#🍥|anime
1 messages · Page 81 of 1
looks like a typical child "BUT MOOOOOMMM" in a car seat
after getting denied mcdonalds icecream
no ice cream because the machine was broken
ooo there's a generation idea
"broken mcdonalds ice cream machine"
hmm my controlnet is a little wonky.. I think I need to re grab the models/yamls
that doesn't look like mcdonalds
Maybe she's the one that always breaks the ice cream machine https://cdn.discordapp.com/emojis/1093689156280000524.webp?size=48&name=Anime_Think&quality=lossless
here's today's Tifa.
Needs headpat
any of ya'll running a 4090?
Same
one of these things is not like the others https://cdn.discordapp.com/emojis/1093687535533830157.gif?size=48&name=SataniaLaughing&quality=lossless
👵😃🥒
anime line art style, 00110001 00110110 00100000 00110001 00110110 00100000 00110001 00110110 00100000 00110001 00110110 00100000 00110001 00110110 00100000 00110001 00110110 00100000 00110001 00110110 00100000 00110001 00110110 00100000 00110001 00110110
Everyone in the background looks the same 
or maybe it's the 1's and 0's... shrugs
(Beautiful young woman with vibrant medium long pink hair and a cyberpunk outfit with, she has visible augmentations on her hands, face and chest,:1.255) solo, [Gearpunk::0.10], [Cyberpunk::0.40], [Diesepunk::0.45], [Steampunk::0.55], [Futurepunk::0.65], (highly detailed beautiful face with smooth contours:1.18), , Electroluminescent Wired to cybernetic sci-fi horror machinery, Tesla Coil, Electric Arc, in gothic architecture [dark fantasy|cyberpunk|futurecore] space cathedral, (clean pastel art style:1.27), (digital art:1.20), (2.5d:1.15), (splash art:1.10), cinematic lighting, cinematic angle
Or for img2
(Beautiful young woman with vibrant medium long pink hair and a cyberpunk outfit, she has visible augmentations on her hands, face and chest,:1.255) solo, [Gearpunk::0.10], [Cyberpunk::0.40], [Diesepunk::0.45], [Steampunk::0.55], [Futurepunk::0.65], (highly detailed beautiful face with smooth contours:1.18), stands in a neon lit cyberpunk city at night, [dark fantasy|cyberpunk|futurecore] space cathedral, (clean pastel art style:1.27), (digital art:1.20), (2.5d:1.15), (splash art:1.10), cinematic lighting, cinematic angle
Or for img3
(Beautiful young woman with vibrant medium long pink hair, intense vivid violet eyes with subtle cybernetic details and a darkwave futurepunk leather outfit and jacket:1.25), (she has visible augmentations on her hands, face and chest:1.15), solo, [Gearpunk::0.10], [Cyberpunk::0.40], [Diesepunk::0.45], [Steampunk::0.55], [Futurepunk::0.65], (highly detailed beautiful face with smooth contours:1.18), stands in a neon lit cyberpunk city at night, [dark fantasy|cyberpunk|futurecore], (clean pastel art style:1.25), (digital art:1.20), (2.5d:1.15), (splash art:1.10), cinematic lighting, cinematic angle
All with Negative prompt: low quality, no quality, bad quality, low resolution, lowres, normal resolution, no detail, low detail, deformation, text, title, Description, logo, watermark, image a rtifacts, cropped, cropped face, cropped upper body, cropped lower body, spaghetti body, siamese twins, conjoined twins, conjoined body parts, big breast, massive breast, major breast, big tits, massive tits, major tits, hentai, eroge, nukige, pornographic, porn, XXX, erotica, loli, lolicon, skimpy, slutty, nude, nudity, naked, neg_hands, bad_hands
big breast, massive breast, major breast, big tits, massive tits, major tits LMAO
puts all the negs in the positive
Yes cause some models are so fucking focused on having jugs with their own orbit
😄
o.O ok.. I can't share that
seems it like's th is face.
Girlish Photo:1.3) of (Technical illustration:1.3) anime line art world/town, random Technical magazine detailed scamatic,Highly Detailed,(Warm Colors:1.3), masterpiece, best quality, fine art, ultra detail, 8k, rtx, hi res, cinematic lights, photo realistic, concept art
Negative prompt: an6, kkw-extreme-neg,Amateur,Low rated,Phone,Wedding,Frame,Painting,tumblr,watermark,signature, lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name
same but vertical
takes a nap
but before I go
Good luck
what's for dinner?
