💬｜general-chat

[narrow jackal]

These laws come from a logical and not an emotional sense.

[slate crescent]

absolutely

[narrow jackal]

As I said copy is not steal

[slate crescent]

I can't shake you enough to get through that this semantic distinction does not matter to anything we are saying

[narrow jackal]

But I consider it "immoral" but not unethical.

[narrow jackal]

Anyways, Have a nice night

[fervent thunder]

do you guys add full/mid body shots when doing embeds or just focus on the face?

[orchid wadi]

Depends on what I am training.

If it is a style I am going for, I add as many differnt kinds I can.

And different skin colours, ages and genders. Just to make sure it doesn't train the subject.

[fervent thunder]

trying out different actor/heros like affleck/batman evans/cap

[orchid wadi]

What is the concept you are trying to train?

[fervent thunder]

ben affleck batman

wasnt really crazy about the movies he's been in, but ive always loved his costume lol

[orchid wadi]

If you want Ben Affleck that is one concept, then batman is another. So I'd train 2 embeddings.

One embedding in one concept. And you can bet your ass there are more batmans than Benny in the model.

[fervent thunder]

well I was just going to use screen grabs from Batman V Superman and Justice League

of affleck in costume

[fervent thunder]

how many steps do you do per image? 100?

[weak citrus]

yep this it this is my place

[orchid wadi]

I count only epochs. I usually get results around 500 epochs and finish around 1000-2000.

If I go past 2000. Then I am doing some funky advanced level tricks with embedding inspector

[fervent thunder]

doesnt the number of pictures you use influence how many steps you do?

or epochs whatever

[orchid wadi]

No... Not really.

I usually have 4-12

Sometimes 20

Never really more than that

If I need more. Then I use something like DB for concept training only. Does the trick

Don't touch the Unet or Textencoder at all. But I need to retrain every embedding for that then. So I rarely bother.

[fervent thunder]

what do you add for the subeject prompts?

just the name of the embed?

[orchid wadi]

The last one I did this morning was just 500 pictures of people wearing underwear. All ages, sizes, races. Why? Because I wanted to correct SEO/Clickbait and make it so that I can have cleaner bases to work from.

The name can be whatever, it can be renamed after.

Init text.. Well sometimes I add some, sometimes I use 0 vector as in leave it blank.

Depends entirely what I am doing.

[astral goblet]

you're describing the anarchist side of libertarianism.

[orchid wadi]

If it is doable with embeddings; I can do it in about 3-4 different ways.

[fervent thunder]

so if i get 10 photos of batfleck what do you suggest i do for init text and vectors

[orchid wadi]

There is no such thing as "nature" or natural order.

Humans broke that since we invented farming.

[astral goblet]

it's a metaphor that some people runaway with

[orchid wadi]

Well what is it you try to achieve?

[fervent thunder]

told ya, ben affleck as batman

[astral goblet]

natural order is as simple as supply and demand. so much food supply? so much population.

[orchid wadi]

My main subjects in high school was Philosophy and Religion, I have read plenty about this.

[orchid wadi]

Explain to me then why the fuck there are people dying of obesity while others die of malnutrition. Even in the same countries.

[astral goblet]

that's not natural and is a meme caused by sugar marketing

humans and our ideas. sure are strange

[fervent thunder]

Americans love their corn syrup. So do Brits lol

[orchid wadi]

Ok so... I wouldn't do those at the same time... But considering that you want to try. Then the concept could be "ben affleck batman" since that is probably in the dataset as it's own.

[narrow jackal]

It's called real world some places are less privileged geographically

[orchid wadi]

Like living down town of one the wealthiest nations wealthiest cities and still being in food desert, but instead of sand it is concrete.

So natural that is.

[astral goblet]

its a blip in existence

[orchid wadi]

Right so "ben affleck batman" then tag on few other qualifiers for init text. Then that token amount as vectors should do the trick. Or just like... 10 vectors and see where it takes you.

[narrow jackal]

In nature you are not born perfect and full of riches, but of poverty and misery.

[fervent thunder]

do you use auto? should i leave it on sampling method 'once' or do determentisic

[narrow jackal]

What you consider good and evil came from the evolution of human thought

[orchid wadi]

Humans aren't born capable of walking or speaking. But I wouldn't call a drunk crawling on the street making random grunts as achieving "natural state of humanity".

[orchid wadi]

Yeah. And shouldn't we be evolving as a collective society? Instead of regressing to a state where someone with bigger stick can take all your shit?

[astral goblet]

evolution doesn't know direction

[orchid wadi]

You think we would have things like painkillers or antibiotics in "natural world".

[astral goblet]

why would it?

nothing natural about directed evolution i assure you

[orchid wadi]

Yes it does. Evolution by definition is to change according to environment's pressure.

[astral goblet]

there's no direction to that

[narrow jackal]

The human being is selfish by nature and if you don't agree, you've already proved my point by the simple fact that your ego acts

[astral goblet]

there's no end goal except propegation

[orchid wadi]

If the environment is changing according to a trend. yes there is.

[astral goblet]

evolution is just what manages to stick around

[winged wedge]

Does anyone know how to configure a good computer to run Stable diffusion

[orchid wadi]

Well you don't need life for that. You need chemistry. We have self-replicating chemistry. I wouldn't call that state of nature.

[orchid wadi]

No I don't think human is selfish by nature. I think that is a dickhead justification to cruelty. Humans have at a neurological level the capacity for sympathy and empathy.

[astral goblet]

life is chemistry that keeps propegating and suceeding. that's all

[orchid wadi]

If I hit my hand with an hammer, and show you a video of it. You will vince and pull your hand in.

[astral goblet]

its this thing humans were able to do called language that started off the evolution of ideas

[fervent thunder]

let's test it

j/k

[astral goblet]

but even that is survival of whatever survives

[narrow jackal]

Selfishness is not linked to being bad, but to your advantage. An example of helping someone, I help because it makes me feel good.

*selfishness

[orchid wadi]

Humans have to do some rather fucking complicated things to justify violence towards another humans. Armies have for milennia researched desensitation and othering of the enemy. But at the end of the day you send boys to war and get broken men back. No matter how much of a monster you tried to make of your enemy.

[orchid wadi]

Have you ever done anything for the common good, even if it means you felt shit?

Many people have.

I have.

[astral goblet]

ulterior motives and selfishness are different concepts

while people might not be altruistic, that doesn't mean selfish

[narrow jackal]

Keep being selfish, because you yourself don't like it, in the end you will feel good

[orchid wadi]

All this nonsense theory about "natural state" and whatever ignores a scientific fact: humans are social species, we do not do good in isolation and we adjust our behavior according to the group.

[astral goblet]

anyways, keep on keeping on. lifes a garden, you gotta dig it. later all

[orchid wadi]

Then again these ideas about this stuff are pre-science so to speak.

What amazes me the most is that these crazy ideas don't seem to update with the science.

[narrow jackal]

In a group and not in the form of a coercive collective.

[fervent thunder]

Hello, guys
I am using Stability API for text-to-image generation and interested in how can I achieve the same results as in the UI? I used the same parameters I guess, but it doesn't work

[orchid wadi]

Doesn't matter. If I force you to be in a jail cell with 5 other people. you will adjust your behavior to those 5 people.

If I make you my slave, you will adjust your behavior according to that.

We know this... Funnily enough from prisons and militaries. They form special social groups because of the conditions they are formed in.

And we also know that those social spaces fuck people up beyond redemption.

[tight dust]

i have simple question: on the webUI which one weights prompts? i know its different on certain things so i am wondering

like is it the (( or the {{ or the [[

[orchid wadi]

Which webui?

[tight dust]

i tink its the automatic 1111 from a google colab or something

[orchid wadi]

The 3 most popular all use different syntax.

Auto's uses () and modifier with (prompt:0.0)

[narrow jackal]

What does one thing have to do with another? If so, killing will not be wrong, as it is just a way to organize and gain power through other groups.

[orchid wadi]

What this means can be adjusted in the settings.

[orchid wadi]

Are you for or against euthanasia?

[fervent thunder]

is it best to train with the base 1.5 model? or do you recc one of the homebrews people have made

or do you only do 2.0 stuffs

[tight dust]

thanks

[orchid wadi]

Every model follows the same logic in training and all training is model specific. Some embeddings, HN, or such can do something, but not what it was "trained" to do.

[narrow jackal]

You have your body as your own, if you take your life it will be in your hand, but I consider it immoral and I would try in every way (without coercion) to stop you from committing this act

[orchid wadi]

If I am 4 limb paralyzed and ask you to stop my suffering. I can not myself kill myself. Would you euthanise me?

[narrow jackal]

No

[orchid wadi]

I am coercing you to kill me, socially.

So you'd let me suffer till I die?

[narrow jackal]

I have choices

[fervent thunder]

so in other words it doesnt matter which model i use?

[orchid wadi]

Would you then keep me alive?

[orchid wadi]

Nope.

[narrow jackal]

Something interesting that you mention is that if the means justifies the end. So killing someone to save millions is no problem for you?

[orchid wadi]

Answer the question:
You wouldn't kill me, if I asked.
Would you let me die, if I asked?

[narrow jackal]

Launched fallacies, the answer would be no

[orchid wadi]

Ah yes... Logic bros and their fallacies. Never able to answer a question properly.

[orchid wadi]

Why do you assume I care about the numbers?

Numbers are meaningless wihtout context.

[narrow jackal]

Coercion is the act of inducing, pressuring or compelling someone to do something by force, intimidation or threat.

