/bbwai/

(4.5 MB, 1792x2560, Inko Midoriya AI 7.png)
This is a thread for requesting LORAs of characters.

I made this thread because I still haven't seen an online LORA for Inko Midoriya on Yodayo, a free gen site.
I'd like to request said LORA.

https://yodayo.com/
I'll allow this and it's a good idea for a thread, but, to make it easier for anyone who is actually keen to take the time to make and train a lora, please provide a catbox'd zip of at least 20 high quality images of the subject in question, to minimize the research time.
>>6206
You got it. I assume you mean images drawn by different artists?

But yeah, I can see canon chubby characters being rejected by people who make LORAs. We could also request OCs who have enough art to train a LORA with.
>>6208
Different artists are ideal for variety's sake, but not absolutely required if one artist has many many examples of one subject.

And yes, essentially the prerequisite for this happening at all is that the subject even has enough high quality art pre-made to even begin training with.
>>6209 I disagree but because I choose to be directly opposed to evil in its many forms and although logicly some things sound good at first in practice nothing's ever ideal and we're talking about art, and a thing that's already a matter of perceptive conjecture.

Experience in life has taught me that most of the time the last thing you would ever want is exactly that, many artists copying and repeating the same thing thereby reproducing the same art, especially if said thing happened to be naturally an error prone process and thus producing a certain degree of unwanted effects in quality within the art over a course of some time, thereby making these unwanted, and nonartistic imperfections to get copied and reproduced alongside with the good art, the result of which being an ever-increasing cascading amount of low quality art aswell as low quality artists by comparison. Art isn't logical.
(601 KB, 512x768, 02956-3372396158.png)
>>6216
It could be that I'm misunderstanding what you're saying, but honestly, I have no clue wtf you're saying.
The other commenters are talking about training a lora on a character from different artists, and you're saying you don't think training with different artists is a good idea because "art isn't logical"? Maybe I'm missing your point, but training a character with a single artist's work will also bake in their art style instead of learning the character specifically. Variety keeps the AI honest so it only learns what makes the character look like they do. Correct me if I'm wrong.
>>6220 Yes. It's a bit like the difference between good art and bad art, but I won't take up too much more of your time on such a meaningless and petty thing that commercial art is. The crux of my post was that although variety's good for many useful things in our physical world it's also true that for most things in general variety's not what you want. What's ideal has nothing to do with variety, not even in a consumer driven economy, which this market is most definitely not because it's barely a free market as it is.

Speaking about artists specifically, I said, ideally you want natural born artists and innovaters, not groups of sheep fashioned to copy and reproduce the newest trending and highest grossing art products.

The thing is that you can't ever have both! And because of greed it's doomed that evil triumph. Because of that great art and real artists will go extinct eventually and without real artists to keep them propped up the commercial side will follow suit until there is no longer a mass demand, desire, or need for art products.
I'd love a lora for Kazuu, Kurokazenosora, and/or fugubarakun (bbw and inflation) but I could probably try making one myself
>>6220
Ignore the schizo
>>6226 I'm not a schizo but I too like most am a lover of the spice of life. While I usually agree that more is more, art's 1 of those things that I am no longer convinced benefits from variety.

I believe that art benefits from itself first followed by the artist.
>>6225
I think I understand your post, but, Anon, OP just wants a lora so he can make fat of shes of his waifu. This has nothing to do with commercial art and is entirely off topic.
>>6229 Did you read? My response was to an opinionated post about variety, art, and art idealism.
>>6220
>Variety keeps the AI honest so it only learns what makes the character look like they do.
You're wrong and the AI can't understand what art "is", neither what makes art good art. Contrary to that, an artist need not ever know whether what he's making is art or not, and still he may know and feel that the art he's doing's good art.
>>6220 The AI is professionally dishonest, if you want to be technical about it. It uses math to show you facts presented in such a way that you may find acceptable, in a purely processive way of thinking about the art.
>>6232
>the AI can't understand what art "is", neither what makes art good art
I never said that it did, nor do I think that this distinction actually matters to the point I was making. You can see I was responding a post that seemed to argue against varying the training set to many different artists, so my advice was that by having a diverse training set, it "learns" what the character is more accurately and doesn't "think" the art style is important.
>>6220 Actually the thing that I can find fifferent between this post and these AI art is that this post is written by a human with no intention to decieve or appear a certain way.

