/b/ - Random

Quality is not guaranteed


New Reply[×]
Message
24.3MB total
Tegaki
Password
[New Reply]


If (You) don't like something then post what (You) like instead


ClipboardImage.png U A
[Hide] (968.8KB, 1422x773)
I play around with AIslop, specifically Stable Diffusion.  I have attempted to make a LoRA file for a certain focks character.

I am dissatisfied with the results, to put it mildly.  You might even recognize her. 

I am using Automatic1111 and Stable Diffusion 1.5.  I trained the LoRA file with kohya_ss.  Source images came from the places you'd expect.

I have access to SwarmUI with SDXL, Flux DEV, and 3.5 available on the back end, but I don't know whether any of the tools I have can make a LoRA for any of those.  And the quality of what I get from them is very disappointing, given how much newer they are than 1.5.

Does anyone else around here mess around with image generation?
Replies: >>585 >>673 >>743
8d9a50dffbea8f668d651a13396147dcf834cdc200509b7c1696426ddc34409c.jpg U A
[Hide] (147.7KB, 832x1216)
generate more wholesome averi picrels or gtfo
Replies: >>583 >>584
>>582
this but pepper instead, averi already get's enough attention here
Replies: >>584
ClipboardImage.png U A
[Hide] (885.3KB, 1419x814)
>>582
Quality is not guaranteed.

>>583
You don't recognize her, then? That's supposed to be Peepor.
Replies: >>620
>>581 (OP) 
Are you just new to training loras, or new image generation in general? Cuz my armchair diagnosis of this pic is that you set the cfg too high.
Replies: >>589
Stable Diffusion's still the best open-source model?
Replies: >>589
ClipboardImage.png U A
[Hide] (1.1MB, 1414x823)
>>585
Both, really. Well, I've been doing it on and off for a while, but I don't claim to understand any of this stuff very well.

With the LoRA I trained, I typically run it at a strength of about 0.85, and I normally leave the cfg at 7, the default setting.  These are from a big test batch, one with each individual SD 1.5 checkpoint file I have.  

I am used to having to gen big batches and pick and choose. SD 1.5 has always had serious problems with missing limbs, extra fingers, extra ears, and so on.  People say that, for example, SDXL and 3.5 are better.  But with 1.5 the community support is there, there are checkpoints and LoRAs and training tools and information about using them, that still haven't appeared for 3.5 or Flux or SDXL, or maybe I'm just not looking in the right places, or devoted enough time to research.

>>586
I honestly don't know.  It's the one I've learned to use, though, for some values of "learned."  Grug click orange button, Grug see fuzzy lady.
not an expert but what do you hope to achieve once you master the art of ai slopping
Replies: >>591
>>590
presumably, making videos of hyper-realistic pepper burning spaghetti sauce at 3 am
Replies: >>595 >>598
>>591
What not pepper hitting the nae nae?
Replies: >>598
focks_by_Norman_Rockwell_00837-2508059266.png U A
[Hide] (1.7MB, 1280x1280)
>>591
>"Why are you burning spaghetti sauce at 3am?"
>"Because I've lost control of my life."

>>595
Better that than getting nae nae'd.
>>584
cute!
Replies: >>636
Spoiler File U A
(514.1KB, 1132x1294)
>>620
She a peepor. She peepin. It just what she do.
Replies: >>638 >>657
>>636
>s*x
@floof balete dis
Replies: >>639 >>641
>>638
Wearing sox isn't against the RUURUURUURUURUUS.  https://youtu.be/J548XM9-jKU
Christian_website.png U A
[Hide] (630KB, 1132x1294)
>>638
fix'd
Replies: >>642 >>657
>>641
much better
Replies: >>646
>>642
I am in awe of your l33t editing skills. Can you teach me, Sempai?
Replies: >>653
>>646
nigga that anon just found the right font o algo
>>636
>>641
also is pepper actually short I always imagined her as a bit taller than average, like 5'8"
Replies: >>660 >>665
9b8d8720fb67005d71aa334e2f17f5959a0d64e0c5b1b52958117171a22d9987.png U A
[Hide] (37.3KB, 800x600)
>>657
She is canonically short, yes. I forgot where but I think I remember seeing that she's even shorter than Averi (5'5)
Replies: >>661 >>663
>>660
>41 mins later
>file pending
fire all jannies
Replies: >>662 >>664
>>661
brutal
>>660
wild
77a944599fc1da6e9f59c87b94f3c804ffc6d9396ad393761accc3d5971ab510.png U A
[Hide] (192.4KB, 498x357)
>>661
g-g-gomenasai!
Pepper_is_short.jpg U A
[Hide] (58.4KB, 800x600)
>>657
She is a smol focks.
my oc is pepper but 5'8
Replies: >>669
>>667
Does she wear nerd glasses and put a zip tie around her ponytail because she lost her scrunchie?
Replies: >>670
>>669
you know what maybe she does
>>666
Replies: >>672
>>671
spooky number haha
>>581 (OP) 
slop containment?

