use the keyword with 1.3 weight and atleast 0.5 on the model.
Doesn't work with any other model, i tryied many, it worked with most of them, but with some of them didn't reach the expected perspective in the generated images.
It's very likely to get a selfie with this perspective since most of the training images are selfies.
Seems to work very well with anime.
Enjoy!
If you'd like to support my efforts and have access to early and bleeding edge builds, please think about joining my Patreon: https://www.patreon.com/betterCuteAsians
//info will be updated later(on monday i think, after i start to upload other characters)
Why my LoRa's? Because it's close to anime style, and this style is ~simple in details and can change without mutating target style more easy than artistic one. And it's hard to get quality anime style pictures from LoRa's with more artistic dataset with gradients and complex lighting.
holding a lightsaber,energy_sword, glowing, glowing_weapon,
Prompt:
darth vader standing in front of a group of storm troopers in a room, holding lightsabe in his hands, walking across a bunch of stormtroopers behind him,
Black_outfit,hood, hood_up, boots, darth vader helmet,cape,armor, shoulder armor, armored boots,
glowing_sword, glowing_weapon,holding_sword, holding_weapon, hood, laser, lightsaber,holding a lightsaber,energy_sword, glowing, glowing_weapon,
If you encounter a situation where the LoRA style cannot be recognized, please consider changing the checkpoint model, especially for some checkpoint models with strong styles. If the output images are quite unstable, you should also consider changing the checkpoint model.
权重建议值：1 (or 0.6~1.4)
基于模型：Anything-5RE
训练信息：筛选出共计278张原画（含裁剪图），迭代1112步，多分辨率智能定向迭代训练。
其他的一些建议：EasyNegative，badhandv4，CreationMix。
Recommended weight value: 1 (or 0.6~1.4)
Based on model: Anything-5RE
Training information: Selected a total of 278 original drawings (including cropped images), iterated 1112 steps, and trained with multi-resolution intelligent directional iteration.
Other suggestions: EasyNegative, badhandv4, CreationMix.
This lora is mainly used to generate characters who are bound by the frame. You can use the keywords "bdsm frame,bondage, <lora:bdsm_frame:0.6>," to trigger it. It is usually used to generate males and it will have a good effect, but unfortunately It doesn't work very well when used to generate female characters. Because I am gay.
Recommended size 640x1024, model: Basil_mix_fixed.safetensor or realMaxV34_v34.safetensors
Weight can go from 0.7 up to 1.0. I recommend 0.9. ControlNet will streamline full body shots much easier, or more complex poses. Works with almost every Style and NSFW LoRA's.
This was requested by Айзек.
The Main triggerword is Slimecore. You can also choose color of your Slime.
WARNING: This LoRA is not super flexible. Slimecore is sometimes missing and somtimes it's more Human then Slime.
Trigger Word:
Slimecore
Helping Prompt: monster girl, slime girl
How to use LoRA's in auto1111:
Download LoRA
Copy the file to stable-diffusion-webui/models/lora
Select your LoRA and use the triggerword
Make sure to change the weight if needed (by default it's :1)
*Show your support and leave a post, comment and review! ⭐
*Custom LORAs available upon request. Just buymeacoffee.☕
*(generation advice below) 👇
🌟Veronica Leal🌟
born on December 17, 1993, in Cucuta, Colombia, is a popular film actress and model. She is known mainly for her acting performances in videos and web scenes. In 2017, at 24 years old, Leal began her career in the entertainment industry with her first shoot for the film studio, Lets-Doeit. She went on to work with several other film studios, including Fake Hub, TMW Network, A.S. Media, and MetArt Network.
Leal has also performed in videos with other famous actresses such as Aislin and Lilu Moon. She has a distinctive figure, standing at a height of 162 cm (5 feet 4 inches) and weighing 47 kg (103 lbs). Her figure measures 32C-22-32, with beautiful blonde hair and brown eyes. She has tattoos on her right flank and right shoulder blade, and her left nostril is pierced.
When not in front of the camera, Leal enjoys spending her free time in nature, hiking, and reading. She is also a dog lover and a swimming enthusiast. Leal maintains a big base of fans on her social media profiles and is active on Instagram under the username Veronica Leal Official.
💡Generation Advice💡
🔧Prompt and Sampling🔧
Most images have been generated with LORA weighted @ .75 - 1.
Most samples done with EdgeofRealism with sampling steps best above 20.
LoRA model imitating Piro Mizu's painting style. It seems very suitable for drawing a tsundere look.
Recommended weight value: 1 (or 0.6~1.4)
Based on model: Anything-5RE
Training information: Selected a total of 348 original drawings (including cropped images), iterated 1044 steps, and trained with multi-resolution intelligent directional iteration.
Other suggestions: EasyNegative, badhandv4, CreationMix.
tsumugivamp, two-sided fabric, 1girl, two-sided cape, vampire costume, red cape, mini hat, pink hair, bangs, black cape, black headwear, purple eyes, skirt, twintails, red ascot, top hat, long hair, white shirt, mini top hat, fang, frills, black corset, large breasts, frilled skirt, collared shirt, long sleeves, fake horns,
Code Vein was a great game with great waifus, among them being best boss girl Invading Executioner, to whom many enraged and pissed-off Revenants died time and time again. However, my deaths to this beautiful woman did not diminish my love for her. Instead, my horny only grew stronger, and so here we are with this tributary lora.
There are 3 versions:
Normal is trained on AnyLora (so it's saturated asf)
Desat is trained on NAI (so it would generate less saturated results)
Free is trained on also NAI, with a diversely tagged dataset in the hopes of making it easier to manipulate the character design. Not sure I really succeeded with that, but it's there if you want it.
TIPS:
All three can generate her outfit by itself through weight lowering and removing her descriptive tags, though the outfit itself doesn't have a tag assigned to it.
I didn't actually intend for this, but adding "3D" to the prompt makes it much sharper and more detailed.
Token is not consistently required, a short description is often enough.
Her black scleras (whites of her eyes) are not consistent, I recommend using this lora to keep it frequent.
Tested on both realistic and anime models. Mildly NSFW capable.
Start the prompt with either: 1girl - Typical anime style. chibi - Chibi style, like her official art.
Then use cbmrk to summon her, use black hair, short hair, black eyes to form her head, use suspenders, red skirt, white shirt to summon her default clothes.
I noticed a distinct lack of 80's characters on here and being a Gen-X'er, I had to rectify that issue. So, I have started my Awesome 80's series with the one and only Sorceress of Grayskull from Masters of the Universe. This Lora was trained on images from the 1980's Funimation cartoon and product images for the WETA/Sideshow maquette as well as the 80's era toys. I did not use images from the live-action movie nor the new Netflix series.
Image was trained on SD1.5, so it is good for most other checkpoints. I would recommend setting the strength to 0.8-1. Highres.fx = VERY YES. I usually go about 1/2 of my initial steps with a denoising strength of 0.3-0.4
Activation word is "motu_sorceress". To get the classic look, use "beautiful woman wearing a blue and white long sleeved v-neck bodysuit, blue and white long sleeved v-neck bodysuit with feather detailing, eagle headdress, orange feather cloak, blue boots with white fur trim". If you leave those out, you get a pretty cool looking if generic superheroine/sorceress. Examples are in the photos.
Good luck, have fun, and stay tuned for more 80's icons!
Тэги: character80'smotuwomansorceressmasters of the universe
Токены: motu_sorceressblue and white long sleeved v-neck bodysuit with feather detailingeagle headdressorange feather cloakblue boots with white fur trim
Lora for generating from Street Fighter in their female form.
Some additional words can help you with generating character closer to how they look in the game: "hair rings" and "halo" for hair, "yin yang" for the symbol on their stomach, "metallic skin", "grey skin", "neon trim" for the body
How are you doing? I hope that everyone is doing great ^u^ I am currently sick at the moment. I am very sorry for not sharing more models lately.
Furthermore....
I have never been this humiliated before in my entire life by a person, let alone a Vtuber. So I was commissioned to make an artwork for her. When I delivered the order, she cancelled the order... I did not sleep when making the artwork for her. So I did not get paid for my work T-T.
I feel duped... and to bend my anger for this VTUBER ! I decided to send the lora of her here so you can create whatever you want with her character. Be it for SFW or NSFW, be it for meme or if you're a fan of Faithybun, feel free to use this LORA.