Of course you have to be omnipotent to ever have a chance at even approximating a human's true intentions so there's not much difference. Clearly, I don't have a girlfriend but if I did I wouldn't trust her even if I loved her so we would probably end up divorced because if I loved her I wouldn't let her go anywhere alone without me.
>>6234
I know. I had a good idea as to what you were trying to say in your previous post. We call that common sense.

Also,
>nor do I think that this distinction actually matters to the point I was making
see
>>6235
>Of course you have to be omnipotent to ever have a chance at even approximating a human's true intentions so there's not much difference.
(563 KB, 512x768, 00142-1116667810.png)
>>6236
>I know. I had a good idea as to what you were trying to say in your previous post
then what are you on about?
>>6234 It's not just art style. It's everything else aswell, which's what I was getting at, but I agree with your post. There's still plenty of room for tools that can make this a real thing though. The downside that I've always said is that there isn't any real advantage over drawing for yourself and making a better looking loli with huge boobs.

At this point it's less complicated drawing your own boobs that look more sexually appealing using a pencil and a piece of paper.
>>6239 I'm on about YOU having too much fun posting too many pics and also about the character in your pic having 6 fingers and nobody caring, but that is the evil babylon that they've created.
(2.2 MB, 1584x1152, 47823-3364584255.png) (1.8 MB, 1152x1440, 47824-3655516277.png) (1.6 MB, 1152x1440, 47825-1366179261.png) (1.8 MB, 1152x1440, 47826-4142117197.png) (2.3 MB, 1440x1440, 47827-1993203535.png)
>>6205 (OP)
i gave it a shot. 115 images of inko scraped from gelbooru and danbooru
trigger is midoriya_inko, and you don't need to specify more than that. Works fine at weight 1, only negative you need, if any, is 'red eyes, black eyes'
Without fat tags, defaults to a chubby mom bod. 'mature female' makes images worse. Affects style. Increasing weight on 'midoriya_inko' makes it get her head and proportions better but changes the style more.
Haven't experimented with weight on the lora itself or with making her shorter
https://mega.nz/folder/XRsGgCCA#82AWCQMCxGyrozM7zYAvqw

>>6216 >>6225 >>6227 >>6230 >>6232 >>6234 >>6235 >>6236 >>6240 >>6241
faggots
>>6256
Thank you! These look really nice and I'm glad you did my request.
>>6256
Highly impressed you did the scraping yourself and took the time to do this. Bravo. There are multiple people on fiverr that charge about $15 for service and speed of this quality and you nailed it. Impressive. Did you follow a specific guide?
>>6272 The wise Colonel's got good taste, but is he psychic aswell? There's tons of programs to download images, but nothing beats the good old fashioned monkey userscripts. There's supposed to be an AI booru but it doesn't show for me on google. $10 says that the best pics end up being futas. Everytime I see a good futa pic it makes me feel young again, like it's the first time I made out with two sisters.
Making out with two sisters probably wouldn't be so great if I didn't love lesbians. That's the thing. I've been thinking a lot about lesbians lately and my research into lesbian AI art led me nowhere.
The Colonel's well aware of my lesbian affection. Does he know that 7 months to this day I promised to stop masturbate more than once a day? The past few days have been tricky. A large breasted lesbians AI thread may prove too tempting. I would do it myself if I knew how to start threads. I love the young girls.
(1.2 MB, 1024x1024, 47858-3502259110.png) (1.4 MB, 1024x1280, 47859-3646482291.png) (1.9 MB, 1408x1152, 47860-3829363078.png) (1.0 MB, 1048x1312, 47862-721504812.png) (959 KB, 1048x1312, 47861-2230417108.png) (1.7 MB, 1048x1312, 47865-1195190995.png)
>>6260 >>6272
glad you like it! Let me know if there's any other issues with it, I'm still learning
Character loras are actually pretty easy to make, just mass download everything tagged '[character], 1girl, solo' with a scraper, and then leave it to train while you make dinner or something. If your character has a lot of results just do 'sort:score' to sort by best and take the first 100-300.
this one took 2 hours. Training settings won't really matter unless you're working with very few images, but I'm using the mintanon's settings https://mega.nz/folder/sShj1LpS#Zh5ZxGO3uwK4QEaZ2X6z1w on 16 repeats & 16 epochs
If you're training style one of the SD autotaggers should suffice but you should still make sure the tags are accurate

