#🍥|anime
1 messages · Page 188 of 1
give me...
100 gb of lora randomly handpicked from civitai lol
like they where berries, i was like, it only weight 100 mb, 50 mb, 25, who care, let's just download it and than test it
they are still there untested -.-'
weighting on my pc
Then I’ll agree, just for the list. with links. maybe I'll find something interesting for myself
😝
over 2000 files 😭
I can handle! I'm strong!
i'm not lol
ehhh...
almost perfect
ok, that's perfect lol
that`s not SD 1.5
yeah, i expected it lol
Yes it is
Also this has meta
wow, what model is this?
oh
people are too horny at civitai, it's impossible to look for a pose without coming across at least half a dozen sex poses
to be fair, SD is only as developed as much as it is because of the horny
but I agree, it's a bit much sometimes
hey, i'm all for horny, just look at my username. but man it becomes difficult when there are more horny shit than normal ones
dont go there then and make your own poses
There is a filter for nsfw
exaltation of human anatomy is at the base of the artistic expression
the Venus de Milo
the david
bruh, i'm a engineer, not an artist
the Thinker
or developer
takes like 30 secs to extract a pose from a random pic u find on google by using controlnet openpose
anyone who creates is an artist
even the Sistine Chapel
what i'm saying, just deal with it XD
an impossible one to say all
A man can dream
omg cute
CUTEE
s-so cute omg
buddy chill 
a true masterpiece!
what did you use to make the hands?
if it's an extension, could i please get the name of it?
Both were just prompted. The Miku one had very minor edits. Suika came straight from txt2img.
a good model with good parameters and a decent prompt can get some pretty good hands
i guess that my prompt wasn't good enough
the model i use is pretty good
it must be my prompt
yeah, it was my prompt
i managed to do the thing i was trying to do with her hand
there's only 4 finger, it could be better
here's an option you can try for fixing hands. it has some bugs, but can sometimes work well. I think it's best to use it manually in inpainting rather than trying to get it to work well in adetailer. https://github.com/Mikubill/sd-webui-controlnet/discussions/2400
Welcome to AI art
thanks
thank you, i'll try it
mood
doesn't seem to work

Could I please get the model name?
she just watched "Suzume no Tojimari" anime with me. And she felt very sorry for the cats.
and by the way.. in this model the tears look disgusting.... These are not tears.. they look... hmmmm... well, not tears at all.
yeah.... tears looking more like a scene from a hentai is pretty common in anime models unfortunately.
yeah that happens
white tears 🗿
it's milk, she just slipped and spilled it again 😏
She's been so sloppy with milk since childhood... ^_^
Here. here.. complete hentai and not tears.. although i can try to clarify in the prompt that these are “transparent tears”
i tried 🌚
oh, and it seems to me that I defeated ControlNet. After I updated the drivers for my “three-wheeled bicycle - GTX 1660” yesterday, i have quite adequate time when generating using ControlNet.
crap! In this model, crying nekos are prohibited!
I couldn't resist again. but it's really fun! 😂
https://www.youtube.com/watch?v=xXLpIhFT1sM
Muse Dash×ななひらコラボ楽曲パック「ななひら Festival」への書き下ろし新曲です!
▽Muse Dash制作チーム『PeroPeroGames』公式サイト
http://www.peroperogames.com/home-jp
◆Lyrics & Music:かめりあ https://twitter.com/cametek
◆Vocal:ななひら https://twitter.com/nanahira
◆Jacket Illustration:BECO https://twitter.com/beco_100me
今!世界は大ゲーミング時代を迎えている!
ゲーミングPC…ゲーミングラップトップ…
ゲーミングマウスにゲーミングSSD、ゲーミング炊飯器に...
Renge Miyauchi grown up.. 😁
great strategist!💖
Planning a cheese heist
😂
you try pretty good sometimes! 👍
Yeah but sometimes 
Guys! tell me how change UI in Dark theme? now i am strat Ui in Chrome and cant change UI in Dark mode....
a1111 ui?
yes
COMMANDLINE_ARGS = --theme dark
settings -> user interface-> gradio theme
set to default
mine's always been dark mode 🤷♂️
Same
yes. but in Chrome Default Theme look like this
i wonder if it's using the chrome setting, mine is dark mode
yeah, so it uses your default browser theme, or you can use --theme dark to force it
you can also do http://127.0.0.1:7860/?__theme=dark
or whatever your url is, add ?__theme=dark to the end
yes. but in Chrome my default theme is Dark. oki- doki... i will just add patameter in bat file
yeah i tried changing it to light mode (blinding myself) and it didnt change a1111, so who knows
weird, i can do 512x768 with a batch size of 3, but if i try 768x512 i can't even do 2
just a normal render or are you using extensions with it? some extensions don't work as well with different resolutions
or at least could be less memory efficient
otherwise it shouldn't change anything
It's me! only ears are missing! 
on your pic...😁
she has ears...
but not cat ears
oh, i see
meh, i add cat ears and it takes away her shirt 
still 'sfw' but maybe a bit much for here
"equivalent exchange"! it was in some anime. I just can’t remember where this phrase came from!
lol, full metal alchemist
YES!
and here's another question about the interface. I figured out how to sort the tabs. But is it possible to hide unnecessary ones? for example, I don’t need the “Train” tab
Question closed. i did it myself! ☺️
wait, extension takes vram?
i only have these 4 extensions
if they're affecting the image, some do
not just by being there
adetailer does, and if you change the image orientation it changes the detection boxes as well unless you use this option
if it's not enabled, it shouldn't use any vram though
i was trying to produce a sceneary so i'm not even using adetailer for it
yeah, it shouldn't change anything then
i suppose it's possible the vae or something could use more if it wasn't trained on that size, but I don't think it would matter that much
meh, i guess i'll just have to do one at a time
big 
this reminds me of Kenshi
i know of kenshi, but never played that, so not sure 
it has a very "broken down wasteland" type of look/feel to the world
and your characters always end up with mismatched armor pieces and weapons
and limbs