😶
j/k i don't car what skin color ariel might have 😄
sebastian is where it's REALY at 😄
crabs are weird
@untold glacier - I am trying this thing out but maybe I'm doing it wrong... it's really really slow. 😐
Hmm, what's up?
Lemme see your settings
Hires fix, and multidiffusion+tiled vae settings. Just send me a pm
I just used the same stuff you had. Seems like the hires image is taking longer than usual.
I mean, it will take longer but it shouldn't be much longer than usual. How many tiles is it saying that it's working on? It'll say something like this:
I've run so many now I can't find the big one. Will need to do a new one. Need a min. 🙂
outpaint fail 😄
Which outpainting method are you using?
I wasn't really outpainting. 😉 I was using just a simple crop & fill. Sometimes i get better results. Just being silly. 🙂
Ah, have you tried controlnet inpainting to achieve outpainting?
... I... don't know what that is. 😄
Let me find a decent candidate to try it on and I'll make a quick tutorial.
Before and after. It'll take a minute while I type out how to do it.
Cool, I'll keep an eye out. I'mma go grab a burger. 😄
How to outpaint using ControlNet:
- Step 1: Open your base image in your image editor of choice
- Step 2: Expand the canvas to the new size that you want. You doesn't matter if the new area is filled or transparent, although making it white will make Step 4 a little easier. Save it as a copy if you would like to keep the original.
- Step 3: In the WebUI txt2img tab, set your resolution to accommodate the new aspect ratio of the image and set your prompt to match what you want in the new area that you're outpainting.
- Step 4: Add the image that with the expanded canvas size from Step 2 into controlnet, set Control Type to "Inpaint", and mask the area that should be inpainted in the Controlnet image field just like you would if you were inpainting. I recommend enabling the Pixel Perfect option. You can experiment with the Control Mode to see what you like.
- **Results: ** #🍥|anime message
@dull jackal Can this be pinned for easy access in the future?
Made another high res husbando.
<@&1060374302433738782> I'd say this might be worth a pin. Whaddya think?
I've been trying to use controlnet off and on for months. I can't even use reference mode cuz it gives me an error that nobody else has experienced. Outside reference mode it's hit or miss if it even does what I tell it to do if I use it with a lora, composable lora and uhh that other thing I forgot what it's called it creates the blocks allowing you to connect prompts with AND.
The real issue seems to be lora compatibility for me
Heard there was husbando posting?
pinned ^^

Nice, wasn't aware of the 'inpaint' control type.. will try this! I've definitely used expanded canvas with normal inpainting, but not this.
hmm, this doesn't work for me @supple raptor . Changing the seed/prompt changes the base image.
Are you in the txt2img tab? This should not be done in img2img.
Yes
Maybe I am doing something wrong
Here's my base image with expanded canvas. 960x400
If I keep the prompt/seed, it's sorta similar.. though with obvious problems with the outpaint.
changing the seed/prompt completely changes the whole image
Here's the controlnet settings.
oh I wonder if maybe I'm missing an inpaint model...
I have an older install of controlnet so maybe I need to pull down more recent models/yamls
Need a pointer. 🙂
Think I found 'em on Civit. Hope this is them. 😐 https://civitai.com/models/38784/controlnet-11-models
STOP! THESE MODELS ARE NOT FOR PROMPTING/IMAGE GENERATION These are the new ControlNet 1.1 models required for the ControlNet extension , converted...
Ok so now I have an inpaint model. Selecting that now retains my original image. But the newly generated stuff just doesn't match at all. 😄
Need to figure out how to get rid of the obvious inpaint lines.