I haven't figured out how the mintanon copies styles so well though. I'm using the same settings and getting way worse results for style for ~100 images on 16*16.
Picrel are an oldschool trinityfate lora a herroroo lora I tried to make (spoilered for furry). neither were very good or accurate. if someone who's made loras could share advice & settings i'd appreciate it
Is it possible to request a Lora for an art style like Cookies-Cat, but if it's just then could I request a Lora of LeShawna from Total Drama
is there an exising lora or textual inversion for force feeding with a tube? can't find any and can't get starlike to generate it without it looking weird
anyone considering a lora for decisive tang?
>>6579 Only if you have a death wish. And this' coming from an engineer.
is a Boberry LORA possible?
>>6256
What program did you use to make this LORA by the way?
Anyone have a LORA for hypnosis? I cannot figure out how to do solid eyes or glazed over...
Suppose there isnt anything wrong with requesting things here. Ahem. Can anyone here render or generate a sumo wrestling vore theme? the subject should be something like a fat female bear, who just devoured her opponent.
I may be in the minotiry here but, I'd like to request a lora for gas/farting. Doesn't need to be the physical action but just the green/brown steam in the air behind the rear, I've been having a hard time using it in a prompt.
Are there any good loras for eating or funnel feeding?
idk how many share the interest, but more vtuber loras would be great! western vtubers as well are lacking. think vshojo and the like. if i have to be specific for a request haruka karibu
>>7245
Where can I find this Lora?
>>7278
It was on civitai, pretty sure it'll show up if you just search "detail"
(2.5 MB, 1440x1440, 48161-938292913.png) (2.1 MB, 1296x1584, 48162-4185914718.png) (1.8 MB, 1296x1584, 48173-1485595076.png) (1.4 MB, 1296x1584, 48174-2216067928.png) (1.6 MB, 1296x1584, 48176-489843814.png)
>>7000
https://mega.nz/folder/zdsCGbxZ#_6TPDH2FOlW0rcE_6iJPcw
Haruka karibu lora. Scraped her booru tag (jesus christ she has so little art) and screencapped her stream a couple of times

Ideal settings 0.5-0.8, I am partial to .7. Activates with 'haruka_karibu'
Seems to work well with "1girl, haruka_karibu, deer girl, antlers, long hair, (green blouse, buttons:1.1), bare shoulders, brown ear fluff, vines, see-through_sleeves"
with "nude, thighhighs, brown eyes" in negatives. Nude is an important negative because it has a bizarrely strong preference for making her naked
Fur collar sometimes disappears but haven't had much luck with 'fur collar' as a tag. Inpaint it on if it's missing, maybe? Same with leg vines
>>7320
Did you train it with the Novel AI checkpoint?
>>7322
That I did, anon
hear me out: we already have a shylily lora, so why not a Bao lora? we need as many whale girl loras as possible. they're whales, they're MEANT to be HUGE. for your consideration, thank you
>>7393
Not only is there a lora for Bao, but some of the already mixed vtuber bbw models include her by default
the only zentreya loras around are of dragon zen, if anyone would like to take a crack at it, a cyborg zentreya (white hair) would be amazing. skin-tight suit im sure everyone can enjoy
(734 KB, 1659x1991, image_2023-06-19_182823023.png) (437 KB, 1536x2048, Fy1Q23haEAA77Nv.jpg)
Here is my request. Agano Mai, she's a chuuba who hasn't debuted yet, so the two best images I have are these. She hasn't posted a reference sheet yet, but in theory, it is possible to generate a LoRA based on only two source images. Additionally, generating ai images could supplement the lack of source images.
>>6206
The minimum number of images needed to train a LoRA is two.
(1.8 MB, 1757x1000, aicandiiref.png)
hello everyone. i would like to request an aicandii lora. candii's twitter has a lot of fanart images for it, and also a lot of VODs for screenshots.
Has anyone tried making a CGhonk styled Lora? They have a ton of pics and vids backed up on rule34.
Anyone know of a Lora for Raiden from Metal Gear Solid 2?
>>6256
These are so high quality
Need to get this thread up and running again
(15 KB, 203x202, OIP.jpg)
What about Toga?
unlikely to be fulfilled but dropping it here. mobility scooter lora for superfats
would anyone be down to make a rosedoodle lora? she has a 3d model and a good handful of fanart, so you could realistically get reference images from many different angles
(290 KB, 1656x1656, RoseDoodle_Portrait.webp)
>>9883
image for reference of who im talking about, searching rosedoodle on google gives you a shitton of images
>>9459
I second this request. As well as propose LORAs for different methods of force feeding. Stuff like feeding machines, conveyor belts, hoses, even by hand.
>>10058
Nevermind, figured out how to make it.
(2.1 MB, 1152x1536, 48428-1205057506.png) (2.1 MB, 1440x1728, 48424-2422062408.png) (2.1 MB, 1440x1728, 48422-2398166369.png) (2.7 MB, 1296x1728, 48448-3927453889.png)
>>9883 >>9884
Still working on it. been busy + just can't get it right
here's some pics though

