#🍥|anime
1 messages · Page 34 of 1
having some difficulties with the white shirt 🤔
@magic wave can you give @dusk seal the upscaler message link pls
upscaler message?
how did you did that ? ahah
?
gn ma'am
pls
give @dusk seal the upscaler!
prompting on aurora + some specific loras
also a husbando lora and the colored inner hair lora
yeah doesn't like a turtleneck underneath a short sleeve shirt 🤔
I couldn't do the detached sleeves well
#🍥|anime message @dusk seal
maybe arm warmers?
def better
what lora you using btw?
concept colored inner hair ; husbandoloraV1 ; 2.5D style
should all be on civ
got one lol
husbando lora?
I only find nsfw one
yup i have nsfw negs and husbando only set to 0.4
also can't find 2.5d style
removed maybe? 
don't have nitro 💀
Its close, nice one 😁


I didn't know that was the original :o

Also that Aurora? 
yup yup
That white outline is pretty cool :o
Soon it'll be Aurora with ExpMixLine2 style 


What you prompting for the white outline? 
(outline:1.2)





Comparing Aurora's clip to my old random merge's clip
O, fair :o
omg
aurora ane clip?
Replaced Aurora's clip with my merge's clip
oh, I like it so much better
But on the male one it doesn't do horns
actually depends on what the prompt is
kim dokja, black eyes, black wings, black pants, collared shirt, jacket, white shirt, necktie, upper body, open clothes, hair between eyes, belt, white jacket, pants, black shirt, phone, 1boy, shirt, black hair, long sleeves, short hair, bangs, wings, horns, male focus, holding phone, artist name, closed eyes, coat, closed mouth, white coat,
Wasn’t there a nsfw sub or am I in the wrong stable
wrong server
This is SFW server
Wrong stable
wrong server
go to unstable diffusion
thanks ma'am
is this face restoration or upscaler
I tried searching unstable diffusion but can’t find it , anyone Inv?
Upscaler :o
put in esrgan folder
aight
pmed you, careful posting anything loli, they will ban you
This time I think Aurora's clip is better
koakuma, touhou, red hair, red eyes, demon wings, hair ribbon, dress, detached sleeves, thigh-highs,
Clip as in text encoder
What extension do you often use to create your images?
I love how this server is less conserving than
the one made to be opposite of conserving
got muted 24 hours for posting long skirt girl at unstable diffusion
lmao
how about this

Training is done, time to test
@lyric zealot

let me know how is it 
Im gonna go get my breakfast 
Will do 

It is scuffed @lyric zealot 
did you really just do that
go right middle it's the best strat here
if the eyes count as O, it's a double win already
Yeah, I think we agree to call it a tie.
ExpiestMix
The movement... the sword... the... HANDS. It's beautiful...
XD
At least you don't have a deformed waifu

t'was a tense battle.
I'll probably sleep while it trains tho
actually kinda good tho
768 takes longer

guess i will make more tic tac toe board for you guys 
i saw potential tho

Cause out of my Sabu LoRAs, I think the 768 res / 1e-4 lr was the softest
Anybody else greeted with an interface like this recently? The heck is going on?
it's really good, the queue got so long that i have to literally order when they just open up their store
did you close your cmd by accident?
Freshly open.
refresh the webpage

No change. 
Time to remove all extensions. 
you are in good hands!
well, before you do that
did you install any new extension?
if you have any of the theme extensions, that would be the first place I would look.
I already removed any newly installed extensions. Or at least I thought I did.
None of those, sadly.
did you update your webui?
not telling you to update your webui btw
Nope
Been like this for a while, which is why I'm so confused.
Only changes I've made have been extension updates, so I'm removing them (backing them up first) to check.
Even with no extensions it loads the same way. Good grief. I'll try rebooting my PC. 
not sure about this advise, but maybe try deleting venu?
can you comment on this? @supple raptor
And when I restart windows is going to have to update. AAAAARGGGGGH 
ggwp