[orchid wadi]

And you still failed to answer.

All I need is a yes or no.

[narrow jackal]

A false dilemma

[orchid wadi]

No it isn't.

[narrow jackal]

Yes, it is

[orchid wadi]

See this is the problem with your "moral framework" It fails in conditions like this.

You can bang on about the natual order, or meaning of the terms, and youl will always fail to answer a question.

[narrow jackal]

@orchid wadi a false dilemma is a logical fallacy that presents only two options or sides when there are many options or sides. Essentially, a false dilemma presents a “black and white” kind of thinking when there are actually many shades of gray

[orchid wadi]

And if I dragged you to the local hospice where people and in horrific pain, asking you why isn't it legal for them to get euthanasia. YOu can bank on whatever the fuck you want about coercion or natural order and it wont make their pain go away.

[narrow jackal]

Do you read what i wrote?

[orchid wadi]

OK. Ill give you a better setting. You must choose whether to kill a person asking for euthanasia, or keep them alive. Or I shoot you.

[tight dust]

what does the (prompt:0.0) max i put my prompt in it with it at 5.0 and it was just a bunch of noise i got

[orchid wadi]

There 3 choices system. They do, you die, or both lvie.

[narrow jackal]

Euthanasia is immoral, but that doesn't mean it's unethical.

[orchid wadi]

5.0 is way too much. 0.0 just means token has 0 value.

[tight dust]

yeah i kind of thoguth so after

but like whatsthe range

[orchid wadi]

Immoral according to whom? IN my view it is immoral for Jeff bezos to exist in a nation where children go to bed hungry.

[tight dust]

is it like 0.0 to 1.0 or something

or am i misunderstanding entirely

[orchid wadi]

In my moral compass it is immoral to own property. You have no claim for anything.

[narrow jackal]

Hahahah at this point you are the aggressor and no I wouldn't kill the person and you could kill me, because I am not a relativist like you.

[orchid wadi]

1.0 is normal

[tight dust]

Ok

so if i put 2.0 would the weight on that prompt be more emphasixef

[scenic shadow]

hello, I'm new to all this AI. So can we use this discord SD here like the midjourney bot? sorry newbie question

[orchid wadi]

Ah yes... You know that wolf hunting you in the wilderness gives no fucks about them being an aggressor. Why do you think I care?

[orchid wadi]

No. Dreaming was disabled for now.

[scenic shadow]

temporarily?

[orchid wadi]

You libby lots keep banging on about these terms. But tell me. Would you live in a world without painkillers?

[orchid wadi]

No idea.

[narrow jackal]

You say that something is not immoral, but at the same time you use the western mentality to say that a hungry child is immoral

[orchid wadi]

What?

You saying that in eastern culture it is totally OK for children to starve?

[narrow jackal]

Moral and immoral came from Christianity. Tribes kill albino children and they don't consider it wrong, what would you do?

[scenic shadow]

ok, so we can't use SD in this discord group? are then there any good advice on prompting, maybe a channel?

[orchid wadi]

#📝｜prompting-help

[tight dust]

@orchid wadi thanks for the help but maybe you guys could put your conversation in a dm or off topic or something

[narrow jackal]

It's not good for a child to die of hunger, but using it as a way to attack another human being is relativism

[scenic shadow]

thank you @orchid wadi

[orchid wadi]

Dude... Zoroastrianism, And numerous eastern relgions predate christianity.

[narrow jackal]

My god in heaven now you're creating a scarecrow

[narrow jackal]

Where did I say it's right?

[orchid wadi]

What the fuck are you going on about? Scarecrows? You mean strawman?

[narrow jackal]

I used your thought that there is no such thing as immorality

Strawman

[orchid wadi]

You said it here.

Moral and immoral concept exist in other culture and religions that predate christianity. Even Ancient egyptians had good and bad.

[narrow jackal]

If a child is killed, it has to do with ethics and not morals

[orchid wadi]

Gods judged your heart.

[narrow jackal]

Now you pretend to be a Christian?

[orchid wadi]

Nope.

Never went to sunday school, but even I know there is one God in christianity.

Sunday school ain't even a thing here.

[narrow jackal]

When I say of tribes (it is not to say that they are right), but to show that morals change from place to place.

[orchid wadi]

Also it hard to take serious moral advice from a book that says slavery is okl. Women are property and gay people should be killed.

[orchid wadi]

Indeed. Finally we agree on something.

[narrow jackal]

@orchid wadi Moral represents the habits and customs of a society, while ethics is rationalized individual moral behavior and a kind of moral philosophy.

[orchid wadi]

Now why is it that you assume natural morality from somewhere? About natural rights, negative rights of whatever.

[narrow jackal]

God

[orchid wadi]

You can't have a society if there is no group. YOu can't have a group if everyone is individual.

From what source should I derive morality to respecting your property?

From what source do you derive the right to claim land?

[narrow jackal]

The bible, but you may not believe in God

And a debate about the existence of God is something that will go nowhere.

[orchid wadi]

Bible says that the lands are fool all people. The bests are to be servants and bouty is for all.

It doesn't say anything about it belonging to a specific person.

I have read the whole fucking thing.

Remember not to mix fabrics btw.

It isn't allowed.

[narrow jackal]

What is Thou shalt not kill and Thou shalt not steal?

[orchid wadi]

Tell me. How many commandments are there?

[orchid wadi]

And how does your moral system justify slavery. Because 10th commandment tells you not to covet their slaves.

[narrow jackal]

In the bible 9 if you are not seventh day adventist

[orchid wadi]

You are wrong. Old testament has 613 commandments.

[narrow jackal]

What fuck

[narrow jackal]

Mosaic laws and what happens when Jesus Christ come to earth?

[orchid wadi]

"You shall not covet your neighbor’s house. You shall not covet your neighbor’s wife, or his male or female servant, his ox or donkey, or anything that belongs to your neighbor."

[narrow jackal]

Where I defend slavery?

[orchid wadi]

You said your morality comes from bible.

Bible doesn't forbid slavery

Actually sets rules about it when and how it is allowed.

[narrow jackal]

Are you now trying to say I'm in favor of slavery?

[orchid wadi]

I think you should go to bed because I am. Tataa.

Read the good book.

Maybe youll learn something form it.

[narrow jackal]

The bible's slavery is different of what we think of slavery

@orchid wadi Masters, treat your bondservants justly and fairly, knowing that you also have a Master in heaven.

@orchid wadi Slavery in the Old Testament Law

Slave ownership was a common practice long before the time the Mosaic Law was given. So, the law neither instituted slavery nor ended it; rather, the law regulated it. It gave instructions on how slaves should be treated but did not outlaw slavery altogether.

Hebrews with Hebrew slaves. The law allowed for Hebrew men and women to sell themselves into slavery to another Hebrew. They could only serve for six years, however. In the seventh year, they were to be set free (Exodus 21:2). This arrangement amounted to what we might call indentured servanthood. And the slaves were to be treated well: “Do not make them work as slaves. They are to be treated as hired workers or temporary residents among you” (Leviticus 25:39–40). The law also specified that, “when you release them, do not send them away empty-handed. Supply them liberally from your flock, your threshing floor and your winepress. Give to them as the Lord your God has blessed you” (Deuteronomy 15:13–14). The freed slave had the option of staying with his master and becoming a “servant for life” (Exodus 21:5–6).

Hebrews with Gentile slaves. When the Israelites conquered the land of Canaan, they were to drive out or destroy all the former inhabitants. However, that order was not fully obeyed, and many Gentiles remained in the land. God allowed the Hebrews to take slaves from among that population: “Your male and female slaves are to come from the nations around you; from them you may buy slaves. You may also buy some of the temporary residents living among you and members of their clans born in your country, and they will become your property. You can bequeath them to your children as inherited property and can make them slaves for life, but you must not rule over your fellow Israelites ruthlessly” (Leviticus 25:44–46). So, the law did allow for slavery.

[quick bobcat]

Man, getting real in here

[narrow jackal]

https://www.gotquestions.org/Bible-slavery.html

@orchid wadi https://www.gotquestions.org/Bible-slavery.html

[narrow jackal]

I say that same for you and don't forget So God created man in his own image,

in the image of God he created him;

male and female he created them.

*the

Remembering that I am not a native speaker and a debate will be difficult due to the language barrier.

[tight dust]

I know it kinda 🙁

[narrow jackal]

What?

[quick bobcat]

Feels like a reddit thread tbh

[tight dust]

i atted one of you saying you should take it to a dm or off topic or somethign but you ignored

[quick bobcat]

People talking past each other thinking there's something to somehow win at.

[narrow jackal]

Sorry

[narrow jackal]

Sorry

[tight dust]

its ok ig

[narrow jackal]

I debated him because he's talking about That there is no natural right

Property, no being killed

[silent sand]

Ive got a quick A1111 question, where should I ask it?

[narrow jackal]

@silent sand https://discord.com/channels/1002292111942635562/1002602742667280404

[lucid storm]

Is there any dreambot working/online?

[narrow jackal]

Lukium swarm

[narrow jackal]

You have to talk for 5 minutes

[lucid storm]

What is Lukium swarm?

