#π₯ο½anime
1 messages Β· Page 273 of 1
You picking these up on something like fiverr?
Then again most vn games are static images anyways
the ones that are considered decent always seem to have some animation involved
Oh no i just know people who make them/have a need for it. Im in a lot of discords and the pay is crap
Mostly because i dont really charge anything unless its not interesting for me to make
@gleaming flame - we need some details on your prompt-fu.
I like to generate ultrawide stuff for desktop wallpapers.
3840x1600 specifically
I'm obsessed with different styles, and always check what can be achieved with each model... and try to memorize it... and img2img and inpaint a lot between models because one can get me a pose, second one big eyes I need (for example), but 3rd model gives the final touches to the style of image... and I almost always make some image postprocessing for colors in editing software... also experiment with terms from classic art...
Are you more of a natural language prompter, or tag prompter?
almost only nouns and adjectives
also I often experiment with order of words in a prompt... especially those regarding style description
oh, and there is one more thing... when I mention artist styles in prompt... sometimes I give a lot of creative freedom to model... and only focus on the mood and style
would be interesting in prompting/workflow on those
3rd one was made with shiratakimixXL_v20, the other 3 with leo anime xl
more good ones...
eh not necessarily... just finding neat ones and curious about some aspects of them
like that last one has an interesting cartoon look
they are rarther simple in terms of composition, its all about the styles
sometimes the framing is neat -- different from the 90% 'portrait' images we typically see
find classic disney illustrators, add a suggestion about composition from a classic photographer... and you get a better composition... usually
if a model recognizes mentioned artists
I think the secret is carrots.
I'm just not familiar enough with various artists or composition styles I guess. π
not sure you can name specific artists in dall3 anyway
Naw I mean in other models.
but sd models usually recognise some of them
and that also takes a lot of research... xD
also recognition of art styles and terms in general is helpful
the noobai models seem to be heavily artist influenced
I like to do the sketchy styles sometimes. (examples)
these are sweet
but a lot of mine end up with a very similar look... you can tell this is a pony model..
the other just used a lora
and i hate using loras π
but that lora got me to that sketchy style...
maybe try something besides pony? I could suggest a few xl models... at least the ones I use
I noticed that
if you want smooth portraits, it does well...
yea, they are good for specific characters
but can't do something like this...
pretty much all anime models do that, you can force it more with like detail loras
yeah... that's simple stuff π
I need more Vertlain prompt-fu
every so often i'll hit a win...
those were that lightningbloom model that someone here made... who was that?
oh it was torqx
4 step xl model
man... I worked my 3060 hard on that model LOL
if you like pony but want more detail you can try waiANINSFWPONYXL, it behaves almost exactly like autismmix_confetti, but has more detials in the clothing and backgrounds without prompting for it
I'm trying to get away from the pony looks and do some stuff that's a bit more interesting, methinks.
not dall3, though, that's cheating. π
I like my unfinished doodly styles...
it's like frank frazetta with a touch of samdoesarts
actually it's a mix of few styles...

which brings me to one of my "tricks": if I suggest artist styles, I usually mention 4-5 of them...
some artist styles combos are amazing
maybe I need some new xl models to fart around with

quack
magic book
magic and stuff flowing around is kinda of random with anime models lmao
I noticed it depends on the type of magic you specify, i had great results with hydromancy, ||succubus magic||, and stormmagic in the way i had intended
gonna type book magic
grimoire you mean 
grimoire are magic books
Would also recommend magic runes/runic circles
Is thhat succubus magic in a book?
Sounds more like a summoning book then
ye just typed it in
Try geomancy grimoire, aged pages, brown cover, gold binding, glowing letters. Idk
Just book magic sounds silly and lacks for a intresting prompt
going to add this to a promtp and see what happens
first result
Hmm looks like electric to me
Or earth element energy
Try adding floating rocks, bricks or any other earth interaction like earth spikes
as you wish... will add more purple
Purple?
yeah.. more purpley magic vs lightning π
Geomancy is earth magic
eh... I don't love it...
high heels with armor and stuff
naturally
lol
@vital raptor
fire cat
witch outfit, standing, (glowing, swirls, colorful electricity, magic, energy, wisps:1.4), open book, fantasy, floating, storm,
i just dumped a bunch of tags