🤝
ggwp

Maybe. Gradio files are inside venv so that could be the problem.
\
ahhh venv yes yes

idk why but flowers 👍
(generated with SD 2.1, but a custom model I trained an released)
1.5 is still way better for anime because more people train for it but 2.1 isn't that bad
lmao
Time for a 5 GB download. 
we must go w i d e r


Somebody give me something to listen to before I beat my head against the wall while the virtual environment reinstalls.
https://www.youtube.com/watch?v=1rnYpm5UHSU not sure if it's your cup of tea but here goes
go sleeeep

you doing generation for 6 hours? 
ahhh, still doing your merge? good luck 



And all it took was taxing my bandwidth. Now to see if my extensions work again.
Also all of you go watch Tengoku Daimakyou, or I'll steal your candies

what is

Is a good animoo
ahhh saw that trailer, was super interested
Can you give me a single sentence description?
Uhhhhhhh
Meanwhile, among the survivors of the outside world, teenagers Maru and Kiruko band together to search for a special place called Heaven, each for their own reasons. Carrying past burdens and tragic secrets, the two hope to find answers to the cruelty they have experienced in their lives and in the world, which still remains in tatters.```
just pretend it's 1 sentence

XD
This is killer btw
the trailer is really well done
I mean, the 2nd song is from the OP of the anime
TVアニメ『天国大魔境』
📌2023年4月1日より毎週土曜22時〜TOKYOMX・BS11ほかにて順次放送中
📌毎週土曜22時〜ディズニープラスにて世界見放題独占配信中
【OPテーマ】
💽BiSH「innocent arrogance」
各プラットフォームにて楽曲配信中!!
▶https://bish.lnk.to/innocent_arrogance
●公式サイト:https://tdm-anime.com/
●公式Twitter:https://twitter.com/tdm_anime
●公式TikTok:https://www.tiktok.com/@tdmanime
【INTRODUCTION】
2024年、世界は崩壊した―
未曾有の大災害から15年。廃墟となった日...
It slaps 
Still listening to
Fair, A New Fear slaps too 
PTW? 
O, make sense xD
Also good grief, how long has it been since I restarted my computer? Spotify looks completely different now.
XD
Same, I feel ya
tho this is my number of ptw list
ptw?
237 animes? 
Why do you have that many 
why
ha
i have no idea tbh
same,, i do not like it lol

It's over
10,000!!!
Well, if Tengoku Daimakyou is not on your list, I will find your house and eat your chairs 
please eat my chair, please
but it's in my ptw list

Well, it better be first 
it is, but i think it's on-hold
Aka, next one you watch 
my currently watching tho

tho
pretty sure Dororo is finished
Well, next one you watch gotta be Tengoku Daimakyou tho, 100%, otherwise I cri
And piss myself
Out of sadness
Ohhhhhhh that one. Yeah, I read the manga until it got weird. Not going back. Didn't know they made it into an anime.
i think i lost interest when i was watching that previously
I mean, it didn't get that weird 
:OOO

the old or latest one
cos theres 2
the og and the remake
the new one
shame
love the op btw
When a guy is in a girl's body, that's pretty weird.
ᴜɴʟᴇss ʏᴏᴜ'ʀᴇ ᴏʙᴀᴍᴀ ᴀᴘᴘᴀʀᴇɴᴛʟʏ
what about
1月30日発売
女王蜂 New Single「火炎」
http://www.ziyoou-vachi.com/
[火炎:フルサイズ先行配信]
1/22(火)0:00~
https://smar.lnk.to/3P8CxAS
「叶うなら、遠くまで。」
【TV放送】
2019年1月7日よりTOKYO MX にて毎週月曜22:00から放送開始
2019年1月7日よりBS11 にて毎週月曜24:30から放送開始
2019年1月11日より時代劇専門チャンネル にて毎週金曜26:00から放送開始
【配信サービス】
Amazon Prime Videoにて日本・海外独占配信
<スタッフ>
原作:手塚治虫「どろろ」
監督:古橋一浩
シリーズ構成:小林靖子
キャラクター原案:浅田弘幸
...
if thats not on your list im cri
Unless you're obama? 
obamaa
sɪɢʜ
And I mean, thats not the weirdest thing I've seen
Pretty tame I'd say