[narrow jackal]

And then lukium will give you a link

[narrow jackal]

Discord's server

[narrow jackal]

I sent for you a message with the link to the server

[lucid storm]

Thanks

[boreal cosmos]

this laptop has a PCIE port at the back, can I buy it to plug in a GPU and run SD?
https://www.youtube.com/watch?v=eVf7NLBj0y0

[narrow jackal]

You're welcome

[wild steppe]

Enough, drop this conversation or take it to the DMs @narrow jackal @fervent thunder

[fervent thunder]

happy to

[narrow jackal]

I stopped by you saw who curse who

*but

[wild steppe]

Please be respectful to each other in the community

[narrow jackal]

Sorry for pin you but could you delete the debate comments?

[narrow jackal]

To make the chat clean without useless discussion

[warm tiger]

beep. quick question. for the prompts. is there a way you can comment out certain wording like in scripts? like how some of them a // in the line makes it a user comment rather than being read by the program?

trying to better sort my prompts into catagories like 1 line for backgrounds 1 for the characters, 1 for art style, etc

[cloud granite]

I seen the messages get deleted

[narrow jackal]

Are you the dm?

[cloud granite]

I wasn't joined in. But when I checked this channel people were insulting just recently.

[narrow jackal]

I didn't curse no one but in the other hand Chaz did

[wild steppe]

Yea I deleted them, they've been timed out. Lets keep religion and politics out of general chat.

[cloud granite]

Good job

[narrow jackal]

I gonna answer again could you please delete the debate comments?

[narrow jackal]

It's useless debate

[wild steppe]

gen chat is clean again 🧹

[narrow jackal]

I still see the comments, I could delete my comments, but others I can't delete only you can do.

I'm not native speaker

[vestal dew]

I've come to the conclusion that the defects SD produces in human images is purposeful. There's always at least one issue, at least one imperfection. When it gets hands right, something else is wrong. Hands right - feet mangled. Hands and feet good - face mangled. All that good - extra limb. All that good - some weirdness in clothing. I generated 300 images earlier. Most had mangled hands, many mangled feet. I thought I found one that was good - nope. Noticed that the skin of one arm was like a skin colored sleeve. Always some problem.

[narrow jackal]

AI is Still learning

[vestal dew]

I dont think so. It's doing what it was told to do, programmed.

otherwise I wouldn't get 100 images in a row with mangled hands, then the next one has good hands...but flat, super wide feet.

[warm tiger]

it blends hair into the surrounding objects a lot too. tech is still pretty young and it only does things as well as it's been trained to.

[vestal dew]

mangled hands, mangled feet - finally get one with good hands and feet....legs have double knees. None of the previous images had that. Only once hands and feet perfect does it mess up the knees. That has to be by design.

[brisk bloom]

it really depends on the model you are using and embeds, which can fix a lot

[warm tiger]

hands and feet are difficult. The AI is basically given all the bits and parts of what makes up a 'thing' but isn't really told how they are put together except by prompts or the images it learns from

and poor training data gives poor results

as do poor prompts

[vestal dew]

About 8 days ago someone said in "three days" he would demonstrate how to get perfect hands and feet every time. Still waiting

[warm tiger]

with the proper model and sampling method and the right prompts you can get okay+ hands and feet about half the time

[vestal dew]

it's like you ignore one important part of my post. In an image in which SD finally got hands and feet right, it messes up knees in a way it hasn't done before. THAT IS BY DESIGN!

[warm tiger]

It's okay to say you don't understand how it works

[vestal dew]

I wanted an image of a baseball player running to first base , away from the camera. I finally got a perfect image that showed the player running to first, and a fielder throwing the ball to the 1B. In this image...ONLY THIS ONE...the running player is, for some reason, naked. Every other image, he had a nice uniform.

[brisk bloom]

put nudity in the negative

[vestal dew]

I got another perfect in almost all respects...but his left arm was a bat

[warm tiger]

take the prompts/seed of that image and rerun it a bunch of times

[vestal dew]

Why would I think to do that? EVERY OTHER image showed him in a uniform

[warm tiger]

or take it to inpaint

[vestal dew]

I did. Never recreated the image. Never could get everything else right again

inpaint never EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEVER works for me

I cannot match what I see in videos

[warm tiger]

yeah and sometimes when i generate images and specifically state "pink hair" 1/100 of them they have blue or green

it happens

[vestal dew]

it either does nothing, or it changes the whole image

right, but watch

you'll get the wrong hair just when everything else is finally right

it's by design

[warm tiger]

just sounds like you're doing it wrong and claiming the AI hates you and is doing it intentionally

XD

[vestal dew]

Right...I am doing it wrong

I get everything right, but in that one image there's a defect that hasn't happened in any other image...that's on me

it's by design

[warm tiger]

so redo it

[vestal dew]

I did

put in negatives. Generated 1000 more images on the prompt...none good

[warm tiger]

rerun the same image with a non-randomized seed. same prompts. set batch size up to like 6 and do seed number+1 or something

[vestal dew]

then did img2img with the best image....always a defect in hundreds of images

[warm tiger]

it'll regenerate that same image over and over and over again with minor changes till you get what you want

[vestal dew]

I got one that was perfect....but in JUST THAT ONE...suddenly the face was mangled

[warm tiger]

so fix the face

XD

[vestal dew]

If I have to fix the face by adding a face, I may as well generate images by pasting together parts of other images

[brisk bloom]

to fix faces you do inpainting img2img, mask the face only and run a few gens. usually it fixes

[vestal dew]

inpaint never works for me

[warm tiger]

^ same thing with hands or feed

[brisk bloom]

yup

[warm tiger]

stop doing it wrong then

[vestal dew]

it does nothing or changes the whole image

[warm tiger]

select to only have it change the marked spot

XD

[vestal dew]

or makes a tiny little repeat of the prompt inside the maasked area

like a tiny mangled person, or just one limb

[warm tiger]

change the prompts for the masked area

[vestal dew]

when I select change only marked spot, that's when it does the repeat the image in the inpainted area thing

[warm tiger]

derp

[brisk bloom]

depending on the size of the head, hand you will need to reduce the denoise right down 0.30 or less

[warm tiger]

i've not even once hat it do that to me

[vestal dew]

I had a mangled hand

masked the hand

[warm tiger]

maybe go watch a video on how to actually use img2img?

[vestal dew]

prompted "perfect hand"

got a giant thumb

[warm tiger]

yeah because an AI cant know what "perfect hand" looks like

XD

[vestal dew]

I tried hand

perfect hand

human hand

[warm tiger]

you're giving it bad prompts

[vestal dew]

all kinds of things to make it generate just a hand there

Im gonna test what you say

in inpaint, with a mangled hand, what prompt will replace it with a good hand?

[brisk bloom]

Mask mode: Inpaint masked. Masked content: original. Inpaint area :only masked

then low denoise

[vestal dew]

done dopne and done in the past

[warm tiger]

also there's a reason a lot of people want strong GPUs to so this. it's because sometimes it takes literally THOUSANDS of photos before you get what you want

[hardy junco]

Hello all, do you know if all extension propose by the original link in extension tab are all safe? ( no virus sure)

[warm tiger]

to do this*

[vestal dew]

so maybe that's it

I have a weak GPU, can only do hundreds when it takes thousands

[warm tiger]

you can do it with hundreds. did you even read pure's example?

[vestal dew]

Yes I did

[warm tiger]

i bet you just ran with the default denoise rather than turning it low

[vestal dew]

No. I watched a video that suggested the same thing

that's when I got the giant thumb

[warm tiger]

i think all those should be 'safe' in that they are not malicious code

[vestal dew]

once a giant leg

once a scorpion for some odd reason

[hardy junco]

i think too but i'm note sure

[brisk bloom]

you dont get giant body parts with low denoise

[vestal dew]

once a tiny copy of the whole image

[warm tiger]

it's a skill issue. that or he's using a crappy model. or both.

[vestal dew]

what model do you suggest?

[brisk bloom]

something is off for sure

[vestal dew]

Ive used the models people suggest

protogens, lucid, all sorts

[warm tiger]

ah see, perfect example. i have "((pink hair))" in the prompts and (blonde hair) in the negative prompts and it just generated one with blonde hair

XD

[brisk bloom]

CFG scale can have an effect on what it returns also

low CFG and it will decide itself

[warm tiger]

maybe check and see the various prompts that model is even capable of using or run the thing using higher sampling stepts

[warm tiger]

yeah there's a randomness factor the lower you turn that

funny enough i've had some of the totally random photos generated that ignored my prompts turn out to be some of the nicer ones

[brisk bloom]

for sure, it can be incredibly creative when it finds the right style

but thats the issue the style is in a mountain of data

needle in a haystack

[warm tiger]

using a LORA or knowing the style of art you already want is a good start

also it can be fun running the exact same prompts through various models or even the same model with different VAEs on it to see the various results

the available facial expressions for a lot of models are.... lacking

[hardy junco]

from this one https://raw.githubusercontent.com/wiki/AUTOMATIC1111/stable-diffusion-webui/Extensions-index.md

[warm tiger]

i mean. if you trust AUTOMATIC1111 any extention listed in their repo should be just as trustworthy

that's the one i use, btw

[coral lark]

Is there an inpainting model that's good at hands?

[dry drift]

I need best stable diffusion model for logos can anybody help me in this

[green sandal]

is there a way i can train my own model when i dont have an adequate gpu? is there a simple online way to upload images to train a model and then generate art? even if i have to pay for it, i cant download sd locally on my computer.

[warm tiger]

yes. not simple though. it'd involve renting a server somewhere and setting it up and running it remotely. but even training a model isn't exactlly super easy

[green sandal]

thats what i thought. thanks anyay

[warm tiger]

you could just rent one

pretty sure there's a lot that offer per-hour rates of use

some of them even use A100 cards

[cursive plume]

hey sorry if this has been answered already, couldn't find an answer, when working with stable diffusion are there any files being saved or altered outside of it's own enviroment?