pocket fox
Replies: >>674
Z1BAWQ0VA7YXDPF4PXG53KXBN0.jpeg U A
[Hide] (199.7KB, 832x1216)
>>673
whyyy didn't the pic work?
Replies: >>675
>>674
generate a pocket kitty pls
Replies: >>676 >>677
>>675
Do this
YA3J4E05XDFCYDSH8HV8B5E0C0.jpeg U A
[Hide] (155.8KB, 832x1216)
>>675
of course civit decides to be slow as a snail when someone actually wants something 
i should probably set up some local stuff but i don't caaare enough
Replies: >>678 >>683 >>809
ClipboardImage.png U A
[Hide] (1.2MB, 1024x768)
>>677
I'm trying to get a kot in a giant pocket, or perhaps a very small kot (kotlet?) in a regular sized pocket, but SD 1.5 does not recognize the prompt. I'm trying img2img now using pic related as a starting point.  So far it isn't looking promising.  Lots of kots wearing big shapeless sweaters and hoodies, none in an actual pocket.  e6 has the "sitting_on_hand" tag but only a few images have it. The checkpoint files for SD 1.5 that were trained on data from e6 mostly only recognize tags that have at least 2-300 images associated with them.

>>677
What a pretty kitty.  How did you manage to get that?  Is there a "tiny character sitting on giant hand" LoRA you're using?
Replies: >>679 >>689
>>678
wai-nsfw-illustrious-sdxl
no Lora
I'm pretty sure the active ingredient is:

((micro)), (micro girl), pov hand, human hand, ((sitting in hand)), holding micro girl, legs dangling

as with anything, play with the stress/weight/emphasis/whatever of each tag to change the focus, and some of them may not be strictly necessary
Replies: >>687
>ai thread
you know, i have seen ai artists be very passionate about their work like how artists are with their artwork

for artists, i can understand
for ai artists, i dont get it. (no respect ofc)
is there more to it than just getting the right generator and writing the right prompts
Replies: >>681 >>686 >>1240
Spoiler File U A
(1.7MB, 1857x1040)
>>680
I don't call myself an AI artist.  I'm a guy who clicks the orange button to see fuzzy ladies.  I don't call what I do "art."  I speak for no one else here, of course.

I will admit to hanging around the drawthread and occasionally I am tempted to shitpost, but picrel is probably a little too spicy for /anthro/ anyway.
i completely forgot /b/ existed lol
>>677
Gem, I love her so much. Could you try make her eyes green? I don't know how hard that would be
Replies: >>684
61a5c256ec68a0f882b07dd8bb48601b858c37b8b6150e3d661563c6c48537d5.png U A
[Hide] (928.5KB, 820x1203)
>>683
Replies: >>685
>>684
>bucket tool
Lmao i feel retarded for asking
>>680
saying "ai artist" is already giving a lot of respect

bit of a tangent, but in general i don't like it when ppl get really far up their own ass about stuff
i feel like i see a lot of internet artists turning out jaded and burned out when they get into art as a community rather than a hobby
Replies: >>687
>>679
Maybe it's something SDXL can do that 1.5 just doesn't understand. I'm not getting anything remotely appropriate so far.  "POV" in particular seems to make 1.5 draw sexually explicit poses.

>>686
Did you mean "commodity?"  Because I agree, if that's what you meant.

This is yet another digression.  What you do to earn a living, you will hate.  You WILL hate it, every second of it.  You will hate the people you work with.  You will hate the people you work for.  You will hate every second of what you do every day.  It doesn't matter if you're a stockbroker or a farm laborer.  "Working for yourself" is bullshit.  Unless you're a trust fund baby with a gorillion dollars in annuities that pay you whether you get out of bed or not, you've got a boss, and you've got people you have to deal with on the daily while you're doing stuff for him, and you will hate every second of it.  "By the sweat of thy brow shalt thou eat bread, til thou returnest unto the ground, for out of it wast thou taken.  For dust thou art, and unto dust shalt thou return.  Dust shalt thou eat for all the days of thy life," can you dig it? This is what they call "the human condition."  Be glad you aren't digging ditches.  You may not hate your life as much as those guys do but at least you aren't destroying your body and bound to end up walking with a cane with destroyed back and knees by thirty.

College recruiters tell you it doesn't matter what you major in, because "your time is now" and you can "follow your passion," and if your passion is lesbian dance therapy theory, you are gar-um-teed a big fat paycheck doing it the instant you graduate with the sheepskin that has that stuff written on it.  They're salesmen working on commission, pimping you to the people they work for, just like military recruiters.  They're paid to promise you the moon and stars.  It's all bullshit.  Most college grads don't work in the fields they studied, and more college grads work at Starbucks than in STEM.  

I promise, there's a point.