Can someone else explain for me lol
sadly nope

Samurai Champloo
is gud
intro is gud too


i did
I think I started that one but then lost interest
wow
Im still kinda upset i got spoiled about what happened
i know we aint friends but we cannot be friends anymore
XD
it was good if you get past the slow start
I mean, dunno, I forgor how it went or why I dropped it

I think I didn't xD
like its silly for like 60% of it
i think you will enjoy that, given what i know about you
tho i don't really know you

then Yato gets serious and i cant continue what i want to say witout the fun police involved
dmannn


What you know is that I like Tengoku Daimakyou, so you need to watch it to get a idea of what I like

big chad in the house
I imagine its the Redo of a Healer thing? 
i actually never seen that
there's definitely nothing about redo of a healer there
I'd say
Don't watch it
I'd compare it to-
Well
There is something comparable to it
Kinda
Xd
The start of goblin slayer
ahhh
the only anime my irl friend recommended, is that 
XD
I mean, I'm like inmune
And its not as bad as goblin slayer
Goblin slayer I felt like, my soul die
To all who asked, presenting: Obama switched bodies with a girl--in Air Gear.
This one is a little less worse cause of the context
lmfao
obamaaaa
So I mean
I feel like casual
I would recommend it to someone with like a strong mind
Same kinda, I don't count em, but its probably not that many 
I mean, i just use MAL to keep track of what i've watched and stuff
sus gives me mayochiki mc vibes
i dont know that one
I think that my stats are missing like 1k episodes since I'm rewatching Detective conan 2nd time atm
Fair 

Eh, it just means you wasted less of your life than I did. 


facts 

Probably not much since I haven't updated mal properly in like +2 years
i really enjoy tracking anime stuff with MAL
Same, but anime-planet
Now I know where that one 10/10 vote is
I had like a brain aneurysm for like 30 straight seconds
should have done the same with manga, but was lazy, then one day, my app got deleted and i lost all data

I saw tis, and I was like "Monochromatic... wait, no... Mono...chromatic?... Mono..."
@trim current XD
For like 30 seconds before I remembered the word monogamy
But for now I just imported my Mal to 9anime and let it track automatically

yes i am monochromatic
WHAT? YOU CAN DO THAT?
XDDD
i am colorless
you can do that?

its a western labeling thing dont thinktoo much about it 
Fair 
I haven't rly checked but I noticed that it remembers what episodes I have watched and since it let me import Mal
that's not what i meant but... starts searching on civitai
.
AND a petplay one too


Nice
i actually got a pet the waifu lora
Ayyy :D
Well, too bad, I'm gonna use it on your waifu now


👀
ALSO, I might as well stay up at this point
do it cos i wanna see it as well 
what time is it for you? 
i respect it 
Well, its 8 PM, its pretty early
But I wanted to get some sleep in to play with a aussie friend later 
notes 11 hours behind

what's playing game when you GOT A WAIFU TO MAKE
True 
Well
Waifu to steal

too much talking here, so here's one waifu
That emote is pretty well done, I like that there some space up and down to give the illusion of shake 
we might need to chill with the meme and stuff
people viewing this page be like 
Tru
XD
We should AI generate the meems

I think it's fine, unless people are smothering others' content or convos. It is the anime channel, after all.
Sicne we need more waifu here I'll throw old image
true, just afraid we might be deterring new people from entering 
More like general-v2
How about some REALLY old images?
I'll throw something in too 
Lemme see what some of my oldest animu are
Gun Kanna 
one of my first few old gen
When we first got composable diffusion:
Tis first ever thing, I think
I can see where your heart was at least... lol
Kinda good gun
That'll remain nicely unspoilered for me. 
ʏᴏᴜ ᴛᴀɢɢᴇᴅ ᴍᴇ sᴏ I ᴜɴsᴘᴏɪʟᴇʀᴇᴅ ɪᴛ 