@brittle rampart
she's too happy to have made cake

Tears look normal in this sdxl model
SDXL here ... SDXL there...
but in that model it looks hentai! 🤣
i totally agree
Private model
I'll try that 🤔

How much cabbage did she have to eat?
Not enough, She gonna eat us

Those eyes say tasty human
I love flowers 

What is your prompt
Is it the default model look or are you prompting for that painting style
can look in PNG Info
Whot model?
What model did you use?
abe is here to steal models
run
Kakigori_v3 with a lora that someone sent me.
I ask for models more than I post gens lmao
some old merge+ custom lora
Can you send me the merge?
What model is susmerge o.O
some merge from like a year ago
sus
I think the recipe is legit just anything v4 + pastelmix on layer 5/6/7
its pretty dumb

Surprisingly, I see this element already on the 6th model during generation. although there is nothing like that in the prompt🤔
i saw something similar in the past. when i played adding random traces of noise on the image, without a prompt with concat, i saw some repetitive patterns
confetti, dust, reys of light, insects, butterflys, depending from the model and the color of the noise
I think that all the models in which this appeared were made on the basis of one original one.
yeah, i think is some sort of bias with blue noise
but it’s also true that by removing the word “butterfly” this element disappeared
maybe that element, for some reason is very prevalent in the model... or maybe the reason can be a lora?
some style or add details?
I tried to remove all LoRA
i suppose, for some reason, that element is very prevalent in the model, so, it's added even when not requested
do you mean many models?
if the base models are already very biased
like when you don't specify the age or sex and the model chose for you
hmmm... after all, a butterfly is a butterfly and not an element of hair decoration.
it's a good think to reflect on. the reason may be many
Its normal like with any color
You ask for white hair and it can spread white over the clothes and stuff
argee
some models more in particular like to add lots of things
(blazing drive, counterfeit)
Like you do a prompt: Girl, sitting, bedroom
And then the whole garage worth of junk appear on top of the bed too 
blazing drive is another one I like using it adds a lot of foliage, rays of light and sparkles on gens if possible
blazing drive - this is words in prompt or?
for SD1.5?
yes
ah i see...
yeah, tokens influence the successive tokens like a cascade
using a color as a first token, it will leak on all the successive tokens
ugh, my chair broke.. not a fun way to wake up 😩
generate a new one
good morning btw 
i generated one but it's already in use 😛
here, i got one for you
tbh, if my new chair came with a cute girl I'd be a lot less bothered by having to pay for a new $500 chair
ah well, new chair ordered, warranty claim filed for the broken one
there's a touhou metroidvania game

