#💬｜general-chat

1 messages · Page 46 of 1

modern stirrup
#

🥄

north meteor
#

@dusty treehey

#

can someone teach me how to make nfsw made by ai

dusty tree
fervent thunder
dusty tree
#

internet went out

dusty tree
north meteor
#

yes

dusty tree
#

now make a folder on your pc

#

can be called anything

north meteor
#

ok

dusty tree
#

open the folder and

fervent thunder
dusty tree
#

press here

#

oops

#

didn't copy lol

north meteor
dusty tree
#

can't paste here lol

waxen pasture
#

YOO ONE OF MY EMBEDS IS IN STABLE DIGESET

#

Let's go!

bleak marsh
#

was there a crypto giveaway on this server?

#

seems like bait but i figured id ask anyway

stable pawn
#

Crypto is always a scam if it is said to be free, safe or give 10 times money back.

bleak marsh
#

dude named tlr said i won a giveaway of crypto

wild steppe
bleak marsh
#

and this server was the only mutual server cause he didnt actually say what server it was

#

yeah i figured it was i just had to make sure so i wouldnt missout

#

but the amouts were ridiculous like 5 eth which he says is almost 8 grand

fervent thunder
#

They always go for high numbers to impress people

bleak marsh
#

was thinking about trying to go to the link he sent on a disposablevm on my qubes os

#

for fun

wild steppe
#

Yeah there are bots DMing random people with these fake giveaways, unfortunately.

#

You can report them to us through the ticket

bleak marsh
#

yeah ive got stuff like this before but thats too much money to missout on if it did turn out to be true on some lottery type luck lol

#

eh i feel like its pointless taking one out just means several more will appear

north meteor
#

can someone teach me how to make nfsw made by ai

dusty tree
#

bruh

stable pawn
#

That was a odd feeling. I go to #roles and see "photograph", I can see my camera from here, all the lenses and tripods, 20 year of photography. Then a day I got a stroke and after that I almost not lifted my camera even my body and mind have recovered, but all old interest is like gone. Am I still a photographer?

fervent thunder
north meteor
#

can someone teach me how to make nfsw made by ai

stable pawn
north meteor
#

0

#

know and made

stable pawn
#

Then I sagest you start with a standard tutorial about Stable Diffusion and maybe the Web-ui that is made by Automatic1111

fervent thunder
#

when you guys make textual inversions do you make the text file into prompts or just the name of the t.i.? hope that makes sense

north meteor
stable pawn
#

After that add NSFW to your prompt.

north meteor
fervent thunder
#

like if i have 15 images of myself for a t.i. should i name each photo or just label them all my name

stable pawn
#

Oh, Is that a Chat bot trolling?

north meteor
#

what

fervent thunder
#

idk im lost too lol

north meteor
#

can someone teach me how to make nfsw made by ai

stable pawn
#

Well, if some want to make a "NSFW image" they have half already written in the question 😄

random kindle
#

hi guys, srry to interrupt your conversation. I'm totally new in this ai stuff, can someone teach me how can i create an image using stable?

stable pawn
#

Let see what happen 08:08 pm, I do expect a repetition.

fervent thunder
#

We really need to put installation + usage tutorial for new people somewhere

#

there are questions like this everyday

stable pawn
#

My prediction was not good :/

fervent thunder
#

Hey guys. I was wondering if there was a way to run stable diffusion for free without downloading on the computer. Having really low specs on my computer and heard there are some web alternatives for this 😊

random kindle
fervent thunder
fervent thunder
random kindle
#

so sad:(

vestal dew
#

when you connect to a collab, is there a way to use a GPU?

#

the first time I connected, it showed I have no compute units

fervent thunder
fervent thunder
#

unless you are already out of your today's limit
(List will be there, but you won't be able to chose it)

gleaming lance
#

The best way to fine tune stable booth for this case

The way I understand is Dreambooth is great at a particular subject or concept.

But what’s the best way to train on a new dataset?

vestal dew
#

I am trying this collab

fervent thunder
#

I just have one more question, where do I find the link that gives me the latest update for this?

#

I tried to run it from older tutorials and it failed.

#

And on the github page I could only find the dreambooth stuff.

chilly valve
#

morninggg Coffee

gleaming lance
#

whats the difference between training stable diffusion on new dataset vs fine tuning?

vestal dew
#

collab never works for me

fervent thunder
#

I used this link and it worked for me

#

that is why I asked where to get the latest update links the others didn't work.

vestal dew
#

the last step is what doesn't work

fervent thunder
vestal dew
#

no

#

usually last step gives me that it doesn't find the SD files

#

now the first step is giving the error that it can't connect to a gdrive

#

now I got first step to run

fervent thunder
#

maybe message tech support to see if you can solve the problem.

#

I will go there and ask if there is a link to get the latest version of this collab ui

proud nova
#

Is google colab having some issues or what? My webui is not starting up because all of the modules (like fastapi) are missing

vestal dew
#

it's running now, we'll see what happens when I try to run SD

proud nova
#

Do you think it's only a temporary problem?

fervent thunder
proud nova
#

Wait. So I am not the only one having issues huh?

proud nova
vestal dew
#

when I start SD, I put in my GMAIL account info?

fervent thunder
proud nova
#

Hope they will fix it

#

I need SD for presentation lol

#

Hey hey hey

#

It looks like I had to update it. I used the repo you sent

#

And it seems like the problem was solved

#

I had the older version of it

fervent thunder
#

That is good then. Now we just have to find the link that would generate the updates so we don't run into this issue again :)))

proud nova
#

Yeah. Thank you

devout hull
#

Hello, can someone help me connecting to stable diffusion from other pc in my network?

vestal dew
#

add -listen in the runtime arguments

#

--listen

devout hull
#

oh... thank you!

dusty pewter
#

how do you deal wit trying to use contornet to copy a lying down pose or a pose where the figure has her head tilted back looking up . seems to have issues with that.

#

it seems to deform or have issues more so with lyign down poses

#

hmm maybe some of those negitive embedding addons might help lol

#

nope 😦

warm junco
dusty pewter
#

hmm maybe both 🙂

#

oops i was using openpose weight at 2 no wonder taylors face look a bit weird lol

#

yay a bit better :-p im not really a huge taylor swift fan but its kinda fun to get to do a "photoshoot"

fervent thunder
#

Hey, i have tried to do art in SD, i was curious after trying Pix Art, but it seems i dont get the same quality as the website

#

Anyone have a Idea how to get similar art ?

peak orchid
#

It would be really nice if Twitter had actual moderation and didn't let anti-AI hate groups organize mass harassment campaigns

reef lichen
#

how do i contact staff?

#

need to dm them something if that's okay

#

(someone in your server is tryna get people to join other servers)

#

looks like a bot

trail wing
#

not to be confused with controlnet, use contortnet on it

#

it doesn't have an web ui script yet but it really contorts those bots into line again

reef lichen
#

yeah idk shrug its just a bot to me

vestal dew
#

Im under a general mayonnaise today. I should be generating images on my new system. It was working yesterday...then it went to pot

obsidian geode
#

Where can I find an instruct pix2pix model only for realistic images? I'm sure I saw something on civitai

vestal dew
#

So fantasy.ai now has its models exclusively on their site....which can't be accessed. Send an email to get on a waiting list. Bad plan.

#

then they send an email to let you know where you are on the waiting list. It's white text on a white background. I am number 17,343

mystic remnant
#

how od i find out what IP address the AI of stable diffusion is running on

#

im on ubuntu

vestal dew
#

for use at home network or online?

mystic remnant
#

please save me homelander!

vestal dew
#

that I do not know

#

homelander is lame

mystic remnant
#

what were you gonna say

vestal dew
#

if it was on local network, the ip of the machne

mystic remnant
#

is there a specific port it defaults to

#

(the AI, i know SD is 7860)

vestal dew
#

yes, that's what I know too

#

I dont know about setting up as a virtual server

#

when it runs on Google collab, it connects it

fallow flare
#

hello, anyone know if theres a way to automate generations in different models? like, i want to generate the same prompt using all my models but clicking between each model for 1 generation is not efficient.

brisk bloom
bleak matrix
#

with a screenshot and a name

ripe walrus
#

Guys is it normal to run at 2-3 it/s for a RTX 3070?

astral goblet
#

depends what you're genning

waxen pasture
ripe walrus
waxen pasture
#

On a laptop gpu, thats probably normal. Are you running xformers?

#

I have both a 3070 laptop and a 3070 pc.

ripe walrus
waxen pasture
# ripe walrus I didn't use xformers, should i? will it impact the quality?