was wondering why my prompts were failing, swarm after pc reboot lost my settings from last time lmao
mine came out weird-
swarm doesn't save prompting settings unless you make a default profile
it useually remembers my last run, prompt and sampler
it kept my loras in the selected field but not the sampling
by default swarm should
least mine is without me needing to do anything
pick an image from history and reuse params
oh well its a quick fix anyways. just click on last image > reuse parameters > change prompt
witch prompt was fun to get working, it spazzes out with some tags reducing the image quality by a lot so you really gotta pull some tricks here n there
and yes the smile is intentional
orb = pondererd
using the orb pondering as a init you get some nice results
last witch for me for now
furry 
Nah, pokemons 
mind is a garden...
seconds before tragedy shots?
Noooice
Smol look
π
yue has voice now 
Grabby 
I used these to get it. "vaporwave, neon light, neon line art,"
@rancid remnant face if the loras turn out bad
Even worse
All my lora training
Just
Reset?
Went from 11 out of 20 epochs, 6/20, and 3/20 to all of them being 0 out of 20

Bet, thanks
Blonde hair and white hair women 
White hair and black hair women 
Black hair

Through unknown magical means she has managed to turn her beloved dragon into a marketable plush unfortunately for potential buyers she has decided to keep him for herself


I want to see the rest of the products
π
Sorry one time limited edition no more in stock 
But Booba

gotta catch them all?
clearer toy look for the dragons
so cute
To use again noobXL, it says I should not use _ but use \ but it is tiresome to copy paste the backslash because it is not in the keyboard. Is there a quick and easy way to do it?
\ alt + 92
92, like I press alt and 9 and 2 or...?
press alt and the numbers 9>2
things appear 
β like many symbols
needs to be the number pad
Check out this extension:
https://github.com/DominikDoom/a1111-sd-webui-tagcomplete
Is there one for comfyui? I have installed the pythongoss one but it is with _
If something exists for A1111 then comfy must have an alternative of some sort
I see I'll check
still using ashii at this age. i'm so proud of you 
What is ashii

ashii code is it's all that nosense you do with alt+num pad
ops, it was written ascii
it was used a lot in 2000 while coding
i played a lot of tricks on irc when i was young lol
very sweet lol
Indeed


evol 
Imagine a YouTube channel
A_R_P_I_7_EDITZ #π₯
ummm... basic small image was generated with dreamArtFusion... then upscaled and "animefied" with yiffymix, and finally with those big eyes in place turned a bit more realistic again with dreamArtFusion... With a lot of experiments along the way π
no loras
basically realistic image > anime > a bit realistic again at higher resolution
still not exactly the style I'm searching for, but looks nice I guess
delicate for a change
flowa
Looks like... DBZ inspired or something ?
hair and earring reminds of it
Nahida
I had to google "Green girl genshin impact" to remember her name lmao
your image reminded me of her
(Still wish they didn't kill off her older form) 
Is that her older form ?
I am a genshin enjoyer
I have no clue about the game 
nooo, I meant my image, is it her older form ?
I just typed "nahida, from genshin impact"
Nahida is a reincarntion of Rukadevata
I wonder if Illustrious knows her older form, might try to do a gen of her at some point
try using Rukadevata, instead of nahida
we have a delay, images got too big for this server

May I dm you lmao
yeah

Nahida

It's just a simple prompt, I used image to image
Something like full body white background wind magic...
π
Is it a base style ?
Ok I remember now
slightly aged up

i got a bunch of smaller ones but i didnt feel like they were worth upscaling/face segmentation
You guys upscale 
especially when some of them say genshin impact in the reflection
yeah, only by 1.5 tough to also refine the image a lil
3rd is nice
prompt is slightly modified but still same seed n all (so its mostly the same) (also used a slight 0.1 variation seed strength)
prompt positive:
HDR, high contrast, masterpiece, best quality, amazing quality, very aesthetic, absurdres, newest, scenery, (volumetric lighting), photorealistic, detailed, nahida_\(genshin_impact\):0.8, simple white_background, wind magic, green magic, wind, looking_away, green eyes, lips, pointy_ears, sandals, grown_up, medium_breasts, cowboy_shot, <segment:face> HDR, high contrast, masterpiece, best quality, amazing quality, very aesthetic, absurdres, newest, scenery, (volumetric lighting), photorealistic, detailed, nahida_\(genshin_impact\)```
prompt negative has some things i wont mention in here lmao
tried the grown up version
but doesnt work
the tag nahida is too strong, all it does it stretch her legs and make her slightly less "loli" what i wasnt going for
like this one but alas
last 2 i do of her (first runs with upscale)
Can't do much more than make her bigger
But what 
ponytail on the wrong side
Just type ( child, loli :1.5) 
nah this part is the problem
pony tail and that other green part is are on oppesite sides
problem would persist
nice try fbi
ah yeah but my main negative prompt has
Oh you want her to be big

this model i use has some tendacy to generate lewds so i just always blacklist stuff like that
just to prevent any accidents

