#🍥|anime
1 messages · Page 237 of 1
hmm interesting flat image but 3D poster
tho I think someone cut her hair off
oops
have a cookie
What are you doing!?! I can't eat that many cookies!
huh... ok
Fennec?
got me
shrugs
I didnt' prompt for it but I am using a merged with a furry model so .. yeah
haha
it's early grammer turn on is not
ok back to my NOMD model 😄 anime
hehe
dance floor
Cookie dance
Why is she holding one on her hand and her mouth
I'm ready

Precipitation in the form of cookies is expected!
yes... not the same Gummy bears anymore
I'm in the hospital right now and have nothing to do except my laptop.

Why hospital 
I got into a car accident, I'm not really hurt, but I'm in the hospital for about a week to get insurance compensation. lol
take this chance to check yourself well
thanks ~!!
Little kittens go help
Get well and get well soon! Take care of yourself!
thankx mate !
dall3 still king of comic style...
big purple guy!
The style of the Egyptian pharaohs is missing
ok thats not as easy as I thought. still trying
we wait
This is not the right Thanos! he's wearing heels! The galaxy is in danger!
ok moving on
change his hair colors, his dressing in jeans and that's me from young age lol
when i was very young, i was such a thug lol
making mistakes teaches a lot 🙂
A mature miku dancing in a gotham ball room. or running away not sure ...
to many arms
changed it to flatter color scheme.. trying different things
😄 had to do it
strawberry!
Gothic style
.
When angry, the Loli Gods are very scary!
now i'm just going random.. 😄
enjoying a day off listening to music and doing art. so relaxing
and just.. wow
what ya think?
Vibe transfer is so fun to play with
then it is advisable to show the original source so that everyone is imbued with the functionality of the new feature for NovelAI Diffusion V3 called "Vibe Transfer"
Huh? 😭
With Vibe Transfer, you can upload an input image that will be used as a sort of inspiration for generations while enabled.
If you're asking which images were used for that then I'm not exactly sure where I found them but it was basically just guts from berserk and some dude from dark souls I forgot the name 💀
Can't remember where I found them but I deleted them cause I got bored 😭
oki-doki😉
Nevermind, found them on danbooru instantly just by searching for "full armor", I love the style of the dark souls one
nice
Rawr
this seems to be a weakness of adetailer
just like sometimes seeing hands in hair
i just kinda stopped using the hand model
all it does is make fucked up hands look fucked up in a different way
i've had it turn hands backwards with the knuckles on the palm
i basically only use person and face now
there is no alternative. even if you connect controlNet with HandRefiner
sometimes this combination works and sometimes it ruins it
if i don't get good hands in the initial generation i just don't use it
my workflow is this:
generate a batch of images without hires fix or adetailer enabled > pick the images i like and regenerate them with hires fix > send them to the img2img tab to apply adetailer, running adetailer 3 different times on the same image and picking the best result
little vampires ?
dark gothic lolly!
others are too monstrous. like a night horror...
stare
for what?
Nothing, just a staring image looked fancy 

scary?
she'll eat me like a delicious strawberry
she can! my workout!
Keep posting images they're amazing to look at🙏

thats a bad kitty!
looks complicated
i love coke much more than pepsi lol
but the blue Cola emblem.... 😁
@dry pecan by the way. I wanted to ask everything. Do you get groups of characters simply through prompt tokens?
There are a lot of misspellings, but at least the prompt is possible to some extent.
Trying out T-Ponynai 5.5 :o
Coot, also hai soooosiii 
very cool gen 
Thankeee, what you been up to lil soosi? Hows stuff been? :o
been doing great. Was on and off about SD, and mostly playing games