"Make your avocation your vocation!"  Like the guys who like computers, and write code for a living, for and with morons who tear down everything faster than they can build it up.  Like the guys who like computers, and build and fix them for a living, and bring computers into their homes that have cockroaches skittering out of the PSU and nicotine-scented dust bunnies clogging all the fans.  Drawing art on the Internet for a living means either you don't get emails for months on end and never know where your next meal is coming from, or you get emails from people who want to pay you to draw their Jeffrey Dahmer masturbatory fantasies of the cute characters you created murderfucking one another and painting the walls with gore and spooge.  And it's the only way you can pay your rent and buy food.  You will hate it.  You will hate your customers.  If you work alone you'll hate the noisy neighbors and their loud goddamned dog.

In a way I feel worse for Internet "professional artists" who live on the proceeds of their commission work than I do for the history major who's trapped delivering pizza to hoodrats for a lifetime.  Because I think the artist's pain is greater, and purer, because every single day in order to survive he has to shit all over everything he thought made his life worth living.  The guys who dig ditches don't have that particular anguish.
Replies: >>688 >>801
>>687
no i meant community
>>678
there comes a point where it's simply going to be faster to draw it yourself
Replies: >>690 >>801
ClipboardImage.png U A
[Hide] (1.5MB, 1024x1024)
>>689
You aren't wrong, but at some point I started to take it as a challenge.
Replies: >>801
PHKT40BD80Q28XG11BN41G5K50.jpeg U A
[Hide] (168.8KB, 832x1216)
>>581 (OP) 
something i did a while ago
Replies: >>745 >>760
>>743
Very cute
why is this thread here and not on /anthro/?
Replies: >>750
>>745
People don't seem to like AIslop very much there.
Replies: >>760
>>750
Who cares what other people think, do what you happy. Also gem >>743
i think it's better for this to be on /b/ because of how much faster /anthro/ is
also thx 4 teh kind words evry1
Replies: >>786
ClipboardImage.png U A
[Hide] (828.6KB, 1671x691)
Stable Diffusion just doesn't want to render a certain snoot correctly. And I can't unsee a certain scene.

>>764
:3
16bf21cf602acbcfefca7244fbc1c947cb80475cdcb818058fe9f0a4aa3b3161.png U A
[Hide] (1014KB, 1200x1600)
>>687
I usually never post in this place (due to the cuckflare captcha), but your post deeply resonated within me. So I needed to respond it. When due to shear curiosity I read your post, I felt identificated with everything you said, like a direct critic towards me. But I didn't felt ofended, as the polar opposite I feel vindicated. So thanks fur that wall of text anon.
Your explanation of the job market as >"you WILL hate it, every second of it" 
is painfully accurate. I personally been "trying to earn a living" as you say, fur around decade now, and the phrase "Be glad you aren't digging ditches." hits specially hard. Cause I used work in jobs like that kek in such labor jobs, physical explotation was the norm, and if you dare to complain they would just say "if you dont like it, then just leave and go pick another job". Looking back at such conditions is obvious now that, I should have left sooner when I had the chance, but the past is the past and no one can change that, Now my knees and back will never furgive me fur such transgressions lol. 
After that I became one of those "guys who like computers" you talk about and got an office job, and yeah the expected happened, from moving a lot to not moving at all, made me feel nervous and stressed out all day, plus the countless hours making shitty excels and reports ended up making me feel mentally exhausted all day to the point of burnout. 
In the meantime I discovered the whole "AI thing" when models could only hallucinate some psychedelic dogs and write utter nonsense which allowed me to cope with that job. I really though AI was going to be a true "revolution" that would "change the world" lol. But Its been years and nothing has significantly changed, I still heard the "AGI next week" and "democratization of art" every 5 minutes when talking to a techno bro though. I guess they lied too and we'll have to wait another 5 years fur truly useful models that dont need whole servers to run. 
>Drawing art on the Internet means either you never know where your next meal is coming or you get emails from people who want to pay you to draw their fantasies of the cute characters you created murderfucking one another 
Yes, you hit the claw on the head. I recently decided to explore the "art community" and is exactly that. Some months ago when I decided to give it a try, and started to draw Due to shear desesperation lol . And the only communities I could find, were either filled with starving self-hating crabs or absolute degens without morals. Its a shame that art has become nothing more that a "commodity" fur these type of people. I always thought art was about communication, about manifesting a message, and I still do. But everyday it feels harder to connect and to interact with other people.    
>I feel worse fur Internet "professional artists" 
Now talking about "those people", i dont feel bad fur them, they decided to cross the limits and made their hobby into a job they'll furever hate. By selling their values and morals fur some quick bucks, they have condemned themselves. That's why you should never monetize art, because then it becomes a product. Therefure, I prefer to remain humble in my ways, and avoid those behaviours, because a simple mistake could ruin a whole drawing and a wrong step could ruin your whole reputation.   
After all of that, I still would like to say that we shouldn't just give up, why? Because someone has to plant the tree in order to eat the fruit. Some sacrifices are necessary fur a better future. Remember that poverty and ignorance are natural "human conditions" after all.
 