One of earliest (if not earliest) composable diffusion tests.
tbh I used to delete liberally, so I don't have much of my earliest gens
last ping i promise 

it's not that bad 


The type of horror I hate the most (though I hate all) is body horror.
body horror? 
Lemme give some earliest NON horror gens:
cats
Damn that 6th one is cool :o
First testing of prompt editing
"a comic book cat, intricate details, sharp claws and sharp teeths,, Catstravaganza"
model to come
I love the middle one, especially knowing that there was no actual cat that was forced to wear a ridiculous costume, lol.
I love it too but too close from the original submission on that one
ahhh that boto submission right?

Was wondering where it was from 
I put lots more example, lot less anime, in here : #🏞|general-with-images message
Wow. That's a lot of cats 
it's the "complete" edition of the SDArt model, with 16 PoW and the pictures people gave in #1099245477007990825
still open for collection by the way, for next version in some 1 week or 2
Yeah, I want that black text sooner or later. My effort has been displaced for now.
I really liked the "creature" one this week end
I didn't do POW cause it was a pretty hard theme, I did give COW a shot tho
you will get it one day, i believe 
i should stop getting lazy...
Oh I have confidence I will. I believe I have the necessary knowledge and tools, I just need to do some proper R&D and planning to win.
and also to give a unique twist to it

Standing out among those who stand out is always the trick in creative endeavors.
exactly, that's almost why it's almost always better to look at other submissions before cooking up your own
Ok. Now to see if it works properly... 