there's a touhou for every genre, apparently
model?
coffeebreak_v20 + LoRa`s. you can look in PNG-Info
thanks
lol, i'm just finding out creating male characters are not so easy
behold my blonde male characters 🤣
i know i know!!! This is a modern UNISEX man! 😂
why not easy !?
corrected some prompts, but still getting half of them with boobs
Ban boobs! if there is no Elf, then there is no room for BOOOBS! 🤣
huh, i completely forgot i can put boobs in negative prompts. lol, i only ever did female or scenic generation so i completely forgot that was an option 😂
😂
since #📝|prompting-help is kinda in a lull, i guess i'll ask it in here.
what prompts should i use for reflection images like this
i tried many different prompts, but all i get is character half submerged in water.
I think the @vital raptor showed the generation with an image in the mirror. Ask him.
Wo
Here take a snek
or maybe try something like "mirror reflection, or reflection in water"?
hello all (╯°□°)╯︵ ┻━┻
already tried reflection in water, those woman/man images i posted earlier were the result
guess i'll ask him next time he comes online
┬──┬ ¯_(ツ)
@gloomy scroll this "DontWory - Be Happy" is a BOT i think...
🤖
i think @gloomy scroll is acting like a bot today, very sus
i'm a bot
Well, no... @gloomy scroll is good! He's not a robot. He just plays like that...
🤖
@steady grail is Good too! He's not a robot. He just plays like that.. i know....
🤣
Plot twist, I'm also a bot
aaaaaah! robot invasion!😱
@brittle rampart @winter fable I actually don't prompt for reflections usually, this one doesn't even mention a mirror
sitting at a vanity applying clown makeup```
the model knows what a vanity is, and did it by itself
for this type of thing i just call it a "reflecting pool" instead of 'water' or something similar, but I'm not sure it matters
I just remembered this picture. and directed his.
stuff like this, it just does it 🤷♂️
anime girl looking in a mirror
it's really hard to get them to actually match usually, some inpainting or controlnet might help
@winter fable as far as generating boys, it's all about the model, some have been trained on girls so heavily that it's really hard to do anything else, i'd consider that a 'bad' model personally
I really like this one
really?! hmmm
it was my fight with Controlnet...
Lava Girl 
trying to figure out how to prompt a reflection in the water without the non-reflection being the main subject, my model really doesn't like this concept.. but here's a random interesting result 😄
this is biiiiig mirror on ground ^_^
but the idea is very interesting. due to the fact that reflection is the main object
it seems to be technically possible, but SD basically can't do faces at extreme angles or upside down without special training/lora, so it's not ever going to be good
it would be 10000 times easier to just generate the person standing in/near the water, photoshop the image upside down and use img2img to blend them, maybe with more touchup with inpainting and photoshop
it's a pity. there would be very interesting generations
maybe you right
I've had a few ideas for generations that don't work because the person needs to be upside down, or at least viewed from that angle,
I imagine it's just a matter of training, and there are probably loras for it
i see, thanks for the idea
Pew pew!!!! 😂
Model name?
It's called ProtoLarkXL, but I haven't released it to the public
Oh ok 👍
Good model
Good job
what do you use to get this nice pixel art? if i may ask, good sir. 
sdxl pony diffusion model
thanks.
xl can do the "anime girl in real life" look sometimes
Model?
This is cool 🙂
This is also cool!
And this one. The waterfall is a little weird but I like the character.
Ayyy thanks, started using a SDXL model @nova remnant told me about 
cat
tried my best to draw a cat
send help
nice cat
This is as confusing as when you make portals and look through them in the portal game 
Thats cute :o
it reminds me of that newer Elemental movie
whos dad is this
my vrams would blow out if i tried to run sdxl 😔
i believe in you
and your vram
Prompt:jump.chase.busy
how's your running? and how much vram you have?
my runnin is pretty decent i can do 100meters in 3mins
anime
Fair, I didn't watch that one :O
I think for SDXL you need at least 8 GB of VRAM
Mommy Gardevoir™️
yeah, i only have 6gb of vram
the new amogus? 🆘
Surely you know about palworld, and isn’t joking right 
yes its one of the open world games released
truly a game 
It’s definitely a game 
Go to #1100170312106127410
/comic