[static gazelle]

ive seen mention of "merged models" is there a place i can learn more about this or how to go about merging models?

[brave matrix]

hey guys, very curious what you think about my newest project: https://youtu.be/ecw39LWezUM

InfiniteConan

First episode features Seth Rogan and Norm Macdonald, all the visuals are SD

script is chatgpt

[young ivy]

is it turtle-tts or coqui-ai?

or a commercial one?

[astral goblet]

looks like low denoising on an actual conan episode of seth rogan

[brave matrix]

its actually a 3d animation that then uses img2img

[astral goblet]

thats some cool generative workflow. way better than the seinfeld go

[brave matrix]

Yeah it allows for more custom results

[astral goblet]

consider some triumph cameos

[glad pagoda]

im stuck on the installation

help

[glad pagoda]

venv "C:\Users\Elite bk\Desktop\Nueva carpeta\stable-diffusion-webui\venv\Scripts\Python.exe"
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Commit hash: 3e0f9a75438fa815429b5530261bcf7d80f3f101
Fetching updates for Taming Transformers...
Checking out commit for Taming Transformers with hash: 24268930bf1dce879235a7fddd0b2355b84d7ea6...
Cloning K-diffusion into repositories\k-diffusion...
Cloning CodeFormer into repositories\CodeFormer...
Cloning BLIP into repositories\BLIP...
Installing requirements for CodeFormer
Installing requirements for Web UI

[wide juniper]

What happened?

Your configuration specifies to merge with the ref 'refs/heads/master'
from the remote, but no such ref was fetched.

exit code: 120

stderr:
Exception ignored in: <_io.TextIOWrapper name='<stdout>' mode='w' encoding='cp1252'>
OSError: [Errno 28] No space left on device

Launch unsuccessful. Exiting.
Press any key to continue . . .```

[warm junco]

Your Hard Drive is full. Make sure it has 15-20gb Space left

[warm tiger]

any suggestions on how to best get 1 character using prompts? SD seems to want to generate 2-4 people despite the fact i have tried ( solo ) and several other variants of "1 person"

[wide juniper]

Oh wow it needs that much. Thanks

[vale spindle]

For human inpaint, what setting would i want to use to get rid of glossy skin

Like, that weird ultra-polished skin that makes the photo look obviously fake

ill link a photo in a sec here

https://imgur.com/a/roVyZBy

[wide juniper]

I have been trying all sorts of models . A lot of the images come out blurred , and people often suggest using 'vae-ft-mse-840000-ema-pruned'. i have 'vae-ft-mse-840000-ema-pruned.ckpt' sitting in the same folder. Am I doing it wrong? Do I have to rename the checkpoint file to be the same name as the models I'm using?

[fervent thunder]

work on your prompts

[astral goblet]

now that ive determined i dont' need to replace my psu yet, i'm going to get a 2tb ssd i think

maybe 2 tb ssd's

[viral stag]

Put it in the models/VAE folder. Then you have to activate the VAE in the web-ui settings -> Stable Diffusion -> SD VAE.
Don't forget to "Apply Settings".

[brisk bloom]

Also in Settings > User Inferface, add ,sd_vae to the Quicksetting list option

this gives you a VAE dropdown list next to the models dropdown selection

[wide juniper]

Do i have to change to the file type to .vae.pt?
Also, should my SD VAE to Automatic? I don't want to specify the VAE because I'm using Anything v3's VAE for when I use the Anything v3 model

[warm junco]

@wide juniper Go with the settings pure_fire provided above then you can select the vaes easily

Dont rename the vae just copy it into models/VAE folder

[viral stag]

You can set a default VAE in settings. Also there is an option that allows you to automatically override the default VAE with a model's custom VAE (if that exists in the same folder and has the same name)

[wide juniper]

Thanks all

should i rename my Anything-V3.0-pruned-fp16.vae.pt back to .ckpt and move it to the VAE folder too?

[warm junco]

No .VAE.pt goes into the models/stable-diffusion folder

[wide juniper]

Ok so I added ,sd_vae to the Quicksetting list option. Clicked Apply Settings but no additional drop down list can be found. Where is this supposed to appear?

And why would I want to choose the VAE? Isn't it best to leave it to be chosen automatically?

[warm junco]

Restart the webui

Then it should appear

[wide juniper]

Ok

[warm junco]

Its good to have the setting in reach to quickly switch from autimatic vae to a specific one like the mse-840000

[wide juniper]

How do I know when I should switch to it though? When my results look bad?

[warm junco]

Yea if you get pale images for example

[wide juniper]

Is there any harm if I choose a VAE when it's not necessary?

[warm junco]

No harm but the image could look worse

[wide juniper]

So there is some kind of guesswork involved yeah

And it's not like there's 1 vae around

[brisk bloom]

just leave it on Auto. if you think the image isnt good change it to none or another in the list

very easy to test

[wide juniper]

Ha. I just proved that wildcards don't work on negative rpromptas

prompts

positive is sky, negative is clouds. That works

But when I put clouds as a wildcard, and trigger it on negative, clouds remain

[tardy musk]

what i need to do?

OutOfMemoryError: CUDA out of memory. Tried to allocate 3.32 GiB (GPU 0; 6.00 GiB total capacity; 2.14 GiB already allocated; 2.04 GiB free; 2.20 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

[wide juniper]

buY bEtTeR GPU

I still get it at 24 GB VRAM when I get too greedyu with my batch count and image dimensions

[warm junco]

You need to edit your webui-user.bat and add --xformers --medvram behind Commandline_ARGs=
Then save and restart the bat

[tardy musk]

My PC is: iCore 7 , Gtx 1660 Ti

[tardy musk]

: is not an internal or external command, operable program, or batch file

[warm junco]

?

[tardy musk]

the error:

[astral goblet]

lol yeh me too. 16gb gives lots of space but sometimes i'm like "i can do 4k" and i cannot

[wide juniper]

I wonder if there’s any use in going beyond 512x512, or (768x768 for 2.1). I used to think SD would draw more detail on far subjects, like faces, but I can’t tell

[astral goblet]

i make great desktop wallpaper resolution gens. once with a low res pass for consistency. 2.5x the resolution on 2nd pass

[wide juniper]

And I think doing upscaling is the second step ultimately ends up with the same result, no risk of SD giving up

So what’s a .vae.pt? I only have 1 for Anything V3. It stays in the same folder as the model, and there’s no need to select it specifically in the webui right

[astral goblet]

it's only as same as you allow it to be. i like to denoise with around .55 so that it adds a ton of detail

[warm junco]

For me upscaling with 0.5 gives me less Details in images :/

[astral goblet]

i get that when i use karras samplers with hires fix. not sure why that goes on

[warm junco]

Im using euler a

It smooths the image but it lesses the Details overall

Maybe i should try other upscalers

[crimson anvil]

hello all, how i start generate image at dreambot channel. i dont have permission

[wide juniper]

How many VAEs are out there?

[flint fiber]

Countless

It’s like how many model are out there or how many embed

[wide juniper]

Ugh so the only way I can tell which one is which just by checking the civitai page if the output is blurry, right?

[flint fiber]

Or you can just use model that doesn’t need VAE

VAE are essentially filter

[wide juniper]

Hm filters are traditionally enhancements though. Aren't VAEs fixes that come paired with models?

[warm junco]

Most models have their Vae already included. They dont need one

Anime models mostly need vae files and they always Provide some or mention the one to use

[brisk bloom]

with 24gb vram, whats the largest image size you can hires to - without using other upscalers?

[fervent thunder]

hey guys I've just released my first checkpoint data

https://civitai.com/models/6984/czeromixv6

please check it out and let me know what you think ❤️

[warm junco]

Nice work! Gonna test it later today.
But make sure to credit the creators of the models you merged in!

[fervent thunder]

will do, I only uploaded it for testing but a full readme and better examples and tips will follow shortly ❤️ please show me or PM me anything you find out about it 😄

also its all the same liscence

non modified standard one used on civitai

[warm junco]

Ah nice

[pale blaze]

Can i ask that where do i put the checkpoint file in this repo?
https://github.com/huggingface/diffusers

[warm junco]

@pale blaze I dont understand the Question sry

Do you want to download models/checkpoints?

[pale blaze]

i already have a checkpoints file, where do i put it in here to run it locally without downloading the model online?

[warm junco]

Checkpoints go into the models/stable-diffusion folder

[pale blaze]

Oh, when i clone it it doesn't have this folder, i also need to change the name to models.ckpt right?

[severe zealot]

i cant to seem to git pull update my Automatic 11 11 to get pix2pix, can someone please dm where to find the correct repository? im using this one and its not working https://raw.githubusercontent.com/wiki/AUTOMATIC1111/stable-diffusion-webui/Extensions-index.md

[brisk bloom]

https://github.com/Klace/stable-diffusion-webui-instruct-pix2pix.git

[warm junco]

Do you want SD local then you have to use a tutorial to get started

The diffusers site is not a webui for creating images

[pale blaze]

yep ik, i just want to put it the model.ckpt and run the file with my prompt

[raw iron]

Where are channels for dreaming?

[warm junco]

Which webui are you using?