I wished all the best for her in life and I hope she would treat other people nicely.
If you encounter a situation where the LoRA style cannot be recognized, please consider changing the checkpoint model, especially for some checkpoint models with strong styles. If the output images are quite unstable, you should also consider changing the checkpoint model.
权重建议值：1 (or 0.6~1.4)
基于模型：Anything-5RE
训练信息：筛选出共计208张原画（含裁剪图），迭代832步，多分辨率智能定向迭代训练。
其他的一些建议：EasyNegative, badhandv4, CreationMix。
Recommended weight value: 1 (or 0.6~1.4)
Based on model: Anything-5RE
Training information: Selected a total of 208 original drawings (including cropped images), iterated 832 steps, and trained with multi-resolution intelligent directional iteration.
Other suggestions: EasyNegative, badhandv4, CreationMix.
My first Lora) I wanted to make it better, but it's really hard to find a good source images.
I recomend using weights somewhere between 0.6-1.
Tags
There are should be some differences when you'll use this words (you doesn't have to use it):
tiptoesFuck, from side, from front, from behind, bent, standing, penis
Also I recomend using:
sex, standing_doggystyle
And for the negatives:
Loli, dwarf, hug, 2girls, 2boys, threesome, yuri
There are lots of arm grab and hair pulling in the source images. But for some reason i couldn't been able to invoke them.
Versions differencies
v1 is generic version
v2 New verion, which is slightly better with limbs, charactrers bodies and position. (I haven't used upscale on the examples for this version)
v3 Just works better, but affects style much stronger. I don't know why, but it works bad in anything v4.5. You can use weight 0.8-1 for anything, and then use imgToimg for details. Worked out for me.
v4 Now it gives you more control but also it's much harder to trigger. HiresFix sometimes better. Sometimes not)
*still trying to make it work as intended, so I hope I'll make a better version someday
Good results are unstable, you'll get a 2 good images out of 10 or less. So it needs some patience) Also ImgToImg randomly fixing some strange limb connections.
Enjoy)
Feel free to ask any questions here in comments, or in my discord channel.
Also I am making games with AI arts. They will be free in the future, but if you want to participate in making or have an early access - you can support me on patreon)
Also my wet hair Lora was banned here, I dunno why. You can download the last version of it from my patreon for free.
Тэги: pornposessex from behinddoggystylehentaitiptoesdd
My first Lora. Works fine with weights around: 0.7 Trigger is CassCain Prompt example: CassCain, cassandra cain, batgirl suit, 1girl, black hair, brown eyes, looking at viewer, mask, short hair, face scar <lora:CassandraCain-01:0.7>
Sit down ladies and gentlemen, we're going to take gooood care of you.
Download the LoRA of your choice and get ready to rediscover your hair!
PLEASE PAY ATTENTION TO LoRA WEIGHT!
At a certain level your character might get changed to get the face of the model from whom I took the hairstyle from. Be mindful of the weight and the model will only give you the hair and leave the aesthetics of your character alone.
I made this LoRA to test how I can create one with just few images.
I started only with 22 images, some of which are quite low quality. I trained a LoRA on these images, generated 15 good images, painted over weird errors and then put them together to train a new LoRA on the 22+15=37 images.
Looking at the tests on four different models, epoch 14 seems to be the most consistent, so that is the one I upload here.
Activation tag is just Roberta. No need to specify blue skin or black hair, but adding furry and animal ears seem to help.
You can get girls change into cuter dresses with this LORA, These dresses are probably Japanese style, but I'm not sure because I don't know much about girls' clothes.
Don't set lora's strenth too high or the imgs will go wrong.
"dress" is the trigger word to activate the lora, and you can use your imagination to edit this word, such as "plaid dress", "printed plaid dress", "printed blue/yellow/white/green dress", "colorful dress" etc.
Thanks to the help of 搞事德·乔斯达 and the inspiration from one mega's owner (sadly I do not know his/her name).
感谢搞事德·乔斯达的帮助以及mega盘的一位大佬带来的灵感（可惜我不知道TA叫啥）。
This lora will bring you super cool Prosthetic Quadrupeds/Mechanical Beasts. Also, it is another Mecha Musume lora.
该lora将会带给你极其酷炫的四足机械兽。此外这也是个机娘lora。
If you want to get a Mechanical Beast Girl, you should lower the strength of this lora a bit like setting it around 0.9, add strength to 1girl and describe the girl's head's features like the color of her hair and eyes.
Sizes of imgs might have influences on the output, it is suggested to set width larger than height.
图片尺寸对结果可能有影响，建议将图片宽度设置超过高度。
"all fours" and "crawl" will help with the pose, just input one or both of them. Well, if you just want a standing girl, that's also ok, just input "standing" instead.
"all fours" 和 "crawl" 对姿势有帮助，输入一个或是全输入比较好。不过如果你想要个站着的机娘，也可以，输入standing就行。
Suggested strength: 0.55-0.75 should work with every prompt. The LoRa works with other character LoRas really good in my testings.
Suggested prompt: "<lora:selfspread:0.65> raw photo of woman (selfspread), (taking a selfie in a mirror):1.3" than adding description of woman + usual qualitypromptwords.
The given activationword is "selfspread" most of the trained images was with girls spreading legs in front of mirror so this post produces the best images in my opinion.
By adding the word "mirror" generated picture will show the borders of the mirror, personally like it gives it a more realistic look. For different poses i usually use "kneeling" "legs up" "standin" or "squatting."
This LoRa doesnt give good results with Anime models! And was made and tested with realistic models and a single comic model.
Hope someone can use this. If you like share your results and leave a comment. Have fun.
Sincerely invite you to try this Lora. I am eager for your feedback and comment. It would be great if you can share the beautiful images generated by this lora. Thanks~ The training expectation of this LORA only contains the dress, not including other elements such as faces and backgrounds. You can mix this LORA with elements (for example : face) of other lora or checkpoint to create a complete picture. If this LORA pollutes other elements, pls let me know by comment. I will try to fix it.
Recommend LORA weight : You can try 0.7 first. If you feel that other elements are polluted, please reduce the weight as appropriate. It is not recommended to exceed 0.85.
This is Zelda's short-haired version in Tears of the Kingdom. If you prefer the long-haired version, please refer to my other model.
https://civitai.com/models/69139/the-legend-of-zelda-princess-zelda
Recommend Lora weight(推薦lora權重/おすすめの Lora ウェイト)
0.4~0.7
Recommended prompt(推薦提詞/おすすめのプロンプト):
<lora:zelda_shorthair_v5:0.6>,zelda\(princess\), 1girl, solo, breasts,short_hair,
You can choose not to include the tag for 'short hair', but if you do, it will be more stable.
achieve the same effect as me, make sure to enable 'Hires.fix'.
If you enjoy my work or models and would like to commission me for training assistance, feel free to contact me through Twitter or Facebook. Love you all, and I hope you have a great time!
⭐Linktree -> https://linktr.ee/1zgame
⭐Twitter -> https://twitter.com/love_1z_game
⭐Instagram -> https://www.instagram.com/1zgame/
⭐Facebook -> https://www.facebook.com/1zgame
Тэги: characterfemalethe legend of zeldaprincess zeldazeldawomangame characterゼルダの伝説tears of the kingdomゼルダ薩爾達ティアーズオブザキングダム王國之淚
Make sure you are aware on the usage instructions of LORA
Sampler: DPM++ SDE Karras (Recommended for best quality, you may try other samplers)
CFG Scale : 5 to 10
LORA weight for txt2img: reccomended 0.3-0.7.
Keys: long_ponytail_hairstyle
Very simple to use, just use the keyword, maybe with a weight of 1.3.
Many of the images are generated using controlnet poses packs that you can find in my page.
Works also with anime.
Enjoy!
If you'd like to support my efforts and have access to early and bleeding edge builds, please think about joining my Patreon: https://www.patreon.com/betterCuteAsians
Rin Kanzaki from Isekai de Cheat Skill o Te ni Shita Ore wa, Genjitsu Sekai o mo Musōsuru.
The examples are with weight 0.9 - 1.0
Triggers with rin, ((rin))
If you have advice on other LoRAs tell me and I will do my best to accommodate!