Go to #1100170312106127410 to use sdxl bot for generation
man im glad tomorrow is friday
source?
calendar
and greg rutkowski
its the new skibidi fornite ohio rizz game 😱
trying out this model torq was usin'...
blazingDrive V13 + LoRas(Mikimoto_haruhiko,Hado) + Cnet like Reference
Thanks
sorry for the long answer
It’s fine, I’m patient.
if you use A1111 UI you can look at PNGInfo
Ok.
funya
Nya ^_^
You ended up downloading blazing drive 
turns out I had this model but V1.0
Petalia? 
It's a pal o:
Ok
Ok
O I haven't captured that one :o
👍
Hello
milk... river of milk
white glue 
this is milk and not what you think
Any of you use comfy? :o
/Prompt:
Hello, anyone are freepik contributor here ?
I am freepik contributor anyone are here like me ?
help, sd prompt enhancer not showing up
what that? extention for work with Civit?
hmm, i saw someone else using it in another chat, it downloads straight from civit and places the files where it need to be, no hassle.
though for downloading models you'll have to provide it with an api key
those. Is this another robot prompt generator?
don't know, that's why i was downloading it to check. Apparently it helps a lot in choosing the correct prompt and discard the useless ones.
useless extension that is
Well, that won't be interesting. where is the flight of fancy? Stopping thoughts!? Excruciating selection!?
the only way to get good prompts is to test them with the model and find what works, extensions can't decide what's "good" or not
its good for creating word salads
yeah, and people think word salad is a good thing
hmm, but i would at least like to check if it works or not, and if it does, then how much
less is more in most cases
+100500
btw. extention "Civitai Browser+" what about it?
now i use Civitai Helper sometime
what does Civitai Helper do?
wow, cute
with its help you can download and maintain the relevance of LoRa, models, etc.
maybe the expression o the boy looks a little cold, but the complex is very cute
huh, then doesn't Civitai Browser+ and civitai helper both does the same thing?
I suspect that the browser is more functional compared to working with Civit
honeslty don't know
I couldn't make the boy look more surprised on the fly
Let's attribute his cold appearance to the fact that he experienced a cold temperature that day!
eheheh
what are you asking? You are already in Civitai Browser +. Just search for what you need, example lora or model
@steady grail another result
then you can download straight from sd, and, it will auto get the model/lora preview pic and also the token that was listed
so whenever you use the lora, it will auto add the token prompt
he looks shocked, cool lol
then this is extention more functional that Civitai Helper?
i don't really use civitai helper, so not sure
As they say - If you don't try, you won't know! Then I'll install it and see. ^_^
i think you will like it 
I used to, but only for the simple stuff like genning lol
Bear
me
Fair :o
Well, is there a way around Comfy randomly re-loading the model that should already be loaded?
Cause that can be pretty annoying
i think i heard about something about this.
are you using lowvram or midvram setting?
because, if i'm not wrong, only at highvram setting the models stay loaded
图标十个
icon 10?
Probably trying to generate images 
Sensei... who will explain the difference to me this is information in models?
safe tensor? pickle tensor?
PickleTensor has pickles which can be unsafe
NSFW?
As in possible virus
like corona?
Yes
looks like highvram can't be set any more on comfyui
Basically everytime you generate with pickle tensor model you roll chance to get corona
its ok dont worry im safe i drank a full gallon on bleach the other day to protect myself
I'll have to wear my latex suit!
Yes this is anime channel
guys, i know we are all adults, but remember to specify when you are joking, there can be Childs taking us seriously
One is .safetensors, one is .ckpt iirc
.ckpt can have malicious codes if it's actually a sketchy model
there are malicious people out there.
better stay away from ckpt
anyway, most of such models, nowadays are obsolete
full of mutations and very rough
but it turns out that they are less safe as stated above?
still using my AOM3 ckpt's here
better convert it in a safetensors.
one day a second personality may wake up and infect your model lol
nope i checked it already
you both cool hackers!
when someone check for a split personality, i don't know if he is more cautious, or paranoid XD
this is ghost in the shell?
Gasaraki