This is at 1 weight
Ok. So something in one of these two is screwing it up.
Anyone update their controlnet recently?
Time for another 5 gb download
not me, i don't really update stuff now, too lazy to fix anything if it's broken
This is at 0.65
Controlnet as of 16 hours ago was fine. I don't use canvas-zoom.
Im gonna try like, epoch.... 10, maybe its doing a similar thing as my V3 Aroma LoRA was doing
Thanks.
I'm used to having to narrow problems down from years of ᴛ ʜ ᴏ ʀ ᴏ ᴜ ɢ ʜ ʟ ʏ modding games like Skyrim, but it's still monotonous.
e10 is still trash
I think I'll retrain 512 but at lower dim
blood
I'm sorry, but at this point what you have shown is poor quality anime
Maybe you need more time to train the model or more dataset
In any case, progress is good
i like this one :3
no problem, it's not an anime model, but a more generalistic one, and is trained on 512 on those samples. also, 20 steps no highres fix
I don't think I'll get anywhere close to the quality of main anime model. it performs better in other categories, but it still has quite some samples in the dataset, all coming from this channel
(plus the base I trained this on is 1.5)
I take it that the model is aimed at a realistic image?
it's aimed at community art.
the dataset is composed of all the pictures people submit in #1087493421209485393 and #⭐|pow-info contests each week.
https://civitai.com/models/44230
I'm working on updating it with better captions, and the latest events. those were the sample tests I got that seemed ok while preparing the example picture for the model update
evenin'
Anyone been looking at this new "vlad" repo?
Been away for a few days again, so catching up. 😛
My first model merge, has about 6-8 different models, used with supermerge. Has a heavy pastel focus, prompt (((digital photo:1.2))) to help offset...
i uploaded it yayy
UMBRELLACORE
boy what the hell
it is absolutely vacuuming up my ram
i probably need to restart my PC to clear my ram
Yes, I've switched to Vlad. I had too many problems with A1111 and haven't had those problems now
Interesting. You like it so far?
Yeah. The layout is similar, as it is a fork after all. It comes installed with the bread-and-butter stuff like ControlNet, Pytorch 2.0, etc. I was trying to install GroundingDINO but kept failing in Auto1111, but it installed without a hitch in Vlad. I've only tried it for a couple of days though.
What do you miss most from Auto's thus far? Anything that you'd love to see in Vlad's that isn't there yet?
Cool
I didn't use too many extensions and tools, so I am not missing anything so far. Maybe someone who uses many tools will miss something? I am a simple guy
I getcha. Yeah I have a ton of extensions installed. I guess they don't work with Vlad's then.
Have you tried?
im amd so im guessing wont work for me
I haven't. I've only heard of Vlad's and was wondering if it would be worth it to switch now that as of today, ControlNet stopped working for me with Auto's.
This was the main reason I made the switch
Wait... so your ControlNet stopped working too?
At some point it didn't, then it worked, then extension install didn't work. Too many problems in general
Also, I tried to make it work with Pytorch 2.0 but it ended up not even running ControlNet because the VRAM use was so high that it couldn't even run it. But with Vlad it worked just fine
The UI/Gradio stopped working for me today, I removed all extensions and when that didn't work deleted the venv. Then I place all extensions back in pieces until controlnet then it stopped working again.
Upon installing controlnet fresh, I see that it failed to install properly due to what it says is permissions issues, but its not due to file permissions from what I can see in the properties of the folder.
Sounds like something's going on.
So technically the Controlnet worked, I just wasn't able to run it. For example, I did 2 batch of 2 with 1920x1080 Multi-ControlNet with Pytorch 1.x. but then after upgrading it took 64gb VRAM. And of course I don't have even half of that
But now I can't use Vlad for a week or so because my sister is in quarantine so I can't use my desktop. So I will use this time to learn about other stuff and write scripts. I want to try making a consistent manga with Autodesk Maya, Blender, and Stable Diffusion.
Sorry to hear that. That's an interesting project idea though--I hope you're able to make it work!
Thanks. I basically use rigged 3d anime heads and attach them to bodies and snap clothes onto them. It took a while to make but can be kitbashed fast now that it's done. Then extract line-art using the Pencil+4 add on for Maya (just a few days ago it came out for Blender) and generate a depth map. Then Multi-ControlNet the entire scene from the line-art and depth map. I intend to use GroundedDINO+inpaint tool to correct for color consistency
Two Gods, from my story
How can you make them consistent when the model is relatively complex (ie compared to simple cel shade) and has more chances of inconsistency in detailed areas like designs and patterns?
Lots of specific prompting, cherry picking, and patience, lol
I've found also putting their names in seems to actually help the appearances, slightly. Even if hte names aren't related to anyone the models' been trained on. That could be placebo though, lol
Red eyed guys name is Fyroris, God of War and Strategy, blue eyed guy is Haldor, God of Knowledge and Wisdom. Despite this, Fyroris is cool, friendly and collected while Haldor is boastful, proud and cocky
So they are trained? LoRa or TI? And I think the name thing actually does help consistency. Have you heard of the specific name trick? And yeah good idea on the personalities, they should be distinct and varied usually
I dont have a lora for them but i'd like to
just haven't got around to it
First time in this channel I’m seeing more text than images
You should've been here for the convos earlier today, heh. Though there were loads of memes.
You can use specific first and last names and sometimes it is consistent. In that case, you can keep using them for a consistent character. You can also add an ethnicity and even gender swap so you keep the consistency but have a completely different character
ooh fun
Ye. I figured out the ethnicity part, lol
I made a Lora for my friend’s fav waifu and he was ecstatic lol
surprisingly decent result even with only 20 images
i think its funny though that the characters I made I later decided were all family ended up looking similar, and I cant' tell if its just the model or if I got lucky, lol.
If I wanted to make a character LoRa who is short, how could I use openpose in a 3d application? Because the skeleton would be of a normal sized human
I think anime in general suffers from same face syndrome. So if you get the colors the same/similar, they'll pass as family
Any good ways of prompting a character with vitiligo?
oof... that one might be tricky. If there's not a lora for it, i'd try something like (presuming anime prompts) "vitiligo, dark-skinned male/female, two-toned skin, dark skin, white skin" or something like that
love this
wow very nice
Anyone tried HorridMegamixV1? It's supposedly made in a way to make better hands consistently. I didn't get a chance to try it too much but a lot of the hands were bad and once in a while they were good. So maybe I didn't try it enough
I'm curious about @south sable 's checkpoint as those hands are A+
:3
yeah that's still really good
RevAnimatedV11 is the base image used, then I upscale using various other anime models :3
Negative embeddings too?
I wonder if hands come out better if they are hand poses that have the fingers more spread apart. Butz a small amount of 2d skill really goes a long way. Artists should be happy that their skills can put their work far ahead by being able to fix messed up hands and eyes and artifacts
have you tried the newer RevAnimated @south sable ?
https://civitai.com/models/48579 promising
how are you guys getting nice quality when upscaling like that
not yet but i will give it a go
model?
it's hard to believe this is generated from RevAnimated since it looks so different, but love the look
aurora1
the base image is V11, then I send to img2img and use SDupscale with something like Flat2D
I thought esrgan 4x anime was like the main thing everyone used to upscale anime
I use Fatal Anime 500000 G for upscale on most things
NyanMix is a good model too
oo looks nice
I think models can basically be categorized into several groups. Everything these days is a mix of something else and they're very similar to one another within those groups
There seems to be correlation between castles and crosses. I think it's because many castles had crosses because they were European
So it comes out in the outputs
best i can get when upscaling, still feel like can be better when looking at others
is a pantsu shot nsfw ?
When is it not? 
if it's for scientific research
I mean
I'm researching controlnet depthmaps
got good result but it's pantsu