Evil 
i fogor was it here yesterday where the pondering the orb trend was happening
Just go for NSFW: 1.5 negative
π
accidents happen
Lol
anyways heres two witches pondering an orb
interesting model this one... took me a moment to get my prompts right, but effects are quite nice I think, maybe a bit generic(?)
Famous hair over eye
I mean in between
And hair intakes
That's a funny tag but a thing
the others look better but theres a thing in them i cant put my fingers on
maybe its the skin idk
No, just looks like rubber without all that much detail
guess it depends on the prompt
but my biggest issues are the eyes and not the skin to get right
I like when a merge has a good style
oh hey its uhh
Need to try more of them
cappy
MITA
crazy yandere girl
no cappy isnt yandere
yeah mita
i havent played
ive just seen stuff about it here and there
Cyberpunk?
Definitely not cyberpunk
is it a vr game?
Lmao
no lol
the second girl he posted is from cyberpunk
horror
lucy i think
Lucy is from edgerunners anime/series
yeah based off the cyberpunk game
Cappy is from a unassuming horror game
just the model doesnt get the tear right
and i am too lazy to edit the tear
Funny her face looks more 2d than the rest 
stuff that isnt eyeshadow is just balls
It's less trained
okay but at least thats easily editable
i could do it rn for you ong
still cute tho 
i got a pretty good built in editor in swarm
i just throw open photoshop ngl
i can even add in other characters this way lol
i trust myself more than ai to do that stuff, mainly because i know what im doing in photoshop
swarm confuses me π
takes slight getting used to but once you do its soooo good
Now I see it too π¦
forge and a1111 feel outdated to me ngl
i liked it but i found myself preferring forge
easier to udnerstand, also they layout is way easier on my eyes
holy shit im blind and swarm is small and all of its settings are super packed together in a wall of words
cant see shit
slight inpaint took like 10s (model change)
Sorry, it's just what's been bothering me the most about illustrious

i got more but cant post them
still, percentage of usable images once prompts are set right is really high, so I might finalize them in another model
in a sec but i will show you the power of swarm first
just yeet in a real image over a already generated image
god himself cannot stop me from using forge
granted i the placement doesnt make sense and the creativity is really high it will just make it into something else
someone already pitched swarm to me
i tried it
like i said, it wasnt bad, i liked it
i just like forge more

fair
also i can dm you the other ones that arent suitable for here
if you got that dm scanning disabled
Swarm Supremacy
dont drag me into yalls UI war for i care not about supremacy but rather comfort and ease 
one day I might try another UI... but for almost 2 years I'm invoke user and it works the way it should and that's enough for me
real
except the 'works the way its supposed to' i cant get anything to work right 
wdym the previous 2 looked american cartoony
?? what did
soooo swarm
.

was that not the intention?
the girl has a certain skin shine that doesnt belong on a certain spot
i fucking HOPE they look like americna cartoons, i spent weeks making a dataset 
but the other one could totally be from a cartoon
where??
if u want american cartoon style ill just send you my family guy lora lmao
o wait nvm its a light. i thought it was one of those shiny skin pimples lol
i wanted specifically the ultimate spiderman style
loks good
so i made it
nice
tyty
ehehehe lois im an ai generated image
also i prompted it in my enviorment
poor petah
o newports were still iin the prompt box
LMAO
still looks cursed
very
but somewhat more petah
the feet are scaring me
my loras are moe style oriented

like it prolly doesnt know peter or lois well enough to gen them but i got the aret stryle down
oh yeah just petter griffin in te prompt this time didnt generate his pants, yeah completing the prompt better
i train all of my loras to include the style and all my favorite characters
so it can do the style, characters, and characters in other styles
cos like i never go for specific ppl when i gen
how do you prompt for peter griffins chin
lmfaooooo

christ
aint no way
chin needs to be shiny,
give him a beer
lmao
man do NOT use peter griffin, imminent hug and biting own lip + pov
you finally meet after a long time
enough petah
π₯²
Grabby ears 
You're fired
Can't be fired if there isn't place to get fired at


if you're going to use Forge, you should at least switch to the Panchovix' fork, reForge as it doesn't contain sketchy obfuscated code like the main Forge repo does https://github.com/Panchovix/stable-diffusion-webui-reForge
tbh i use forge, reforge, and a1111-
i use stability matrix to manage them all
a1111 is shit BUT it has the least amount of issues with region prompters
forge is what i use most often
reforge im trying out so i might replace forge with it soon
the only regional prompting i find to be worth using is invokeai's implementation. it's way better than any of the other options, but vpred models are broken for the time being in their UI so it's unusable for me until they merge the PR that fixes that.
re make
kinda looks like a person wearing a fursuit/mascot costume now. the artstyle is pretty cool though
I fully switched to SwarmUI, myself.
full swarm and comfy, started making my own workflows
and im back on linux cos fucking windows is ass
Penguin is fired by fox 