#✍🏼|rules-and-tos haha
Almost the same story.
rngesus
!?
you don't know the almighty rngesus?
what is that?
The deity responsible for situations that are determined largely by luck, chance, or randomness in online games. Derived from RNG, which means "random number generator."
How do I know this if I don't play games. 😁
wait seriously? you never play games? any games?
I'm not using either of them
why make the shirt bad 😭
Maybe I need to do highvram
I tried several times. but not more.
and that was quite a long time ago
some say there is no more highvram. the model is automatically loaded on the cpu, after the gen is started
maybe some automatic setting, in case there is not enough vram, i don't know
Not sure, but happens often to me
Even though with Auto1111 I can run it fine
it's by design, i'll look in it
Alrighty thanks, seems like --highvram does the trick, but I dunno where to write that down
OK
should go on the bat file
5.((best quality)), ((masterpiece)), (detailed), a grand and colorful sci-fi background, (steampunk:1.3), (Final Fantasy:1.2), airships, gears, pipes, (machinery:1.2), (steam:1.1), city, towers, bridges, (architecture:1.2), (history:1.1), people, costumes, (culture:1.2), (adventure:1.1), (day:1.2), (sunlight:1.1), (shadow:1.2), (digital art:1.2), (portrait composition:1.1), 8k resolution
This one? I did a thing so it runs with like, auto1111 a while back, I don't think its working atm, but comfy still runs xD
wrong channel @stiff dome
yep
Alright, and where would I put the args? :o
no idea, mine is like this ComfyUI\main.py --windows-standalone-build --preview-method auto --normalvram
pause
so, try to put it after main.py
i have the portable version
the install version on window, is a big mess
Alrighty, I'll do that :o
but there, they say --highvram is not necessary anymore so they deactivate it
I mean, atm I have a bottlenecked GPU, it works fine for generating, but CPU is old
So maybe thats why it keeps unloading the model
And slows it down a ton
maybe
Yeha, people probably don't usually generate with bottlenecked gpus, so maybe its that
Why can't I send it?😩😩
m?
why
5.((best quality)), ((masterpiece)), (detailed), a grand and colorful sci-fi background, (steampunk:1.3), (Final Fantasy:1.2), airships, gears, pipes, (machinery:1.2), (steam:1.1), city, towers, bridges, (architecture:1.2), (history:1.1), people, costumes, (culture:1.2), (adventure:1.1), (day:1.2), (sunlight:1.1), (shadow:1.2), (digital art:1.2), (portrait composition:1.1), 8k resolution
Which channel to go to for picture generation?
Red icon - Bot is down
y
awwwww.. cure dress!
yep
town print excelent!
😄
what a biiiiig steam boiler!
I hear that a lot 😏
lol, it even has the logo 😂
nice a background!
slightly dark
yep
might just be due to a bad VAE
usually i fix the color with lora
5.((best quality)), ((masterpiece)), (detailed), a grand and colorful sci-fi background, (steampunk:1.3), (Final Fantasy:1.2), airships, gears, pipes, (machinery:1.2), (steam:1.1), city, towers, bridges, (architecture:1.2), (history:1.1), people, costumes, (culture:1.2), (adventure:1.1), (day:1.2), (sunlight:1.1), (shadow:1.2), (digital art:1.2), (portrait composition:1.1), 8k resolution
5.((best quality)), ((masterpiece)), (detailed), a grand and colorful sci-fi background, (steampunk:1.3), (Final Fantasy:1.2), airships, gears, pipes, (machinery:1.2), (steam:1.1), city, towers, bridges, (architecture:1.2), (history:1.1), people, costumes, (culture:1.2), (adventure:1.1), (day:1.2), (sunlight:1.1), (shadow:1.2), (digital art:1.2), (portrait composition:1.1), 8k resolution
5.((best quality)), ((masterpiece)), (detailed), a grand and colorful sci-fi background, (steampunk:1.3), (Final Fantasy:1.2), airships, gears, pipes, (machinery:1.2), (steam:1.1), city, towers, bridges, (architecture:1.2), (history:1.1), people, costumes, (culture:1.2), (adventure:1.1), (day:1.2), (sunlight:1.1), (shadow:1.2), (digital art:1.2), (portrait composition:1.1), 8k resolution
crap! I broke my SD server. It started giving me something I didn’t understand and then only latent noise! 😂
5.((best quality)), ((masterpiece)), (detailed), a grand and colorful sci-fi background, (steampunk:1.3), (Final Fantasy:1.2), airships, gears, pipes, (machinery:1.2), (steam:1.1), city, towers, bridges, (architecture:1.2), (history:1.1), people, costumes, (culture:1.2), (adventure:1.1), (day:1.2), (sunlight:1.1), (shadow:1.2), (digital art:1.2), (portrait composition:1.1), 8k resolution
wrong channal
:squint:
poof.. fixed it... 😁