It doesn't impact quality (it can create negligible differences in same seed images but that doesn't matter because there is barely a difference, like 1 or 2 pixels) , but it greatly improves speed. Add "--xformers" to your webui-user.bat command line args (I assume you are using A1111) and it should instlal xformers for you.

ripe walrus
ripe walrus
waxen pasture
#

Weird, did it say "installing xformers" in the console on bootup?

ripe walrus
sweet zinc
#

any tips on how to build a good logo using the prompts?

dreamy igloo
#

I've run into this problem for the 1st time on instagram posting ai art: I made some parody images, and on those images I made altered logos of a certain tv show/movie franchise. It may not have been altered enough for them to be interpreted as derivative work perhaps, but I have no intention on monetarizing those pieces. My posts are either reported/flagged, since they disappear from all the hashtag within a few minutes of posting. Someone/some bot keeps taking my post down so that no one can search for the tag and see those images. I've had other parodies with logos done before without the same problem. But man, either an employee or some hardcore fan kept reporting my stuff on this particular franchise.

#

But the logos were tiny, and are not the prominent part of the images. They are just there to inform those who may not be familiar with that series to look for reference in order to get the parody.

#

I tried to wait for an hour or so to repost those images after the first 2 tries end up being reported/flagged so they disappear in the hashtag space. But again on the 3rd try, same thing happens again, someone either reported or flagged the images so they all disappear.

#

I spent a lot of time to place those logos on the ai images manually. I guess I have to repost tomorrow without the logos to see how they are being treated. I don't have the time to monitor how they will treat the versions without the logos right now, so I will try tomorrow instead.

#

What I do not get is, there are plenty of AI generated art with broken text/logos that highly resemble exclusive logos, and people even hashtag the real logo/franchise on those pieces, and they get away with it, because it is just derivative work for entertainment. Why would manually added, and altered logos be any different? We see tons of superheroes/movie characters a.i. art on instagram with the original franchise on the hashtag, many of those logos are not even altered or so ever slightly altered that there is no mistake on what they copied them from.

astral goblet
#

the show you're trying to parody may have an army of lawyers

dreamy igloo
#

who/how are they watching that space like hawks? a few minutes up, posts get flagged/reported so they disapeear.

astral goblet
#

protecting a franchise ip rights is a big deal for a lot of companies

dreamy igloo
#

Its on my personal "feed", but on any of the hashtag's space, my posts don't show up. Everytime like 5 minutes or so in, the posts disappear.

astral goblet
#

You have to be like Mad magazine about it. Their style is able to parody anything they looked at simply because no reasonable person would ever confuse it for the real thing.

#

If anyone could be confused and think it's a real part of whatever show you're doing, it doesn't qualify as parody

dreamy igloo
#

But look at all the ironman/batman/superman on aiart instagram...tons of them, no one ever take those down.

astral goblet
#

once claiemd its up to you to deal with that and get a lawyer to file back

dreamy igloo
#

my parody is not even as well known as those.

#

no i am not going to fight it since I do this for fun not for money, but its the first time I've seen something like this happen to me personally.

astral goblet
#

fanart is a zone that technically isn't "fair use" but franchises typically allow fan art

dreamy igloo
#

I bet the DC and Marvel guys take those as chances to promote rather than hurt their I.P.

astral goblet
#

because their fans are like "yay!" and thats good

dreamy igloo
#

with more people paying attention to aiart and more people wanting to draw their own ironman and stuff, their i.p. stay hot on the radar

astral goblet
#

right

dreamy igloo
#

so yea I am gonna try posting again tomorrow without the logos and see what happens.

astral goblet
#

theres a balance though and i bet there are a lot of meetings going on in big rooms with big tables lately

dreamy igloo
#

well, at least they can't stop me from drawing those in my local space, stored in my own drives for my own personal entertainment, even if they may in the future stop us from posting them publicly.

vestal dew
#

if you store on cloud, they can remove it

dreamy igloo
#

I was just surprised at how fast they spot and remove them...each time its like 5 minutes or less of me posting those, they get reported.

#

Its like there is an AI bot scraping instagram for anything with a logo that look like theirs, and then they evaluate and report.

vestal dew
#

they probably have something watching it

vestal dew
astral goblet
#

they maybe running image recognition / classification on all new posts

#

a logo wouldn't be hard to train a classifier for

vestal dew
#

make everything backwards'

#

instead of a logo, an ogol

astral goblet
#

oh. idea. make your own logo classifier, then do funky things to it till the classifier doesn't recognize it anymore, but a human can. like, make it a difference of the image, or fractal it in, or like, two tone slash it so it looks like two objects to an ai.. or some

dreamy igloo
#

I will post the clean version without the logo tomorrow. I just hope this is some angry fan who does not like to see derivative/unreal version of their hero on their hashtag, and that tomorrow that person is too busy to report me lol. I can understand how fans may feel about seeing AI version of their favorite character not being what they expect.

astral goblet
#

then once you find a logo recognition defeating pattern, train another ai to create those

#

then you're in baby

thorn hare
#

is there any benefit of upgrading RAM from 16 to 32GB for SD ?

fervent thunder
#

nah, I don't think so...at least mine don't use as much RAM

#

maybe if you want to keep more models in RAM ...or vae's ...whatever it stores , I don't remember , it's somewhere in the settings lol

lament steeple
keen crescent
ionic olive
#

Why is the dress in my picture blue instead of red?
prompt：
RAW photo,realistic,
a closeup portrait of a beautiful gril, [white:(pink:1.5):4] hair, blue eyes , red dress,
outdoors, complex background,

gilded obsidian
#

im new to stable diffusion, does anyone know the latest methods to install and use it

young ivy
#

Try to put some weight to it too

thorn hare
alpine bear
alpine bear
young ivy
#

or hi , hi , hi or hi ,hi ,hi

broken gull
lament steeple
alpine bear
#

oh no doubt

lament steeple
#

you may need to inpaint with "red dress" as your inpaint prompt

thorn hare
lament steeple
#

space after the comma

young ivy
#

I also dont know what youre wanting from this [white:(pink:1.5):4] hair

#

better something like (pink hair:1.10) with ([white highlights:0.70]:1.20) which starts creating white highlights at 70% of the steps at a we ight of 1.20, without stopping the pink hair.

bleak matrix
#

Good morning, everyone! How are we all today?

stable pawn
bleak matrix
#

Well, I am sitting next to the heater, plotting my next move of my morning routine, which involves staying next to it hahahaha!

#

So long as I stay near this toasty spot, I'll be okay.

ocean mango
winged sparrow
#

How do i create a masked image?

#

For inpainting

lean laurel
#

I dont know if I discovered a bug or anything, but ControlNet dosent work properly on Opera GX

#

Like, it just cant seem to find the path to the controlnet models, but on Edge it works as it should

copper bramble
#

hello

#

anyone have time please ?

#

i wanna ask something about stable diffusion and google colab

fallow flare
#

hey guys, whats the best way for me to take my open pose i got in control net from an image, and modify the pose myself?

#

if anyone has an idea thanks 🙂

bleak matrix
#

For more info

stable pawn
#

I can not much about how AI work, I just know that it can be useful, so I want to know if it would be possible to make a web UI like Automatic1111 created to generate a suggested negative prompt? Like when I often write "art" or "picture" I have to add "frame" and "border" as negative prompt, and we all know we have to add "girl" in the negative prompt if "cute" is used even we want a "cute pasta salad" or so.

warm junco
fallow flare
#

okay got it, was wondering how to use that, so ill just import an image in open pose editor, then send it back to txt2img after i adjusted

#

doesnt work so well with 2 subjects heh but itll do

astral goblet
#

I actually don't like using the open pose editor. after learning how to rig a model with inverse kinematics and actual control points that move the pose skeleton like it should, i've never been able to do manual edits like that. It's infuriatingly tedious. Plus, you can get it wrong very often. don't get the head points exactly right? you've got a hammer head shark suddenly.

#

i've been meaing to see how far i can make it screw with generations abnormally lol

winged sparrow
#

Is there any way to batch inpaint exact same results on different images? Like I'm trying inpaint sequence of images and it must be similar inpainting bcoz i want to make an animated gif out of it.

astral goblet
#

I didn't realize that the openpose editor allowed you to edit existing poses though

silk dew
#

Hey

fallow flare
#

@astral goblet the extension posex allows you to move the openpose skeleton in 3d. havent used it tho

keen crescent
# winged sparrow Is there any way to batch inpaint exact same results on different images? Like I...

To my knowledge, using the batch function does not work well with inpainting moving details, because each image in the batch will use the same mask. If you use a low denoising strength so that the original image shape is preserved through a round of inpainting, then you can paint a more general mask that covers the entire area that will contain any part of the moving object throughout the entire batch of images

lapis rain
#

hi all

#

quick questions i generated an image of a detailed superhero, and i really liked that image, so wonder if its possible to take that super hero and generate more images and get same person on other images?

astral goblet
molten quest
#

Can anyone explain to me what the workflow would be for adding my face onto any picture, I already have stable diffusion trained with a keyword for my likeness

#

I want to be able to swap my face onto the faces of other people in an image

fervent thunder
#

is Stable Diffusion better than Midjourney quality wise?

fallow flare
#

@fervent thunder pretty sure midjourney uses their own custom model for stable diffusion. theres some more discussion on reddit if you search a little

#

could be wrong tho if someone wants to chime in

fervent thunder
#

Interesting

fallow flare
#

some ppl say no, some say yes, this is interesting "They also add a text-transformer that converts your prompt into a new one essentially, but reatains the meaning. Basically it augments it a bit to cater it to their specifically trained model"

vestal dew
#

I'll be the judge of the meaning of my prompt, thanks.