how about you?
Thats good! And das fair, SD3 came out but not much going on until someone makes a good model from it I think :o
And I've been kinda on and off with SD too, mostly when I do do something or testing stuff, mostly been playing too, finished the Elden Ring DLC yesterday :o
Yup, SD3 was okay imo. But i think a lot of finetuners are worried about the license. And civitai also removed SD3 models due to that as well, so not sure how widespread SD3 will be. And for elden ring DLC, i heard a lot of people complaining the difficulty boost
Aniother T-pony pic, face a lil funky :o
not a souls gamer myself, so not really sure about that
what's change from v5.1 to v5.5?
I mean thats true, hopefully something good comes from it though, maybe one day
hopefully 
Difficulty boost was rough, I'm a uhhh I guess I could say avid modder of the game, and they definitely fucked up on that xD
Style changed quite a bit, and I think better limb stuff :o
Yeee 
how bad was it? If the main game is like a 5 in difficulty, how's the dlc?
okay, i just downloaded it, gonna test it out

Any tips for trying to get into elden ring for the third time now? 
are you into souls game? 
I'd say the DLC is like a 10, its way too hard imo, its like not fun difficult, there are ways to beat it easily ofc, but I was often forced to change builds to beat some stuff :o
cos im not, so don't really feel the hype for elden ring
Not really, I suck too much and get annoyed quicker than others probably
But I still wanna try 
wow, okay, that's actually crazy different then 
Well, if you're struggling with the difficulty, play INT, so sorceries, and use spirit ashes :o
wait for black myth wukong 
I wouldnt say I struggled with the difficulty, just struggled with keeping my interest in it
Or jump attack strength builds, those are pretty strong too
If it gets too hard just go explore since it's open world
I mean, what Vil said, but also I guess if you wanna get into it, look for the why in a way, like at least for me, Elden Ring inspires me and gives me ideas whether its like for writing or art or just general concepts, besides the fun gameplay :o
I think im kinda on the same boat as Boto, as in I don't really find Elden Ring that interesting, or just souls game that interesting
And like for me when the DLC trailer dropped for example, I didn't care too much about the stuff in it, but I was more focused on the VFX and I was like "damn, those visual effects are pretty dope... I bet I can do better, nah I'd win" XD
Wukong seems dope, though as same as the DLC, I was more focused on the VFX and animations when watching the trailer for that too XD
Das a pretty big improovment :O
I guess why im interested in Wukong is because I'm asian, and probably can relate to that story more
since the tale of Sun Wukong is always on the tv, even when i was young

Fair, I mean I think the only game I've been like really big about and fell in love with is Fear & Hunger, dunno what that says about me
xD

I really wanna like and enjoy it 

XD
It is a really good game though
The dev is a madman, hes so good
And hes like a single guy I think, he outsources some stuff I think, but mostly does stuff himself
well, i've learned to not force myself on stuff i don't like. But i did try it for a few hours before quitting it
WELL
Maybe give it a few more hours, you might eventually like it, or if not, then it's basically obvious that you won't like it no matter what
I have 2 chars with 1 hour each, both stopped pretty much at the same spot in game 
The 2nd game, Fear & Hunger Termina, I like that more than the 1st, the 1st is good too but its a bit too difficult for my taste, though both are gamified CBT if you don't know what you're doing, but they're great games to play with a wiki at hand
What spot did you stop at? :o
Walked out of the gate, ran around for 10 mins or so and thats about it
Well, I think you still have yet to like properly give the game a shot then :o
Its like monster hunter for me, but I played that for like 10 hours before I felt like it was a job rather than a game
I will for sure
Is the ultra wide support nowadays?
Monster Hunter can be a bit of a grind, though for Elden Ring you can actually go relatively linearly and if you follow just the main path with some side exploration here and there you should be able to complete the game :o
Uhhh, good question
xD
Though theres probably a mod for it if it doesn't have it
But I feel like it does
Im currently playing Ghost of Tsushima and that game just sucked me in and doesnt let me go, thats what I am looking for in a game
Just looked it up and it seems like it supports ultrawide :o
I mean, exploration on your first run is great in ER
really really fun game. But no idea why, i started losing interest in the DLC 
On my first run I ever did of the game, I have 189 hours on that save
xD
Though I have 1808 hours on Steam but thats cause I mod it, and I probably have more cause I often launch it offline and I dunno if that counts on steam 
I have like 80 hours in AC Valhalla and that already feels like a drag
I mean for me the game didn't really feel like a drag cause its like a lot of fun stuff you can explore :o
While true, 80 hours is just too much for my working human butt
If I get to play like 10 hours a week im busy for 2 months with that
I mean, I think you can complete ER in like 40 hours if you really wanted to :o
O that is pretty significant :O
both's good imo 
Yeah I mean, you can definitely do that :o
Depending how linear you go with the game you can take very little
Yeee I think so too :o
But I like to explore 
It really sucks not having as much time as you used to be
I mean, I dunno if this works like this in the US, or wherever you live, but one of my mom's friends usually sometimes I think like once a month goes to donate blood to get a day or a few days off from work xD
Germany, I have 28 days a year paid time off, just doesnt feel enough
Especially since it takes me an hour to work and back another hour
With 1 hour of break
Youre outside for like 11 hours