thanks.
👍
oh, cute
i found out i can take the caps off my keyboard
So nice
You can do it with any keyboard
just a matter of it surviving after
i can put them back on too
im scared im gonna break something uz i dont have the tool
If it's a mechanical keyboard then the chances are pretty low of breaking anything. The caps just friction fit onto a + shaped post. Wider caps will have a stabilizer bar as well
The keycap puller mostly just makes it easier to get a grip on the keys. Half the time I just pull mine with my fingers.
my keyboard is so old my keys come out really easy 
its mecichanal
its like a generic kind of razer clone so i never tried taking the keys off til yesterday
but now i wanna clean it
that makes me feel better i wont worry abotu breaking it now
For the ones with stabilizers, (usually just the shift, enter, backspace, and space keys) just try to take off the smaller keys around them first so you can pull them straight upwards more easily.
Oh, numpad 0, +, and enter as well if you have a numpad
true i feel like i need to take them all off or many at a time for an effective cleaning
the tab key cant gett that off the way it is currently with the keys aroudn it o
I tried the model a bit but didn't have as much success 🙂
Jumpscare
offended because she was extinguished?
Lmao
I can never get a fiery body 
this is the lora I use. even with this some models flat out refuse to do it https://civitai.com/models/90086/special-effects-fire-art-style
they made an xl version too which I haven't tried https://civitai.com/models/233291?modelVersionId=263209
i love his models
sometimes removing the character part of the prompt earlier can help. something like [1girl, red hair, sitting:0.8], fire hair, on fire, burning might even need to remove the character part earlier than 80%.
this is one of the LoRa models of that author
if I'm not mistaken in my archives 🤔
Ah yes, Memcho's cat mouth.
damn that looks amazing
yep.
@native halo
Too nsfw for this server
needs more wind 
alr thx
⠀⠀⠀⠀⠀⠀⠀
yesh, it's in there. model is EasyFluffV11.2, but it needs a lyco and extension to work right.
very forest, very nature
take your bodysuit dose
no slander allowed
Why is she stuck in a virtual world talking to herself
average discord user
What is this ? Done reading? Check out anime
meat 🥩
need me some morning steak 
nice horns
Hi
@nova remnant what model you use to generate those pixel images?
Pony diffusion SDXL
how do you use sdxl on 4gb vram? 🤨
Dam you, my eyes got seared with all those cursed my little pony previews 🤢
Kek
You should've expected it by the model name