Lmao
Good red 
BEST RED
animatediff?
good stufff
@gloomy scroll @green temple she bite 
closed source π

Gotta need that prompt for the last one chief
sec
if you want the style
its from a lora i just made
im about to post it
but this image should have the metadata
Style and prompt, love the astethics, but ill dl the style from civit if you post it there lol
No rush, im ony way home first
aight bet
civit
this one came out nice
i train all my stuff on civit, laptops too weak otherwise id use kohya 
thanks friend

Ill make some examples too once im home
.7-.8 if you dont mind some weird shit but want more of the 'sketch' style
Just got glass fibre today, waiting to see how it turns out
Gonna delete my unused sdxl models to make space for more illustrious
real
i did that with all of my sd1.5 models and loras
like 400 gigs of stuff gone
poof
still have no room 
I still kept my 1.5 because i had like 200 style ones
nice - more leo?
i have around like 3k
had*
now i just have like 30 loras
two models
ill download more at some point
Nice, i had ones that werent on civit but just thru the special boards one
im still new to illustrious ngl
Shady megaupload
real
I prefer illustrious over pony ngl
i hated finding shady but good loras but never being able to find them again
oh 100% same
But pony just has betteroras for now
i just make whatever loras i cant find for illustrious
And a working controlnet open pose
base images were from Leo... it's great for making enviroment more lifelike than just a background for a portrait
i prompt and pray
i use extensions for it but theyre shit and almost never work
Same but some more imaginative poses really need it
so i keep genning until one DOES
lot of stuff needs a sort of openpose/regional prompter ngl
wish whatever NAI was using to prompt multiple people would come to open source stuff-
yeah I like it... would be a good model to create a calendar I think
Swarm has regional prompting 
Yes
can i use loras with it
I mean yeah you should be able to
i suppose i can reinstall swarm
granted i still have to mess with it more
i like the weird eyes
i prompted for them
it can do normal eyes
i just think weird eyes are way cooler 
Minecraft with shaders 

Did I mention that I'm rather impressed with this model? One take is usually enough xD So far 90% of images are usable... hard to choose best ones
then again... It's just anime girls and some background...
Need to put more (ultra detailed, many sharpening, 4k, 8k, 16k, 32k, beautiful)
(probably not gonna help)
Ye anime models don't think too much about the images
nah, I'm ok with the results, untill it tries to make 3d with flat face xD
runs away
tbhkiro is the trigger word?
ye for the style
gotta mess with the style a lil on my end
it currently does ONE character
but i think thats just me using a different checkpoint

my loras work fine on a lot of different checkpoints
i test that before i post them-
what model are you using
noobaiXLNAIXL as a base test lol

(she's not insane)
mol is very cutesy
thats what they all say 
hmmm can i see the full prompt + negatives
hmm i did move the lora from the text box into the lora selection
ill try to force it into the prompt
what are you using to generate?
swarm lol
hmm moving the lora into the textbox has some difference
def
think its close enough
add (sketch)
How do I generate images on here?
here? hmm we have some oddballs throwing in random prompts in general chat till someone runs it locally
but generally you either use a website or on your own computer
this would have been cool if the knife wasnt floating-
you dont gen images HERE
sketch added the missing deets ish
but yknow like he said, on a site or your computer
try making the weight higher if needed maybe?
oh yeah the regional prompting thing we talked about
oh shit
pretty nice UI and if you click add it just adds the nessacary stuff into your prompt box
i see BOXES
i love the ones where you just move boxes
its so much easier-
should be the standard

tough since i selected impaint i should use it when impainting not when generating
But it says on the website that the bot can be accessed here. In the server.
not my server broski i wouldnt know

so give me an example of it working?
something simple maybe, just like two girls standing next to each other or smth
adding non impaint it does
so its generating in a region instead of impaing a existing
hmm need to check that

using my own prompt for this one tho
using my weird prompt to make more weird girls
but i went back to the spiderman style for now
i want to see how it handles it