TL;DR: g{The grass always looks greener on the other side of the fence. Because everyone lies.}

>>689
>there comes a point where it's simply going to be faster to draw it yourself
Maybe, maybe not. I will try it though. Using >>690 as reference I did a drawing in around ~30 mins. This is probabiy the worst thread to post drawings but whatever lol
ClipboardImage.png U A
[Hide] (45.3KB, 1167x444)
>>801
At least this thread has been pretty comfy.
Replies: >>809
>>801
>due to the cuckflare captch
Name one alternative captcha that is free, actually works, and isn't a FOMO PoW open sores garbage and we will switch to it today.
Replies: >>809
>>801
Kitty gem with sovl
Replies: >>809
9109c662-691d-4204-9dd2-2cf9b865025b.png U A
[Hide] (1.4MB, 1024x1024)
>>802
>At least this thread has been pretty comfy.
Yeah, I agree, most of the AI threads on other sites always devolve into drama, this thread seems very chill. 
>>803
>Name one alternative captcha that is free, actually works,
I dont know a lot about captchas, so correct me if am wrong, but I heard that cloudflare is only necessary if a website is under attack. So I ask you ¿Is antroo. foo receiving DDoS attacks? if not, then maybe using a common capcha like BotDetect or something similiar would do the trick.
>and isn't a FOMO PoW open sores garbage
Kek, theres a lot of projects like those, but they may be useful under the right circumstances though, like "go-away" or "anubis" that are used against bots and Ai scrappers.
>>804
>Kitty gem with sovl
Thanks, your kind words mean a lot to me!  
>>677
>I'm trying to get a kot in a giant pocket, but SD 1.5 does not recognize the prompt. I'm trying img2img now using pic related as a starting point. So far it isn't looking promising.
Img2Img is good when you have the poses or the depth of an image, but it sucks with real photos and complex scenes. Have you tried using a tagger instead? I took this >>801 drawing and a got the prompt: 
solo, slit pupils, traditional media, green eyes, animal ears, monochrome, fangs, looking at viewer, :3, in container, greyscale, box, no humans, sketch, smile, animal ear fluff, 1girl
 
It's a promising prompt but the closest I could get was this "chest cat" lol. It look like a mimic chest trying to trick you lol
Replies: >>817
>>809
The issue with PoW "captcha" is they don't stop spam. PoW is only effective against DoS, requiring massive amounts of computing power. But for spamming it's a totally different vector to attack you only need 5-10 IPs flooding posts to make an image board unusable. A spammer's machine is just as capable as a users and they could easily solve 10 PoW captcha every second to continue spamming. This is all a educated guess, not exact numbers, but I'm just putting this into perspective that if a user can solve a PoW so can a dedicated spammer do it 10 times every second. Spamming is a type of DoS and PoW is only effective against brute force DDoS.
Spamming is only stopped by a mechanism to separate machine from humans, and PoW does not fit that requirement. You essentially pay a price to post on computing power, not be stopped completely (unless you break the captcha with vision etc) To solve a PoW 10 times/s is cheap, therefore it's useless for forums and to deter spam.
ClipboardImage.png U A
[Hide] (1.7MB, 1024x1024)
So, what if Norman Rockwell painted a certain fuzzy housewaifur?
Replies: >>1192
ClipboardImage.png U A
[Hide] (839.6KB, 640x640)
Krystal by Rockwell looks like Alfred E. Newman.
Replies: >>1192
3cea870dfbf6de23a1be5a4ade1ad0f4da582844ff90786414b99408517877ee.png U A
[Hide] (1.8MB, 1280x1280)
She calls herself Pwny Express and she is very l33t.  Say something nice about her.
Replies: >>1192
34f8a15bceffdbc14d8edf21bbf180e00698b4e32b5a20a2f86aaf73b9909c7a.jpg U A
[Hide] (100.3KB, 796x418)
>>1191
>>1177
>>1123
>no sisa in this thread
Replies: >>1219
>>1192
I think the nerdy horse girl is kind of cute in a Tomoko Kuroki kind of way, but I'm weird like that.
Replies: >>1226
Untitled561_20250603161312.jpg U A
[Hide] (1.2MB, 1536x2076)
I do too on occasion, but I'm lazy I use Pixai

Sometimes I do have to edit the pictures manually but otherwise it's pretty good
Replies: >>1224 >>1225
>>1220
>no piercings
that's what we call based
>>1220
Maybe it's the lack of a snarl, but she looks a lot younger.  She's not underage b&, is she?

Also, this is the kind of pedantic autistic bullshit fur which /b/ was created. I must lecture.  This is going to be a lot of tl;dr bullshit and you probably don't care, so feel free to scroll past it.