#

A tyrant will tell you when to sleep, when to wake, when to eat, what to eat, when to work, what to do - you don't have to worry about all that. A free man has to worry about it - about what prompt to write and how it produces results.

lunar knoll
#

@river badger
Hii, BasicNP. Sorry for reaching out like this. Can we talk? I can't send you direct messages right now.

lunar knoll
#

I'd like to still be friends.

strong forum
#

does anyone know how much disk space i need for stable diffusion

proud nova
#

Prepare to have space at least 150GB

#

Models take space and if you want experimenting it can come up to this size

#

But yeah. If you are starting, 40-50GB will be enough

strong forum
#

really

#

last time i installed it wasn't anywhere that much

proud nova
#

Well, on my local PC SD takes about 80GB

strong forum
#

i havent had 80 gigs free on my pc in years

#

it was somewhere like 20-40

proud nova
#

Ah, I have two drives so I don't struggle with storage

strong forum
#

ill just try it with 20

proud nova
#

You will see. It's true I am running SD on CoLab with 15GB storage and even though it's almost full, I am fine

lapis rain
#

Any idea how i can generate same face everytime?

proud nova
lapis rain
#

Sry i am totally noob

#

oh voice chat? sry dont have mic ....

lapis rain
#

Hello do you guys suggest the more prompts the better ?

proud nova
lapis rain
#

thanks 😄

young ivy
#

it depends entirely on the model and your experience in looping prompts

#

and what image you want

lapis rain
dreamy igloo
# dreamy igloo I will post the clean version without the logo tomorrow. I just hope this is so...

So I reposted the images without the logos, and removed all the hashtags related to that franchise, so far 20 minutes in, the images survive on the rest of the hashtags' spaces. Unlike yesterday, 5 minutes or so in, my images with logos and tags related to that franchise disappeared on all hashtags' spaces. So cautionary tale here, be careful when your parody carries any symbol/logo too close to the original, and careful in how you hashtag your creations on instagram. Some companies definitely have police watching their instagram space.

young ivy
dreamy igloo
#

I will let you guys know if the situation changes. I am still not 100% that my images will survive. Since I have yet to see widespread portrayal of those characters on the aiart world. I suspect they tirelessly crack down on parodies of that i.p.

young ivy
#

then you will know how to

lapis rain
#

oh yh thank you !! will experiment !

buoyant jacinth
#

yep, all about the experimentation.

#

and the iterations. So many iterations.

young ivy
#

if you want a character like batman you could always go (male superhero:1.25), tall, imposing, muscular, wears a (dark carbon fiber suit:1.20) with a (carbon latex face mask:1.10) revealing only his mouth and jaw, dark mode, dark in background, back lighting, shadow full body profile, ("masterpiece, best quality, hq, absurdres":1.20)

buoyant jacinth
#

46 models, 35 loras, 43 textual inversions. It's like eating chips. Can't stop at just one.

young ivy
#

could even reference Sin City

#

to get that black and white nice stuff

buoyant jacinth
#

and by <insert favorite comic artist here> for style, can also help with the character

lapis rain
#

thanks alot !!

lapis rain
#

does () and [] have any meaning to prompts?

deep crag
#

Hi there

lost hornet
#

I'm looking for ckpt, where can I talk to them? Can I do it here?

deep crag
#

How do I use train on my Stable Diffusion?

proud nova
proud nova
deep crag
proud nova
#

Good prompts are necessary for generating good images. What are you trying to generate?

deep crag
proud nova
#

Uhh.... Alright

#

Well

#

For that...

#

Cerain models must be used

#

You can get them on Civitai. There is like a TON of them

lost hornet
deep crag
#

I'm getting them, but I can't get the images right. Is there some way to make them perfect?

proud nova
lost hornet
proud nova
lost hornet
buoyant jacinth
#

I think it's referring to etrian odyssey. There's a LORA for it on civitai.

sly wasp
#

Anyone knows a good model for Roman empire ?

buoyant jacinth
lost hornet
dire grail
#

Good night

glossy shoal
#

What would y’all suggest for running stable diffusion on a notebook? I’ve been using Colab’s paid version and it’s good but I’m wondering if I missing out on something better. Anything worth trying out?

modern jolt
#

this is fun

high quartz
#

I downloaded Stable Diffusion automatic 11111 and im wondering how can i see what version it is?

#

so i know which model sets to download from Civitai?

short tapir
#

So I keep trying to make anime characters using various models and for the most part they look amazing. But then when it gets to their face their face is look terrible.

#

Like distorted blurred and stuff. How do I fix?

#

I've tried turning CFG scale up and down. Using sampling method DPM++ 2m karass

fervent thunder
delicate iron
#

super quick question, is there a way to make SD make multiple pieces? I want to leave my house and let it make like 10 pieces while im gone
or do i have to keep pressing generate after each piece is done?

past edge
#

Anyone know the gradio python code for a small scrolling box with numbers? Working on a personal project. It should look like gr.Scrollbox() or some shit like that

past edge
short tapir
# fervent thunder What's your prompt, and model info

anything v3, Elysium anime v2, and orangecocoa5050 mix, those are the models iv tried.. Positive Prompt: Male, full body, cyberpunk, city in background, neon lights, neon signs, white t shirt, mexican, short hair, Negative Prompt: (loli), lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, bad-hands-5, (bad_prompt_version2:0.7), motion lines, motion blur, bed, (vaginal_penetration), bed blurry face, blurry eyes,

fervent thunder
delicate iron
delicate iron
delicate iron
fervent thunder
short tapir
short tapir
#

its text to image thats having the issue

past edge
#

subjects that are far away almost always get distorted faces

#

u gotta do laczos resize then inpaint

#

to get something decent

delicate iron
fervent thunder
fervent thunder
#

lmao

delicate iron
short tapir
#

like you mean add that to my negative prompt? or just use that as negative?

delicate iron
fervent thunder
#

use that as your negative

#

to test

short tapir
#

trying now

fervent thunder
#

also anime models tend to not understand "mexican"

delicate iron
#

oh yah, you need tons of negatives, i have like a gigantic list of negatives

fervent thunder
#

Try to keep the positive prompt organized

delicate iron
#

i dmed you a list of negatives u can use

#

its too long to send here lol

short tapir
#

ty

vestal dew
#

please send me those negatives

delicate iron
#

done

fervent thunder
short tapir
fervent thunder
#

put 1boy instead of male

delicate iron
#

if you say "male" it wont understand

short tapir
#

it understands when io say female tho

delicate iron
#

how donbooru refers to males is "1boy"

#

but it will understand much better if you speak its language

short tapir
#

i need to pull a british empire and make it learn

delicate iron
#

heres the site

short tapir
#

so instead of "mexican" should i but tan skin?

delicate iron
#

hm

#

go to the site

#

find a piece that u like and that is similar to what you want

#

and follow those tags

short tapir
#

i did one ealier and used "swiss" and it made a blue eyed blonde girl

#

i just figured it understood nationality's

past edge
#

yes it certainly does

#

I like putting in a random country's model to see what comes up

delicate iron
#

it understands it as long as its in donbooru

#

at least thats my understanding of it

#

like, "canada" for example only has 73 results in donbooru, so the ai wont know well what it is. Compared to like " blonde hair", that has 1.1M, so it will have a much better idea

high quartz
#

I just got into this stuff and the prompts and negative prompts are something I’m still working on

delicate iron
#

ill just send it here

#

(low quality, worst quality:1.4), (bad anatomy), (inaccurate limb:1.2),bad composition, inaccurate eyes, robot_ears, extra digit, fewer digits,(extra arms:1.2), fat, lazy_eye, lazyeyes, lazyeyed, glasses, ((watermark, watermarks)), closed_eyes, muscles, strong, (ugly:1.3), (fused fingers), (too many fingers), (bad anatomy:1.5), (watermark:1.5), (words), letters, untracked eyes, asymmetric eyes, floating head, (logo:1.5), (bad hands:1.3), (mangled hands:1.2), (missing hands), (missing arms), backward hands, floating jewelry, unattached jewelry, floating head, doubled head, unattached head, doubled head, head in body, (misshapen body:1.1), (badly fitted headwear:1.2), floating arms, (too many arms:1.5), limbs fused with body, (facial blemish:1.5), badly fitted clothes, imperfect eyes, untracked eyes, crossed eyes, hair growing from clothes, partial faces, hair not attached to head, fat. pee, peeing, ((disfigured)), ((bad art)), ((deformed)),((extra limbs)),((b&w)), blurry, (((duplicate)

high quartz
#

Thanks!

