#🍥|anime
1 messages · Page 272 of 1
on my system it was like 3-4 hours of waiting for everything to be indexed. you can pick individual files to index to start with though so that could be done overnight. also, if you aren't a hoarder and/or don't store shit on a hdd it would be much faster
i found out how to generate the images but
i figured you recommended it because it could help me with what i was needing

so how do i do...that
i just loaded every folder i had
at once

yeah, give me a minute to load it so i can take some screenshots
bet
but its lagging a bit
but i bet its because its loading all the stuff from the folders
is there a live preview option on this 
yeah, it reads the metadata on all of the files so that it knows what they are and which model architecture
shit. gonna be a bit for me. i fucked up the venv
well lets hope quick resume for downloads exists-
venv files are crazy
i remember using a1111
and having to delete that shit all the time
it was the only thing that consistently fixed my stuff when it broke
i use stability matrix now though
easy install
looks just like a1111
really fast too
like
daammmnnn
i thought i was going to be able to switch completely over to python312, and removed 310 but forgot that doing so would break everything that was using 310 until i rebuilt venv
should be quick once everything has settled. it may still be indexing files
two things
- is a live preview an option and 2. why does it keep setting up a queue of 100 images

just realized your pfp is padoru omg
i made some padoru pfps
i'm pretty sure i remember there being a live preview. 100 images sounds like batch size got changed. it's been a little while since i've used it so i don't remember off the top of my head where those things are
nice!
alright ill see if i cant find it
i got mine from danbooru
i just find images and edit them into pfps when bored sometimes
my pfp is one i also got bored and made
this is the one i used https://safebooru.donmai.us/posts/3334881?q=padoru_(meme)+kirisame_marisa+
yeah idk what this is doing
or how to change it
or if i need to
i found a setting that looked like live preview but it was already on so

yeah 
i'm just waiting for the installer to finish updating and fixing my broken install then i can check it out with you
bet
cuz it doesnt seem bad so far
i just
cant find out why its reading 100 prompts from one prompt

oh?
it fixed itself?
i changed SOMETHING in the prompt
i forget what
it's definitely different than a1111/forge/reforge. took me a little bit to figure things out even with being very comfortable using comfy, swarmui, a1111, forge, and reforge
i just removed some stuff i didnt want there anymore
and now its like oh okay 1 image it is

i think it was tweaking for a sec
<3
gonna turn this into a pfp let me cook
i haven't decided what i'll make my discord pfp yet.
ok, so this is the basic setup for regional prompting. i'm gonna need to do some more digging to see if i can force invoke to use vpred models correctly because that didn't seem to automatically work for this model.

wheres the hires fix option, and adetailer 
das'sa cutie.
hires fix is going to be done on the upscaling tab, so you'll gen your image then send it to upscaling. adetailer is more complicated. you'd have to build a custom workflow for it in the workflows tab, basically like you would in comfy. rather than that. I would just recommend manually inpainting the faces when it's actually necessary
it wont let me upscale-
ehhh id rather go through the effort of making a workflow if thats how i need to install adetailer
for upscaling, make sure to drop the image into the image box on the left
i think this is my issue
oh, yeah if you're trying to use controlnet and it doesn't have one, that would cause errors
ah does it just use the same controlnet models as what im already using
the ones i used in controlnet in forge
yeah it should. you'll just need to make sure it got the metadata for them correct. go to the models tab and make sure that they're set up right in there
yep, it's looking for a controlnet tile model for sdxl which you don't seem to have
rip. vpred seems to just not work for sdxl models in invoke. last thing for me to try is converting the model to diffusers first
it's a different type of model. until noobai there weren't really any noteworthy sdxl vpred models. without proper support you just get noise from vpred models like this
oh tragic
whats vpred
a different type of model yes but like
how is it different
im trying to pass time waiting on this download my internet is ass, my bad g
okay its installed
i can upscale now
basically normal models use epsilon prediction which guesses the image based on the noise. vpred, or v-prediction does the opposite, it guesses noise from an image. the end result that users see is that vpred models support a broader color spectrum
now i just need to install adetailer
sounds oddly specific