Make sure the sampling method is the same as the original image
I've noticed some times when I switch from txt2img to img2img it keeps the default euler a sampling method, and I usually use DPM adaptive and have to change it manually
So-Bin model?
me?
cool horns...
hmm back from work. whats todays challenge?
spots the yellow jacket kill bill type girl. hmmm
hmm got a lot noiser with trash, interesting...
hmm it's not really liking the face. odd
there thats better 😄
attemptes to create an image using just a white image, and inpaint :D
mutters about inpaints lines. blah
hmm didn't work so well
chibi!
it's resisting
evenin'
imma see if i can make chibi
huh it's not changing. wtf
not enough weight 
tries again

mutter

mini me
hmm.....
thats better
forgive the fingers
What model do you use?
ok great picture but thats not a library
hmm I wonder
oh I misspelled it as collage insteade of college
doh!
mutter
hmmm
hmm failed successfully??
well ok then. hmm umm not sure what warning to put
||(masterpiece, 1girl, solo, beautiful eyes, hyper detailed), flower hair ornament, red flower, butterfly on hair, fox ears, white hair, blue eyes, short hair, single braided hair, (Shirakami Fubuki:1.1), maid outfit, hanfu, black sash, huge sleeves, long sleeves, (red flower) hair ornament, blue neckerchief, blue neck ribbon, black_boots, ahoge, white thighhighs, short sleeves, (black sleeves), hair clip, standing, gunatyou, gun, aiming at viewer, holding gun,||
changes to the new anime merge made and tries again
Anyone know a model or loras for fake anime screenshot
oooooh

need some old english type
latent, animesharp, or fatal
i use lanet (nearest)
Do it. and report back your findings.
Latent, AnimeSharp and R-ESRGAN 4x+ Anime6B
Kind of scary with that blank look
time to sleep, good night
Damn it I've gotten really rusty when I haven't barely prompted at all for a few months
man I love going back to old stories I've written and putting in a paragraph or 2 from certain scenes. Stable Diffusion is really impressive at understanding it
i've tried to come back to gens but everything is so much slower for no reason
genning is around 10x slower, and i get so many errors
this just comes out of nowhere
you try restart computer?
@gritty patrol have you tried pip install -r requirements.txt in command?
in the main folder
it seems like either a compatibility issue with lacking something or having something outdated. Or a corrupt venv
still happening
delete the venv folder and run webui-user
it will download a lot again
but if there's a corruption in there
it'll fix it
artistic filter in clip studio paint is nice
oh the background is seethrough on the line-art
hah
beginning to understand control net now
Backgrounds
What's the checkpoint?
thighs
What's that model
Really liking the eeyes
custom mix I might upload later
though I think the eyes are a result of the lighting
Looks great
thanks 🙂
I'm a sucker for great eyes lol
Huge
i need to work on my mix
very nice
Something's been bothering me that I can't unsee: I thought it was just me but in some of the photos here there's excessive hard lighting in the forehead areas. But when I use a lora it doesn't seem to happen much at all, no lora and it happens often.
Vae or models made no difference
I see that a lot with anime style checkpoints in general.
guess that means nobody found a solution, at least outside of inpainting
i don't really understand what you meant

looks like a white headband lol
well while that wasn't in my prompt maybe it just inserted one
this thing?
nah
little weird white highlight
top right
am I just being like ultra nitpicky?
can you like circle it, cos i have no idea what you are talking about
even zoomed out you can see it clear as day
the highlight is too hard for the rest of the softness of the color tone
inpainting was also messy
yea that one is quite a bit cleaner
yup, its definitely a style if you like or if you dont
pretty sure it is a limitation that isn't solved yet .. the AI algorithm don't really understand how to do that kind of lighting as there isn't enough data on it (inpainting and hoping it does the correct skin tone is your best bet)
You are probably right, it would explain why I rarely see it in loras cuz loras are providing it info on it
have yet to find a good prompt for anime generations with good 1:1 reflections of the character 😄
It's probably a symptom of the same problem that noise_offset is meant to solve. Basically the AI is attempting to make the image at approximately 50% average brightness which results in some areas being brighter or darker (usually brighter) than they should be.
sad Nico
That composition actually looks good and competent, is your stable diffusion broken?
only the stablest of waifus for me 
exactly who wants a waifu that's just going to fall over all the time?
only silly people
extra legs = stabilization
extra arms = superior hugs
extra fingers = idk... I haven't solved that one yet.