Fair, that does suck :o

Then you get home and your GF wants attention, you eat something, tada maybe one hour left for yourself
Shes not into games, but im trying to get her there
She wants her own PC soon
ER probably isnt the best game to start your gaming career 
cozy games like stardew valley
how to genarate image ?
Imagine starting playing with your s/o and your first game is GTFO, that would be brutal XD
the easiest way is probably using a cloud service, but if you have a decent enough computer, then doing local generation might be better (and cheaper)
She wants to play sims cause her friends told her its good stuff lol
And she tried warframe cause she saw me playing it
But shits overwhelming for newbies
On that one both seem pretty similar :o
It do be, I've got uhhh
637 hours in warframe
Still got no clue what I'm doing
xD
sims 4 is fun, can agree with that, but unless you intend to spend like 1000+ dollars on all expansion/dlc then it's better to be 
Snek girl

Start with noita (definitely meme)
👍
Stoopid emote what is that
left one is wearing a bra, and have better shirt design
right shirt looked better to me
And is bra on the prompt?
nope it's not
thats good,always remove the bra
i meant the red lining thingy
Steven is here 
feel this nuts 
snake lady
where hires fix??? 🤔
Rookie numbers 
or you ⛵
this one is fun.. i like the colors
cute - crazy hair haha
hmm what is it with CivitaI archiving sd3 models. wtf?
guess it's not officially released yet. oi
Its because of the license
like boto said, it's cos of the license, they explained it in a post, im linking you the followup of the post, so you can read the first one if you want as well https://civitai.com/articles/5840/update-on-sd3-on-civitai
Was looking for that but you were faster
reads thru it.. oh boy .. are they insane? has their legal writing team lost their minds?
hmm looks like they are caving to some gov oversight or threatened to be sued to into non existance... shrugs. oh well
not surprised
Your reward
like the saying goes "people will F! up any good idea.. give it time"
or is it more few who misuse it or someone thinks they might or may already be.. ahh the paranoid, get ya hands of what I think is i mine or what should be mine, crowd
sits in corner and sighs
shrugs. oh well shrugs. wateva
i think it's a combination of a lot of things
it usually is. but anyhow back to art
some happy
i'm gonna ejoy all this while I can.
about time for me to start new training for new work project. sigh, gonna be a long day
but before I got , one last one
take this girl instead 
This doesnt look right 
dangit...
Here, have a Tifa.
Holy shit lmao
😛
Why is your book so big
it's a cold-swap external backup drive
C: is just OS/apps. D: is steam/games. E: is external cold swap. F: is my data drive (docs, photos, etc). G: is hot-mirror of F. R: is AI stuff. S: is VMs and Blender.