Follow me if you are interested in any of my other LoRA
If you enjoy the work I've done you can show your support by buying me a coffee!
Тэги: animecharactergirlloraisekai de cheat skill o te ni shita ore wa, genjitsu sekai o mo musōsurui got a cheat skill in another world and became unrivaled in the real world, toorin kanzakikanzakirin
Make sure you are aware on the usage instructions of LORA
Sampler: DPM++ SDE Karras (Recommended for best quality, you may try other samplers)
CFG Scale : 5 to 10
LORA weight for txt2img: reccomended 0.3-0.7.
Keys: ponytail_weave_hairstyle
Very simple to use, just use the keyword, maybe with a 1.3 weight.
As you can see many of my images have a perfect hime cut hairstyle not showing the fronthead, in those images i used my model Long Hime Cut Hairstyle in inpaint, first i generated the images as i wanted with ponytail weave model, then i sent to inpaint, masked the fronthead, checked fill checkbox and wrote in the prompt only this: "(long_hime_cut_hairstyle:1.3),<lora:long_hime_cut_hairstyle:0.5>" and then generated to have a perfect hime cut hairstyle.
Remember if you are inpainting an image after the other that before sending to inpaint your image you have to close the last image from img2img or in the new generation it will remember the old masked part other then the masked part of your new image, doing a mess.
Also many of the images are generated using controlnet poses packs that you can find in my page.
Works also with anime.
Enjoy!
If you'd like to support my efforts and have access to early and bleeding edge builds, please think about joining my Patreon: https://www.patreon.com/betterCuteAsians
Make sure you are aware on the usage instructions of LORA
Sampler: DPM++ SDE Karras (Recommended for best quality, you may try other samplers)
CFG Scale : 5 to 10
LORA weight for txt2img: reccomended 0.3-0.7.
Keys: very_long_hair
Very simple to use, just use the keyword, maybe with a weight of 1.3.
Many of the images are generated using controlnet poses packs that you can find in my page.
Works also with anime.
Enjoy!
If you'd like to support my efforts and have access to early and bleeding edge builds, please think about joining my Patreon: https://www.patreon.com/betterCuteAsians
Necro Sketcher is an LoRA model, specifically trained to emulate the gothic, sublime aesthetic of the iconic game, Darkest Dungeon.
This model was rigorously trained using a dataset comprised of game screenshots from Darkest Dungeon 2. This ensures that each piece of generated art maintains the distinct, haunting aesthetic that fans of the game have come to recognize and love.
darkest dungeon (I recommend using these word in every prompt), portrait, full body shot, 1girl, 1boy, male, male focus, female, woman, hood, mask, hat, eye path, the grave robber, the occultist, the flagellant, the man-at-arms, the plague doctor, the vestel, the highwayman, the jester, the hellion, house, architecture
Make sure you are aware on the usage instructions of LORA
Sampler: DPM++ SDE Karras (Recommended for best quality, you may try other samplers)
CFG Scale : 5 to 10
LORA weight for txt2img: reccomended 0.3-0.7.
Keys: long_hime_cut_hairstyle
As you can see many of my images have a perfect hime cut hairstyle not showing the fronthead, in those images i used my model in inpaint, first i generated the images as i wanted, then i sent to inpaint, masked the fronthead, checked fill checkbox and wrote in the prompt only this: "(long_hime_cut_hairstyle:1.3),<lora:long_hime_cut_hairstyle:0.5>" and then generated to have a perfect hime cut hairstyle.
I didn't do that for all the images, some of them are just generated, but they are not perfect hime cuts. So, remeber to inpaint if the generations is not how you want it.
Remember if you are inpainting an image after the other that before sending to inpaint your image you have to close the last image from img2img or in the new generation it will remember the old masked part other then the masked part of your new image, doing a mess.
Also many of the images are generated using controlnet poses packs that you can find in my page.
Works also with anime.
Enjoy!
If you'd like to support my efforts and have access to early and bleeding edge builds, please think about joining my Patreon: https://www.patreon.com/betterCuteAsians
(As Requested) Trained on 54 photos with ChillOutMix, work best on DPM++ SDE Karras sampling steps 20-30 CFG Scale 6. weight 0.8-1.0 is recommended for better results.
lightweight LoRA, works on any generator (easydiffusion,stablediffusion etc). share your results.
My Ko-Fi Shop is now open. Get more Epochs of your favorite model, new unpublished and premium models, as well as exclusive image sets here : https://ko-fi.com/meantweetanthony/shop
Thankyou for downloading. Please leave a rating and comment if you like my models. also leave a comment on what model would you like to see next. Please consider to leave a tip if you like my work 😌☕️ enjoy
Disclaimer : Usage of my LoRA models is 100% under your own responsibilities. I bear no responsibilities on however the models are used and it is beyond by jurisdiction.
D'Artagnan: "Votre excellence, si vous ne souhaitez pas que je me rebelle contre vous, alors faites en sorte qu'il y ait plus d'honneur à vous servir qu'à vous combattre."
Ainsi parlaient les vrais hommes de jadis.
D'Artagnan: "Your excellency, if you do not wish for me to rebel against you, then make it so there is more honor in serving you than fighting you."
(As Requested) Trained on 54 photos with ChillOutMix, work best on DPM++ SDE Karras sampling steps 20-30 CFG Scale 6. weight 0.8-1.0 is recommended for better results.
lightweight LoRA, works on any generator (easydiffusion,stablediffusion etc). share your results.
My Ko-Fi Shop is now open. Get more Epochs of your favorite model, new unpublished and premium models, as well as exclusive image sets here : https://ko-fi.com/meantweetanthony/shop
Thankyou for downloading. Please leave a rating and comment if you like my models. also leave a comment on what model would you like to see next. Please consider to leave a tip if you like my work 😌☕️ enjoy
Disclaimer : Usage of my LoRA models is 100% under your own responsibilities. I bear no responsibilities on however the models are used and it is beyond by jurisdiction.
Apple Daily (Hong Kong) was a Hong Kong pro-speech freedom newspaper/ media. It was shut down by the China Communist Government by accusing them violating the Hong Kong National Security Law in 2021, right after the Anti-Extradition Law Amendment Bill Movement.
They have nostalgic CG illustration to assist their news reporting. Here is a lora to micic and revive their unique visual style.
Suggested weighting at 0.8-1. It seems not obedient to prompts, likewise those naughty Hong Kong people go against their great grandfather. Just let the lora flies.
Private Trading and Public Trading of this LoRA Data is NOT allowed.
Please use this LoRA Data at your own risk. I can not be held liable or responsible for any damages caused by using this data.”
This LoRA Data may not be used for any works which contain negative content toward the original product, individual, &/or group.
This LoRA data is made for personal use only. Please do not post images that expose the lower half of the body or suggest sexual depictions on sites that can be viewed by an unspecified number of people.
AI images created from this LoRA data are not permitted to be sold in any form, whether digital or printed. Please do not use this LoRA data for commercial purposes. A subscription type like PIXIV FANBOX is also useless. It is also not good to post a URL link of a paid service on a page with a healthy AI image to guide it.
I'm not releasing LoRA data to fill your wallet
Use the Lora at 0.45-1 weight for optimal results. If the parts of the variable suit are excessive, please lower them
Required prompt: leon fou bartford\(mobseka\), black eyes, short hair, antenna hair, bangs, black hair, hair between eyes,
close up
I used google translate. I think there may be mistranslations, but please accept that I am serious about writing in Japanese below.
The image source and training code come from the Internet, and the model is only used for exchange of scientific research interests. If there is any infringement, please contact to delete, thank you.
LoRA model of Reina Hidaka (日高零奈) from media-mix project Denonbu (電音部).
This model is based on animefull-latest; the trigger word is mimito sakurano (denonbu). Considering that the hairstyles of character has changed to short hair, here are two sets of trigger words for hairstyles: very long hair, twintails, OR short hair.
The LoRA weight of samples is 0.75; you can also try other weights.
Note: I am not using the official LoRA loader, so you should use the additional network WebUI plugin to use my LoRA setting. Moreover, due to the small data set, it may be difficult for the character to change clothes.
If you see anyone selling images that generated by this model on Pixiv or other platform please message me.