Model?
extra grip definitely 
Aurora with <lora:loraslayernoiseoffset:0.45> <lora:pen_sketch_v1.3_dim_16:0.8>
oh aurora? I have a lot of trouble getting good stuff with that
i mean when she holds pen and such 
Aurora is one of the easiest models to use for me. I find that it prefers a simpler negative prompt than many other models though. This is what I use: negative_hand, (KHFB, AuroraNegative), (Worst Quality, Low Quality:1.2), border, nsfw, skimpy, grayscale, multiple_girls, (watermark:1.2)
ahh I probably need AuroraNegative, is that built into the model?
That one and KHFB are both from the embeddings folder in the model's repo on HF https://huggingface.co/SweetLuna/Aurora/tree/main/AuroraEmbeddings
You can grab PNG info off of one of my images to see what I'm doing. I pretty much always keep the same CFG and sampling steps and just change samplers when I find a prompt that doesn't work as well on DPM++ 2M Karras. For Prompting I mainly use booru tags, but do sometimes include some natural language with it.
Did Marissa fuse with Utsuho or something? Got that three legged thing going on.
You should inpaint crow wings.
@wispy canopy Why didn't you ever tell me that DarkAlfa got an update 
I've been using 1.0 this whole time.
😮 yeah...i did update to 1.1 a while ago xD

which improved the general image comprehension
too late, I've already moved on to generating Reimu
nice look
I see that it doesn't make as good hands, or at least you have that note on your Civitai?
i think i noted that in the v1.1
hands aren't as good
as you are right to notice, v1.0 has REALLY good hands for some reason
Well, time for me to test it then.
I only found this out because I was gathering images together for writing a review, lol. Only to find out it'd be for a PREVIOUS version.
v1.1 seems really good to me
I got to hand it to you on that one
It's looking much better, the checkpoint seems to just be extremely sensitive. It made the image monochrome when I used the word "vibrant" though, which is funny.
i really tried getting the v1.1 hands right, but whatever i did, couldn't get the block weighting for the hands right without screwing up the rest too (at the same time)
and after trial and error for couple of days straight, i kinda gave up and just did the "best possible" merge for 1.1
BTW someone over in the Touhou server made a debloated fork of Umi-AI with all of the embeddings removed. He also added a bunch more wildcards to the normal list that Umi-AI comes with. In case anyone is interested: https://github.com/Tsukreya/Umi-AI-debloat
I wish we had the ability to set the hires fix model as a different model than the current one. That'd solve so many issues--getting a general layout for what we want and then focusing on the details with the upscale.
yeah, i'd rather do the upscale with v1.0 and do the initial image with 1.1
not sure why it went monochrome but that looks nice https://cdn.discordapp.com/emojis/978879333248667718.webp?size=48&name=AYAYAYA&quality=lossless
v1.0 has a really clean artstyle
Is this more usable than dynamic prompting? I.e. is there a reason I should switch?
You do have the ability to do that, you'd just have to either use ComfyUI or do the upscale+img2img portion of Hiresfix manually in A1111. It would considerably slow down generations to have it happen automatically though so I don't really expect him to add that option.
I feel like just doing post-processing upscale is a bit janky in contrast to hi-rez fix
hey nubby since you are here, how does segment anything works?
Ignore me, I can't read.
I don't really know. I don't usually use wildcards and keep dynamic prompts disabled unless I need to use it because it interferes with the prompt blending script that I use.
I can't read again. What I read as ComfyUI, I interpreted as CozyUI. 
Here you go if you want to check it out https://github.com/comfyanonymous/ComfyUI
What are the benefits, you feel, of switching to comfyUI as compared to A1111 and Vlad-Diffusion?
I spent so much time customizing 1111 I can't jump ship anymore. I just gotta wait for an extension to fix the problems I have with it.
I had completely forgotten about that one. I tried to get it working a while ago and something went wrong and I never came back to it. As I understand it, you use it for inpainting (either on the img2img tab or txt2img with the inpainting controlnet) and it uses one of the segmentation preprocessors on the image combined with some sort of vision model that allows you to use a text prompt to tell it what you want to mask. Its sort of like a more flexible and more accurate, but much more cumbersome version of Adetailer with the biggest difference being that you have to supply an image as well as a prompt for what to mask.
ahhh, okay, so basically dont use it 
It's probably very good, but I don't know that it's worth the effort. If I really want to inpaint something accurately I'll just load up Krita and use the extension personally.
ahhh gotcha
to his reflection
what models/prompts do you use to get those crazed evil expressions?
i have a solution.....