At least youre organized
yep, it's a system that works for me
would like to build a new computer once the new CPUs and the nVidia 5000s have launched
Im pretty sure ill stick to my 4090
So bored
have another Tifa... she helps boredom.
I do have a lot of Tifas...
But most of that used disk space is models, not images. 🙂
I use my program to help cull my images. 😄 :: shameless plug ::
what kind of model does that?
SD3 
Most of my gens lately are SD3
Its just too fun to use 
Food


The first thing that caught my eye was that something different
the colors are amazing on sd3, specially black 
but the hands 
Just don't gen hands I guess
👍

what model was that? i absolutelly love the colors
pretty sure is a pony
maybe confetti?
Should have the workflow in it, I dont know from my head
I think Autism Mix and some loras
let's try it out
lurks as he's in training reviewing info about client
i tried
Keep going im also digging with it 

@dull jackal 


squish
Protect your cheeks
cheeks squish ?
would
im innocent
Running a few grids to check some model and Lora combinations. I'll circle back
lucky gen
some more Tifas for ya
was pairing a few different models and loras
👍

@dull jackal @wooden coral more gifts
Which one of you can get it first 


best miku ever
indeed

Huh, so many scams today
probably people clicking on them, or just the person scamming, activating the script across multiple compromised accounts
Those gullible folks
best one 
part of prompt: Triple super imposed exposure, overlapping plasma ghost layers
yeah just throwing in random off the wall things
is there a website out there (not an extension) that will just pull a set of random tags?
there are extensions for that yes
found this.. created my own generator. 😛 https://gist.githubusercontent.com/bem13/0bc5091819f0594c53f0d96972c8b6ff/raw/b0aacd5ea4634ed4a9f320d344cc1fe81a60db5a/danbooru_tags_post_count.csv
gah that was a long boring day
but pretty much got paid to do nothing cause my creds aren't setup yet.. oi vey
but i'm not surprised.
anyhow... I need a nap.. thud
What pony variant is good for making more western styled comics?
probably a lora for that
Base pony can already do that no
Well I added toon, cartoon to the end of my prompt but outputs look like trash. Maybe I need to add more words
Try a specific artist or something, toon/cartoon is not a style
Something like Pixar usually works because that’s a pretty solid style
The source tags are generally pretty vague anyway
pony furry and anime works pretty well
But using cartoon is about as specific as using anime, it’s not a singular style
‘Retro comic’ generally works because it’s more specific style
But also, using toon/cartoon to try and get a western comic look doesn’t make sense. Prompt for what you want.
Working on a Solo TTRPG Campaign set within a Scifi/Cyberpunk like world, found one of my protagonists
That's really cute, what model is that?
autismmix_confetti
I'll have to take a look at it, thanks
What kind of dipfake is used for anime characters?
what letters did he get wrong!? 🤔
What deepfake techniques are used with anime characters to preserve their appearance?
i don't think i understand what you meant. People usually train lora for anime characters, or just if it's a well known character, then they are probably already trained on the model
I've been drawing the character since the beginning
I don't want to train lora from 1 picture
ahh, you want to generate your own original character. Well, the only way is to probably draw a few more picture of your character, and then train a lora of it. The other way is to try your best to describe your characters as best as you can and generate a few images of your "character", if it's close enough, then train them
just go for fox girls 
but there is another question. How can I find out if the specified “embeddings” are used during generation?
isnt it on the neg prompt ?
right. but for example how can I find out if it works or not?
Embedding is not necessary only on negative prompt. But you should be able to see it in your metadata under TI hashes or something
One easy way to get if the embedding is affecting the generation is to do a plot xy, same seed, one with the embedding, one without
yeah right i remeber now
To check*
@wooden coral
Is there a possibility that during generation embeddings are not loaded in some cases? for example, one generation provides information in the metadata that embeddings were used, while another generation does not contain such information..
If it’s not loaded, then you might have to look into your cmd, maybe it’s not compatible or something?
there is nothing about this
Is the embedding named properly? And are you using the correct embedding for the correct type of model? Like SDXL embedding can only be used for SDXL models
I only use versions for 1.5
Is it only that embedding not working?
Now I noticed that the methods sometimes contain and sometimes do not have this information.
information about Lora is always present there. and information about embeddings that are indicated in the positive prompt.
Was there a p i n g?
Loving it

If you're using a1111 (and possibly forge) there was a bug that made it so some information didn't make it into the metadata even if the embeddings were still working. I don't know if they fixed the bug or not, or if that's the problem though
but a big thank you for this! Indeed, using Forge, one gets the impression that this information either ends up in the metadata or not.
oh nioce
but what she is thinking about is kinda dark
I have that look on my face sometimes 🤣
That's the you're about to die look 😆
aww
Those are very vibrant and intricate. I like her tattoos, horns, hairstyle, etc
They all got the killer look somehow lol
booba
O! Red Hat!