Sry i still dont get what you trying to do 😅

[pale blaze]

wdym webui? i just clone their code and then i want to use the checkpoint file with the python code, currently this doesn't use the checkpoint file, instead it download it on internet
https://github.com/huggingface/diffusers
My code:

import torch
from diffusers import StableDiffusionPipeline, EulerDiscreteScheduler

model_id = "stabilityai/stable-diffusion-2"

# Use the Euler scheduler here instead
scheduler = EulerDiscreteScheduler.from_pretrained(model_id, subfolder="scheduler")
pipe = StableDiffusionPipeline.from_pretrained(model_id, scheduler=scheduler, torch_dtype=torch.float32)
pipe = pipe.to("cuda")
pipe.enable_attention_slicing()
prompt = "A astronault riding hourse"
image = pipe(prompt).images[0]
    
image.save("astronaut_rides_horse.png")

[pale blaze]

I'm sorry, i'm very new in this

[warm junco]

Ahhh so you using the cli version local and want to use an other checkpoint

[pale blaze]

Yep

[warm junco]

Can you Screenshot the folder structure?

Most people here (me included) use Automatic1111 Webui so it differs from yours

But i think if you find the stable-diffusion-2 model file. You can put your model in the same folder as it. Then change the name of model_id

[pale blaze]

Oh got it, thanks!

[warm junco]

Np but i would recommend you to use the other tool for convience! Also more features and easier support ^^

For technical Questions or Installation help there is #🤝｜tech-support

[sleek osprey]

I think distilled SD is just a figment of all our imaginations

[severe zealot]

bump

[gusty oasis]

But hey, atleast SDA is a possibility now that TensorRT shows to be working on windows

[elder cradle]

Actually, you'll only be able to use models which are:

• Hosted on Huggingface
• Supports Diffusers (If you see the "Diffusers" tag, it is supported)

[sleek osprey]

whats SDA

[gusty oasis]

Stable diffusion accelerated

[sleek osprey]

u got any source or something

[gusty oasis]

https://github.com/chavinlo/sda-node

[sleek osprey]

link to an article perhaps

thanks

[gusty oasis]

There's a ui for it but it's still linux only

[sleek osprey]

looks like its archived already

[bleak matrix]

Good morning, everyone! How are we all today?

[gusty oasis]

Just pain

[bleak matrix]

L i f e

[warm junco]

Hey

[bleak matrix]

Sup? 🙂

[warm junco]

Watching Detective Conan in Lunchbreak ^^

[bleak matrix]

Woo! I'm just waking up.

[warm junco]

At 2pm ? 😄

[bleak matrix]

And it's time for more SD.

Nah, it's 8! And it takes at least an hr for my meds to kick in lol

[warm junco]

Oh okay then good morning

[bleak matrix]

What ya havin for lunch?

[warm junco]

A Bretzel 🥨

[bleak matrix]

I approve this message!

[warm junco]

😄

[bleak matrix]

Excellent choice all around 👏

[warm junco]

Haha thx today is a good day just 1:30 hours and im back at home to generate images ^^

[bleak matrix]

Yesss! Score! I am so doing the same. Once I find my tuna snack.

Got anything in particular planned?

[warm junco]

Yea gonna try out some models, test a few Art Styles (popart, retro, low poly etc), and then i ask ChatGPT for some prompts 😄

Need some new ideas i can try

[bleak matrix]

Awesome!

Sounds like a blast

[warm junco]

Yea Hype!

What about you?

Big plans today ?

[toxic agate]

how to use the ai

[bleak matrix]

I am working on a bunch of stuff for guides!

[warm junco]

Ouh nice we need guides!

[warm junco]

Local or in web?

[toxic agate]

i cant find it

[warm junco]

The discord bot is offline

[narrow jackal]

This discord was closed for a while but you can use lukium swarm's server if you want to use the stable diffusion 1.5 inpainting

[warm junco]

@upbeat gorge hey any updates on the App ?

[silver wigeon]

Hey everyone! I just got my new Macbook Pro M2 and am finally looking to intall SD locally. Are most people still using Automatic 1111? Where can I download the latest version? Would be very grateful for help to point me in the right direction.

[upbeat tangle]

guys we can combine models?

@warm junco hello, do you know bro?

[empty barn]

https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Installation-on-Apple-Silicon

[silver wigeon]

Thank you!

[peak orchid]

Has anyone seen the "stable attribution" site created by the anti-AI art folks that is currently making the rounds online? It that pretends to decode the exact training images used to create an output, but that's impossible so its a total misinformation laden scam: https://www.stableattribution.com/

[warm junco]

Yes with the Checkpoint Merge tab

[upbeat tangle]

how can i dowland and use it

[steel glade]

There's no stablediffusion bot here? Which will generate?

[warm junco]

Its build in in the webui under the checkpoint merge tab

[steel glade]

Sorry there is no stablediffusion on the phone 😂

[warm junco]

there is an app but only for iphone

[steel glade]

Understood thank you

[merry solar]

I've been trying out Stable Diffusion 2.1 on my local Apple M1 using Core ML Stable Diffusion. The quality on all the test images I try generating is so bad, and I have no idea where to start with why? I'm not just referring to the limited resolution of 512x512. I'm just wondering if there is something I should be trying to adjust or tweak that would help it try harder? Attaching some of my output tests.

I usually let it run 50 steps.
*also tested the Swift Core ML Diffusers mac os app.

[warm junco]

the images depend on the models you use, also for Portrait an resolution of 512x768 can help

For a local install i recommend this video over any outdated guides you find:
https://www.youtube.com/watch?v=VXEyhM3Djqg

if you have any questions feel free to ask in #🤝｜tech-support

[upbeat tangle]

how can i install it if you have time can you help me?

[warm junco]

do you use automatic1111 webui?

[upbeat tangle]

yes

[warm junco]

then its already installed and usable

you can find it under the Checkpoint Merger Tab

[upbeat tangle]

when i paste checkpoints here

it will work?

what about vae's

[warm junco]

you can inbake vaes to a model there too

[upbeat tangle]

in file?

[warm junco]

?

[upbeat tangle]

i cant found

[warm junco]

in the webui

not in the files

its a feature

not a folder

[upbeat tangle]

ah okay

thank you

[warm junco]

no problem

[sly arrow]

First of all, is this the right channel to ask this, or should I move it somewhere else?

What's a good workflow for training a textual inversion? Is there any good settings for any of the variables, or possibly good settings for the first steps NNNN steps, then switch to other settings?

As a textual inversion templates I started out using a lengthy prompt trying to cover everything not to learn, basically ignore my dataset being renders: a 3d render [name] of [filewords], ray traced, an ambient occlusion render, ray tracing, a raytraced image, generative art, physically based rendering, Beeple, 3 d model, a 3D render, photorealism, computer art, plasticien

The appendix is unwanted words that Img2Img interrogated from the training images btw.

But as [name] and [filewords] got lost in the prompts, now I use:
a [name] of (([filewords])), ambient occlusion raytraced render

Sometimes it approaches a correct solution, but then seams to wander of. Is this normal, is it a give it more time scenario? When do you know that the training is going haywire?

if training is looking weird, do you alter / replace the training images and continue? Or do you start over?

What's your best practice's? Thanks!

[warm junco]

yes easily

[upbeat tangle]

Can we choose which model to use with which ratio?

[upbeat tangle]

Is the other model similar as the line goes to the right?

[warm junco]

wait

Here is a quick explanation of the Merge tab:
If you have 2 Models (A, B) always use Weighted Sum: It will merge them with this formula where
M is the Slider Scale: A * (1 - M) + B * M
It will merge the two with the Percentage of the Slider (0.5 = 50%).
If you have 3 Models (A, B, C) always use Add Difference:
It will merge them with the formula where Slider is M: A + (b - C) * M
In Fact it will erase all of the C Model of your final Merge.

Interpolation is only when you have 1 model (its for making ckpt to safetnesor, baking in a vae or just renaming).
ckpt and safetensor are the output files you can choose for the model.
Choose dont copy config if your models are based on 1.5
also dont check fp16 if you dont know what it does.

[wide juniper]

Do 1.5 models only go up to 512 by 512?

[warm junco]

no

you can use any size your gpu can handle

[wide juniper]

I mean details-wize

As I understand it, a 1024 by 1024 image on 1.5 is just 512 5512 scaled up in post

[warm junco]

no 1024 is not upscaled

[wide juniper]

So what's the meaning of 2.1 being at 768 times 768?

[warm junco]

the pictures used to train were at 768x768 resolution

so you can get better pictures at higher resolution

[wide juniper]

Yeah, which brings me to the futility of trying to generate pictures above 512 by 512 in the first place

[steel glade]

Does anyone have the source code of Different dimension ME where it is possible to generate an image with objects, not just a human face?

[ornate snow]

what is the command for making images?

[warm junco]

the bots are offline

[pulsar canyon]

Hey where can I announce a release of a new model?

[astral goblet]

https://youtu.be/fTqgWkHiN0k prtetttty nifty stuff

[warm tiger]

depends on what you want it for

but typically, yeah, it's pretty good

[sinful mango]

Raiden warned about AI art censorship (partially AI generated codec call).

https://youtu.be/-gGLvg0n-uY

[astral goblet]

anything v4 is a troll model somene made. they merged a bunch of porn and other anime models into anything v3, then stole the name and said it was v4. the people who made v1,2,3 are not involved with it.

it works because they didn't train anything and merged popular models together, but the authors hate v3's popularity for some reason and want to troll against that
@crude elk

i don't think they have the technical capability to inject viruses into models. they're just some kids that want to troll weebs

if you use the model and like it okay, but av3 is still super high quality and hard to improve on

there are some new anime models made for 2.1 generations too that are looking really nice

the benefit there is that they're not based on stolen works. All the 1.5 anime models derive from stolen/leaked novelAI models

[fervent thunder]

How would you go about using a child drawing and turning that into actually you know a decent picture?