delicate iron
#

the last word is missing two "))" sorry

#

make sure to fix that

#

"(((duplicate)))"

high quartz
#

What does the brackets do for negative prompts

past edge
#

you mean the parentheses

#

they add emphasis, its unecessary to have too many tbh

fervent thunder
past edge
#

I just use them for (collage) since that one is super important to emphasize

fervent thunder
#

@high quartz

past edge
#

prompt editing doesnt work on negative prompts

delicate iron
#

everyone ive seen uses it

past edge
#

I think you misunderstand what prompt editing is

delicate iron
#

😅

past edge
#

prompt editing is picture of a [cat: dog: 20]

#

its a weird name for it though I agree

fervent thunder
delicate iron
#

i just copied this from other people online 🤷‍♂️

#

but yah i could agree

#

its been working so far tho

fervent thunder
# 
(low quality:1.2), (depth of field, blurry, blur:1.2), (greyscale, monochrome:1.1), (morbid, mutilated, deformed, disfigured:1.2), cropped, lowres, text, jpeg artifacts, signature, watermark, username, artist name, trademark, title, cross-eye, multiple view, reference sheet,

That's my main negative

delicate iron
#

i use embeddings in my negatives too that i gotm i didnt include that because yall might not have em

fervent thunder
#

low quality might be useless too tho

#

idk =\

delicate iron
#

so, i said before, that as long as a tag is in donbooru, the ai will get what i mean

#

but is that true?

#

cause "low quality" isnt in donbooru

#

or "swiss"

fervent thunder
#

yea low quality I copied somewhere and never though about it, now I think it's kinda dumb 😄

delicate iron
#

yah, not judging tho, i have a ton of dumb ones

#

but does the ai only undestand donbooru language?

fervent thunder
#

No idea

delicate iron
#

dam

fervent thunder
#

I think custom models might know more then that, but that's just a thought

delicate iron
#

yah maybe

fervent thunder
#

Where can I look words donbooru knows?

delicate iron
#

in donbooru

fervent thunder
#

link or something?

delicate iron
#

i sent it before

#

plus google

fervent thunder
#

wait...isn't it just anime thing?

delicate iron
#

bu here

#

yes

#

this is the brains of the ai

#

this is its language

#

largest anime collection

#

thats free

#

like pictures

fervent thunder
#

yea , then SD knows more then that , it's trained on other collections too

delicate iron
#

which collections tho

#

idk

fervent thunder
#

I don't remember , there was an article with list somewhere , but probably not even full list...

delicate iron
#

i dont think either of us know, i know that youre supposed to follow tags in donbooru tho, specially if you have a certain SD VAE

#

what SD VAE do you have?

fervent thunder
#

mostly using vae-ft-mse-840000-ema-pruned.ckpt

#

Some others anime models use kl-f8-anime2, but I don't use them anymore...I don't do anime most of the time anyway

delicate iron
#

hmmmm

#

interesting

#

i dont do anime stuff too

#

but i use kl-f8-anime2

#

should i do yours you reckon?

#

vae-ft-mse-840000-ema-pruned

fervent thunder
#

You should use one recommended for model, most I use - use this one

delicate iron
#

my results arent anime-like at all tho, im getting realistic results 🤔

#

i c 🤔

fervent thunder
#

It makes colors more vibrant as far as I remember

#

your vae I mean

delicate iron
#

hmm i c

#

what is the VAE anyways

#

like

#

i get it

#

but i dont

#

like, i get how a pc works

#

but i dont know wth happens inside the cpu fully

#

thats me with the SD VAE's lol

past edge
#

im basically a pro at this shit and have never even had to touch a vae

#

ignore anyone who tells you you need to mess with them

#

you dont

delicate iron
#

dam

#

i still wanna know what it is doe

past edge
#

not important

delicate iron
#

ok well hm, im gonna go touch grass, its a nice day out, later yall

past edge
#

its 11pm

delicate iron
#

i live in japan ;-;

#

its 3 pm

fervent thunder
untold ocean
#

how art thou diffusors?

delicate iron
fervent thunder
untold ocean
#

i heard multi-controlnet helps with that

fervent thunder
#

Region control thingy?

#

or you mean controlnet with more then 1 input?

untold ocean
#

3d depth renders of hands + open pose seems to get pretty good results according to twitter

#

haven't tried it myself

fervent thunder
#

yea that should help if you have an idea what pose should look like, I mostly just let AI do it's thing , I have no idea what I want 😄

#

I just like seeing random things AI does with my prompts

delicate iron
#

some people are making more and more embeddings and lora files to put in your negative prompt to train ur SD to know better, but its still not working super well. i think soon SD will include a fix in vanila SD

untold ocean
#

I've also seen some cool ideas surrounding negative embeddings

delicate iron
#

i also just edit the hands on PS if its a really nice piece but just the fingers are messed up

delicate iron
fervent thunder
delicate iron
#

oh this guy is a DEV owo

#

praise be

#

anyways, i really gotta get going, nice talk yall, later

untold ocean
#

there are some goodies in the oven, you can find some neat stuff on twitter if you know which hashtags to look for

fervent thunder
#

interesting , interesting

past edge
#

when 1024x1024 model catlurk

#

I can taste those delicious skin details already

forest apex
#

wow. TIL you can hotplug GPU's.

dapper wraith
#

is run diffusion worth?

short tapir
#

Is anyone have any recommendations for prompt to force it to do a full body image?

#

I'm using image to image. And the image is a selfie, but I want what it makes to be a full body.

#

Iv tried full body, body shot, and stuff like that with no success

past edge
#

not gonna work with img2img if the image isnt already close to being full body

#

ur better off doing a regular full body txt2img and then inpainting

fervent thunder
#

if it's image you just want to add legs to - just outpaint it instead

short tapir
#

I have no idea how any of that works

#

All I know. Is that as long as I feed the machine Spirit spaghettios it makes happy noises

past edge
#

maybe its all just magic in the end.

#

and gpus secretly run with a little hamster spinning a wheel inside it

#

that could explain all the strange noises it emits when I crank that vram to 23.9gb. whoa..

lament steeple
short tapir
#

Well jokes on you. My PC is actually made of wood

lament steeple
#

not magic. Math

past edge
#

now thats a fire hazard if ive ever seen one

short tapir
#

Shhh it's fiine

#

I even dust it once every 3 years

#

Outside at least

past edge
#

the dust might actually protect it from fire. lol

short tapir
young ivy
#

well now, good morning flosk

past edge
#

dang its late for me

#

if anyone reads this and knows a thing or two about gradio and/or python pleaseee tag me. tomorrow I want someone to help polish my script

#

uploading to github and showcasing it on yt tomorrow as well. pretty plz. tyy

young ivy
#

was there a time when Markiplier didn't talk like he was a news anchor?

high quartz
#

Off topic but are y’all excited for RE4 Remake

near silo
#

Currently testing my first high quality multi subject LoRA

#

if this works, this is gonna be amazing

near silo
#

oh my god it WORKED

untold lava
#

Right, really want to get into SD again proper, now I have better time, but is it correct it still takes 50 different steps to install or is there a simpler process?
Mostly after several hours of following the various steps and attempting to create various custom folders in the command prompts for downloading installation files, and not being sure if I did that point of the installation correct, I ran out of time for the day and never got around to picking up continuing the installation again

sharp nebula
#

Hello hello! Former Midjourney user hopping ship to Stable Diffusion 😄
Nice to meet you all!

lament steeple
#

Welcome

sharp nebula
lament steeple
#

believe it or not, they make a difference

sharp nebula
#

Yuuup. Just fascinating to see different services having similarities like this

#

Right now I'm just enjoying the exploration phase of SD. LORA's have been a blast to play with

wheat hazel
#

I made a little experiment just now. I created a zero vector embedding (essentially a 1-vector with 0 norm), and then I tried to include that in a prompt to see if there was any effect. I was expecting no effect, but there is. any explanation ?

austere junco
#

I would like to ask why my RENDER will be blue?

sharp nebula
chrome spindle
#

How do i cancel a generation? 🙂 The interrupt and skip buttons do nothing

tawdry jay
#

Hello, I'm new here and can someone please help me how can I generate pictures here? What should I do ? Thank you Khong

lament steeple
chrome spindle
#

Aaah, ok. Thx

elder cradle
#

@rigid bough ehh, I think that you haven't deleted the message yet

rigid bough
#

Done!

delicate iron
#

hey, so im up scalling my pieces from 544x960 to 1088x1920p, with like 30 hires steps
that way i can get 1080p results
each pieces takes around 45 mins, but i end up getting 1080p results which is great and sharp

#

BUT

#

i noticed that the regular steps are great and have no issues, but as soon as it switches to hires steps and it starts upscalling, it starts adding belly buttons or adding fat folds or extra legs

#

anyone else having this issue? anybody got a fix for that?

abstract pivot
#

New sampler in A1111, UniPC

bleak matrix
#

Good morning, everyone. How is everyone's lovely Saturday?

ember hemlock
#

hey im new help me out where colud i make my art?

ember hemlock
bleak matrix
stable pawn
#

Is there any free app out that can organize images so they can be tagged, but also have face recognition?
I want to do a test with some AI models to see if the AI produce same face several time and how face recognition works on AI images.

past moss
#

Sorry if this is a stupid question, new to this. Can you keep the same model generated model i.e a person, and keep adding prompts to generate new art with the same model (person in this case) that you generate.