And me behind
Is inpainting with controlnet better if the image is bigger to begin with?
All else being equal
I could be wrong, but doesn't this initially depend on the "Inpait area" mode?
Or do you mean the Inpaint preprocessor?
I don't know. I make the renders in blender but I can choose how large I want the render to be. Then I inpaint each part individually to make it look 2d. So I guessed that the bigger the image, the better
In any case, if the picture is larger it contains more information about the details. unless of course it is detailed in the render.
cube - on render in any size just cube
Zan, does their Grease Pencil work so poorly in Blender?
I don't use it really but I do use the function that adds lines to the edges of objects to make it look more 2d. It runs pretty slowly but it's a big help
Apparently this is some kind of addon?
No it's included but there is an add on which probably does it better. Let me get you the method I use
Today I will show you how you can make Line art in Blender
Join my Discord: https://discord.gg/CkxUfeFXha
for business: theanimationdudee@gmail.com
through a modifier. It's clear.
everything is brilliant and simple
Yes. I use kitbash 3d stuff so when you add lineart to buildings like that, it adds a huge amount of lag
kitbash 3d is not familiar to me. so I'm sure you figured it out. and sadly this method doesn't work well
Yeah it's ok if you want to do it to 1 building but more is tough. I haven't tried a different add on, maybe it would be better
cute
she looks funny with those emotions on her face.
her hair is like mine when waking up

I get just as confused
I believe the only difference to do it with controlnet is that it can take the whole image into consideration without affecting the full image after
and you can do other controlnet stuff with
I have tried controlnet inpainting. I thought either with or without controlnet, it can take the entire image into consideration?
By the way, why exactly the preprocessor Inpaint? why not LineArt?
or am I misunderstanding?
I forgot exactly why you would use controlnet instead of normal inpaint, I think its just because it takes advantage of the generation options, you're kinda of generating the image but making it a fixed thing like inpaint
and it can resize the specific part you inpainted on a higher resolution image back, so you don't have to generate the full resolution
I think thats something you can't normally
I think lineart is something like depth, you use it to generate something according to an input, inpaint controlnet does inpainting
I'm completely confused about what workflow we are talking about....
have an apple confused funya
apple back
I think preprocessor is needed to convert the image to something useable for controlnet?
With controlnet you can keep the same composition
right. you can, based on your render, get a sketch, so to speak, which will then be the basis of the generated image using a prompt
Segment Anything+ Grounding Dino is great to use as well; you type in what you want and it will auto mask what it recognizes in what you prompted in the image
but I couldn’t figure this out. I couldn’t understand how it works and how it should work... 🥴
You mean how to install it or how to use it?
how to use. and what exactly the workflow itself should be
You have to use it in img2img. And you have to check the box to use Grounding Dino. right or left click to put red or black dots on or around what is needed to be recognized or not recognized (to make it easier for it to know what to mask). And then type in the prompt (ie if you want to change hair color then you type in hair) and then click the button. Then it should give you mask options. Select the mask you want. Then you can prompt and inpaint it at the top like regular img2img
That's what I tried to do. but I remember it worked rather slowly and, moreover, the result (apparently due to the mask) was very unsatisfactory.
Do you want to send the image you're trying to me? I can give it a shot and let you know how it worked out. Haven't used it in a while since I was experimenting with other stuff but I can try when I get home
I want to find the image with which I tried this extension and the result itself. If I find it I'll show you.
K
you didn't play with luma?
not much interest in playing wth it myself 