What are the best sampling methods?
What's the solution?
hello
hiyo
lyco testing weee
I think this one is worthy of my "enhancements".
go for it 
hehe add chibi and well...
there we go
interesting
sings along ~how bizzarre~
Honestly sharpening and Vignette is pretty much cheating.
nice nice nice 
hmm ok then
hmm lets try that with my other model 😄
not what I wanted but ok
ooh fail
Nico heard you're not a fan of her
She ain't smiling that's for sure
crazy anime girls are cute and cuddly. A fine example of 2d > 3d
😄
anybody know how to get diffenrent mask then surgical/medical mask
negatives really dont work
ok more anime
now thats interesting
hmm ok..
ah ha thats what was going for
not perfect but ...
haha
uhh wah!
vampire
hmm am I the only one awake?
a drawing of a girl drawing a drawing
hmm interesting
I think she wants lunch
hands.. almost .. almost
hmm 5 fingers and a thumb.. don't look bad really
now thats anime...
hmm not sure whats going on here
Stand before she who is stone
See? No difference! This is why I always use (Greg Rutkowski:69), (masterpiece:420).
yup, maybe this model just wasnt trained on the concept of contrast
nijijourney

I want those Niji gens to be SDable 
Well, the style mostly 
They got some kewl styles
i think the models they have are huuuge
I need all the niji styles as LoRAs :o
buy 1 month and create your own 😄
I mean
I follow a IG page that makes stuff with Niji, I tried making a LoRA, but I grabbed too many pics
XD
not even your own?
I mean, why would I make pics when I can just grab some
I've seen this style so much lately, what model is that?
ExpMix_Line_V2
:O
O there is V3? 
Luna knows me well 
i'll look at it, thanks y'all
yup yup 
Well, yeah, but you told me to make pics with niji :o
Wao :O
ye xD we drive in circle now anyway 😄
Well, then why would I use the pics of my failed LoRA xD

have you tried Pika's model? https://civitai.com/models/47067?modelVersionId=71733
She grew up 
I haven't, I've had it on my list of things to download for like a month tho xD
she sure did 
V2 is really nice. V1 had the style, but it was wasn't very flexible and had some anatomy problems.
Ohhh, fair then :O

I have no idea why she is squatting in like 3/4 of the images. I have no pose specified.
Fair, looks pretty tho :o
I... I think it just made her a little person...
Shes smol :O
Windows wants to update to Windows 11. Should I do it?
i don't think you should
haven't really heard good things about it
Yeah, I guess 1 am probably isn't the best time of the day to start that process anyways.
true true true
uh... miss? don't look behind you, but you might want to start running.
This one trained decently I think :o
ah dang
O, good nite :O

Seems decent enough :o
Also yall recommend any anime? I wanna watch something but I dunno what
what types of anime do you like?
Also damn my OC turned into a mommy using Pika's model
Well, I like fantasy stuff, though I have been watching Oshi No Ko cause its really good :o
Goblin Slayer,
Hai to Gensou no Grimgar, Vinland Saga (not really fantasy, but it's good)
Though also I think I still like Aurora More :o
Ohhh, arent all those pretty dark tho
xD
Yeah I guess they are. Let me look through my list for some lighter stuff.
It's occurring to me as i look through my list that I watch mostly dark, sad, or romantic stuff XD
O damn, fair then XD
Doctor vampire Nico will see you now
Don't ask why her previous patient never walked out
Have you seen BOFURI? It's not technically fantasy, but it might as well be. It's lighthearted and chill.