But, do demons have free will?  In the Bible it says that demons are angels who chose to rebel against God.  As such they do indeed have free will and they choose to do evil deeds.  But in the pop-culture theology of Current Year, the concept of "demon" carries the connotation of a being that is Always Chaotic Evil All the Time.  They always do the worst, cruelest possible thing possible.  They are, in fact, pretty predictable in this regard.  They may speak, they may solve problems, they may make and carry out plans, they may communicate and cooperate, but if this is the only manner in which they behave, is it the only manner in which they can behave?  If so, are they truly sapient in any meaningful sense?   Is there an "I" inside their heads at all?  Do they have minds like ours at all?  Or are they just p-zombies, just NPCs that were compiled with the box fur "gratuitous sadism" checked?  If that's the case, if they have no free will and are not capable of change or choice, are they truly evil at all?  Or are they just really destructive and dangerous natural phenomena, like a storm, or an avalanche, or a school of piranhas in some tropical river, and not moral actors at all?

I have seen some critics get completely wrapped around the axle about Hazbin Hotel and Helluva Boss, calling it "demonic propaganda," because one of the premises is that demons have minds that are human enough to have choices in how they act, human enough to be able to choose.  There are references in some Helluva Boss/Hazbin Hotel fan wikis to a hellhound character named Larissa ("yet to appear in the show," whatever that means, yet listed as having a voice actress) with a very gentle, shy, and anxious personality, who is a pacifist who loves animals. And she's listed as being voiced by Andrea Libman (if you know, you know). 

And that seems to >imply that demons, at least in this fictional conception, have moral agency and are capable of choice.  This seems much more consistent with the Biblical material from Ezekiel 14, Isaiah 28, and Revelations 12, than this modern conception of demons as being flesh-and-blood murderbots that do what they do without any choice or agency at all.

Of course, the modern conception of demons makes an exception fur Satan.  In current pop-theology Satan is the one God can trust to be Hell's warden, the one who can be "trusted not to be trustworthy,"  and capable of choice where other demons are not.  This is a recent gloss on the traditional concept, which holds that Satan is, rather, Hell's single most dangerous prisoner.

Yes, I know.  This is shit no one cares about.
>>1219
https://youtu.be/bDZ52NdwTNE
ClipboardImage.png U A
[Hide] (3.5MB, 1792x1792)
I too, am prone to enjoying a bit of genning.
>>680
there can be more to it than just writing a prompt and hitting button but at the same time, 99% of ai generated content is just that, unupscaled, uninpainted, unedited and uncared fur slop. as a long time enjoyer of photobashing, i get to photobash with what i gen and it gives me joy, it's fun and feeds in to a hobby i've had fur a long time. not that i consider it "art" either.
34ee859d7a0c0e4919eb04fabe4f5ee3b6ec29d7a148c9a7321045b4139f3b59.png U A
[Hide] (466.1KB, 512x768)
I found another online AI image generation tool:

perchance.org/image-generator-professional

I don't know what its limitations are.  I know it doesn't seem to be able to do anything over 512x768 in size.  There are drop-downs with dozens of styles with which to experiment.

It can do anthro women but you must specify "furry $SPECIES anthro woman" in the prompt and put "human" and "kemono" in the negative prompt, otherwise you just get human women with animal ears.  With a somewhat decently written prompt you can get something like this.
ClipboardImage.png U A
[Hide] (123.4KB, 744x612)
Apparently they also have a text generator.  Whee.
download.jpg U A
[Hide] (117.9KB, 512x768)
>>1319
found a furry-specific one on there
the site seems pretty good in terms of free stuff, but like most online services, I can't seem to find any inpainting tools, so you're stuck with those wobbly AI bits in otherwise nice images
ClipboardImage.png U A
[Hide] (774.1KB, 512x768)
>>1319
interesting, and I love your skunk anon
fob6.jpg U A
[Hide] (150.3KB, 512x768)
>>1319
this generator is actually pretty good even with a small prompt
Replies: >>1327
>>1326
>this generator is actually pretty good even with a small prompt
bet
do you take requests?
Replies: >>1328
>>1327
just go to the generator bro 😭
Replies: >>1329
>>1328
too lazy to do that, sorry
Replies: >>1330
>>1329
what did you have in mind
Replies: >>1331
>>1330
bicycle date with waifur
Replies: >>1332
ClipboardImage.png U A
[Hide] (731.6KB, 768x512)
>>1331
best I've gotten so far
Replies: >>1333
>>1332
the bicycle handles look wack as shit doe
Replies: >>1334 >>1335 >>1336
>>1333
yeah it's not perfect, free generator with minimal prompt yknow
Replies: >>1335
ClipboardImage.png U A
[Hide] (774.2KB, 768x512)
>>1333
>>1334
tons of potential doe. give it a try
Replies: >>1340
>>1333
If they're crooked, wack them until they're straight.
>>1335
What is wrong with his face bro
Replies: >>1342
de94a87a28977e1e635b4bfd3fb835bd82bed5889e5798e233775578b1d5fa94.png U A
[Hide] (466.1KB, 512x768)
>>1340
It's a perfectly cromulent face.
Replies: >>1346
490abae02e7942e68a84c2c8bad5b6cbffc7b819f7f51047080965e326094c94.png U A
[Hide] (617.5KB, 512x768)
zOMG! It's the Mugrums!
Replies: >>1344
>>1343
what's wrong with her hands bro wtf
Replies: >>1345
Spoiler File U A
(99.6KB, 734x1129)
>>1344
She's some kind of chimera. Most of her is white fur. Some is scaly. I personally think she looks like she's half aardwolf, half pangolin, but the character lore is open to interpretation.
>>1342
holy gem
00109-2516996087.jpg U A
[Hide] (381.6KB, 1792x1792)
Replies: >>1350 >>1355
>>1347
Bully Fox!  Did you create a character LoRA?
Replies: >>1351
>>1350
nah, i didn't. this is the one i used; https://civitai.com/models/1585122/v-bully-fox-fur-illustrious-request
00018-320327221.jpg U A
[Hide] (422.3KB, 1792x2240)
>>1347
rather dog/wolfish (which is based)
Replies: >>1360 >>1362
>>1355
this AI slop is bretty good ngl
Replies: >>1362
00003-1614264723.jpg U A
[Hide] (234.6KB, 1792x1792)
>>1355
>>1360
:)
>>1319
if you want to get in to ai generating "seriously" i'd recommend looking up how to run it locally. nvidia cards can virtually do it out of the box, amd card need some hoops jumped through. anything from and above a gtx 1060 6gb vram can pull it off. a good "beginner" card would be the 3060 12gb. 200-400 bucks depending on region.
Replies: >>1365
Spoiler File U A
(1.6MB, 1280x1280)
>>1362
:)