would anyone have a "sad" prompt?
If its sexualized, then probably not.
Plenty of bikini pics iirc, so if its tame then it'd be fine imo 
it's just a casual pic bro
Post and we'll judge 
Can inform on deleting or not
Let me put it this way. If she's kneeling down and blushing with a major cameltoe, then it's unlikely to pass the test. Post it and find out.








Fiiiish post the pic, i seeeee you
🐟
Send it to dm's then 
Honestly, i really cant decide. Anime enough to be here but real enough to be kinda sus >.<
Send my way please?
What I said to fish: "Yeah, I'd think that would be fine. There's no nudity, and it's not very sexual. It's clearly anime style. There shouldn't be an issue."
You see much more with swimsuits anyway.
which keyword you use for upper-torso images?
🧐
booru tags = upper_torso ?
As for others, shoulder's up, or chest up is what i'd try.
oh I see what you're asking now
sometimes portrait will do it, especially if you use square aspect
the more i try the more the ai says "no, fuck you"
I better mosey
i've no idea what's happening
¯_(ツ)_/¯
the same prompt is generating this
does anyone use opera for stable diffusion i just tried it out and would highly recommend
I didn't come here for advertisements. 👅
But seriously, I used to use firefox for SD then switched to Edge so I could keep my main surfing on firefox.
I've been hearing good stuff lately about OperaGX but haven't tried.
i use google and would get crashes saying there isnt enough vram even tho i have a 3080 and with opera it doesn't happen had to shill
yea one of my homies told me a couple days ago and its pretty insane and it uses same interface as google too
Yeah I never use Chrome. Far too restrictive and (last I saw) was a RAM hog.
yea i have 48 gb of ram and it always uses 5 gb
but i always have a lot of extensions but opera has the same extentions and uses 1/10 of the ram
sounds like im getting sponsored rm
It does, lol.
But yeah, that's why I do firefox and not chrome. Heck, even Edge is better now.
is there a site to see prompts i wanna test different prompts with a new model
You can find prompts on the example images for the models/networks here: https://civitai.com/
Civitai is a platform for Stable Diffusion AI Art models. Browse a collection of thousands of models from a growing number of creators. Join an engaged community in reviewing models and sharing images with prompts to get you started.
In fact, now that they have an image category you can just go to https://civitai.com/images
Browse the Civitai Image Gallery, featuring AI-generated images along with prompts and resources used for their creation, showcasing the creativity of our talented community.
I used opera for a very long time before they moved away from presto to chromium engine a decade ago or more. I still don't think they have feature parity with what used to be. Torrent client built in, etc.
it was a ground up rewrite and when they launched it I don't think it even had speed dial yet
Looking through some of OperaGX's features, it actually looks pretty neat.
I use brave
Yo!