Oh and I need to ask, could you give me the prompt for this? I love it so much
Sure! Give me a minute to find it.
twitchemote, 1girl, yandere, constricted pupils, shaded face, crazy eyes, yameroyandere, crazy smile, (open mouth:0.7), upper teeth, pink frilled dress, frills, sticker, white background, simple background, <lora:EmoteMakerLora:0.8> <lora:yameroyandere_v10:0.8> <lora:lighting-locon:0.3>
Negative prompt: AuroraNegative, KHFB, bad anatomy, disfigured, deformed, malformed, mutant, gross, disgusting, out of frame, nsfw, poorly drawn, extra limbs, extra fingers, missing limbs, blurry, out of focus, lowres, bad hands, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username
Steps: 20, Sampler: DPM++ 2M Karras, CFG scale: 7.5, Seed: 3505763780, Size: 512x512, Model hash: 1b5f8211ec, Model: AuroraONE_F16, Denoising strength: 0.51, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: 4x_fatal_Anime_500000_G, Discard penultimate sigma: True
You're awesome
anime art is super attractive when you add some crazy to it
Oh, I'll watch that one :O
And I mean, it seems pretty fantasy to me xD
Yeah I think there's maybe like 15 minutes of the show that takes place outside of the game.
Fair xD
took me about 1 hr to upscale omg
what video card you using?
I used them to upscale to 2560p
but plus the controlnet
3070ti laptop
control net might add a lot of time yeah
at least initially when loading and processing
why control net when upscaling?
This was your problem right here. You were swapping out of vram Lower your tile sizes for both Tiled Diffusion and Tiled VAE next time.
I dont know why the glow could be pretty good when I add lineart anime
just tried to see if it would be more glowing
I see
try a lora/locon with offset noise and if you want a smoother more digital 2.5d effect, adding line art to negative prompt can help
hmm elemental ratio in supermerger is fascinating
giving like precise control over what block and lane should be set to what ratio
attn1, attn2, layers, ff net etc
I've been obsessed with anime art my whole life and now that AI is throwing anime art in my face 24/7 I realize that I can't get enough of it no matter how much I see
I let my laptop keep generating anime art while I'm doing works
I sometimes spend 12 hours a day prompting and regenerating
got my Discord bot loaded to make AI art, I'm still making it at work whenever I can play wi th my phone
lol
I found it
ahh yeah the API gives you what the json payloads should be it's honestly very easy
I think I'm going to make a cenobite lora today if I can
I had to look at it so you have to too
If it's something you can get good quality screenshots from even if there's zero fanart you can make it happen.
I trained this off straight anime screenshots cuz the anime was unpopular and nobody made 'any' art of it wasn't even an artbook
https://civitai.com/models/68355/mona-teppen
Anime I thought deserved better. Teppen—!!! Laughing 'til you cry. My goal is to get most of the girls from Teppen made as soon as I can figure out...
and it came out really well
cursed hand but this is still super cool
using LLM to write prompts for you >
nice cropping 😒
love the girl in the forest
meta-human creator is amazing
is controlnet impossible to use in 6GB vram?
thought u were playing starfield for a sec
i remember there's an option for lowvram
just tick on that
heh, I wish
It's possible with 4 GB, but I had to use everything possible to lower memory usage
--lowvram --opt-split-attention --always-batch-cond-uncond --xformers
if it doesnt work even with Low VRAM enabled then chances are that yeah wont work
What is control net?
Dw
https://github.com/lllyasviel/ControlNet
https://civitai.com/models/38784?modelVersionId=44716
the first link is the extension, i recommend reading it, the second link is where to get the models for it and what they do,
they offer multiple purposes and its really helpful depending on your needs.
a quick screengrab from the github to quickly visualize a purpose
Is it a tool that creates images but uses ur pc to render them?
Ah wow can it colour grade?
uh im unsure what you mean by that
oh im looking it up and mm
i never used it for that purpose but im sure someone has
Is it like stable foundation that basically creates something from a prompt or is it used to make already exsisting images different/better?
Ah okay
Ahhh so you can basically add an image and use a prompt to edit it? Cool
Cool cool il check it out
Btw for this i used prompt: solarpunk art ghibli style (ecc ou can put something youd like to focus)

i used to get really good gens but now the image is slightly uncanny
very cute
Holmes or Wattson? lol
Such beautiful hands
Got that perfect rng seed
"what do you mean you can't generate hands"
use that luck and buy a lottery ticket before it's too late
just realised i mistyped
inb4 1girli was the solution all along
Gotta love it when the model got hands
I guess I'm saving this merge
my friend just sent this in a group chat
hahahaha
oh man if those actually work
wtf is ths
both are creepy non the less
pssst https://civitai.com/models/92845?modelVersionId=98971
my first lyco is out 
https://github.com/KohakuBlueleaf/a1111-sd-webui-lycoris Trigger word: Tiefling This is my first Lycoris! I created three versions that resulted fr...
they didn't have a pony... sad
tries to make MLP
waits for horror
ok.. not bad...
the ai went nuts.. wtf?!
ok i'm not sure what I want to know but on point for pinky... sigh
well.. it's better
well.. ok then...
yeah some of these I can't share 😄
lives new meaning to losing the plot
gives
hides
pink going on a trip..
laughs.....