Mile? :O
painting style
Here's an apple for you!
I'll put it on my pizza
not enough, make her bangs stick out to the other side of the image

this generation takes like forever 
he is hard and cold. and there are sharp parts too
I'm brave, I can do it
whew that was a long day
1st lcm
im2img changed to different sampler and some extra style change , came out good
it's like it eroded
sunset or sunrise
super cool oni horns
resistance against fire: -50
this one looks a lot like midjourney
is midjourney?
All SD3 
Mikoo :O
great stuff, can someone explain me the difference between midjourney and stable diffussion?
midjourney is online
sd is offline
very simplified explanation
midjourney makes cool stuff, stable diffusion makes lewd stuff
midjourney is a specialized , supposedly highly trained and has a certain look to it. and it's usually paid for more features and is only online.
and very censored against anything remotely even slightly lewd
to add on, Midjourney is close source, and SD is open source
Casual Hu Tao sama 👻🔥
i hate how it always tries to make the characters leaning on something
i also can't figure out how to make it not do this with skirts ^
it's like it can't decide if it's a skirt or shorts
pleated is the kind of skirt i'm going for
i mean the fold inwards outlining the crotch area
This is very out of line with the theme of the channel. I think it’s worth looking for another place for such work.
holy crap venom booba
The tounge is a little fucky
and the boots don't look right.
This is better
O! red hat again! 🙂 now the cat will say that he is in the background! 😂
Didnt say perfect 
I honestly don’t even remember what kind of “paws” he has 😁
lacking the face 
Always on the edge 
Not creepy at all
I have something even worse 
The fuck is that 
well model horny
Ill remove it to be safe 
I think it was the tounge
it was an expansion!
something wholesome to clean your eyes
Cant do sorry
My new fav thing to do, venom cross overs no one needed
still
tho

Really was a tongue btw
Absolutely no clue whats wrong with this 

Cant get more venom in there 
All that's left is the Forbes cover!

Perfectly timed post from @wooden coral 🤣
Let's not talk about hidden context 🤣

Oh, these genetic experiments!
A gift that keeps giving 
I won't sleep well now...
No worries i can make it worse
🫣
and now after another 20 years!
-
saw really fire pic
-
done by a public model without shenanigans just txt2img
-
no loras no nothing
-
and of course the author hid the prompt
😃😃😃😃😃
It's all cabbage!
but they look the same as they did 20 years ago!
wow cute
I have a question.
Adeteiler - parameter "Mask erosion (-) / dilation (+)"
in which direction and how much to “twist” the values?
I mostly leave it at -32 or -48.
sometimes I notice that the result before Adetailer’s step looks better than after his work. with the "Inpaint denoising strength" parameter having a value of no more than 0.34
i have it at a low 4... 
As I understand it, this controls some kind of applied mask?
what strong “arguments” she has! 😉
from what i understand, it slightly merges the border where inpainting happens and the part that it dont
still
ok was the sampler .

those. Is it like blurring a mask?
I got mistaken for a girl
yup, kinda bluring border of the mask, so it's not just one square
no problem 
Ready to bite all the girls 
So the positive value is the sharpness of the border?
bad cat!
i guess im wrong
I saw this article, but alas, my knowledge of mathematics is the same as my ability to dance in ballet...
ok then
1st: normal
2nd: postive value
3rd: negative value
so basically increase/decrease "the amount of pixels" into or out of the mask
better
math
Let's run funya, smol likes math 
Wait no running, I'm staying

Now make it 2D
those. if "on the fingers" if I want the mask to be applied to a larger area, I must specify a larger negative value.
Colors
But i love how good the generated images you do sleepti keep that up🙏🏻
most of the time, using adetailer for hands isn't gonna help much
yeah, I know. What about Persons?
It's an image on screen, can't be more 2d than that

i mean its really hot the character is hot and all... Yeah hot

That adetailer will do a full person in the image
It's good when you're trying to do characters far away
A full body far away from the "camera"
yeah! only in this case? does this not partially correct the proportions if the character is close?
You can use that way but I don't think it will help much
@lyric zealot 