————————————
跨媒体企划《电音部》的角色樱乃美美兔的LoRA模型
本模型基于animefull-latest训练，触发词为 mimito sakurano (denonbu)。考虑到后期角色的发型变为短发，此处提供了两组发型触发词：very long hair, twintails，或者short hair。
This LoRA can be applied on almost all checkpoints.
The recommended weight is about -2.0~2.0. The higher the weight is, the brighter the images will be. (while the negative weight can decrease the brightness)
Stripping nude weighted weight, hairpin recognition is not very good yet. AI has recognized it as an earring, and may refine it again when it is drawn.
Тэги: character
Токены: animal ears,green eyes,red eyeshadow, bangs,tassel, red ribbon, brown dress,white dress,earrings,jewelry, bracelet, ponytail holder, tail,fur-tipped tail, clothing cutout, brown hair，1girl
(As Requested) Trained on 54 photos with ChillOutMix, work best on DPM++ SDE Karras sampling steps 20-30 CFG Scale 6. weight 0.8-1.0 is recommended for better results.
lightweight LoRA, works on any generator (easydiffusion,stablediffusion etc). share your results.
My Ko-Fi Shop is now open. Get more Epochs of your favorite model, new unpublished and premium models, as well as exclusive image sets here : https://ko-fi.com/meantweetanthony/shop
Thankyou for downloading. Please leave a rating and comment if you like my models. also leave a comment on what model would you like to see next. Please consider to leave a tip if you like my work 😌☕️ enjoy
Disclaimer : Usage of my LoRA models is 100% under your own responsibilities. I bear no responsibilities on however the models are used and it is beyond by jurisdiction.
26th Release, 鏡流 - Jingliu From Honkai: Star Rail - 崩壊スターレイル.
Put everything in Filter for seeing all images!
Click on Show more for Weight & Trigger recommendation, Also click on About This Version at Right Corner for more Info.
Warning: It's the Very Very first BETA Version, So It's actually Hard to Use, i will absolutely Update when More Original Art & GameScreencaps are released.
It's was a Challenge + A Request, It's the New Character's from Honkai: Star Rail, it has leaked & I tryed to create a Lora of it with Extremly Low Datasets, i used the same Technique as Sushang & Cocolia Lora, tho this Lora since it's the First version he's Inconsistent with Official Dress.
About this Lora:
Lora trained on 44+ Images at the max quality possible. (Atm she's have like 4 Art at Max, I used the same Technique as Sushang & Cocolia Lora).
Trained at 512x Resolution.
NAI Used, Work with Majority of the Anime Model's.
Flexibility is ''Ok''
Inconsitent with Official Dress (Not Very Precise), Will be far better when more Art are Released.
For Prompts & Weight:
Flexibility is Ok. (See examples images).
About Trigger Words & Prompts: JingliuV4 (For Trigger) & for prompts just refer to examples images.
Weight recommended: 0.6-0.8. Actually it's work at Weight 1, but i don't recommend for Quality.
Leave a Review & Share you're image in comments if you appreciate this Work :).
Тэги: animecharacter崩壊スターレイルgame characterhonkai star railgame1girlstar railhonkai: star rail崩坏星穹铁道スターレイルjingliu镜流鏡流
I am not the original creator of this, I've just reuploaded it due to popular request. If the original author wants it removed, or if the site mods removed this for some specific reason, let me know.
Princess Zelda from Tears of the Kingdom. Trained at 768x768 using in-game cutscenes.
KEEP THE WEIGHT AROUND 0.75.
I just beat the game and HAD to make my own LoRA, I don't mind if there already are others. I just love the game so much! The ending was incredible and I hope you enjoy the game as well.
Anyways, this LoRA turned out BEAUTIFUL. I used a lot of images with GREAT lighting so that Princess Zelda SHINES beautifully.
I ONLY used images where she is wearing that new Zonai dress we see her wearing since the beginning of the game.
Here's a list of words you can add to your prompt straight from the training dataset:
jewelry
earrings
pointy ears
bare shoulders
necklace
braid
strapless
crown braid
triforce
strapless dress
collarbone
hair ornament
outdoors
bracelet
alternate costume
alternate hairstyle
thick eyebrows
Тэги: characterthe legend of zeldaprincess zeldagame character
downgrade the cuteness of No.177 Girl by merging No.169 Girl
No.169 Girl is boyish
add a bit more natural style LoRA like Dr.She - Dr.She is a nice example. it is not so pretty and not so cute but well made.
add high-quality LoRA a bit like as KoreanDollLikeness. it makes LoRA more robust, it has cleaner skin nice smile, etc.
you may need to use OUTALL weights preset only - so many low-quality LoRAs have not optimal IN* weights and it makes trouble. simply use only OUTALL weights to solve this issue.
the obtained and used ratio is No.177 : No.169 : Dr.She : koreandolllikenessV20_v20 = 0.5 : 0.4 : 0.15 : 0.02
No.177 and No.169 are remade its dimension to 128. (the remade dimension version is named NO177, NO169 in each case.)
use only OUTALL weights like as "<lora:NO177:0.5:OUTALL>
<lora:Dr.She:0.15:OUTALL>
<lora:NO169:0.4:OUTALL>
<lora:koreandolllikenessV20_v20:0.02:OUTALL>"
In this case, It is hard to use merge several LoRAs in 1-step and reproduce the same result. so I use "merge to checkpoint" method supported by the SuperMerger extension
fill in the LoRA form: "NO177:0.5:OUTALL,Dr.She:0.15:OUTALL,NO169:0.4:OUTALL,koreandolllikenessV20_v20:0.02:OUTALL"
output file name as "MERGE1"
select checkpoint-A "v1-5-emaonly"
then press "merge to checkpoint" button and you can get the "MERGE1".
select checkpoint-B "MERGE1", output file name as "new-lora"
then press "Make LoRA (alpha * A - beta * B)" button and you can get "new-lora" (alpha == beta == 1)
to make <lora:cutieMixGirl_V1:0.6> is the default, scale the LoRA by 1.7(1/0.6=1.666...) with the "same strength" option.
now the cutie Mix Girl V1 obtained!
In this case, only "OUTALL" weight blocks were selected to reduce possible interference by Lora "IN*" weights. (OUTALL is defined by Lora Block Weight extension as "OUTALL:1,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1" that means BASE,OUT03-OUT11 weights are used.)
How to use
recommended ratio is 0.5~0.7 like as
<lora:SweetyMixGirl_V1:0.6> (0.6 is designed ratio)
lower ratio weights could be a good starting point for non Lora-friendly-Models.
the higher ratio could result in overfitted behavior. in this case, lower your ratio.
What's the difference
What are the advantages of this unlike anything other lora girls?
it is not a trained LoRA, it is a mixed LoRA.
its recipe is open, anyone who wants to change the mix ratio and wants to get proper results can change their recipe based on it.
The model in the image I used is a woman in her 70s and 80s.
RoLA's strength is recommended to be 1.2 to 1.4.
When using other LoRA models, lowering the strength of that LoRA will give good results. For strength, 0.7 is a good seasoning. Of course, depending on the model, there is no need to lower it.
First ever Lora I ever made and decided to make it for one of my favorite fictional characters. I do intent on updating this in the future because I'm trying to learn. This was purely made for fun.
Found best results using strengths of 0.5 - 0.8. Higher than that looks deepfried which is my mistake. Keyword is "Magik," followed up by "midriff, gloves, thigh highs" for the black outfit and "yellow and black jumpsuit" for the classic x-men fit. Though "midriff" works fine by itself as well if you want to play around.
Her "midriff" outfit, natural look (hair, face, etc.) and the headdress are the most consistent in generation while the "black and yellow bodysuit" and sword are hit and miss. Naturally tends to lean more towards realism as well for some reason but I found adding "comic style" helps a lot. Go figure.
The use of this LoRA is prohibited for illustrations where the generated results appear to be, or would be, underage.
Whether the generated illustration is R18 or not is completely irrelevant. It is prohibited even for all-ages content. "It is troublesome to judge whether it is OK or not one by one! It would be easier to ban it uniformly."
Even if the actual age is 495, if the generated result looks like a minor, it is not permitted.
In the unlikely event that I discover a violation of this rule, there is a "possibility*1" that I will take action as I deem appropriate, taking into account platform, maliciousness, etc.