I have SD 1.5 with Automatic1111 installed, also SDXL, 3.5, and Flux. A while back I got this by prompting fur a fox woman in the style of Phil Foglio. But I still look around fur online image generators.
Replies: >>1367 >>1368
>>1365
understandable. this is my favourite current model, has a lot of flexibility with artist tags and loras if needed.
https://civitai.com/models/1280186/stablemondai-sdg doesn't use "human syntax" just tags, ie; solo, furry, female,
leg-59-gaseous-form.jpg U A
[Hide] (339.9KB, 672x936)
>>1365
Wait, this Phil Foglio?
Replies: >>1370
ClipboardImage.png U A
[Hide] (214.3KB, 824x834)
>>1368
Yes. SD 1.5 knows certain artistic styles if you prompt fur them, or at least the checkpoint files do.

I can't recall whether he's ever done much anthro stuff.  She does have the big googly toony eyes and the, well, huge tracts of land, that his female characters tend to have.

Which art styles it knows depends on which ones went into the checkpoint file.  Easyfluff and Queasyfluff, fur example, were trained on e621 data, and they definitely know what to do if you add "by Waspsalad" or "by Baron Engel" to the prompt.
try to make the AI gen a meme/cartoon with as little as guidance as possible and see what it makes
cheeto4.jpg U A
[Hide] (133.6KB, 512x768)
aiGODs I beg thee for one AI gen of a chubby tanuki lady being used as a lap pillow, extra wholesome, and make her blush
Replies: >>1385 >>1387
>>1384
purrgem wrote this btw
but yes, please make this a reality
Replies: >>1387
00134-2799867301.jpg U A
[Hide] (409.6KB, 1792x1792)
>>1384
>>1385
men want one thing and it's [spoiler]fluffing based[/spoiler]
Replies: >>1390
>>1387
Gem so pristine it fooled a nufur
Replies: >>1391 >>1392
f571735042dba2b0fe0c3ab3351786110a2763b574245b905467b018682e97bf.png U A
[Hide] (3.3MB, 1280x1600)
>>1390
When I click the orange button, sometimes something gud comes out.
Replies: >>1392
00169-2589141818.jpg U A
[Hide] (481.3KB, 1792x1792)
>>1390
:) actually reminded me i haven't genned my raccoon girl fur a while so have this :)
>>1391
i fluffing love mashing the orange button.
Replies: >>1395
ClipboardImage.png U A
[Hide] (841.4KB, 640x800)
>>1392
She's purty! Is there any character lore?

This is a sort of "artist's conception" of a character from a short piece I wrote a while back.  I am hesitant to talk about it in /anthro/ because the story is NSFW and I think it wouldn't be taken well.

Her name is Rose, or at least that's the name she uses.  She is a lady of negotiable virtue, a bit older and heavier than the norm fur that trade.  She is articulate, cultured, flirtatious, obliging, and  f l o o f y .  She is from another timeline, with very different prehistory but very similar recent history, where locally the year is 1953 or thereabouts.
Replies: >>1397
>>1395
ah, no lore as of yet. my other two characters i like to do, a rat and a bat, do have some lore though. maybe if i get around to posting a gen of them here i'll do some lore if i get asked.
Replies: >>1398
8930ae1780787dc00a39c66ed4cb81110e271c70f8685203cab955aea5447e95.png U A
[Hide] (402.5KB, 686x386)
>>1397
Gens? Moar liek gems.