how do you put extensions on invoke
tldr: epsilon prediction produces images that trend towards very average brightness, with muted colors. vpred +zsnr addresses this shortcoming to produce images that are much closer to the intended colors.
not having hires fix is kinda sad though ngl, i prefer not having to go through 3 tabs to make max quality images
that's why most models struggle to produce dark or very bright scenes
here are a couple of examples of the latest noobai vpred model vs the latest noobai eps model
but im also not seeing a way to browse through your loras
oh the right if definitely brighter
yeah, brighter and lower contrast
not seeing a way to view my previously genned images either on invoke
makes the colors much more washed out

so, invoke has 2 options when generating. you can send the image to your gallery, or to the canvas
if you send to gallery, it's saved and you can reload it later
wheres this gallery at
if you send to canvas, you can continue working on the image, such as inpainting in layers, before actually saving it
man inpainting is lowkey balls for me
gallery is a tab in the right pane in the "canvas" tab where you generate at
neve rfigured out how to make it work consistently
so i just prompt everything out
and use adetailer
@slate narwhal I really like this style
and the above too
Is there any similar artist style In noob @rough oxide 🥹
I think torino aqua could be , but I don't know any other ones
you can do pretty much any style but uhhhhh idk a way to make that specific style.
Maybe cogecha could work with some style mix
I should try later
What pony model is good for making a more simplistic style? The simpler the better
define "simple"
for what model too, SDXL? PONY? illustrious?
For pony. Something that doesn't have gradient shading, but defined shading (ideally a 3 tone color scheme). Like mistoon_anime
just a sec
checkpoint right?
try this one
Oh, nice! I'll try it tomorrow or so. Thanks for finding it 😊
Merry Mikumas @finite herald
how do you do hooman?
Miku-Mas!
cute
thx
❤️
steal your smol
@dull jackal free gift 

Do not click on that link until a staff member deletes it
don't click on it after they delete it either 😏
don't click their link, click mine lol
blurry smol 🥲
give glasses to your AI
@green temple new character unlocked: bnnuy smol
what're these things? sum1 tryna advertise their bot or something?
that person never posted in this server before either
maybe midjourney commands idk
saw that one right b4 i went to eat dinner, love the coloration on that one.
Thanks! I'm not sure if noob has this artist, but if it helps, the artist's name is Tokkyu and the styles I posted in those messages are from a Lora I created in their style (the Lora didn't turn out that great, but I can share the download link if you want). From what I've seen in your recent messages though, you've got styles that are just as good, or even cooler and better. 
oh the styles i asked mimzukari about ?
yay it`s in the aritsts 
there`s a similar style called ittokyu too so it was confusing 😂
Ah, nice that you found Tokkyu! Yeah, I can see how it could be confusing with ittokyu having a similar name. 😎
with some artists 👍
I always love hair like this
🎄Merry Christmas, ya'll! 🎁
idk why just wondering when i post some cool ai art that doesn't involve children will get no reaction
lots of us just post here often and thats big part of it!
Is cfg rescale like completely needed in noob vpred?
Like is the model just not work well without cause its been hard to make imgs fo me
its...a problem. allot of community servers are rampant with it and it's just disgusting.
No sir its not 'children' its 'loli' /j
smol hater @green temple