The "possibility*1" here is not a common sense possibility. The "possibility*1" is within the scope defined at the end of this document.
As a general rule, I will deal with violations that I myself find according to "possibility*1". In the case of violations against the rules of each platform, please deal with it according to the rules provided by the platform provider, as it will be the platform's problem.
Important things to prevent problems
Only for the above violations, even if you guys find them, please avoid pointing them out directly. It may cause trouble with users.
What? Did you get into trouble with the violator by pointing it out directly? I don't know, solve it by yourself. Even if you ask me for advice, I will throw your message in the trash.
Purpose
This LoRA is designed to generate illustrations dressed by mature women and to enjoy the gap between them! No other use is envisioned! Well, even a good-looking gentlemen will work. Maybe.
How to use
Trigger words: kindergarten_uniform
For "big breasts~huge breasts" adult women: 1.2~1.3 is recommended.
1x is recommended for flat chested adult females.
Optional word: school_hat, bag, school bag, [color]_skirt OR [color] skirt, pleated_skirt
Negative (Oprional): animal ears
Deprecated in most cases: white_background
possibility*1
The definition of "possibility*1" is "n" within the following range.
1 / R3 ≤ n ≤ 1
Note that R3 refers to the value stored in register 3, not R*3.
Theoretical machine executing pseudo-assembly code: it has infinite computation speed and registers (R1~R8) capable of storing infinite bits.
Well, it doesn't matter if it is finite; a register that can store G(64) bits should be sufficient.
The specifications of each instruction are as follows.
Instruction: LOAD Rn, x
Summary: Loads the value x into the register Rn.
Instruction: LOAD Rn, Rm
Summary: Loads the value of register Rm into register Rn.
Instruction: CMP Rn, x
Summary: Compares the value of register Rn with the value x.
Instruction: JZ label
Summary: If the results of the previous CMP instructions are equal, jump to label label.
Why my LoRa's? Because it's close to anime style, and this style is ~simple in details and can change without mutating target style more easy than artistic one. And it's hard to get quality anime style pictures from LoRa's with more artistic dataset with gradients and complex lighting.
Trained on 35 photos with ChillOutMix, work best on DPM++ SDE Karras sampling steps 20-30 CFG Scale 6. weight 0.7 is recommended for better results. try to experiment with different weights that works for you.
also, it works together with all of my other LoRA models. have fun trying to mix & match
lightweight LoRA, works on any generator (easydiffusion,stablediffusion etc). share your results.
Thank you for downloading.please leave a rating and comment if you like my models. also leave a comment on what model would you like to see next. Please consider to leave a tip if you like my work 😌☕️ enjoy
I generally do not support the use of this model to generate any adult content. When using this model, users should abide by local laws and regulations and shall not violate any other's right to reputation, privacy or portrait.
This is a collection of Jujutsu Kaisen characters made for fun (and also fulfilling a couple requests). Most of these models comprise mostly of screencaps.
Your characters may have insufficient training data! Click here.
Weight: 1.0
The results of your generations are dependent on the additional LoRAs, weights, and models you use, so it may not work or come out as consistent as my sample images. Other reasons may include me using xformers, token merging, and a modified DPM++ 2M Karras sampler.
ABPOPA on russian read as Aurora XD This is my friend's OC. I made lora just for fun. The dataset used only her art made using a neural network in an online service. I'm not sure that lore works at least to some extent, neural arts are hardly suitable for learning. But as a fact - I was able to make the lore of the character without real art and without artists.
Henyo ٩꒰｡•◡•｡꒱۶ !! I bring to you a cute snake girl hehe !! I hope you enjoy !! <3
Requested by @Cinders thank you so much for the request !! I really appreciate it ( ˘͈ ᵕ ˘͈♡) I also really appreciate the continued support and suggestions from you !!
This Lora works with almost every anime model and semi realistic models !! so go wild with it !! ^^
Her snake tail is included in this as a heads up, Sometimes her tail doesn't show up as well sorry :<
I recommend keeping the weight at (1-0.95) and using the negative prompts as needed !! Lower weights tend to lose her likeness !
Also use trigger words slit_pupils, scales, if your having trouble getting all her features and clothing !
Please use Trigger words ((naked)), (nude), nipples if your having issues with getting good NSFW generations (specifically with her breast not having nipples) <3
I had so much fun making the Lora and generating the images I hope you have as much fun as I did ⸜(｡˃ ᵕ ˂ )⸝♡ !!!
Please be kind with this Lora ♡ Also sorry I type wayyy too much lol
Тэги: animeleague of legendscharacterwomansnakespirit blossomspirit blossom cassiopeiacassiopeia
This is LORA based on Game Icon Institute_mode. Welcome to join the QQ group: 489141941 to discuss and comment, which is convenient for model development. This version is the mass production version of the Icon Academy model. You are welcome to use and comment for better optimization in the future.
Kitana Mortal Kombat 1 Character Lora |ALPHA|. I recommend using 0.8~1.0 strength.
Why is this an alpha? Due to the limited amount of scenes and imagery there is of Kitana for the upcoming game, this model was only trained using 13 images or really 4 segments from the trailer. The model works well, especially with how well it's able to guess the clothing as there isn't really a detailed "showcase" compared to the other characters.
Use the trigger word: kitana
I created a Lora request form. Please do not submit anything that breaks the tos. Thx!
This is a LoRA of Catherine Zeta Jones, when she was still young. It is is based on Stable Diffusion 1.5, and should work well on models like Realistic Vision.
Zhanshuang Nuo Keti model, the default weight is 1, and the recommended weight is 0.8. At present, the model is still being optimized. You can save it and wait for the subsequent optimization and update version. It is recommended to use the main model:
More of an art style than a concept Lora, it really likes skulls. Which is good because it's all about the sugar skulls. Doesn't do amazingly with animals but people and stuff seems to work pretty well.
Single keyword: sug4rt3ch
Best for a more toony style but will work photographic
The LoRA model imitating the painting style of Unkoteru's has a very strong texture.
Thank you for your support. I will work hard to provide higher quality models. If there are any shortcomings, please forgive me. Let me explain in more detail.
As shown in the preview image, the characters drawn in the style of Unkoteru generally have a flattened face. When the weight is negative, the function of whitening and slimming the face will be activated. I believe some people will find this useful (haha). But if you don't like wide faces, you can use other LoRA models to compensate or moderately reduce the weight of this LoRA model.
This model is relatively complete, and there should be no problem with its style being recognizable as LoRA. However, if you encounter a situation where the LoRA style cannot be recognized before or after, please consider changing the checkpoint model, especially for some checkpoint models with strong styles. If the output images are quite unstable, you should also consider changing the checkpoint model.
Most of the 2D LoRA style models I trained have not been tested with 3D checkpoint models. However, if you are strongly interested, it is worth a try.
If you are using badhandv4 (Textual Inversion) and the generated images have limb deformities, temporarily stop using it to improve this situation.
Training information: Select a total of 863 original drawings (including cropped images), iterate 863 steps (1 step per image), and train with multi-resolution intelligent directional iteration.
Other suggestions: EasyNegative. In addition, if there are limb deformities, try drawing again with a different seed.
if becomes unstable on far distance shots use strenght range 0.8-0.95
for CloseUps you can leave value of 1.0
She is NSFW, all dataset is nude, if she got clothes is because you prompted it!
Yulia Nova actually is 40 years old, and at this time no recent photo of her,
this LoRA enhances the period when she become famous. (at 18yo)
Yulia Nova was a university law student who entered a "Russian Beauties Contest" sponsored by photographer Satoshi Kizu in the summer of 2000. She was chosen over one hundred other applicants that fall. Her first three videos were made near Christmas 2000, at age 18.
and she was the wet dream of my father, so, that is why this Lora is present, dont blame me, i'm a woman and I just publish the perv dreams of ma'fatha, k?
LoRA trained to generate gym storeroom backgrounds, together with characters in different positions and poses, or just the background itself.
Please rate and leave a like if you enjoyed it! And be sure to follow to check out the upcoming models.
Trigger: gym storeroom
Optional useful tags: basketball, volleyball, balls, etc. Be creative with your prompts and post them here!
Additional info
I make these models for fun and for the challenge and release them for free for everyone to enjoy, for non-commercial use (see license). The one thing I ask of you is to provide rating and feedback! That helps me understand where I can improve and helps me stay motivated to produce more content!