>rat

Is she ebil?
Replies: >>1399
00113-1723588393.jpg U A
[Hide] (359.9KB, 1792x1792)
>>1398
her name is Libby (short fur Elizabeth) and she is a narcoleptic alcoholic wistar rat who is very nice :)
Replies: >>1406
>>1399
>wistar rat
I had to look that up.  Good lord, no wonder she's an alcoholic.
Replies: >>1407
>>1406
wistar rats recognise their favourite lab techs are, are happy when they are handled, exhibit care and delicateness and will not attempt to cause harm even if they're rewarded fur it, though this goes fur all rats. they're just sweet little guys really.
Replies: >>1415
>>1407
She seems nice enough, though the alcoholism is a bit off-putting. I wonder whether the drinking is one of her coping mechanisms fur being a survivor of lab rat stuff.
Replies: >>1416
>>1415
i don't wanna bang on about rats too much but some studies have shown that lab rats will prefer to choose alcohol over a social interaction with a cagemate. it's more of a parallel to that more than anything beyond that. it's similar to my bat character who has an ear tag paralleling the fact that bats get tagged fur tracking/health reasons.
Replies: >>1421
>>1416
Fair enough.
57faede8013aa8414b32c605860c205f690adf6697a5efca9808be261cc26382.png U A
[Hide] (420.7KB, 512x512)
Beavis & Butthead, but in the style of Grand Theft Auto game art.
ba71c02a03fd74654cf969ec2581ba5f5a2a15b251f85617fd918c53ea81325f.png U A
[Hide] (1.6MB, 896x1200)
This one took many, many, many tries.  It's a certain fox girl, as she would look if she were drawn by Moebius.  I am very pleased with how well it came out.
Replies: >>1442
>>1437
would
Replies: >>1449
aedd8f0e74fa0bfff475bf138dce889cc906b2be0d9caab509ad70f7d11316c7.png U A
[Hide] (382.3KB, 1416x2000)
>>1442
I am never quite sure whether the character is supposed to be really attractive, or just a tad goofy-looking.

I like her fur a number of reasons.  I like anthro fox ladies.  I've always had a thing fur redheads.  And she's an easy-going goofy stoner chick, a type of woman to whom I've always been particularly drawn. (no, it never works out)
Replies: >>1466
>>1449
What's it about stoner chicks that makes them really attractive?
Replies: >>1472
>>1466
https://tvtropes.org/pmwiki/pmwiki.php/Main/ManicPixieDreamGirl

Or, rather, the fantasy is the Manic Pixie Dream Girl.  The reality is that she smells like weed all the time, doesn't have any interest in guys who won't buy weed for her and get high with her. If you spend much time with her you may begin to notice that she's angry, maybe at the world, maybe at Daddy, and the dope is one of her coping mechanisms.  Some of these women aren't just into weed and you will get to witness all manner of self-destructive behavior and maybe even Borderline Personality Disorder.  Not all of them are like this but they are not always especially functional.  Maybe the sane and functional ones are such unicorns that they might as well be anthro fox girls too, given the odds of ever seeing one.  ("I reject your reality and substitute my own!")

Or maybe the problem is me. That's always possible. When the common factor in a long series of negative interactions is me, maybe there's something wrong with me, and not with everyone else. I have to be mindful. It's easy to start thinking of yourself as the Main Character, isn't it?  Maybe I'm not even the main character in my own life.
Replies: >>1473
>>1472
Very interesting thoughtful post anon, but I didn't read any of it I still want my stoner fox gf!
Replies: >>1474
6c48d1e08783b8a4fb4fb1a9529dcf1cd6053c3911464e62a6a908a1a6518643.png U A
[Hide] (2.4MB, 1536x1536)
>>1473
Maybe there are more than one.  (2023, colorized)
not morshu.jpg U A
[Hide] (1.3MB, 2304x1792)
was going to be a morshu edit originally but it genned with her mouth open unprompted so i stuck with it
enjoy plump robin i guess
>>1483
best Robin
>>1483
fatGEM
>>1483
I need this so bad
>>1483
I don't know, guys. This is the Bully Fox we're talking about. How can she bully you if she can't catch you?
Replies: >>1488 >>1489
>>1487
If you're not expecting it, she can fall on you like a thwomp
Replies: >>1491
>>1487
she's not fat, that's just her winter coat setting in
Replies: >>1490 >>1491
>>1489
Floofy!
fox has instant floof.gif U A
[Hide] (311.6KB, 326x464)
>>1489
It's a fox superpower.
>>1488
check 'em