obviously well moderated servers like this is fine but in he case of community servers like nubbys and lottalewds server.
Ill never understand the logic of "its okay if she has an underaged body shes 1000 years old" a quote i saw from one of sed servers.
First off, you know damn well that my server has rules against lewd images of minors because I've removed your images that violated it before, which you protested vehemently at the time.
Second, you seem to have an issue with letting things go. You apparently dislike my server because you verbally attacked another user about the content they were posting in a SFW channel. Content that has also been posted in this very channel multiple times without a problem. And you got called out for being an asshole.
Third, don't come to me in DMs one day asking for help and acting friendly then try to trow me under the bus in one of your vendetas a few days later. It's a bad look.
it's ok if it's me doing it? 
hello guys
I gueess the shoe fitted
If youre this upset over facts then I guess It must come from a place of truth
Also get off your high horse?
No, just didn't think you were quite that much of an asshole before.
even being an a**hole is a form of art lol
Idk I call out sus behavior that counts as verbal assault lol
socials are so spicy as always <.<
if we were in real life i would have offered to you two a beer and joked about it 😦
iz 🎄, offer them a 🎁
but on socials we are all on our own <.<
Christmas should be a day of love and peace, but today it's already the 2° discussion i observe powerless <.<
marry Christmas guys
There isn't joking with Ellie they are just like those twitter terminally online people you can't talk sense with those guys since if they hate you they will keep bringing up what they hate you about even tho it doesn't have anything to do with subject of discussion and not respond and repeat same thing over and over again until you give up on answering them or multiple people call them out
usually a beer is enough to melt even the harder people, but i know what do you mean and it's pretty saddening
too many meaningless divisions
ah youre the one who constantly makes lewds of your loli oc Yue was it?
I know people dont like being called out on their bullshit
But that doesnt bother me
You still haven't answered me what does this have to do with me posting chibi miku meme?
Because you took something out of context purposefully and still perpetuating it as if you have the moral high ground.
You constantly misdirect when accused and called out of your sus behavior like Just now
What did I take out of context?
Well here's your context
Because I like
or do you have another out of context ss to share
lolis?
exactly
thats suss
lol
ty for proving my point
How is that out of context ss?
@trim current can i ask what are you trying to achieve? it's Christmas and you are bothering people in a channel that is usually very fliendly and serene
You are one that brought lolis to memes that didn't have lolis
Lol keep deflecting
I don't see what I did there wrong you just started flaming me for no reason
you pedo
pretty tasteless
Like this is what terminally online people on twitter are like they just can't answer to one question and go on about why they hate others
slandering a person without proof is a crime, do you know that?
for your sake, watch your own words, or one day you may end in a bad situation for a very petty reason
people lives can be destroyed for such stupid reasons
notice how i didnt call out specific names until they came out after me and i only mentioned the servers lol
that's not the point. you are acting as a menace
protect them kids man
wrong or true you may be, that's not the place or the way to do things
can we all just take a deep breath and go back to wholesome anime? 😉
totally on the same page. good gen
thanks - hunyuan knows at least a bit of miku it seems... the hat/hair glitch in the beginning was not prompted for 🙂
Drinking is bad

Eat cookies instead

oh no they called out my sus behavior better block them and pretend it doesnt exist
lol
🤡
BTW everyone, for reference lolicon was something we were not allowing in this particular server and have cracked down on numerous times: #🍥|anime message (it's also pinned here in the channel)
i had mentioned several times and there appeared to be a lack of it for quite some time, up until recently where there has been heaps of it.
i'm not here to ethics battle, only here to remind that it's not for the server, especially when it is all that is posted majority of time.
thank you & happy holidays everyone
literally the smallest most undetailed dataset youve ever seen

idk why but it works the best for me
i went back and made a new dataset with LESS images
he came out kinda fruity