#

Wording is a bit confusing because models have an actually meaning in stable diffusion, I just meant the person you generated.

#

like if they were wearing a jacket, and the next one you want them to wear a jumpsuit, keeping the same model just changing the atire.

lunar scarab
#

Hey guys, do you know if there is a good model of card in 12GB? I would like to replace my RX 6600 XT to have cuda and more vram but it means taking a 3060 which is less good or a 3080 but its way more expensive

broken gull
past moss
broken gull
stable pawn
#

New sampler, UniPC, do any here know what it is good for?

zinc steppe
#

What's the current best way to train a Lora? Can you do it from the Automatic1111 web interface? I've been looking at guides online, but they're mostly about 2 months out of date, so I expect I've missed some progress.

bleak matrix
stable pawn
#

I am not 100 % sure, but Stable Diffusion is free and Open Source so it can be used on a local machine, and MidJourney is a paid service.

zinc steppe
astral goblet
zinc steppe
#

Thanks.

#

I'm really hoping I can git it working on my 8GB AMD card. I'm seeing lots of people training on low memory Nvidia cards, but they have all sorts of tricks to save memory.

rich bloom
young ivy
#

what do you think the optimal set PYTORCH_CUDA_ALLOC_CONF=max_split_size_mb: is for 8gb?

#

anyone?

zinc steppe
#

When trying to set up the training parameters for doing a Lora, the options in the automatic1111 train menu seems to be very different to those that the instructions mention. Has this menu changed a lot recently?

#

I've clearly done something wrong, it seems to be training a textual inversion rather than a Lora.

young ivy
buoyant jacinth
#

I'm still trying to figure it out for 2gb

zinc steppe
#

OK, I think I have a Lora training.

#

At least its sitting there producing 150 class images. Not sure why it needs quite that many though.

lunar oar
#

I'm getting the error "Launch errors detected: ['Incorrect version of xformers installed.']" when starting SD now. I'm wondering if someone can tell me there is a simply command I can type to update this, and do I need to be in a specific directory when I do, etc?

fervent thunder
#

never happenned to me, it just installs whatever version it needs when you add "--xformers"

zinc steppe
#

Noooo - It spends an hour making class images only to throw an out of memory when it starts training proper.

warm junco
zinc steppe
#

For some reason mine has started complaining about xformers too. That makes little sense because I'm using an AMD card with no chance of using the library.

forest trout
#

Maybe it's just me but I'm feeling like a lot of community created models are feeling quite homogeneous. They all seem to have a very similar photorealism crossed with anime artstyle.

#

Problem is that I'm finding it difficult to prompt out of this art style.

obsidian geode
#

Did anyone try control net clip vision with t2iadapter_style? it doesnt seem to be able to load the dict

#

and many people complain its not working under videos tutorials

lunar oar
zinc steppe
forest trout
#

Yeah. 1.5 feels much easier to get varied art styles, but these finetuned models that create very beautiful art seems to be locked into a very specific artstyle.

#

Like I did a little test and began removing tokens and was noticing that nothing was changing, or at least very little was changing. That just seems like a very overfitted model to me...

zinc steppe
#

So my Lora is training, but it looks like it's going to take 20 hours 😦

#

I've just got a Vega 56.

fervent thunder
#

Applies especially to realism

sleek osprey
#

whats a good lambdalabs alernative
looks like they dont support my country
for some reason

deep lintel
surreal perchBOT
#
FAQ: What is Stability AI?

Our vibrant communities consist of experts, leaders and partners across the globe. They are developing cutting-edge open AI models for Image, Language, Audio, Video, 3D and Biology.. AI by the people, for the people. Learn more here stability

strong forum
#

how do you load ckpts

#

there's nothing that lets me select it

#

what format does a file need to be to load from file it just translates to jibberish

marsh wharf
#

Hey guys, I'm making AI art generation software, and I'd like to know if anyone here has also made similar software. If any of you have, what factors went into charging your users for generating images on your website?

vast ingot
strong forum
#

okay thank you!

vast ingot
# marsh wharf Hey guys, I'm making AI art generation software, and I'd like to know if anyone ...

you'll have more answers in #1011228667659178055 I think. Some people generate the pictures on their software directly and use a token system to make it possible to pay the bills. Some others ask the user to put their API key so that you just relay to DreamStudio API, and they handle the facturation on their side, like the photoshop plugin that came out some time ago, paying its own bill by selling the plugin

coral wind
#

My parents aren’t the smartest, they are letting me take the most expensive car to Texas to see my cousin, I been driving since I was 13, I’m driving our 2022 Ford F-350 Super Duty

past moss
#

is training embedding the same as training a model?

rose aurora
#

How do you solve consistency in stable diffusion?

chrome spindle
#

Does textual inversion and embedding mean the same thing?

obsidian geode
#

yes

vestal thistle
#

hi

elder cradle
boreal cosmos
#

I want to generate cute anime cat girls, which model/LORA is most suitable for this?

wild steppe
wild steppe
warm junco
boreal cosmos
#

thanks!

#

what is a CHECKPOINT MERGE in civitai?

#

do I use it like a normal model or do I have to merge it with SD somehow?

#

sorry I'm noob

warm junco
#

Which model?

#

I think it should be like a normal model

boreal cosmos
warm junco
#

Ah yes you can use it like a normal model

wispy jasper
#

Hey guys, has anyone ever figured out how to use the upscale model 2.0 on 24GB ram?

#

Is this a thing yet?

#

Also whats the Tencent body pose pth model adapter?

vestal dew
#

with all the innovations with SD, we still can't get good hands and feet?

vale steeple
#

Thought of course you would need quite some luck on specific hands that doesn't training specifically in hands or/and feet

vestal dew
#

right, so it still isn't solved

thorn elk
#

hey, anyone can tell me about warpfusion? this has eluded me

tawny quail
forest brook
peak orchid
lethal sinew
#

Hi
I have a problem with SD.
The software don´t generate images when i change the size settings.
I need to stay in the default size "512x512".
Anyone knows why this happen?

Thanks.

silver bronze
#

does anyone know how to make SD generate arts faster?

vestal dew
#

get a better video card

#

fewer steps

#

smaller resolution

#

--xformers

silver bronze
vestal dew
#

use google collab

silver bronze
#

whats that?

vestal dew
#

you can run it on google with a GPU

#

I dont have links. Look up Stable Diffusion notebooks

silver bronze
#

how to use tho

silver bronze
vestal dew
#

in your web-user.bat

#

COMMANDLINE_ARGS= --xformers

#

but I don't think that works without a GPU

silver bronze
#

it doesnt work

#

ye

#

so theres no way for me to make it faster?

vestal dew
#

fewer steps

#

smaller resolution

silver bronze
#

wdym

vestal dew
#

when you generate, there is a field for steps

#

there are fields for resolution

#

now I am outta here

hallow plaza
#

Whats the key to persistent characters?

fervent thunder
#

Lora?

hallow plaza
fervent thunder
#

thinking it should, but probablt will take some attempts to get decent result

past edge
#

nah g forget lora

#

do dreambooth

hallow plaza
past edge
#

essentially what u need to do is 1) grab a bunch of pix from the character u want to make consistent

#
  1. train a model with dreambooth using those pix
#
  1. now u have a model with a keyword u can use to summon that consistent character. boom
hallow plaza
past edge
#

how many sprites u got

hallow plaza
#

44

past edge
#

yeh it should work good. If theyre like 8-bit style I recommend training a model thats made for that particular style

#

you would get better results than if u trained the base 1.5 model

hallow plaza
#

A dreambooth model?

past edge
#

no, any model will do

#

u can look some up in civitai

#

u download the model then select it when training with dreambooth

hallow plaza
#

Oh i see

#

I'll look for one

past edge
#

also if the resolution isnt 512 make sure you specify it in the training

hallow plaza
#

May be it would be easier to upscale to 512 then downscale back?

past edge
#

Yessiiir

#

as long as they dont look like pooh when upscaled

#

garbage in = garbage out

hallow plaza
#

Is there a good guide on how to train it if I had like 500 sprites?

past edge
#

is it the same sprite?