Img2img?

[sick wave]

Greetings, anyone's got a good tutorial to get started? I manage to install everything, I can write a prompt, but I want to understand how to write a prompt and how to find the right keyword for what I'm trying to do.

[fervent thunder]

Try #1014939219904450590 and #📝｜prompting-help may help too

[forest apex]

dude that is gold

[sick wave]

Thanks, I'll see there

[sinful mango]

ty, it was inspired by the famous "Selection for Societal Sanity" codec call from MGS2.

[forest apex]

how about some shorts clips? 🙂

no no no no, I was just... calibrating...

[sinful mango]

hahaha. watch until the very end for another little joke

[cunning carbon]

How would one combine multiple features from different images, into a single output image? Something like this: https://huggingface.co/spaces/lambdalabs/image-mixer-demo. Do you have any other suggestions that are similar to this?

[astral goblet]

Thats very cool tech!! i'm not sure how that'd be used outside that ui though

https://huggingface.co/lambdalabs/image-mixer this is the model card, instructions to load it locally there

[cunning carbon]

Is there any channel dedicated for dev stuff? I want to write my own model, and would like some help if possible.

[wild steppe]

#1011228667659178055 and #🔧｜finetune perhaps?

[cunning carbon]

Thanks

[sudden ruin]

someone is spring cleaning channels right now

[warm junco]

yea the tech support channel has vanished xD

like the change for dreamstudi channels

[tidal bough]

Everything (and more!) will be back with you really soon! Please stand by...

[still glacier]

I was about to ask XD

tech support's support

[warm junco]

my fav channel bring it back 😄

my whole BetterDiscord theme crashed because i was in the channel as it vanished xD

#🤝｜tech-support is back 😄 nice !

[unkempt slate]

I want to train my own Stable Diffusion model but the SD-WebUI is broken for me in the way that no extensions work.
Does anyone know any website or online service where I can train my own model? I'm willing to pay as well

[willow coyote]

[REDACTED]

[sudden ruin]

would help if you link the image

[willow coyote]

[REDACTED]

[REDACTED]

[sudden ruin]

#🏞｜general-with-images

[willow coyote]

[REDACTED]

[willow coyote]

Ofcourse, my bad!

[icy lake]

why were all channels archived???

[spark dust]

when are bots back

[subtle sun]

does this server update mean i can use my amd gpu now

[dull swift]

Im feeling kind enough to let you test starboard on me

[charred mesa]

um interesting changes

[dull swift]

Im to kind

Star me for

[gentle otter]

new icon sucks, will be even harder to find in my server list

[charred mesa]

WHAT IS TH

bruh it's a big purple lens

I just saw it too

[icy lake]

WHAT THE

[gaunt meadow]

Pog Eleuther is here

[dull swift]

Star this and il tell you

[alpine vault]

asking for stars 🤓

[charred mesa]

^

[gaunt meadow]

They do text generation

[sick fractal]

I think there was stable diffusion update to 2.2 model but not, another unnesessary ping

[charred mesa]

stop asking please, EARN it

[grizzled harness]

hmm, I am quite interested in this. Any more info to share?

[rustic isle]

curious, what bot's being used for #1072233397746929755

[dull swift]

Gotta make sure it works tho

Ofc

[red oxide]

Can you guys not remove emotes on announcements plz

Rude

[gaunt meadow]

Usually it’s carl

[charred mesa]

could be dyno 🤷‍♂️

[dull swift]

[red oxide]

I know ow you guys removed the thumbs down and clown emote

Put it back

[severe trench]

where's the anime channel 😡

[charred mesa]

there was one?

[celest oyster]

N

I

G

[severe trench]

😭

[celest oyster]

H

T

[charred mesa]

#anime

[celest oyster]

guys

[charred mesa]

I'm not getting anything

Maybe it's hidden

or is sadly archived and gone forever

[grizzled harness]

They re-arranged everything

[severe trench]

all the generation channels are gone

[grizzled harness]

there's one in "styles-genre" now

[severe trench]

they might've done a little goofy and messed it up

[grizzled harness]

it's all under 'dreamer communities'

[severe trench]

oh

[gentle otter]

read the announcement

[wintry stream]

https://discord.com/channels/1002292111942635562/1072013871730131004

[grizzled harness]

they're threads or whatever... I dunno, I don't have a degree in Discordology so it's gonna take me a few to figure out this mess

Didn't realize the way it was before was broken....

🤭

[wicked hollow]

The heck is up with the server icon now?

"Welcome to the Abyss"?

[red oxide]

It stinky

[grizzled harness]

hmm.. not April 1st..

so I dunno 😛

[wicked hollow]

Once you go AI you don't go back? Lol

Careful that the AI doesn't stare back into you?

[timid gate]

did they kill #anime

I might just lose my shit+

[charred mesa]

was there something in it that you wanted to retrieve

[wicked hollow]

Wait they killed #anime?????

[charred mesa]

it's in #1070829982252798043

and in a specific thread 🤷‍♂️

[wicked hollow]

Not a fan of that. It was popular enough to have its own section. Super popular in fact.

Anime is eternal, after all.

[grizzled harness]

agreed

it was the most active community

[wicked hollow]

Omnia you're up. ⬆️

[grizzled harness]

I liked that channel and I haven't even seen anime. 🙂 Best place to learn about new techniques!

[grizzled harness]

hm?

[quaint lark]

Hi

[quartz loom]

yeah it's hilarious

[quaint lark]

Is stable diffusion free?

[quartz loom]

yeah

[jolly prawn]

Which is better..stable diffusion or mid journey?

[sudden ruin]

if i have a photorealistic portrait i wanna post, does it go into #1072016199837290536 or #1071937700778741860

[gentle otter]

so extremely subjective and complicated that any answer would just get you further from the truth

[jolly prawn]

I was using mid journey bots but I need to pay its subscription monthly

Does stable diffusion does it for free?

[gentle otter]

yes

[jolly prawn]

For lifetime or for limited time period

[gentle otter]

forever

[jolly prawn]

Demn

[gentle otter]

but you gotta install it

[jolly prawn]

That's why my friend is richer than me

[jolly prawn]

Yeah sure

[whole swallow]

what is the server icon supposed to be

the original eye was better lol

[jolly prawn]

Hey man...can I install it on my android or iOS
Or just it works on PCs??

[gentle otter]

no pc only

[warm junco]

there is an app for ios

[gentle otter]

you need a good graphics card

if there's an app it's just a wrapper for a web interface of someone elses installation

[warm junco]

No

[gentle otter]

the power of SD is that you can host it yourself and do whatever you want with it, and not be subject to sharing resources

[warm junco]

it renders localy

[gentle otter]

if that's true I'm amazed

[warm junco]

yea its true

Here for IOS:
https://drawthings.ai/

[jolly prawn]

It's the same version as that of Pc

Or there is a difference

?

[warm junco]

its different

[jolly prawn]

How much

[warm junco]

It can just generate images

Automatic1111 webui has the most features, like im2img, inpaint, outpaint, extension support etc

[fervent thunder]

yoo whats good

[jolly prawn]

Oh

[fervent thunder]

thats a gross oversimplifcation and not technically accurate my guy

[violet glen]

??? #anime is gone ???

[fervent thunder]

saying that 2.1 is my next target ;D

[teal lark]

hi , how can I start generating image ?

[astral goblet]

easy to test by seeing if "best quality, masterpiece, animefull" create results. the nAI keywords. Find one popular anime model that doesn't get triggered with those keys

[fervent thunder]

my model

[astral goblet]

also there's a few common seeds that always make novelai results

[fervent thunder]

and my friends from civitai who made it off his own art

well I don't consider that stealing, if an artist is born blind, they are truley creating origional art

you might get my point

[astral goblet]

yours has 1 review.

[fervent thunder]

v3 is the last version that was official im not saying that

yeah just uploaded it but it's not any different youll be able to specifically force somethiong to create a result that looks like something else yeah

thats why I'm also moving onto 2.1 I just dont agree with your concept of stealing

[astral goblet]

200 downloads is decent but i'm not saying all models now am i? and i said how it's easy to test.

My Guy

try to keep up

[fervent thunder]

or the court mandated clickbait non technical understanding of it

[astral goblet]

the novelAI model was a stolen release. they were keeping it in house, and the leak was done illegally. Open source licenses doesn't mean it has to be released publicly. Only that if it is, it has to be under the same license

[fervent thunder]

I am coming from the standpoint that training something using images that are not stored is not illegal

[astral goblet]

i'll play around with 1.5 anime models, but because of teh nAI issue tainting that crop so hard, it's entirely segregated to pure play and nothing i'd ever use in real work

[fervent thunder]

in the same way an artist can go to an art gallery see something and be inspired by it

[trail wing]

did they just wipe out the image posting themed channels...

[astral goblet]

you're unaware of how the novelAI model came to be publicly available then I suppose

[fervent thunder]

but yeah theres no need for them to lie and be dodgey so att he same time I am agreeing with you I just dont think it's stealing

[astral goblet]

Why would you accuse me of lying? And for that you earn a block. Childishness is childishness.

[fervent thunder]

huh?