I accept suggestions or requests here or on Civitai discord server.
You can also check my more polished images and upcoming models on my pixiv profile.
Finally I've figured out how to train a Lora Model! This is just a test I've made using 14 images but for the first try it did a pretty good job is you ask me. To activate you need to use (ohwx girl) because that was the token used by the youtuber which tutorial video I was following. Was trained with childrenStories_ToonAnime version so that should be the one that works best.
Prompt: 1girl, (masterpiece,best quality:1.6), scenery,night sky, open sky, stars, star light, milky way, starry sky, black_skirt, black shoes, full_body, green_eyes,necktie, pleated_skirt, dark blue hair, school_uniform, short_hair, short_sleeves, sweater_vest, white_shirt, short red hair, full body, standing,<lora:VioletEvergarden1.0:0.8>
Negative: EasyNegative, bad anatomy, bad hands, text, error, extra leg, extra finger, deformed fingers,extra person, extra character,forest, trees, objects, covered sky,
Hi, this is my first try to make a style LoRA first run (v0.2) with only 17 imgs dataset (just for testing) new version (v0.3) with 30 images (all manualy captioned) next version (v0.4) with the total 116 imgs dataset (manually/automatic captioned) is coming (13050 steps and my RTX 3090 says it need 4 hours..)
I am a deviant gentleman of culture. I haven't been able to reliable add fertilization x-rays to my images(all the LORAs are crap). So my thought process went like this. What if I make a LORA for inpainting and manufacture a clean dataset? Well this is the LORA for inpainting. Only remains to make a shitload of poses with different characters and inpainting on them to make a good dataset for the concept LORA(but that is for the future).
I seem to get good results using: <lora:fertilization:1>, fertilization, ovum, (sperm_cell:1.2)
Use latent noise or latent nothing otherwise the existing stuff will get on the way. I recommend to use it at denoise 1, if you want to lower it to get transparency YMMV.
Also be mindful of the space you give it, if it too small it might omit something.
In the gallery i will add the originals and the inpainted versions of the examples.
I'll add the dataset as it might be useful in the future, also the tag summary but i doubt that one will be useful as i pruned most of it as they were false positives.
*Edit: if you use it in a full image it works like a weird background.
*Edit2: Working on a V2, will probably consolidate everything under Fertilization as depending on the inpaint shape you can get both ovum and sperm or just one of them which can get annoying specially for batch inpaints. My other gripe is the size of the stuff in the inpaint, If I am not convinced with V2, I will tweak the dataset to make the images take less space of the training area to see if I can get it to try to be more compact.
Ladies and gentlemen of the press, for the last time, as the police commissioner, I assure you that there is no vigilante stalking the streets at night! There is no masked boogeyman riding a black Pontiac Trans Am running over corrupt priests and politicians, alright? I repeat, there is no vigilante!
Role of Arknights,and it's the multires noise version.
Weight 0.8~1.0 ,and you can ADJUST prompt to reductive character
TRIGGER:
1.Tremble Cold: whispercold;
2.OG Costume: original outfit;
3.Trigger for change clothes: headw,hair over one eye,
All memories become foam on the waves in the end. Whether suffering, happiness, inevitably, quietly, pasts I do not want to forget will shed themselves from my side.
Vladimir Korenev - actor known for his role as Ichthyander in the Soviet fantasy melodrama "Amphibian Man 1961"
"Amphibian Man 1961" movie with English subtitles -
Lora #6 - Поскольку я сделал lora на главную героиню, то решил сделать и на главного героя картины - Ихтиандра. Lora чисто для тянок (если они тут вообще сидят - видел только одну, значит будет подарок для неё) ,поэтому не уверен в скачиваемости данного творения
1)лучше всего использовать тэг (better use) "beautiful 20 years old boy",не "man", а то будет выходить гигачад (если только вам нужен гигачад)
2)"dreamlike blue eyes" - для его глаз, зачастую "blue eyes" перебарщивает с голубым цветом, а этот вроде как делает то что нужно (если есть хорошие теги для таких мелочей, делитесь в комментах)
3)Для костюма - "silver bodysuit" (могут возникать мелкие проблемы из-за малого количества кадров в датасете, в полный рост их вообще не было)
4)Тэги для подводного костюма/tags for diving suit- (wearing silver diving suit, wearing silver helmet, wearing swimming glasses:1.2)
С костюмом "морского дьявола" нейросеть слепила что смогла. В фильме костюмчик выглядит немного крипово, отчего и результат соответствующий, я использовал тэги с небольшим усилением в 1.2
Очень осторожно используйте тэг "wet hair", "wet skin" и все что связано с "wet" (ну вы поняли) - lora иногда чудит и перебарщивает (тут наверное только силу тэгов понижать)
На последнем фото магия inpaint в действии, т.к. stable diffusion не умеет задействовать двух разных персов одновременно (даже textual inversion не идеален увы)
Все основные советы по получению приемлемой картинки в описаниях к моим предыдущим моделям
Тянки, пишите (надеюсь положительные) отзывы / Куны, не кидайтесь камнями. Никакой лгбтqa+ пропаганды, товарищ майор - у нас тут все гетеросексуально, все случаи смены ориентации после просмотра фильма совершенно случайны
p.s. Еще раз спасибо за лайки, 1000 скачиваний для кого то мало, а для меня прям ого-го, все таки кому то еще нужна советская культура в нашем быстроразвивающемся технологическом мире. Хочу пересмотреть на неделе фильм "Чародеи" (потянуло меня на новый год), так что если не будет таких больших трудностей то сделаю lorу на Алёну Юрьевну))
This LoRA is intended to produce images of crowded underground nightclubs and raves. I'm reasonably happy with the results even though it can be a little wilfull; it seems to want to create certain room layouts and resist showing people from the front if used by itself. However, your mileage may vary.
May work well in combination with character lora's.
For good results, prompt for things like crowds, the color of lighting, style of space (ie; underground, industrial, etc).
This lora is made for creating our beloved malkuth! it DOES have a triggerword "malkuth" but it's effectiveness leaves a lot to be desired your base prompt should be 1girl, brown hair, brown eyes, headband, <lora:0.9> while it does work with other strenghts, 0.9 seems to produce more consistent results
Anastasia Vertinskaya is an actress known for her role as Guttiere in the Soviet fantasy melodrama "Amphibian Man 1961"
"Amphibian Man 1961" movie with English subtitles -
Моя 5ая lora, простите что пропадаю (у меня все таки отпуск))). Не планировал ставить на поток модели для sd, к тому же из-за того что я любитель, времени на них уходит много. Еще обновил описания для прошлых моделей (мне посоветовали сделать описания на английском для продвижения культуры так скажем, простите что такой слоупок, не думал что занимаюсь таким важным делом)))
лучше всего использовать тэг (better use) "young woman"
"neon lighting" (underwater) под водой довольно круто выглядит, особенно с lowra (напоминает чем то bioshock)
Все основные советы по получению приемлемой картинки в описаниях к моим предыдущим моделям
Дальше текст только для тех кому интересно как я рвал одно место...
1)Начну с того что у актрисы казалось бы главная роль, но экранного времени не так уж и много, а кадров которые пригодны для датасета и того меньше (постоянно в кадре лишние люди, либо поза слишком сложная). По итогу мне пришлось столкнуться с тем что за весь фильм на актрисе было всего 3!!! похожих наряда (купальник даже не в счет, в датасете только 2 кадра где хоть что то можно рассмотреть, настолько все мимолетно). На ум пришла идея использовать кадры из фильма "Алые паруса", но там такая же беда, поэтому я не стал усложнять.
2)Основная проблема модели - сразу говорю, модель далека от идеала, в частности вы можете заметить что при генерации то и дело на ней будут появляться серьги и цепочка. Опять же - датасет очень мал, и везде, ВЕЗДЕ на ней были серьги и крестик. Мне пришлось редактировать в фотошопе каждый кадр и стереть их всюду где я только смог. В прошлой версии на ней был неснимаемый крестик, именно поэтому мне пришлось полностью переделать модель 2 раза, чтобы выложить лучшее что только можно. К тому же возможны мелкие проблемы с глазами (это увы нехорошее свойство всех lora). Все таки легче когда на руках миллион фотографий какой нибудь знаменитости, есть из чего выбирать...