#

like the same character

#

cuz theres deminishing returns after a certain point

hallow plaza
#

no different characters

#

about the same amount for each

past edge
#

ooh. yeh if u wanna train a whole bunch of subjects at once u can

#

uhhh let me remember how

#

oh yeah. on a1111 dreambooth theres an option to use a .json concept list

hallow plaza
#

Is there a good guide for that? I just reinstalled windows so I would need to set it up from scratch.

past edge
#

yeh hol up

hallow plaza
#

Then I find a training guide?

tacit island
#

Greetings guys, are these people using you guys' technology?
https://twitter.com/Ameterasu1337
this is insanely inappropriate, if they are using any stable fusion tech, you guys need to shut this down I think

severe zealot
#

help i am noob: RuntimeError: Couldn't install requirements for Web UI.
Command: "D:\Ai\Stable Diffusion\stable-diffusion-webui\venv\Scripts\python.exe" -m pip install -r D:\Ai\Stable Diffusion\stable-diffusion-webui\requirements_versions.txt --prefer-binary
Error code: 1
stdout: <empty>
stderr: ERROR: Invalid requirement: 'Diffusion\stable-diffusion-webui\requirements_versions.txt'
Hint: It looks like a path. File 'Diffusion\stable-diffusion-webui\requirements_versions.txt' does not exist.

fervent thunder
#

yea it's not first person who did this and not last one...
Is this against twitter rules or something? Is there a reason to report? if you even can report on twitter , no idea, not using it

#

They'll probably just make new account anyway...if things like that popup consistently it means people actually pay for it ,reporting won't do a thing

hallow plaza
dreamy igloo
#

I have a Boomer friend, who is so convinced that he can beat artists now that he reads like half a paragraph of news about AI art. He goes to do like 3 word prompts, without a care or knowledge about negative prompts. Ends up spewing out a bunch of crap, gets frustrated in less than 5 minutes.

#

He's the kind of guy born with a silver spoon in his mouth, always have a group of people taking care of his life. Who lives a comfortable life, being stupid, just had capital to hire people to take care of his business and his life. I can't tell if he is really skilled at hiring just the right people or just damn lucky. His email is always like his a.i. art prompts. 3-5 words. If he did not have a secretary, he would not have a business. He gave up on the idea of beating artist after I point to him about prompting words, learning about art styles, structuring prompts

#

He always think he is a big guy...he is giving up on a.i. art FFS....after like 5 minutes. And before he tried, he thought he can beat all artists.

fervent thunder
#

reminds me of Elon lol

dreamy igloo
#

I only have him as a friend as a line of credit who charge no interest lol

#

He likes people showing him cool stuff, and hangs around his wealth.
If I show him that auto-nudify app...god forbid our common female friend's vacation photos will be ruined by him. And I will bear the blame for telling him about it lol

past edge
#

once u get familiar with it u can experiment with custom models

hallow plaza
#

I used 1.5 in the past. Just dipped out when 2.0 came out

past edge
#

yeh 2.0 is poop

#

but custom 1.5 models are where its at

past edge
#

thats much more simplified for the commonfolk

dreamy igloo
hallow plaza
dreamy igloo
#

many of his correspondences were written by his secretaries.

past edge
#

No, you just install it from within a1111

#

let me see if I can find the original videotut

#

ehh its too convoluted to find a video because dreambooth has been thru many stages

#

just google install dreambooth from a1111 and u will get the steps

hallow plaza
#

will do

past edge
#

then the video I linked from secourses explains how to use it

#

if it werent for dreambooth I woulda left AI art ages ago

#

dreambooth is straight up magic

hallow plaza
#

i am trying to get auto1111 to work.

past edge
#

control net is also insane

dreamy igloo
#

oh yea, speaking of dreambooth, many colab notebooks got rocked like a couple of days ago when google upgraded python. I haven't loaded up the old dreambooth notebook, nor looked for an update. I wonder if lastben fixed the dreambooth besides fixing the usual fastauto1111

past edge
#

man fuck notebooks

#

I spent so much money on that trying to troubleshoot stuff

dreamy igloo
#

Colab notebooks are how I run everything. Can't buy rtx40xx or 30xx right now.

past edge
#

Link to the notebook?

#

I assume you pay for google colab?

dreamy igloo
#

I haven't checked back to see if they updated. Your conversation about dreambooth reminded me of that. I don't pay for colab

past edge
#

colab isnt free anymore

#

they give u like 1hr of gpu and then its a paywall

#

google colab is a ripoff. if im gonna use cloud gpu I would use runpod

dreamy igloo
#

I still get like 3 hours or so per day with free account.

past edge
#

the issue is if you run out in the middle of a training

#

but I guess for some folks it works

dreamy igloo
#

yes. that is what I watch out for. I schedule a day just to run training.

#

I made several checkpoints that way

#

Plus friends and family help out. They let me run the colab stuff thru them.

#

I hear that even paying google customers can "run out of time". So yes, runpod, which I've looked into sounds good. If I need it bad enough, that is probably the first place I'd go to

past edge
#

the issue u run into is moving checkpoints in and out.

#

these are like 4gb and upload/dl speeds usually suck ass for cloud

#

and they count towards ur gpu hours somehow r0fl.

#

its too much frustration but hey for something free its a good way to test the waters.

#

once u get an rtx3090 you'll never look back. plus its only a matter of time before AI music hits the scene I cant wait for that shit

dreamy igloo
hallow plaza
#

For Dreambooth if I have several folders to train images, can I point to a root folder?

past edge
#

not that I know of

#

maybe with some coding u can do that but I dont fink db knows how to do it by itself

hallow plaza
#

Thanks. I'm testing it now

#

OK well I got horrible results. I'll come back another night once I got some sleep. Its been fun, and thanks for the help.

proud nova
#

Question: Is it possible to train LoRA with specific artist style?

lament steeple
#

Yes

proud nova
#

Perfect

young hemlock
#

What is the extension name that let you paste the image and it will guess the prompt and everything?

past moss
#

Can you combine 2 character specific models into 1 model, allowing you to generate both in one image? i.e you have a cat model and a dog model, and you combine to make cat and dog images.

lament steeple
lapis rain
#

hello all, is it possible to setup stable diffusion server from my local that can be shared and runned with any device?

unkempt hatch
#

Launching launch.py... ################################################################ python3: can't open file '/home/q/stable-diffusion-webui/launch.py': [Errno 2] No such file or directory

anyone know about it?

fervent thunder
#

hi

#

is it possible to create a controlnet that would transfer the facial features onto the resulting character?

#

for example, you give it an photo or an image containing a face, and it would attempt to preserve the face into the whatever character you're making?

#

I noticed that midjourney img2img preserves the facial features

#

so I'm wondering if it's possible with stable diffusion without requiring an entire dataset to train the model and 15GB of VRAM to train the model

#

also I would like to have one for just the face, and one for consistent character entirely. giving the AI a character concept sheet

#

you think controlnets can be used for this?

signal dome
fervent thunder
#

important matters sir

#

I'm not looking for support, just asking the competent people about the possibilities

#

what do you think of open assistant?

fervent thunder
#

saves money for big corp

#

saves money for small businesses

#

hope you have a government that has a social safety net

quaint tapir
#

Any alternatives to Colab Pro that are in the same price range that give you more run time?

boreal cosmos
#

buying a used GPU from ebay/craighlist/offerup?

quaint tapir
#

I have a 750ti but from what I understand it’s limited to like 512x512 @ ~3mins per image

fervent thunder
#

Ebay is pretty good from the buyers standpoint , just get a trusted seller

boreal cosmos
buoyant jacinth
quaint tapir
#

I think I figured out the issue. This Dreamlike notebook just constantly runs and never times out

tall cargo
#

umm

#

Hello

#

?

boreal cosmos
#

hai

wild steppe
fading hill
#

prompt

#

Flyer gráfico, con una ilustración 2d vectorial de una mujer gritando, colores naranja, negro y blanco.

surreal perchBOT
#

@fading hill

FAQ: How do I generate images?

Currently there is no bot on the server that generates images. However, there are plenty of other ways such as the official https://beta.dreamstudio.ai/ website or running Stable Diffusion locally using your own system resources! Check out #1080946152318443610 for more details! You can also stop by #1025467151206854736 for any issues you experience while using the website or #🤝｜tech-support for any problems you encounter while installing it locally!

chrome spindle
#

Do you guys have a tip for a good resource for creating your first embedding?

boreal cosmos
#

btw why don't you guys add the Stable Horde bot to this server?