nah its fine
first gen of v3
ill let it cook
gen 3 but also with my style lora
real shit
this is gorgeous, what model are you using?
didn't get your ping, but it's novelai
ohhh that's nice. I need to pickup a novelai soon, I really like it's recent multi character setup
I think you can imitate that style with controlnet reference:
to a degree
love controlnet reference <3 but yes, I know :) I tend to lean towards more 2000s styled art usually, but was intrigued by the really high fidelity of the images. makes sense it was nai
I think nai (and midjourney) utilize something akin to ip-adapter for stylization in their image inference pipeline + prompt adjustments
yep, I was using the vibe transfer image while making that image
a.k.a image reference
if you've ever used local, how do you find novelai's new multi character features to be vs like, regional prompting n such
maybe a little bit better than regional prompting
I was thinking it would be, it seems easier to to do proper interactions with
I haven't played much with regional prompting yet but plan on doing so more soon
but there are some cons, such as the image getting corrupted or glitched
yea i find if you have too many chars for the resolution it might start to decimate image quality a bit unless u can make them all fit properly.
but hmmm, vibe transfer? did they release that for v4 already?
Nope
to be fair doesn't feel like it even needs it right now with the amount of style control you can get from even low occurrence tag artists
v3 i used vt like mad
v4 i don't miss it at all so far
Most of my styles don't work on v4 
the ones i've tried all work >_<, when they updated it a few days ago it felt like the styles got even stronger too.
i think if you had like a specific style stack, it's just "different" because there's so much more training on it now basically
Yeah little bit different
are you use any strong line art artist in that stack tho
like all the artist tags i like just got stronger
like 2x more influence than noob
I did some testing but didn't get my styles to work
i think you can prompt it more like noob too without using artist: <name>
when i removed all the artist:
altho it is still more that kinda illustrative style for this one
wait i think underscores
aren't in this either maybe?
this is as far as i get with those artists tho, ye.
i think sho (sho lwlw) wasn't written properly in prompt (?)
v3 seemed to still show their arts effect
sometimes
But yeah that v4 style isn't anywhere near this
don't take what i say as bible but i think you don't wanna use underscores for this one, for sure.
as for the artist: thing , probly a toss-up
it works, but so does just putting the name in w/o that now
way stronger than naiv3 was
so kind of more like noob style prompting
as for style, it's gonna be completely diff just bcz... my headcanon is kind of since we can do things down to like 15 tags now sometimes
i think elements are just way more powerful than they were in v3
even if they were trained properly in v3 from having tons of tags they're probly just way stronger now in v4
but given the model doesn't work with nsfw i'd assume that any artist that has primarily nsfw wouldn't work
(i don't think that applies to this situation)
just new things to test/learn. i'm wondering how it's gonna turn out in the end
styles i like to try still work, a lot better than before even
mine r simpler tho ;_;
only like 1-3 max
all i could maybe say is start with a base maybe and make a new stack but uhhhh
given we had an update 3 days ago and it's still a preview
might be better to wait for it to be finished since things could change
or even for when they release the eventual full
yesh, i type way 2 much
If it was nubby, be would've ended in 2025 to type all that
small food
i guess i'm more of an experimenter type
as long as i can get something that looks good i likes it + i like when effects are more known/stronger/etc
i love how there's so many more usable tags now like caramelldansen & i am the king, anime screenshot is so much more powerful and looks like 1:1 to anime now
i never really setup like large stacks of artists that i defaulted to but there's some in the back of my mind i sometimes add when not randomizing xD
No wonder it takes time for @supple raptor to write long texts with short hands like this
lol
yesh, it hard to reach all the keys
meanwhile dusknoir fat fingrs it.
Why eyes weird 
nyo
it's like hitting random button in an mmo you never kno what ur gonna get
some chars just need diff artists to look way bettr xD
Ugy
*but not always. some tags like ^_^ @_@, facial expression stuff in general uses underscore still Lol
Aoi ogata looks super strong on that

2 stronk
i dun like 2.5d that lean into 3d stylez.
that much
it leans towards wlop a little
All ugly 
I'm using riz aoi too
@rough oxide
It's probably the quality tags stuff

Idk what I'm using or you
i let artists carry and rarely use quality tags
except in very specific situations
and even then i usually prefer just letting artist carry prompt
Ye
but i do use "worst quality, lowres" in neg
That's why your images more ugly

and less "masterpiece" looking
cuz all "masterpiece" ai art looks the same to me
xD