I was clearly talking about novelAI

but whatever man all you do is cry about "stealing " in here anyway

back to doing something worth the time :L

there is a legal dispute specifically around the training data for 1.5 beacuse some people deem the way the ai was trained unethical, often due to a misrepresentation of how the software works

[untold ocean]

#1070829982252798043

[fervent thunder]

that's all, this guy gets super touchy about it for some reason, ignore it , better spent time doing something creative

v3 imo, as the dude was saying 4.5 isn't made by the same people, some one took 3.5 and added more to it ( probably a bunch of nsfw training )

the real POTG is to download both and experiment

[sterile gust]

We are getting "ethical" I guess

[brisk bloom]

ooh... lots of changes to the server while I was asleep

I like the new name. The icon though??

[carmine horizon]

A quick question if anyone is able to give advice about this !
I'm trying to.. understand if some models are more prone to head-cropping than others, and if I'm not mistaken, based on what I could read, it would look so.
So... What I'm actually trying to do is to train an embedding/LORA for a style, a kind of sketch drawing. I have 20 images of the fitting style, and everything is going fine except.. yeah.. I'm getting cut-off head quite often, no matter what prompts (positive and negative) I use. So my question would be, with another model, would I maybe have more luck ? And another question would be.. are some models more "fitting" for certain styles ? Like, I suppose training a sketch drawing embedding using a photorealist model could be meaningless ?

I'm not sure I understand everything correctly, so thank you for your patience !

[buoyant juniper]

What is stable Japan

[willow coyote]

How does one generate voices? Can't see none at Replicate.com.

[woeful tartan]

deepfloyd as main image creator team probably?

[ocean epoch]

Once its out alot of stuff are gonna be outdated

[woeful tartan]

they saying soon..lol. i hope so

[ocean epoch]

I hope its possible to finetune that

[austere summit]

anyone knows if runwayML gen-1 will be open source?

[trail wing]

if it is, damn!

but that's gotta be a vram gobbler

[sweet perch]

Why the name change?

[ocean epoch]

bcs they feel like it

[sweet perch]

Slay

[blissful zephyr]

what is pow?

[trail wing]

somebody announced something and someone got their knickers in a knot, that's my hunch

[fervent thunder]

What even are the other projects stable is working on?

[warm junco]

Picture of the Week

[red oxide]

I thought it was prisoner of war

[narrow sluice]

I got confused cause they changed their name and icon

[tidal bough]

check out the #1072229057707659404 channel for details and links to their socials and servers! lots of cool happenings happening over there ^^

[narrow sluice]

And really reorganized their channels

[charred notch]

what is deepfloyd??

[slow thorn]

hello everyone. a few days ago I learned about unstable diffusion. since their suberddits have been wiped and I can't find them here either, does anybody know where that clique gathers now?

[tidal bough]

#1072220168534642768 and the announcement post have some good info for finding your way around! Let me know if there's anything specific I can help with though!

[frosty turret]

is it just me somehow or have the art channels went poof?

i'm so confused right now

[wintry stream]

A lot of channels were recently shuffled around - you should be able to find most of them under the "Dreamer Communities" category!

[frosty turret]

why tho, they where fine? why break something that isn't broke?

[wicked hollow]

You can read the announcement for more info here: #📣｜announcements message

[graceful elk]

the desire to be innovative and make things worse, like a nintendo game sequel

[wicked hollow]

Looking forward to BotW2 then?

[frosty turret]

I agree, no need to innovate whats already working, add to it.no need to rearrange and confuse the great community. good way to have NO community

[graceful elk]

that will probably be one of the rare exceptions, hopefully

[tidal bough]

We've restructured them into forums to help keep things organised and to easily expand the categories with new threads going forward! The old dreamer communities are still in the archive as read-only channels with all the contents intact; I'll add you the community archive role so you can access them too!
Have a browse of the new forums and drop any feedback in #1002293361526460608 if you'd like please! We want y'all to have the best experience with the communities and we're absolutely open to making changes to achieve that! 💜

Apologies the role isn't already up in the Roles channel; we'll have it there shortly ^^

[frosty turret]

forums was the worst Idea discord could impliment.. ever.. th is is not the web and doesn't need to be like it. at all!

excuse my bluntness but this was a error in judgement and frankly did not need to happen.

makes about as much sense as a screen door on a submarine

[maiden crystal]

Are there any artists here, that are genuinely upset or concerned about Ai art affecting their profession?

[frosty turret]

I can't even find'em

[wicked hollow]

I'm divided on the issue. It looks terrible, but the idea works quite well. I just hope it promotes activity rather than curtails it.

[frosty turret]

so.. yeah.. disgruntled, sigh

[cloud granite]

I also got one question, how do you disable image posting in a certain text channel on a Discord Server? I wanna do that as well.

[tidal bough]

no problem at all, thanks for the feedback! They do take some getting used to! If you're used to the old channel format and sidebar, you can "follow" your favourite threads and pin them to your sidebar, and you can expand the window to a full-screen channel kinda view too

[frosty turret]

I loath forum layouts on platforms that do not need it. thought the net had gotten past that.. oi vey (facepalms)

[graceful elk]

id imagine jobs like coding itself, will be hit the hardest once ai gets rolling

[tidal bough]

We'll absolutely be listening to all y'all's feedback, keep it coming in server suggestions please! ^^

[frosty turret]

what is this the 90's again?

[tidal bough]

If you have the perms in your server, you should be able to access the channel setting permissions and switch off uploading files for whichever roles you'd like

[cloud granite]

I own a Discord Server yes

[frosty turret]

sighs.. well this server just ... no words.. no more words.. i'm just at a loss now.. not sure what else to say. politely anyhow so i'm refraining

[wicked hollow]

Well, 90s didn't have the immediacy and presentation that forums do with discord. Not a huge fan thus far, but regarding the lack of space that we had before in the thread list, I don't have an alternate solution that's better. Do you?

[frosty turret]

sure do put it back the way it was. solved.

[wicked hollow]

and regarding the lack of space that we had before in the thread list? How do we solve that?

How do we have tags for things with such a system? Or visual representation of threads?

[frosty turret]

people learn how to scroll and use sub-channnels. right now I don't even know wher the channels wnet, a more clear explanation and listing of channels and how to get to them would be nice.

[wicked hollow]

"a more clear explanation and listing of channels and how to get to them would be nice." I agree with this, though I think it has been stated to a lesser extent.

[frosty turret]

sorry my adhd and other things has me going in circles in frustration

[charred mesa]

😟 💬 They removed the anime channel!
😟 🫱 🗓️ 1984

[wicked hollow]

Lol.

No

[charred mesa]

Are they archived?

[wicked hollow]

It's here. https://discord.com/channels/1002292111942635562/1072013871730131004

[frosty turret]

wait now i'm seeing archive communities

[wintry stream]

Keep in mind we're welcoming all constructive feedback for changes so we can continue to improve & find the best solution for everyone!

We do need to give the new changes a bit of breathing room, but don't get caught in personal disagreements just because people prefer things in a way different than you do. We want to hear from everyone no matter what their preference is 🙂

[wicked hollow]

"constructive feedback" Being the key term here.

I'm sure most like Kumoa love the server and want what's best for it.

[tidal bough]

@cloud granite @charred mesa I've added you both the community archive role so you can view the archived channels! The old dreamer communities are there as read-only with everything still intact ^^

[frosty turret]

I'm trying to be conscrutive. but... so frustrated and confused

[cloud granite]

Thanks

[charred mesa]

Woah woah I was actually joking on their behalf, but thank you anyway!

I'll check out what I've missed

[wintry stream]

Since we're giving it a bit of time to let people get used to the changes, there is absolutely nothing wrong if you wanted to give yourself some time to evaluate the new look & feel before taking a definitive stance on it - there's no rush!

Then you can 100% bring your finding and feelings/suggestions once you feel confident in your evaluation

[frosty turret]

so.. how do I find these new "forums"? i'm so confused

[cloud granite]

I was talking about turning off the "Attach Files" option in my own Discord Server, yet I still see the "plus" button on the chat input.

[wicked hollow]

Under the Dreamer Communities section.

[wintry stream]

Make sure you have them selected in id:browse then they will appear under the "Dreamer Communities" category

[maiden crystal]

anyone know of a actually good free ai upscale software? Ive been using upscayl. its okay but seems not to allow you to set upscale amount, and seems limited. but its free. wonder if there is a better way

[twilit knot]

The No Text To Speed Vid was banger

[wicked hollow]

Outside of SD? Cupscale.

[tidal bough]

hmm, could be tied to your role as the server owner maybe? You can right click on one of your server's roles and view the server from that perspective, and see if image posting's disabled for your members

[maiden crystal]

ive heard of it, but ill give it another look thanks

[cloud granite]

I could try

[smoky tiger]

Is stable diffusion any good on AMD Gpus?

[wild steppe]

Can you visit this link? #1070829982252798043

The new communities should be visible for everyone by default

[charred mesa]

I recommend chainner

[maiden crystal]

oh i remember Cupscale, i think the tree link thing i forget what this is called, i found it totally overkill for just upscaling a image.

[faint plume]

1v1 me on chess

[maiden crystal]

Wait nevermind

[frosty turret]

ok now there's just to many things to keep up with

[maiden crystal]

i was thinking of a different app i tried last year.. Ill give cupscale a try

[frosty turret]

sighs

[charred mesa]

Why the L reaction though he just dec-

[faint plume]

I meant it as "your loss"

[charred mesa]

ahhh

[frosty turret]

for someone with ADHD/ADD whatever you want to call it this new format with previews is distracting and I can't stay focused

[cloud granite]

ok

[faint plume]

Sorry for the miscommunication

[wicked hollow]

It always takes a bit to get use to a new format. I'm giving myself time to like it as well.