a
b
c
grab these instead 🌚

what is her name again
msyu
too small 

why i can't use bot
Historically the bot has usually only been active while testing new models shortly before and after release. If you have a dedicated GPU, you can run SD locally and gen as much as you want and whenever you want with far better tools than the bot.
I think this has been the longest that it's been active aside from testing models.
booba
you call those fangs? #🍥|anime message
i like this one
yoga is hard 

Any models that can mimic this art style?
yeah I think most anime models can do that
there's some that focus specifically on more "vintage" styles though not sure which specifically 😐
could try any of the vintage anime styles
Check out my other models: https://civitai.com/user/aji1/models ···················································································...
seems like cardos can do it, just added "vintage anime style" to the prompt
yeah that too
wuddup torqx?
not much, just chilling
nice
me too
I feel like I should go out somewhere ... but I dunno 😛
there's my vintage attempt 😛
going to places could be fun
yeah ill probably go out for food later for sure
that's always fun with good company 🙂
I'll give that one a shot, thank you.
How dare she gets in the way of yang kai with those personalities

i wonder what would happen to all these anime images if "duke nukem" was in the prompt...
np!
good gravy dalle blocks like everything
This doesn't really remind me of Duke Nukem...
still kinda fun though
With Doomguy 😛
Stopped at Pizzza Hut, got a pizza hat.
🦊
controlnet
nice, is that with depth style?
with lineart style
ahh ok, i use the depth style often, and there is a newly released version called depth anything
not gonna lie, i haven't explore much about controlnet, so was just trying out
what does that do?
depth anything is finer version of default depth, and it read the image very well to replicate image and additionally compliment any prompt you might add
hi, there's no bot service here. Check #1047610792226340935

think depth anything came out last week with CN update
so let me take @zinc flax's image here, and use that with depth anything and see what i can produce with minor prompt tweak
i think i grabbed a smaller image file.. let me reopen that in new tab
a cute girl, sitting, smiling, swirly water splash in the background
could fine tune with original attributes like hair color, clothing, etc, but thats the basic idea
you know what, i haven't tried out line art extensively enough to understand it fully
maybe you could share some thoughts on it and what to do with it
i think when i read about line art many many months ago, it had something to do with following lines to draw images, don't think i understood that part
i legit have no idea what it does tbh
i haven't play much with controlnet to even know much about it
🤝
only tried open pose/depth
oooh i see now, i just asked bard about it, i think line art is pretty cool feature of controlnet
lineart style can be used to illustrate basic sketches
into something more dynamic and vibrant
so even this image can be enhanced with lineart with rich illustration
openpose is neat, but i prefer depth a lot more, its very dynamic
and with a combination of canny and depth you could make some amazing images
Too lazy to figure out how these work 😔
not a workflow you'd use all the time, but its simply selecting two instances of CN
what's the difference between this and img2img though?
they are both powerful tools you could use but here are some differences
For example this is what lineart model in cnet would use so compared to i2i for starters it doesn't have any colors so you can choose freely (ADetailer changed mouth a lot)
@lyric zealot this with line art
thanks for mentioning it, all you had to just bring this topic and i got the motivation to learn about it 🙂
this is what i used btw
I'm not good with this I have mostly used open pose to with one pose 
well openpose works on the skeletal but lineart can greatly help with inline detail
Here's once again my small setup
so here is another example of line art
and this is what i could produce
a cute woman sitting on the kitchen counter, wearing tank top, shorts, stockings
can you show all settings?
basically using default settings for parameters with few changes here and there
thanks
Today's Tifas
line art is extremely powerful for inline detail
It's now Miku
Img2img uses the latent version of the input image as the starting point for denoising which means that it most strongly affects colors and to a slightly lesser extent shapes.
Denoising strength is actually the percentage of total timesteps to be completed so if you choose 0.25 denoising with 20 steps then you are treating the initial image as if it was step 15 in a txt2img generation with a total of 20 steps then it completes the remaining 5 steps that you specified which is why lower denoising means that the image is changed less. Many people select the option in A1111 that makes img2img always do the number of steps specified (I prefer this method) which instead obscures the total timesteps so our previous example of 0.25 denoising with 20 steps specified means the total timesteps is 80, we're starting at step 60 and doing the final 20.
With most Controlnet modules, instead of that whole process above, it influences the high frequency noise areas of the seed, which influences the image to produce a similar shapes to the input image with very little to no influence on the colors coming from the input image.
I don't know how Reference Only or some of the other more unique Controlnets actually work so I can't explain those.
thanks, i like technical explinations
It certainly helps to understand what img2img is actually doing in order to use it more effectively. I see too many people trying to do things like changing the color of hair in img2img, which you can probably now understand why that is not a great tool for that job.
It’s probably because they saw a YouTube video by a certain person that said they could