For pony tags, I need to do the entire score string or just part of it? (I'm inpainting with it if that matters)
doy
some people use them all, some only use part of it, depends on what gives you the look you want
if you're using pony to inpaint something not pony generated, probably start with it all and then drop parts to see what matches best
if you're inpainting a pony image, just use the same string
I'm back in town! What'd I miss? 🙂 Hope everyone had a nice holiday.
I'm using Redux. I really like this chromatic aberration effect.
Porcupine Waifu
he's a little shy
"random anime manga heroine, epic pose, cover art"
I got inspired by your amazing words
Here's one anime art, epic stuff

neko play
Big

My answer yesterday was too brief, and since the process is quite complex, it deserves a better explanation. I'll try to expand on it while keeping it concise.
As I mentioned yesterday, I'm using Flux Redux, but there's a twist to it. It pulls the original complex effects from the Lora, but if you use the Lora directly with Flux, these effects aren't there - they become reduced or nonexistent. In Redux, the workflow combines the Lora with Flux Redux while also blending the resulting generated image with the input image, influenced by both the prompt and Lora settings. I set the configuration to favor the original image's style, but it still maintains significant influence from the prompt. This combines with the Lora effects, and then the workflow applies an effect similar to Photoshop's multiply (it's actually labeled as 'multiply' in the workflow). Since the Lora has chromatic aberrations built into its training, I think this multiply effect brings out these aberrations and applies them meticulously to each line with different colors, creating images that really get close to the complexity of what a real artist would do.
Sorry for the bizarre explanation, but that's roughly how it works.
For that specific image, I took one of Midjourney's most popular Srefs, did an image search with it, found another image in the same style but with a woman's face, then put it into Redux, made the prompt, and generated based on what I explained above.
Attached are 3 images that demonstrate these complex effects, which could almost pass as an original artist's creation.
This means that this type of Redux workflow opens up a lot of possibilities for exploring styles in Flux Loras. However, when you train the Lora yourself, you know better when the effect is in the Lora or not. For example, the eyes in these images have an immense complexity that isn't in the Lora and wasn't trained into the Lora, and other isolated elements wouldn't add this either. It seems impossible to achieve these effects in any other way - pure Flux, elaborate prompts, img2img, Lora merging, or even Flux with the Lora alone just can't get there.
Okay, no problem - at least it's documented here for anyone who wants to try to reproduce these effects or understand how it was done. I tried and tested even with pure Flux Pro Ultra 1.1 and Redux Pro Ultra 1.1 both with and without raw mode, but the results for anime aren't as good as what I achieved with this workflow. Plus, Redux Pro Ultra 1.1 has the drawback that you can't use Loras with it. And it just won't pick up the style from the image alone (as it should), no matter what. Even though BFL's benchmarks show that Flux Redux Pro Ultra 1.1 performs better than Flux Redux Dev, I still find Redux Dev to be superior.
How can I view the prompt and settings of an image i generated which was Inpainted?
Yeah, it caught that red outline in the eyes, looks pretty cool. Which model?
Okay, nice👍
Love her 
yiffymix XL with some loras btw
slept on this model for some time now... don't know why, I couldn't get anything decent out of it before
I'm not sure too.. I will analyze them in my dm for free
👍
this model is wild... needs taming, but the results are gorgeous. Probably could make it look more "refined", but I always liked the doodly style
Witter
Big eyes like 
Walk the plank now
Cheater
Very cute
Can someone try this prompt 
best quality, solo, (1girl:1.3), live2d, best quality, detailed face, eyes details, official art, highly detailed, effects, magical, official wallpaper, desktop screen, 2d model, (yellow eyes), tall, standing, long high ponytail, (dark green hair:1.2), (sleeveless kong fu shirt:1.2), black pants, fingerless gloves, gorgeous, beautiful, chill eyes, (dragon horns, dragon tail, dragon wings), neck fur, good anatomy, (fur around neck), gorgeous, smile, solo, 1girl, slender, cloth around waist, smile, very long ponytail, full details, wide eyes, very long ponytail, (full body), from front, sitting, rock, white background, [splash art],
here you go
1.5 model. the first on simple workflow, the second, on custom workflow
Did it with animapencilxl too
i just mixed in some stuff trying to make the character stand out
its hard to make a character without lora
@bleak cape I threw it in a model I was using too
no its the same prompt
but i mean i kept editing it a lot
nice poses
rip the face 
yea...
they would not appear unless i mention them a few times
I read this the first time and read it as "without bra"
I have been staring at the prompt to see mention of "no bra" or something lmao
one of the old results i got
like am I trying something NSFW here and I don't know ? 
Lmao
nah its just no lora
I just woke up
wow its like 8pm here
same here
Outpainted the first one a bit
clean
yeah
I really like this model
Looks like it comes out of the ground lol
thats nice
shirt is nice on this one
can you outpaint this one?
sure
does it need the prompt?
vertical
alright
masterpiece, best quality, (red eyes), character focus, detailed face, detailed eyes, shaded, beautiful, very long hair, (red hair:1.1), hair between eyes, ((butler suit)) ,((solo)), fancy, elegant, ((female)), wide eyes, facing forward, serious, mature, adult, black jacket, (red tie), extra ponytail, (open long hair:1.25)
prompt
This one really cool
kinda hard to make pants out of a skirt, or what do you mean
i mean did a pant variation come up
from this
Thats the only version i have
ic ic
I play with pretty high quality settings so it takes a bit
I think I got a version with a shorter skirt tho
nah longer is better
pants even better
superpowers
yea SD isnt great with hands
try using lora for hands
lmao that didnt work
I mean they wear pants
i guess thats true