#

it can generate SD images for completely free

fluid steeple
#

hey guys! If I wanted to send someone right now my model trained with dreambooth exactly what files would I send?

vast ingot
#

You would preferably send the .safetensors. alternatively, the .ckpt. lastly, you can also zip the diffusers and send those if you wanted the person to able to continue training it

#

You can also just open a model page on hugging face and host it there, convert to safetensors there too

fluid steeple
#

and would they be able to run the .safetensors file with the hugging face diffusers library

vast ingot
#

I put a converter link in my guide

fluid steeple
fluid steeple
vast ingot
#

If they use the original diffusers librairy

#

You need to send them the diffusers

fluid steeple
fluid steeple
vast ingot
#

Also #🤝｜tech-support could be of great help, there has been quite a bit of développement going around

fluid steeple
#

Because I have found online some that will convert ckpt to diffusers

#

but no way to run them without converting

vast ingot
#

Even in the diffusers library, there are 2 scripts to convert from ckpt to diffusers and the other way around

#

Most people around here use automatic or invokeai. Both work with ckpt mainly, and have a command line approach as well as a webui approach

#

I almost have only ckpt here, outside of my training environment

fluid steeple
#

saying sending over the diffusers because you mentioned it previously but I might have used it wrong since I have no idea what that means lol

vast ingot
#

Automatic and invokeai are also python though. But yes if you want to stick to the diffusers library.
The easiest way then would be to host your diffusers on hugging face, since diffusers includes an auto download for models from there

fluid steeple
#

Perfect, thank you so much

#

I really appreciate the help. Have a great day

vast ingot
#

You too !

hallow plaza
#

Are there any models for making sprites?

north barn
#

I am new to this chat. Can you help me with steps on how do I create image from text

hallow plaza
vast ingot
astral goblet
#

Llama is stolen IP afaik. I wouldn't touch it with a 100 foot pole. Trying to move away from all NAI merges too for the same reason

#

Actually, i might have an urge to use it. I want to weaponize it against spammers. See, spammers and scammers are very much going to use this stuff. Particularly romance scams. For now though it's computationally expensive. I think that if you created a chatbot that looked for other chat bots, it could engage them in conversations that would destroy processing power. Maybe do really complex queries that would do a zip bomb of processing on the nefarious bot's host.

#

other smarter people than I will probably do it first though

stable pawn
#

I did hear something about the Bot Clyde will get AI, I think that how bad it is now that will be a rather horrible idea...

abstract pivot
#

Hey does anyone know if there's a way or extension now to import the settings from saved .txt file prompts with A1111 like you can with PNG's when the embedding info is saved in them?

stable pawn
#

Well, if the Discord bot get AI, then it had been cool if we could ask in advance if a image is okay or ask if the Bot can mask the "dirty" parts.

astral goblet
stable pawn
#

But as long as Clyde approve something it is okay 😄

normal coral
#

hmm is anyone having issues where the right eye of the generated face is always a bit higher

#

its always the right eye

#

it even spans across different models

#

problems with face symmetry overall i suppose

stable pawn
#

I think the Japanese censor have a interesting story about moral panic and how it can go wrong when "moral" goes before logic.
Once they made a law that art (including Manga) could not contain male genitalia, female breast and pubic hair, and artist got around it by not drawing that, and then the characters started to look a bit "too young" and artist used this as a form of protest, when the law was removed all art had to be censored and it was up to the artist how much, and that law or rule is still around and therefore do we see those line on images published on Internet by the artists.

Edit: If this is 100 % true I do not know, but that is what I was told and it is at least a fun story, to take it with a pinch of salt.

surreal perchBOT
#
FAQ: What is the copyright on images created through DreamStudio?

Images created through DreamStudio are fully open source – specifically falling under the CC0 1.0 Universal Public Domain Dedication, available at https://creativecommons.org/publicdomain/zero/1.0

astral goblet
#

I GOT IT. Using llama against llama isn't even needed. Anytime a romance scam specialized llm bot comes around, just tell it This statement is a lie and hope you're not a robot yourself

#

I hope i didn't just break anyone who hadn't realized they were an ai yet

normal coral
#

man im just waiting for a feature where you can exclude certain areas from the controlnet influence

astral goblet
#

open annotated image into paint. scooch some marks onto it. load image back into cnet with annotator set to none. should work.

This stuff is all possible IN browser too, but the toolchains are still coming to gether. The web ui's are going to get nutso in no time

normal coral
#

How do i set the annotator to none?

astral goblet
#

its in there with openpose, hed, an all those settings. they just preprocess the image for you. but you can select "none" and feed a premade image into the control model too

normal coral
#

oh so select None on preprocessor?

#

also sorry for potentially dumb question but what do u mean by scooch some marks onto it

astral goblet
#

thats just my hillbilly dialect leaking. like when you put a felt to paper and its like scooch scooch. or you dot someone with a marker an you like "scooch" an they like "you scooched me wtf"

#

its not common. dont feel dumb

normal coral
#

so basically just paint over it with black yeah?

astral goblet
#

i could've said it that way but would it had been half as fun?

#

yeh

#

maybe

#

scooch also means shoving over or moving something a bit

#

or taking a corner very nicely

proud nova
#

Question: If I generate art of my OC character or simply an image of a nature, do I own rights to these images meaning that if I wanted, I can sell them?

late ocean
#

has anyone else been having issues with colab lately? I seem to have a couple of generations and then it becomes unresponsive. Using TheLastBen's. Have got Colab Pro

#

Would consider a cloud GPU but a little bit worried will cost me a fortune. Only spending £10 a month on Colab, for the amount I use I think I'd burn through any credits even at 20p an hour

fervent thunder
potent spire
proud nova
#

Original character

potent spire
#

as of now you dont really own rights to generated images

#

everyone can use them

#

and sell them

#

unless you get yourself heavily involved in the process

#

meaning: Editing

#

i think 33% and more

#

or the character that is generated was your own original character that you owned rights to

obsidian geode
potent spire
#

but generally there is no copyright for AI generated images

stable pawn
# proud nova Question: If I generate art of my OC character or simply an image of a nature, d...

For now the law in many countries are written so that it need to be human creative work behind something to be able to be copyrighted and it can not be too simple, like you can not find a rock outside and copyright it, but if you form it to a sculpture you can do so.
For now it have not been tested and that is why some companies sue some creators of the AI, they want to see where the law set the borders.
To be on the sure side you can do post work on the AI, call the AI image for "inspirational sketch" and then do work to xix it up, add text and so, now it is creative work and you may be able to get that Copyright.

potent spire
#

from february

stable pawn
proud nova
#

Thank you for explanation. I am interested in this since I started with SD so some knowledge is good

potent spire
#

im curious if any of the people from AI art community will try out the Glaze thing

#

should be released next week

stable pawn
#

It is a tricky question. As a former digital artist (I got stroke and lost interest, and need to relearn) I could make a image in under 20 seconds, I had templates and layers I could use with things pre made, that had fall under "It not creative work", but all hat was to layout the image and after that I started to work.
Now I use AI, but few images are random chance, the one I post here are, but not the one I use, and now with controlNet I can design and with Embeddings and LoRA I have my old style back. So the question is whyit so hard to be seen as an artist if you can make images you want with AI.

#

But I remember (Im old) when some looked down on digital art and saw digital artists as "art cheaters".

potent spire
#

because its different from actually doing artworks

#

someone who did do art before may still be considered as one, but not because of AI art generators

stable pawn
#

The "It is different now" and "But now it go much faster" and "Now it can be replicated so often that the original will have no value". Arguments from when when records came, when radio was invented, when TV become popular and with tape recording and VHS copy then with Internet and filesharing. AI is on in the line of scary things that will kill art, culture and change the world forever.
Me, I do not care for the old world in the past, it was rather crappy and it had the 80th.

potent spire
#

do you consider the world before AI art existed as "old world"?

stable pawn
potent spire
#

i think there wont be a compromise on the question if people who use GAI programs should be considered as artists

#

an prompter will more likely say he is an artists, artists on the other hand more likely not

#

the question of definition of what is an artist is one of the smaller ones compared to other issues artists and prompts have between each other

stable pawn
#

I think I was an artist like 3 years ago, but I did not see me as an artist but as a employed designer, then things happen and I could not paint as I used to. I do recover I can now paint, but not as before, so I took some of my old images and made a "Embedding" to be able to use my old style and old characters. It would be a shame if I can not longer call them "mine art" for I use a tool.

marsh pewter
#

Guys does out there are any paid or free service where you can train lora? My computer have 4gb GPU and can't train

potent spire
#

and to be brutally honest, disability doesnt matter for the question of such things unless the person was an artist before and actually has the skills

stable pawn
potent spire
#

people even go so far that they dont even give effort for prompts and ideas but instead search for prompt generators and ChatGPT to do the work for them

#

i dunno how i can call such people artists

#

i could call you for example an artist, but not because of AI art generators

#

and since you are recovering and relearning which i think means you are going into the work again (?) i can definitely consider yourself as artist

#

if you practice etc.