>>9474
haven't tried it but there exists one on civitai
https://civitai.com/models/98220

>>9459 >>10208
Don't know of one but hoses are a pretty simple inpaint, just draw it on and add the tag 'tubefeeding' to prompt
You can also try using the symix lora https://civitai.com/models/121416 which can get a good hose more often than normal but it's still gacha

>>10334
I had it rehosted on a mega link w/ documentation, but it seems to be having some sort of service issue right now. here's another gofile for the time being https://gofile.io/d/IUjqMg
Has no activation word, ideal weight 0.3-0.6 (its overbaked to fuck), and has best results on a bbw model. Tends to deep fry normal models
Would it be possible to have a LORA for the legendary bbw porn actress "Charlie Cooper"? I'd love to put up the requested images, but before doing that I wanted to ask, if anyone would be willing to train the LORA for her.
I would love any of Tsukumo Sana!! BBW, SSBBW and even Blob sizes appreciated.. Desperate for some immobile astrogirl :0
(773 KB, 1274x1860, 20230830_062317.png)
Is it possible for someone to make a Lora of Trinity-Fate's character Umi?
I've seen people use an "immobile blob" lora. Does anyone have it?
Are there any attempts to train LoRA on the whole Inflatebooru, successful or not?
(123 KB, 1280x720, 1696305730577834.jpg)
Can you combine lora's or do you need seperate ones? In case of the later I'll need busty ponies.
>>14797
Some places have photos of her around.
I request a lora trained for an average of Imbapovi's + Skie Maree styles.
(11 KB, 155x384, Miko.jpg)
If possible, I'd like a LORA for Miko from TF Prime
>>7652

Do that and I'll quit on the spot, cunt.
(95 KB, 550x510, Luna Loud.png)
Haven't seen a Luna Loud LORA, so I'll request one
(265 KB, 740x950, Sara2.png)
Sarah from Gundam Build Divers

Someone on Pixiv has already made one, but doesn't reply to any of my messages asking how he made it/if he could share.
>>22621
I think she has one. Someone requested her with good results
>>22661
Oh, I didn't know. The LORA must not be on the site I frequent.
(258 KB, 438x1137, Ginko.png)
I'd like to request a LORA for Lisa Silverman/ Ginko from Persona 2
>>9197
>>12207

seconding (thirding?) these anons' requests for mobility stuff
burping expressions lora
just burps in general
(199 KB, 1100x1558, 2-1715276892-o.jpg)
Can anyone make her in this outfit but with a fat belly?
Since someone couldn't done my request because she didn't have one. Can marle from puyo puyo have one?
Someone placed a bounty to create a Lora of Mashiro Kuna from Bleach, can someone please make a lora of her
(1.1 MB, 1152x944, 00015-625579770.png) (1.1 MB, 1152x944, 00023-3918434724.png)
Could someone give an advice or even make a LORA of multiple characters? I have LORAs for multiple characters from anime New game! but i want to make pictures with several characters without separate LORAs, i got only two images being close enough

Back to top