interesting
nice pose
quality increased
🤔
first time seeing legs this buggy
pink
Idk what happened to me, I slept until 3pm 
no evil plans allowed
Smol hater 😱😱
When I inpaint with controlnet preprocessor anyline with misto controlnet model, in the metadata it says "processor resolution 1280". But the image resolution is 1344x768. So shouldn't the processor resolution be 1344?
It's the preprocessor; it just means that it was downscaled to 1028 during processing, can save VRAM and time, but might impact quality a bit. If you want to be completely sure, you can still set the resolution to match the image dimensions beforehand (1344).
Do you know where I can do that (in a1111?)
Should be the rsolution thingy
Thanks. I was looking for it but must have missed it.
She bites back

back of forearm out of view is all 😉
I will post somethign like this on twitter
and be surprised when people unfollow
Why would they unfollow? They don't like that type of art or because it's AI?
white hair a cute
what's the story here
happy new year
that's shronkey
yeah they're quite cute!
I think some people don't like my more joke-y posts
Could be us but there aren't three of you ☹️
feels like im able to look at the chat without feeling gross for a change
gross?
There's a story behind it. A story I don't know
Hm...I actually have no ideas for prompts today.
Anyone know what model this was made with/similar? It was made with AI and a long time ago so a 1.5 model
I have a hard time thinking a model nowadays would look like that x3
Try some drawing style lora
400 step flux cos i goaded my bestie into doing it
So if i remember correctly its about how this chat barely post images of people above 18

Thick 
\
/
this came out cool
Best I can do for a huggie
Spot the difference game
cute eyes
yo
such a cute cat

Tyty
@green temple purp 
Yue looking for someone to rawr at
1024x1024 SDXL images with 30 steps only take about 7-8 seconds to generate
arachne 
i just finished adding her to my soul eater lora 
came out fire

spooder
she heard that

she can witch me any time 
Yue has gift for you
I use eepy girl to block the fireball runs away
Eepy girl gets this
And now it's time for lunch break
@rancid remnant
👍
Now we dance
can't go wrong with emo witches


Imma make some more sfw ones for here 
my dms are open for the not sfw ones-
white roses?
I need to wash my hands from nail dust
its triggering me
fair 
Death witch @rancid remnant 
That does give me some ideas cough cough

it's only a witch if she has 6 fingers 🙂
witchery
turning up the sexy
oooo the style
nice style 👌


scary 😱
Evil silhouette 
nice
yis i rly like flux
oh nice
lol , is that frieren ?
no just a mouse girl
redoes prompt and add's frieren
@green temple Frieren
Is this like that
Starving together (something like that) game

Death starving?
Nightmare starving??
I forgot 
"don't starve"?