it gives random characters some clothes similar to the characters the style lora can creat but i think its still fine it doesnt happen too often unless you somehow use a specific string of prompts thats similar enough to a character by chance-
oh sweet
trying two girls from different series, gotta mess with the strenght slider tho
0.5 was too weak mb mb
did it always work like that in your ver of regional prompt?
2 characters?
ah i forgot to add a <break> maybe thats why it was messing up
dont think its made for two characters at once
egional prompts cannot currently use loras or other global feature changes. This is likely to change in the future.
but forcing there a singular character stands is working as intended:
bwahaha no way it worked, i used impainting but it didnt want to recognise that someone is stansding there
so
just drew her on there
upto rngsus to get the proper character
took a few ish tries but it works decently enough to call it a sucsess
SD is just not made to handle 2+ characters properly
but tbh if i just use "name and name" it works 1000x better or i am just using it wrong
Sucy manbavaran
Hug 
Ah
That's fine
I'll continue to force it to work

It CAN work with loras you just need to get lucky
Anything can work if you retry it enough!
I keep trying to fly off my shed, but I'm running out of legs to break.

but if you keep character details to the minimum and just use their names you get pretty consistent results
My prompts
but sometimes it also mixes the characters and you get something cursed
cant say i hate it in this case
Something
||Would||
not exactly the same style but i think i got pretty close now
also my silly head accidentally upscaling by 2 and keeping the setting on after impainting
so i did it again
1792 x 2304 
the normal 2x upscaler
lets see what abomination i can create
slightly different eyebrows
ill tag it to your civit model page π
really need to remember to turn off the upscaling after inpainting happend again
I have a friend who in comfy has a little setup thing to do warcrimes is the name
it upscales to like 5k resolution lmao
ooo @dusky linden I like those little zombie cat people or whatever
the cat lmao
witch ones lol
its ahri and sucy
comfy can do some funny stuff
poor ahri
this
and this even though it mixed prompts
ah that was just me messing about with 4 word prompts
more words = merge two seperate characters
but its hella random
tough i do notice that a named male character and a named female character have greater sucess then two female
the more distinct they are the better
for obvious reasons i cannot post examples of this
\
Hank Hill
hmm the characters panty & stocking (real names mods pls) works pretty well if combined
surely it works
idk if kissing is allowed but i got examples of that
peter griffin
ugly 
pics in dms
a mix of peter griffin an guts
nu
These are the philosophical questions that plague us, today.
malefunction
@gleaming flame - you duplicate that style outside of Leo online, yet?
which style?
manga authors are insane. an artist just explained in a manga how to make an home made bomb....
you know.. that one you do
I dunno what to call it. π
you had some earlier
I make quite a lot of different images in leo...

but no.
Surely good content for a manga
i feel some info should be restricted
this feels like 20 prompts combined into one
this is a good one
Its just 1girl, tifa lockhart, cat girl, furry, dark room, sitting, one leg over the other, cat tail, angle, cute, from above, cowboy shot, looking at viewer
lmao
and these from earlier today
I feel like I could finally make a calendar for the year and have it printed with that style.
I'm not sure how I feel about Cat Tifa.
@steady grail 
you could use leo for a month? π
creature, pink slime, river, solo, looking at viewer, cat ears, transparent body, fluid body, black eyes, white eyes
i have in mind a new merge, but to that i need a bigger workflow and a lot of testing 
While I like the idea of these online services, some of these fly-by-the-night ones I don't quite trust yet. Seems ripe for problems with protecting consumer payment data, for instance.
Technology
yup, things are becoming smarter
@keen hollow 
i just hope they lower requirements even more. i want ai implemented in games
as cool as ai is i think that would be a horrible idea
Theres some shooter game with AI bots
Don't think its out yet
imagine to play in a game populated by ai
generative ai that is
that nose haunts me
Its not that cursed 
can fix that
Like
If you do more focus face
It does the gradient of siamese better
I posted another somewhere but too much for this server with the face well done 
I inpainted it after
portrait
tbh any extreme close up gets you insane detail
@keen hollow how cursed x3
that's not even Tifa anymore. π
gotta do close up:2.5
What if I giga close up with pink slime creature
i assure you its fine
ugly 
stinky
evil flying @steady grail
since you liked te odd pupil ones, posed some to the civit page (also crazy how some are STILL flagged for review. like ?)
i like to gen fantasy monters
maybe i should focus the next merge on that
it looks funny
OOOOOOO , that one came out great
i LOVE the eyes
im working on one rn
i got a lora that has about 4.6k triggers for monster girls