Does Vlad1111 update automatically?
you see those marks of their face? I have a guess where they came from
MFW I forget to turn off restore faces when switching back to an anime model lmao
evenin'
How can one make the inpaint be filled with the entirety of the prompt? I'm trying to inpaint a pile of bones and skulls but it just makes a few.
If I go above 0.65 Denoising strength the AI attempts to add the entire prompt to the area.
Yeah, generally more denoising is the key. Using an inpainting model can be helpful. You can make your own inpainting models by doing an add difference model with the SD1.5inpainting (A) + (Your model (B) - sd1.5-pruned (C)) at 1.0 multiplier. The merged model must have .inpainting at the end of the filename so it should be something like yourmodel.inpainting.safetensors. If your model is an sd2.x based model then you must change out both of the SD models in the recipe for the SD2.x equivalents.
Alternatively, you can just crudely composite in an image of bones/skulls from the internet and inpaint at around ~0.5 Denoising strength. The ai should blend it in as if it were always there.
Workflow?
I downloaded it and there is nothing you can do about it!
redid some words here and there and used some presaved prompts and style pile scripts for some extra prompt help so.. umm yeah
That's a lot of words to say you don't wanna share workflow lol
<<------
what can I say i'm random
and my adhd brain doesn't do well at remembering everything I do
prompt
tho I created my on merge of three diff models
my thanks
hmm looking at neg prompts seems I didn't change some from prev prompt messing around. ooops
😄
declines to further comment
but lets just say for some reason putting in bathing suit or swimsuit , the ai scripts add underwear , which looks very odd
ok there was more ocmments....
so... Windows 11 appears to have come with one hell of a vram upgrade for me. It got up to 1.5TB just for Firefox before I closed it 😅
hahah
It downloads more vram automatically
how nice
if only....
Can you imagine the shenanigans we could get up to with 1.5 TB of VRAM?
not really
cause I mean.... thats like me wnning a billion dillars
dollors
doll-haires
it's late brainz fading
Could probably run a couple of instances of GPT-4 locally along with DeepFloyd and SD and still have more VRAM left over than we know what to do with XD
i would be happy adding vram to my latop
extrememly rare man gen from a gay
Sure
Prompt: What the fuck did you just fucking say about me, you little bitch? I'll have you know I graduated top of my class in the Navy Seals, and I've been involved in numerous secret raids on Al-Quaeda, and I have over 300 confirmed kills. I am trained in gorilla warfare and I'm the top sniper in the entire US armed forces. You are nothing to me but just another target. I will wipe you the fuck out with precision the likes of which has never been seen before on this Earth, mark my fucking words. You think you can get away with saying that shit to me over the Internet? Think again, fucker. As we speak I am contacting my secret network of spies across the USA and your IP is being traced right now so you better prepare for the storm, maggot. The storm that wipes out the pathetic little thing you call your life. You're fucking dead, kid. I can be anywhere, anytime, and I can kill you in over seven hundred ways, and that's just with my bare hands. Not only am I extensively trained in unarmed combat, but I have access to the entire arsenal of the United States Marine Corps and I will use it to its full extent to wipe your miserable ass off the face of the continent, you little shit. If only you could have known what unholy retribution your little "clever" comment was about to bring down upon you, maybe you would have held your fucking tongue. But you couldn't, you didn't, and now you're paying the price, you goddamn idiot. I will shit fury all over you and you will drown in it. You're fucking dead, kiddo.
A taco bell breakfast quesadilla I'm guessing.
hahaha
🤖
that's probably someone's fetish
gonna be mine if I keep generating those

what model/loras?
such a cool style 
When the hands ain't bad but we needed 2 legs merging with her other 2 legs