Only embeds I'm using are the aurora ones, and rarely badhandv4
That way people can accurately see what the model can do.
Hello, i'm having issues trying to create another character, I want him to have black hairs and light brown skin tone, but whenever I put that in the prompt, it get ignored
Maybe it depends on the model?
I tried with 5 models I have installed, same results with all 😦
This is fantastic, I'm still exploring the model too and don't know much about its capability
@pale moss what tag did you tried?
tag ? prompt ?
Yes
1boy, solo, (dark black hairs :1.2), 5'7 feet tall, ripped, 15yo, bright auburn eyes,dark auburn clothes ninja clothes with black white and beige details, wearing gold dangling, earring in left ear, black stud in right ear, (light brown skin:1.2), high quality, nature landscape background, only one boy
I don't find any tags for light brown skin 😦
it replaced it with "dark skin" but it doesn't work either
Not even dark-skinned male?
let me try
hey no leaking confidential information
Aight, I'm heading off for the night. G'night peeps.

what is it with stable and outputting giant faces in background
Could be bad clip
wow that's rough
A I like that style
just needs some covering and some blood
look here
Those were tests, but now I've gone to real sampler with 30 steps to make eyes for a lora
then I'm gonna take faces and draw in another type of ears and img2img it or control net it and get them drawn properly to then train in a lora
here's better quality faces and eyes, thoughts? gonna run other eye colors like this and then crop into the eyes
then do the same for males
anybody alive?
so many colors
yes, kinda
oh living

I just wanted to post this old image of mine
well technically this is the updated version
this is on the older model version
any ideas where can i get controlnet 1.1 safetensors/
can you guys tell me which models and loras you all are using for anime generations
which is best with torch 2? (vladmadic/automatic)
Sharingan with 10 tomoe
You started it xD
🫡
Do cat girls have 4 or just 2 ears? They're usually shown with cat ears on their head, and their hair always covers where normal human ears would be
I never really had to think about it
but this fucking gen made me
4 ears, they got 4 :o
I mean, I like 4 more :o
I like when I don't have to think about it
like this
i guess addingg anime style at the start gives it the cel-shading im looking for
wojak girl
PNG info pls
How? Anime lineart cn?
lora
Did you try anime lineart controlnet on wojak though?
no, havent used controlnet in like forever
Oh, you need to update and download the new models
why?
Is it not included in the image? I am going to work and can't help until 12 hours later, if I will remember it 12 hours later.
It is probably
I am just not home and too lazy to open a metadata viewer website
Ok I'll do it now
Because there is CN 1.1 out now, major updates
why is it better tho?
It is not THAT better, but the new models are extremely useful and also if you use openpose then you have to update cause it is also far better.
Basically, new models and varied improvements to existing models
Yeah pretty much
I wonder if using a 3d body mesh with whatever type of 3d hair you want, can be a way to make consistent characters quick. Basically use an auto-rigger on it to pose however you want, and get whatever body, head, and hair you like, and auto-skin clothes on it. The get the depth and line-art from it for ControlNet
OH MYYY GOOOOOSH
Enjoy work.
Also damn, I never heard of Sathariel, but it looks so good. Also neggles tested it so it has to be good.

Specific lora or model to get that retro anime look?
no loras used. model is aurora
Satharial tries to imitate Nagishiro Mito style. Mito draws lots of fox girls.
Fox girls very nice
I don't get it
Nelliel D: ?
some of the results from work on on pow
huh, i didn't expect this to work
Smol Sus always has pnginfo

oh wait is it not working?
I still have that stealth metadata extension as well, is that why?
an old one didnt