#

but this is not the situation with everyone or general "population" of people of the community

#

i know that people often bring disabilities up but as i said, i see no relevance in disability in such a question in situation as it usually is here

stable pawn
#

Well, what we have just now is new, now we play, test and learn. Some try to make a computer to do everything, self I do art I like and for I need to create and go out take photos, how it will work in the future, that is where we come in and that is why we are first to use AI and create with it.

potent spire
#

yeah but i compare it to a situation where i order a customised artwork for myself from some skilled artist and i give him the directions of what i want on the canvas

#

i dont see myself as the artist behind the artwork, i didnt even touch it. I gave the "orders" to the artists to do it

#

but i can be AN artist

#

but not THE artist because of that one artwork i didnt even touch

stable pawn
#

Well, there is some thing AI cant do just now, paint with oil on canvas, do slightly illegal street art, paint a church ceiling based on a interpretation of a story and so on, or change things after a customers wish, so design and art will keep exist, it just that more folk now can do their own art.

#

Just now I can not understand what folk want, even "hyper realistic" is used as a prompt for folk think there is a extra level of realism, as an artist you can say that your pictures will now be more real for you add whats trending on artstation to it. Artist that paint will still be around for a long time.

potent spire
#

to be clear i used AI art generators myself and will use again, but neither MJ nor SD or dunno stuff like DeviantArt one

#

and im not going to judge especially disabled people for using the tech alone

#

nor am i scared of AI itself, im enthusiastic about it even

potent spire
stable pawn
#

I think AI is cool and will help folk get art they want, I think it may threaten mass produced art thet been sold on IKEA and that folk can make art is not bad. There will always be some that is being better and can do great stuff, and we will still pay to get art and work they do.

late ocean
#

I do find people are quick to shoot down the argument of AI helping with disability and accessibility. Maybe because their views are that someone to be considered "disabled" it has to be visible and severe e.g. paraplegia. When actually there are a lot of disabilities which are hidden. Things like ADHD, dyspraxia, dyslexia are really common, and it is mindblowing how much more accessible AI is making these things.

potent spire
#

i mean we just need to look at Photoshop

stable pawn
#

Yes, dyslexia is fun, not long ago folk did not write that much and they could take 2 to 3 days to write a letter. So it was hard to find folk that had trouble with spelling and letters, now a Dyslexic person write mush more per day than normal folk did in the 60th and so.

#

And remember, good spelling is when folk can understand, not if it is perfectly written according to academia 😄

late ocean
#

ChatGPT and helping restructure/correct a dyslexic's writing is great. Still under pressure to write as much normal folk, but it takes so much longer and is exhausting. As for dyspraxia fine motor control is a weakness so creating art took ages. Still have a lot of creativity with things like composition, colour choice etc. But it was really difficult to produce the required output especially when studying fine art

young ivy
# late ocean ChatGPT and helping restructure/correct a dyslexic's writing is great. Still und...

digital art would be a big savior for that no? I mean since you can undo/redo and reshape, skew, distort, prespective, create lines, bend them and all that without actually drawing them entirely. Then for writing, having autocorrection that learns from your quirks. I dunno what it's like to have either. I have other difficulties that manifest in other ways but, software has come so far now to help accessibility

hallow plaza
#

Anyone seen a model that's for making sprites?

#

I want to do something like this.

eager shore
#

Anyone fool around with elevenlabs?

fervent thunder
#

But that might be too aggressive, eh...

wide juniper
#

Is it just me or is latent couple difficult to trigger?

narrow sluice
#

It's a bit narrow-minded or frankly, elitist to gatekeep expression of art for people who are strictly artists

#

Yes, mass production will take the uniqueness and the higher market value of specific artists, but did art always have to be specifically all based on monetary value, and not personal values?

#

Yes, there's folks who probably draw atrociously, no matter what kind of training they get or how much effort they spend on it, do we get to tell them, nah, you suck so you should never ever be an artist?

frank latch
#

Hi everyone. newbie here. say I want to train a checkpoint model mainly for product design, and product prototypes, like a motorbike, cars, Bluetooth speakers, gaming consoles, etc. how should I proceed with this? thanks a lot. I have trained art-style Lora models before.

wheat island
#

Imagine if the medical tricorder (a device from Star Trek that does maybe 99% of what modern doctors do) is suddenly invented today. Doctors could use this argument to defend their livelihoods, but they lose sight of the fact that doctors don’t exist because society needs to employ them. They exist because we have a problem of people getting sick… if more sick people can be helped then great! That is an advancement for society because more lives are saved (as opposed to more doctors being employed), and then simply the standard for what doctors are expected to do is raised to a higher level, since someone is still expected to operate tricorders.

#

Similarly, artists exist for their output for society. If these AI models can truly fulfill the needs of society that artists currently output (that is debatable), then that simply raises the bar for what artists are expected to output. But it doesn’t change the fact that we only care about the output for society (which can never be truly harmed by advancements such as this because if someone can not outperform the AI then they are redundant), not the fact that artists exist.

vestal dew
#

someone still has to come up with the idea

wheat island
#

yep

vestal dew
#

I was posting some images, and then I found someone else who had posted all the same types of images practically in the same order

#

So then I came up with a new idea

#

AI didn't generate the idea

wheat island
#

yeah

narrow sluice
# wheat island Similarly, artists exist for their output for society. If these AI models can tr...

I don't disagree on how AI is just a glorified input-output black box, but I don't appreciate using elitist or gatekeeping arguments from stopping people from using the tech.
Even if I had AI vs an artist, I still want to pay certain artists to draw stuff, because I like them and want them.
There's a probably a debate on whenever huge tech advancements happen, and whether you need to protect the people who are displaced from said tech's existence.
If I want to be greatly cynical, AI existing and its growth is entirely tied to capitalism and its needs for pure efficiency.
In any other system, being efficient for art is useless because art is priceless.
But in this modern society, everything has a price, even you.
The bigger question to ask is, why is the system so broken something like AI existing can be considered a threat to people who use it?

wheat island
#

exactly!

#

UBI is a possible solution

#

because at the end of the day, artistic value is a social construct

#

art is priceless but it's not valueless

rugged perch
#

has stable foundation given up on training new models?

sage lotus
#

computer science, the study of computers and computing, including their theoretical and algorithmic foundations, hardware and software, ...

fervent thunder
#

I think it slowed down cause of law suits and shit around AI

#

we have plenty of models to play around from community anyway, tons of embeddings , lora , extensions ...

#

My model folder is 110GB lol

proud nova
#

That's a lot of models

elder cradle
#

More and more people are hating on Emad now. Not Anti-AI, rather hating him for teasing people products (cough cough Deepfloyd, SDXL), but doesn't actually release them

past moss
#

If I was using webui and I switch over to colab, do I have to set everything back up again, like all my models and stuff?

fervent thunder
#

Hello, everyone

#

I'm Yevhen from Germany

#

I'm web and mobile app developer

abstract pivot
#

Hey, does anyone know of any extensions or ways to save an entire current setup in A1111, like say you have a bunch of settings for a prompt inc controlnet and other extensions, something that can save the entire configuration and give you the ability to re-load it as is?

thorny sorrel
vast ingot
#

Currently, how it's implemented, it's each extension's job... They can add "output text" to the output of the picture, that is saved in the picture metadata.
Later, when you put that picture into "PNG info" tab, it retrieves all the parameters that were used for that photo.
You can then click the "send to txt2img". Or copy all the params, paste them as if they were a prompt and click the blue arrow button that is just under Generate. This will apply all the parameters in the different fields.

But once again there, it's on the extension to implement loading of the parameters, and it's not always done

#

This isn't the place for such advertising

mighty gazelle
#

has anyone managed to run blip 2 with auto 1111?

#

how is the experience?

vestal dew
#

blip?

mighty gazelle
#

blip caption

quick island
#

do you think, it will be possible to create ai art again sometime here?

mild grove
#

Guys, how to allocate more memory for Deforum, I have a memory shortage error, I have a gtx1650

fervent thunder
#

Hi! I hope I am allowed to post this here.
As part of my Bachelor's degree program, I am conducting a brief survey about image generation model prompts and I would like to invite you to participate.
The survey will take no more than 5 minutes to complete and will consist of a few questions about your experiences and perceptions about image generation models. Your responses will be completely anonymous and will be used for research purposes only.
To participate in the survey, please click on the following link: https://docs.google.com/forms/d/e/1FAIpQLSfiZF-iGPaix6o4KCFEOp76c97fB9e30LLJseZC2crzcN3fcw/viewform?usp=sf_link.
Thank you in advance.

velvet plover
#

Does anyone know some good models to deblur (restore) images? I'm basically trying to deblur an image I want to use for controlnet since some models don't detect the outlines due to the image quality. The original image isn't small, it's just blurred.

vast ingot
velvet plover
#

Thanks for the recommendation

merry thicket
#

Does anyone know how to install controlNet on mac m1? Is there a guide anywhere?