3)Ну а главный герой фильма все же Ихтиандр - с ним мне было куда проще, поэтому снизу оставлю ссылочку на него (уж не знаю кому он нужен, есть тут тянки?)
Спасибо за !1000! скачиваний, рад что вы оцениваете
Chae Soo-bin is a South Korean actress most known for her roles in: "I'm Not a Robot", "A Piece of Your Mind" and more recently "The Fabulous". I recommend BRA(Beautiful Realistic Asians)V5 as this is the model it was trained on and I got the best results with it. Similar models such as ChilloutMix and a couple others also seem to work fine. Apparently Chae Soo-bin has a face that while very recocnizable is not that easy to train. There might be an updated version of this LoRA some time in the future.
Triggerword: CSBin
The reccomended settings are:
Sampling method - DPM++ SDE Keras
Sampling steps - 35-50
CFG scale - 7
LoRA weight - 1
For LoRA request leave a comment on either this post or one of my other ones.
Silk von Santana from Tensei Kizoku no Isekai Boukenroku/The Aristocrat's Otherworldly Adventure.
Expand with Show More, long but hopefully worth the read.
Slightly overbaked (as I have no idea what I am doing) so use at 0.7 - 1.0 strength depending on the model used. No trigger word is needed.
Recommended prompt:
<lora:SilkVonSantana:0.7>, (masterpiece), best quality, intricate details, ultra details, ultra quality, rule of third, detailed face, detailed iris, 1girl, solo, long hair, purple eyes, hair between eyes, purple hair, hair intakes, bangs
Optional:
Use hair ornament for the crescent moon thing in her hair, if you don't include in the prompt it will appear randomly in some shots but if you don't want it, putting it in the negative will stop it from showing up.
Trained with screenshots of her every appearance in the anime so far using AnyLoRA, so I would recommend using that, though any Anime checkpoint should work fine. All preview images were generated with AnyLoRA.
While not needed, I would heavily recommend using Detail Tweaker LoRA as it is some form of black magic and I rarely generate any images without it. Most of the previews use it.
I generate my images relatively big at 600x960 and then Hires. fix them by 2.2/2.3 to end up just right for 4K screens but I threw in a few done at 512x512 to show what it can do at lower resolutions too.
This was the first LoRA I ever made, I taught myself how to make LoRA to make it, so it is far from perfect. I only made it as no-one else seemed like they were going to, but now one for Silk has appeared, I thought I might as well put mine up too.
I only spent about 30 mins to create this LoRA (10 to take and process the screenshots and 20 to train) so it's bound to be a little rough, but I will likely update when she appears in further episodes.
I have a few other LoRA in the same vein and quality as this, so a favourable review or two on this and I'll likely upload those too.
Тэги: animecharacterfemalesilkthe aristocrat's otherworldly adventure: serving gods who go too fartensei kizoku no isekai boukenroku
LoRA trained to generate school rooftop backgrounds, together with characters in different positions and poses, or just the background itself. I noticed the lack of some backgrounds so I'm preparing to post more LoRAs.
Please rate and leave a like if you enjoyed it! And be sure to follow to check out the upcoming models.
Trigger: school rooftop
Optional useful tags: chain-link fence, fence, building, school, bench, etc. Be creative with your prompts and post them here!
Additional info
I make these models for fun and for the challenge and release them for free for everyone to enjoy, for non-commercial use (see license). The one thing I ask of you is to provide rating and feedback! That helps me understand where I can improve and helps me stay motivated to produce more content!
I accept suggestions or requests here or on Civitai discord server.
You can also check my more polished images and upcoming models on my pixiv profile.
In order to use this LoRA, the following tags must be used:
<lora:kuroo>, black hair, brown eyes
This LoRA is best used at about 0.55 to 0.9 strength (0.8 recommended, at higher strengths it'll resemble the S4 anime style more), and contains three outfits, which can be activated with the following tags:
Volleyball Jersey: sportswear, red shirt, short sleeves, red shorts, sneakers (Please note that the collar might look red sometimes)
School Uniform: white shirt, collared shirt, necktie and optionally short sleeves for the summer uniform.
You can also put blue blazer and sweater vest in either the positive prompts if you want him to wear these or in the negative ones if you do not want him to wear these but get them too much anyway.
Tracksuit: red jacket, sweatpants or red pants, sneakers
There is also a fourth not-so secret outfit: the alternate volleyball uniform. It's pretty rare to get it, though.
Feel free to use this model for absolutely whatever you want, as long as the use is non-commercial and you give credit (unless it is Not Safe For Work, as I would rather be not associated with such content).
Optional prompts (of varying quality and effectiveness):
cartoon If you're aiming for a more flat / simple style. Or, if you want more detail and shading, try putting in negative.
winking One sided wink, like o.<
genki eyes / closed eyes / genki closed eyes Genki eyes, like >.<. Closed eyes, like -.-. Genki closed eyes can also work.
green sweater For the green sweater featured in a lot of her drawings.
chicken sandwich Turns out most models don't know how to make a chicken sandwich (it'll just be a burger or a sliced bread sandwich). Sort of overfitted, they'll be huge a lot, you'll need to prompt around that.
big smile Open mouthed smile.
prone / down lown Sort of going low, like they're sneaking.
ears drooping Puts her ears down / lower. Luck heavily involved.
Lora trained to generate a chocovenus bodytype (based on TasteOfChocklit art).
Definition of chocovenus by the r34 wiki :
"Chocovenus refers to a body type consisting of: - Large (often hyper) breasts - Large nipples and areola - A slim waist - A large (often hyper) ass - Thick thighs"
best prompts:breast, ass, thighs, solo
I recommand using a weight between 0.9 and 0.7 for good looking body, 0.6 or below for a more proportionate body and above 1 to have a body that look more like the artstyle of TasteOfChocklit.
AbyssOrangeMix work the best for this concept (better than Anything v3).
If you manage to have some great results or if you want to suggest some improvement for the model, you can send me feedbacks in the comment.
Very cute maid! Labiata: 1girl, red eyes, long hair, white hair, sidelocks, gigantic breasts, bangs, parted bangs, plump, glasses, rimless eyewear, semi-rimless eyewear NOTES: for her medium hair use: single hair bun, medium hair Maid outfit: labimaid, gloves, white gloves, bare shoulders, black maid headdress, maid, black thighhighs, frills, covered navel, white elbow gloves, short dress, collared dress, maid dress, black collar, black footwear, shoes, black footwear Bikini: labinikini, swimsuit, red bikini, choker, navel, hair flower, micro bikini, side-tie bikini bottom, hair ornament, bracelet, jewelry, armlet, collarbone, heart choker, cleavage, head wreath, skindentation, scrunchie, string bikini, pink maid headdress
Тэги: animecharactersexyfemalewomangame characterlast originmature femalemaidgacha game
Токены: 1girl, red eyes, long hair, white hair, sidelocks, gigantic breasts, bangs, parted bangs, plump, glasses, rimless eyewear, semi-rimless eyewear NOTES: for her medium hair use: single hair bun, medium hairlabimaid, gloves, white gloves, bare shoulders, black maid headdress, maid, black thighhighs, frills, covered navel, white elbow gloves, short dress, collared dress, maid dress, black collar, black footwear, shoes, black footwearlabinikini, swimsuit, red bikini, choker, navel, hair flower, micro bikini, side-tie bikini bottom, hair ornament, bracelet, jewelry, armlet, collarbone, heart choker, cleavage, head wreath, skindentation, scrunchie, string bikini, pink maid headdress
This is a LoRA of the model and actress Austin White. Please be aware that this model was not trained on any of her NSFW content. No trigger word is needed, but I recommend using "red hair, blue eyes" for best effect. Trained on SD1.5 and have tested it on about 8 different checkpoints, and it seems pretty good. Recommend using it at a strength of 1 for best results.
Тэги: photorealisticfemalewomancelebritygirlsrealisticaustin white
trained on about 60 fanart images pretty basic and will mostly just do her standard outfit with no other prompts, but will generate the purple sash on her torso sometimes
A requested character LORA for Shizuku Yaegashi from Arifureta: Shokugyou de Sekai Saikyou. The examples are using LORA weight=1 and highres fix ESRGAN_4x with denoising strength=0.5.