Also, what if both are true? She's not fat, just extra floofy for winter, but she bullies guys by sneaking up and falling on them. But she only weighs like eighty pounds and half of it is fluff. She falls on them and takes a couple of seconds to drift down through the air towards them, landing without a sound.  It's not very effective.
bday.png U A
[Hide] (4.1MB, 1792x2304)
i made the first bully lora with the help of some bros from /sdg/ back when i was trying to learn how to do loras
half considering doing a furshit lora that isn't just some who gives a shit undertale/dr character
gimme suggestions
Replies: >>1524 >>1525
>>1523
it’s scary how well the ai got the artstyle right
even the hands look alright

are we cooked?
Replies: >>1525
f99f0e589e03af5645a0ce4a3eb8893a6e7679726445aaedfa6f1a652a63afba.png U A
[Hide] (5.4KB, 341x291)
>>1523
Hmm.  In no particular order:

Sketchytoasty's fox gal Pepper
SDS's unknown whatsit Morgan
DV's adorkable Buh Dog, Lynn
Napalm Express's spooky yet floofy cryptid waifu, Missy

The advantage is that if you look around Furaffinity, e621, and the /hmofa/ threads on Desuarchive, you can find lots of sample images.  Share your secrets with us, Sempai. I have tried LoRA alchemy with spotty, mostly "meh" results, but I will be the first to admit I have no idea what I'm doing.

>>1524
t. Eliezer Yudkowsky
Replies: >>1529
5a37c852ab86f7a92706b1eff6db5ff16ab6e149e8fde61063642b1b45d94fae.png U A
[Hide] (1.6MB, 1248x1824)
jumbo_09331_.png U A
[Hide] (701.4KB, 832x1216)
>>1525
knock yourself out
here's a guide i made on how to do it with OneTrainer
it's a bit complicated, but the learning curve is only at the beginning - once you get used to it, it's piss easy, most of it is just waiting around
>https://files.catbox.moe/9piyw2.txt
Replies: >>1531
>>1529
Okay. When you make a LoRA for a character, do you have the tools base it on one of the existing specialized checkpoints, or do you have it build it around v1-5-pruned-emaonly.ckpt?  I read that using any other checkpoint meant the LoRA would only work with that checkpoint in the future, so I mostly used the emaonly one.  But I get the same meh results whether I use it or Queasyfluff.

I've been using kohya_ss for LoRA stuff.  My daily driver is a Linux box. There is a Windows 10 Pro N Ameliorated installation on one partition but I don't use it much. Because it's so stripped down it doesn't support installing Python. A bunch of .dll files present in normal Windows installations aren't part of the N versions, Python is dependent upon them, and the missing files aren't available as downloads so far as I am aware. 

Also, while I'm asking dumb questions, I haven't been able to find any real information specific to creating an artist style LoRA, as opposed to a character LoRA.  I've even seen contradictory advice about whether a character LoRA should include the character's name in the tags, or whether doing so would cause SD not to associate the character name with the sample images.  I didn't have much luck either way.

Because I'm working with SD 1.5, I read that you can't make LoRA files for it that have resolution over 640 pixels in any dimension. Supposedly it was created for 512x512 images and trying to generate any images at a pre-resizing resolution much above that results in SD 1.5 becoming confused and filling in the empty space with extra heads, extra limbs, and so on.

I also read about a technique where you train one with 512x512 images then you go back to the images you cropped/resized and get all the same images but cropped and resized to 512x640 portrait aspect ratio, make a second LoRA, then merge the two to get a third, which supposedly results in improved quality. It didn't seem to make any difference when I tried it.
Replies: >>1532
berdmom3.png U A
[Hide] (2.9MB, 2012x1792)
>>1531
>Okay. When you make a LoRA for a character, do you have the tools base it on one of the existing specialized checkpoints
no, i just use a base version of noob, because i find that variations/merges/whatever just make the result lower quality, where as just base noob makes it reliable
the thing with sdxl models is that it allows you to use for instance pony and illustrious loras with your noob model, so really it doesn't matter that much
>Because it's so stripped down it doesn't support installing Python. A bunch of .dll files present in normal Windows installations aren't part of the N versions, Python is dependent upon them, and the missing files aren't available as downloads so far as I am aware. 
im not a linuxfag so i don't know how to advise you here anon, sorry
> I haven't been able to find any real information specific to creating an artist style LoRA, as opposed to a character LoRA.
the difference is in the tagging, for character loras, all you do is tag the composition, what there is in the image, etc (you tag everything that you basically don't want baked into the image whenever you use the lora), the activation word you use to basically activate all the stuff you didn't tag for, i.e. the character
as for a style lora, the only difference is one, you don't need to use an activation word, and two, for all the clothings, background details etc, you tag that as usual, but you add the color to it as well (you shouldn't be tagging individual colors with character loras).
>Because I'm working with SD 1.5, I read that you can't make LoRA files for it that have resolution over 640 pixels in any dimension.
<he still uses SD 1.5 in the year of our lord anno domini 2025
switch to noob, its the current meta
[New Reply]
144 replies | 61 files
Connecting...
Show Post Actions

Actions:

Captcha:

- faq - rules -
anthrochan 1.8.0