[frosty turret]

anyone on the spectrum is gonna have different levels of difficulties with this

[wintry stream]

You may find this option helpful in the form settings - At the top left click on "Sort & View" and change it to "List".

That will make it way more compact & clean so your screen is less "Busy" then you can also sort by clicking on the tags at the top to further specify what you see and make it easier to visually digest

[frosty turret]

k, thanks

[wicked hollow]

I didn't know such a thing existed. This is fantastic.

[trail wing]

do I have to click "open in full view" each time I go to vehicles or some forum like htat

that's a nnoying af

[charred mesa]

Wow this list display is way better 👍

[frosty turret]

ah ha and can open in full view

nice. ok solved. whew. ok my anxiety is going down now

[charred mesa]

almost like if I was browsing channels and then entering one

[wild steppe]

You can follow the thread and it'll be pinned to your sidebar like a normal channel and always in full view, except you choose which ones you want to show up if its a channel you browse often.

[trail wing]

so I do have to click "open in full view" each time...

[wintry stream]

Not if you follow the steps Jae replied to you with

[frosty turret]

seems to be less channels

[trail wing]

see, those steps weren't really useful because one is sticking to the sidebar and one isn't but thanks for repeating the importance of obeying

[wild steppe]

11 more channels in fact, but less cluttered on the sidebar

[frosty turret]

odd i'm not seeing them

eh i'll figure it ou

out

[wintry stream]

That's weird since I was able to get as many as I want - I'll DM you a screenshot of what It looks like for me so we can figure it out

[wild steppe]

Once they're attached to your sidebar, they should open in full view by default

[hasty rose]

Hello how can I general an AI image here? Am new

[wicked hollow]

#1072220168534642768

[frosty turret]

ok got my channels back, except horror, that one freaked me out

sorry I don't do good with drastic change cause's my anxiety to go nuts

there needs to be a FAQ about how to do all this for others. just don't ask me to write it cause I do not explain well.

which is ironic since I now do chat and email support for my job as well as customer service on phone

so yeah this past week has been all over for me

again i apologize for freaking out

reattaches his head

goes back to his dark corner

[tidal bough]

@frosty turret No problem at all, appreciate the feedback! Glad you got your channel setup how you like it; We'll have a look into an FAQ style guide to navigating the forums

[maiden crystal]

the cupescale is missing embedded python stuff, but its stuck at a certain part of install. it wont progress.

says " installed 0 MB if 2500 "

and its just stuck.

Initializing...
Checking disk space before installation...
Available disk space on the current drive: 169881 MB.
Downloading compressed python runtime - 100%
Done downloading!
Installed 0 MB of 2500

[wicked hollow]

Oh yeah, It's been so long since I installed it I forgot there was a trick to it. Lemme see if I can find it in the server history. I think I talked about it on here.

[maiden crystal]

thanks .

[wicked hollow]

Oh wow, looks like we had the author of Cupscale here once upon a time. Still looking for my link though.

[maiden crystal]

Okay thanks. Ill be here just ping me when and if you can find anything thanks a lot.

[wicked hollow]

@maiden crystal Alright, I know there was something specific to the issue but I can't remember the full of it. What I recommend for now is reinstalling from scratch, then when it first starts, (I'm trying to remember my actions from memory here) exit out of the first start prompt that pops up, then click the gear icon at the top right and then in the window that pops up click the AI Settings tab on the left. Then next to "Python Runtime" choose "Use Embedded Python If Available".

After that, install this: https://github.com/pierro42/Cupscale_diagnostics_updater and follow the "UPDATER" instructions in the page.

Disclaimer: I can't remember if you're supposed to install the diagnostics and run it before running Cupscale or after, so you may need to uninstall and reinstall Cupscale to get the correct order to things. My apologies that I can't be more specific than this... the months have addled my memory, and it seems I didn't talk about this issue on this server.

[maiden crystal]

okay ill read this and if all else fails, ill just look for alternative app.

yea seen the comments at github, ppl had same issue. But the uninstall reinstall path worked for some. Will give it a try.

[scarlet tapir]

Hard for me to follow - what is now best way to generate with stable diffusion - dream studio?

[fervent thunder]

What are the API limitations on the free trial? why does it only release the free trial and say nothing. How many images can I generate in this test?

[wintry stream]

The official API uses your credits from the Dream Studio account you chose to use the API Key from - credits show here https://beta.dreamstudio.ai/membership

[old marten]

Why did they have to ruin this channel

Discord room, whatever. Ugh I hate when devs fix what isnt broken.

[silver galleon]

have the genre channels disappeared? nature? architecture? I'm so confused

[topaz cypress]

where can I find other models to base my generation on?

[wild steppe]

We've shifted the communities over to forums, both nature and architecture are under #1070829832151240734. You can view the old channels by grabbing the Community Archive role from #👥｜roles!

[silver galleon]

ahhh thank you

[maiden crystal]

for upscaling, im just going to use chaiNNer . From what I read, it can use a lot of different upscale models that are pretty good. ( just complicated to use ) .
Also using Real-ESRGAN with a custom bat to run it, for basic upscale.
found that bat on a YT video.

[wicked hollow]

Whatever works for ya. 👍🏻

[distant gulch]

other languages produce interesting pictures in stable diffusion

"զդսպքյաֆվկիոժվս" is random letters and it produces pictures of mountain ranges

"ზდსპკჯაფვკიოჯვს" random letters produces pictures of people presumably from the ussr

[thin pine]

It's funny, you change the icon of a server that I'm rarely in or viewing and suddenly I have no idea what that server is about/was about/why I joined it.

Is this the stable horde discord?

[vapid tangle]

I don't really understand the announcement

splitting off into 4 branches?

[wild steppe]

This is the Stable Diffusion server

[wild steppe]

Could you elaborate what you mean by 4 branches?

[vapid tangle]

"explore the Stable Communities!
Harmonai, DeepFloyd, Carper, EleutherAI and more!
" in announcements

It seems like some big announcements, maybe its just me but I'm not really understanding what is going on

[waxen pasture]

I think they are just supporting other open source AI developments

Not necessarily merging with them

It's more or less just to let users in this server know about what other possibilities and resources are out there

[wild steppe]

They're part of the Stability AI eco-system, you can learn more at https://stability.ai/ and #1072229057707659404!

[desert swan]

Hey guys,
I just joined the discord, here is a short intro for anyone who cares:

I have been experimenting with stable diffusion for about a month now.
Just trained my first model with Dreambooth on runpod last week, which has me very excited.
And have been creating some crazy stuff every day since, some of which are pretty cool results (that I might share here soon)

However there is 2 things that are currently unclear to me and I was hoping someone here could answer these:

1. When I installed Automatic1111 I followed the instructions on github, which included a link to huggingface and the 1.4 model that I have been using.

Is there any benefits to using this old model or did someone just forget to update the link to a newer version?

From what I understand 1.5 is the most popular model and 2.0 (or 2.1) is better for some and worse for other things.
If I want to use 1.5 do I just get it from huggingface, put it in the models folder and switch the checkpoint from 1.4 to 1.5 in the webui or is there something else I need to update/change for it to work properly?

2. What exactly is SUPER Stable Diffusion 2.0 ?

Aitrepreneur made a video about it calling it the "most complete version of SD" but it looks pretty much like what I am running. Is it actually the same thing or are there some differences?

Like I already mentioned, I didn't follow the installation instructions from that video but those from the automatic1111 github (and using gitpull for the latest version when launching)

[waxen pasture]

as for question one, do pretty much as you said. Download the 1.5 model from huggingface, plop it in the models folder, and use it as you please. I haven't heard anything about your second question so I can't help you there.

[small tree]

❓ ❓ ❓ Anyone have a good, well-rounded set of positive and negative prompts for 2.1? Something that's been honed to work with various genres and sizes of pictures? ❓ ❓ ❓

Please respond directly to my message here so I can see it later in case I don't get to it right away.

[subtle dock]

been seeing some really cool SD animations on tiktok. trying to figure out how they were made

[fervent thunder]

Wait

Is this an ai porn server?

[thin pine]

Except I have the stable diffusion server in my list. And it's still there.

[desert swan]

Cool, thx for replying!

[wild steppe]

Nope

[wild steppe]

https://discord.gg/stablediffusion

[thin pine]

Oh, I was looking at the reddit stable diffusion server

[wild steppe]

No problem! Announcement post here if you missed it #📣｜announcements message

[thin pine]

Why cause so much confusion by changing the name?]

Doesn't really seem necessary.

[raven thorn]

tf is stable foundation 💀

[soft basalt]

what confusion?

what are you guys confused about???

[still plaza]

Wondering the same thing myself.

where 👁️

[whole swallow]

looks like omegalul

[old marten]

this tbh

and why do we have like 5 new different chatrooms

theyre lowkey tryna kill one of my favorite discords I aint even bsing...

[rapid plover]

Hi eveyone!

[digital kiln]

Yo

[night ermine]

hello

i just joined because i had a question

can i stop the training of my hypernetwork early and try it and then continue training at a later point?

im using stable difiusion web ui

[rapid plover]

I'm new to stable diffusion and generative AI and want to use it to design something IRL by training my own models. Because of the nature of what I want to design I don't want to do it outside of my own PC

I had some luck online with my own images but after that took a break cause life.