Nice, I like the lights.
you're using only dall3?
Dall e is only gpt and bing right
dall3 is my cheat code
😛
I feel like it consistently produces great results, AND, it has the added benefit of me not having to fire up a bunch of crap to generate a quick image. 🙂
I get best results with leonardoAI anime model
BUT it's paid, so I try SD XL models as much as I can
Still unsure if this channel is for web based gens only or also local
just was about to say that
it's the SD server overall, so most images are largely based around SD, but you'll see folks pull from web and local instances. Don't think it matters.
I use online commercial as well as local models... and post both here, from time to time
I don't know if I've tried the leonardo model... not sure what that one is.
Ah fair, i got plenty of anime gens. Might dump them here
I like the online services just because of the ease. 😐
Swarm user for local gens. 3060 12 GB so it does ok.
Ah 3070ti here
sadly 3060 12GB here... but it's enough for XL models
hey, we're in the same club
the 12 gigs is nice
btw my last posted images were made with local models
Sdxl is like 10-13s a image so i prefer that over the cloud ones
Since you get so much more control
sadly no local model comes close to leonardo anime XL model... thats my opinion
that thing is amazing... wish I could get it locally xD
dall3
all these from me in the past 10 minutes are dall3 🙂
Ah looked up the leonardo anime xl, it certainly got a style to it but i get that classic ai look from it
not really, if you use it the right way
Could be, but im a sucker for pony/illustrious
i'm sucket AT pony... 😛
Can get most booru artist styles from it without a lora
illustrious is tough
Hm how so?
would sharing civitai gallery link here be considered a spam? 95% of all my images there are made with leo anime XL...
Wouldent think it would be spam but i recommend using the civit ai green ver
nah you're good... just don't post it in every channel a buncha times
and maybe indicate if it's got any spicy stuff on it
scroll down to see what can be done with that model... and a few loras from time to time
spicy stuff is censored if you're not logged in on civitai anyway
there's some neat ones on here
also quite a bit of sameface.. not sure if that's leo
Looks great but what omnia said
I like this one.
can be useful
I like a lotta these
i look forward to the day when we can actually do something with these creations 😛
no meta posted with the images makes me sad
You don't? I sometimes get commissions from dnd groups or aicg games
the page just keeps going lol
sorry about that... wasted lot of time on AI xD
I'm hoping to use some of that "knowledge" for making some short animations... but have to learn how to make them first
AI is perfect for making props, backgrounds
but animations I want to make myself...
anyway I think that anime model is quite capable...
and has colors like no other SD model I tried
Had to dig for some sfw gens i did. Some wallpaper request i did for someone but never facedetailed/upscaled properly
Rukia bleach
I thought I know her
ME TOO. haha... I made some efforts a while ago to cut way back.
It was like scrolling on a phone or something... quick dopamine hits and addicting.
Or a Ciri close up(no loras) granted face closeups are cheating for detail lol
tell me more about this... not because I want to do it really, moreso just interested in what you're doing
For ai cg games or for dnd?
yes
too addicting... I used to make music or read in free time... now I make more of cute chicks...
haha
Lol, okay for dnd groups i typically make their character in a few states. Healthy, injured, petrified, poisoned etc or whatever the request the dm sent my way as part for his prep
yeah i'd spend time watching youtube and generating images, and look back and think "where did that 2 hours go, and what do i have to show for it?"
people really aren't just doing this themselves? crazy to me.
Poor hardware, bad consistency, lack of knowledge etc
I like this one.. wish the quality came out a little better.
I guess. You can do this right from Windows/Edge now in like 10 seconds. 😛
dall3 can't img2img upscale?
nah, i'd have to pull it into swarm
For ai cg games, well hate to say it but its the stuff too spicy for here but they are low quality, get low ratings on "game" sites etc
Mainly bc people still hate ai cg in games
visual novel stuff?
I created a Visual Novel engine in Unity to utilize the images for fun a while back.
Mostly renpy engine tho
But I'm not a VN player and wasn't really interested in making an actual game with it. 😄
Oh im no vn gamer either
I found that, without short animations and such, it was pretty limiting.