Character trigger: 1girl shizuku high ponytail
Outfit trigger: white shirt white sleeveless shirt belt denim pants black bra cleavage cutout
Тэги: animecharactergirlsarifureta shokugyou de sekai saikyou
Токены: 1girl shizuku high ponytailwhite shirt white sleeveless shirt belt denim pants black bra cleavage cutout
I recommend 0.8-1. ControlNet will streamline full body shots much easier, or more complex poses. Works with almost every Style, every alternative Outfit, Photorealistic Models and NSFW LoRA's.
Trigger Word:
Luigi
Helping Prompt:
mustache
How to use LoRA's in auto1111:
Download LoRA
Copy the file to stable-diffusion-webui/models/lora
Select your LoRA and use the triggerword
Make sure to change the weight if needed (by default it's :1)
你可以切换版本选项卡查看每个服装，右边的《About this version》选项卡有每个服装的一些详细的说明
I don't have the ability to achieve 1 lora, multiple costumes. I'll separate them, so each version represents a costume. You can toggle the version TAB to view each outfit, and the 《About this version》 TAB on the right has some detailed instructions for each outfit
(v1.5 UPDATE): Fixed the Network Rank and Network Dimensions to be 128 now.
Hello there! I've created a LoRA based on 17 movie poster images from the 1980s and 2 epochs!
I've compiled multiple movie posters of E.T., Star Wars, Back to the Future, etc.
I absolutely love the grainy realistic oil painting/aesthetic that 80s movie posters (usually) had and I wanted to see if I can replicate that through a LoRA.
Recommended LoRA weight: ~0.9-1.0
Recommended sizes: either 512x768 or 768x1024 if you want a similar dimension to the actual movie posters
Recommended checkpoints to use: Deliberate and Dreamshaper (but you can experiment with others if you'd like)
Make sure you add the trigger word "80s movie poster" in your prompt
You can also add the keywords "on the top", "on the middle", and "on the bottom" if you want to place certain images in a particular area as well.
It's a helper and may not be very useful. Not all models can generate this kind of images stably (Oh, some models like AnythingV4.5 can). As you can see from the XY grids, it can bring a more stable result.
Besides ”clothes_lift", you can also add "shirt lift" or "sweater lift".
A Canadian comedian, writer, presenter, actress and singer based in the United Kingdom.
There weren't any models of her when I trained this LoRA, but someone else posted an embedding version of her while I was preparing images for my version.
j3n7 or j3n7 woman These aren't 100% necessary but may help the result
TRAINING INFO
Trained on 31 photos (face & full body)
Trained at 768 resolution, so 768x768 inpainting works great
RECOMMENDED SETTINGS
Prompts: "Smiling" and others are trained into the Lora. For a full list of the prompts/captioned used for training, check the Lora file's metadata in your webgui.
txt2img: A weight of 0.6-0.7 will give the best mix of likeness and flexibility.
Inpainting: A weight of 0.7-0.8 will give the best low-noise likeness. If you want the absolute best likeness, switch your checkpoint to SD15 since that's the model on which it was trained.
Face restoration: isn't usually necessary, but I recommend setting your face restoration strength inside auto1111 settings to something like 0.7-0.8 which makes the effect less than the default setting of 0.5.
There are 2 verisons for you to choose. The creative can change different quantities of details in different weights while the just sharpen tries to keep all details of original images.
The recommended weight is about 0.5~1.0. The higher the weight is, the sharper the image will be.
It works much better when generating high resolution images and you can increase the weight to 1.0~1.5 if you make 2K or even 4K images.
（the extra way to use it: If you set the weight to a negative number , it can make the image smooth)
This is a "underwater style" model. It needs more "realistic" and "amateur" reference to force "sex under water" yet, if you have luck, you might produce one or another. ( I couldn't)
If you like what i do, I appreciate it if you support andbuy me coffee!
This version includes 2 LoRa, one for the character+outfits and another for Ayakura Juu's style: https://twitter.com/haino
CivitAI does not allow multiple LoRa in one version so I've put the style LoRa in the "training data" zip file. I've included example images both with and without the style; refer to the prompts to see how to use the LoRa together. All images are generated using Grapefruitv4.1.
The character LoRa was trained on NovelAI with 2000 steps and the style LoRa with 3000 steps. Best strength for both LoRa is about 0.5-0.7. Style does not require trigger words. I find the generations are most accurate when including "sash" and "pouch" tags in the prompt. Also try increasing the strength of the trigger words e.g. (holo:1.5).
The first attempt at backgrounds, excessive hires fix seems to me even a little harmful to this lora. There are no trigger words in this lora.
I made a video on how to achieve the same quality or just repeat my art
P.S. If you want to support me, then you can subscribe to my Patreon and get builds earlier than they come out here (versions on patreon will always be newer by one than here) This will help me a lot in updating the video card over time to make models even better and faster (the main thing is better)
Eta noise seed delta 31337
LORA weight 0.5-0.6
Use Abyssorangemix3aom3_aom3a3.safetensors With these negative prompts: (worst quality, low quality, extra digits:1.4) or (worst quality, low quality:1.4)
just use hires fix as much as your video card is capable of, so as not to give an error, with 8gb of video memory it is 512x768 with hires fix 2-2.15, upscaler Latent (nearest-exact), the picture will be much more detailed. If you want a style close to the original but with more detail, don't use Denoising strength above 0.6, but to avoid artifacts, do not use too low a value, optimally 0.5
CFG Scale - 5.5-7 (for some reason hands are a little better at 6)
The training took place at a resolution of 768x512
Lora trained with Hollowstrawberry's script on 3:4/4:3/1:1 aspects using NAI as base. Dataset pool - 82 images 768px. Generates Morgana, Altria, Lartoria, Lartoria Alter even without tagging them, some artifacts are possible. Definetely overtrained, tends to accurately replicate training image. I prefer using 9th epoch in [0.45; 0.55] weight range on NAI and aom3, didn't test on other models. Lora has no activation words.
Generated characters will always have sidelocks, even males since there is no lockless person in dataset lol, same goes for steam, unfortunately. Sometimes draws bad hands, so watch out and adjust negatives respectively. Genereates good on euler-a, with Steps 31 and CFG 6-8
upd:
changed dim and alpha to 32, added 2 new images to dataset, corrected and cleaned tags, set different lr. looks like some concepts got in as well, for some reason straight-on pov is hard to achieve. 1.0 weight works fine for me, generetes oversize booba, adjust sizes by replacing huge with medium etc.
There are only TIs of him available and honestly none of them were good enough for me, so I trained a fresh LoRA and have decided to share. Trained on the 1.5 base at 1024 res.
Please don't rip this off as your own, or do anything crazy with it. I am not responsible for how you use this. I am not releasing it for professional or commercial use, this is just for personal use and non-commercial research purposes. Please respect him and please share your results!
Use "pedro pascal" as your keyword
This LoRA stacks well on almost all models I've tried, from photorealistic to anime
I had to overbake it a bit to get it to learn his characters, so tone it down if it's coming out too "crunchy" or it's only giving you closeups
Works best between :0.4-0.8
Using a lower value with a higher keyword works well to nail the balance between realism and stylization
I trained this to understand Joel Miller and Oberyn Martell, but it didn't pick up The Mandalorian (see prompts on images for getting them)
SD wants him to have green/hazel eyes, counteract it with "brown eyes" in prompt or "green eyes" in negative prompt
My example images were all upscaled using Tiled Diffusion
Тэги: photorealisticgame of thronesthe last of ustlouactorcelebritymalemanmenrealisticgotpedro pascal
Weight can go from 0.6 up to 1.0. I recommend 0.9. ControlNet will streamline full body shots much easier, or more complex poses. Works with almost every Style, every alternative Outfit (sometimes), Photorealistic Models and NSFW LoRA's. This LoRA is NOT flexible enough. Her Bodysuit will blend in most of the time.
Trigger Word:
Mistral
Helping Prompt: hair over one eye
Optional Prompt:
extra arms
How to use LoRA's in auto1111:
Download LoRA
Copy the file to stable-diffusion-webui/models/lora
Select your LoRA and use the triggerword
Make sure to change the weight if needed (by default it's :1)
