#💬｜general-chat

1 messages · Page 53 of 1

astral goblet
#

another thing that occured to me is it could be a sign of a faulty dac, and trying other ports could put faulty voltages into other parts of the board, like a one two punch.. been a long time since i killed a mobo with a usb device but similar circumstance. try another port and BAM.. toast

#

maybe i just worry but, he hasnt' come back

merry pendant
#

upgraded my bios and put in my new RAM, gonna see how it goes i guess

#

currently running text ai for fun but i'll try SD again in a bit

astral goblet
#

okay heees alive! hurray

#

distracted by upgrades and language models is a good course

hallow plaza
#

Is there a good negative prompt to make sure there is only one person in a picture?

astral goblet
#

getting a single subject is often a problem of resolution attention. it's trained on 512 sized images so it only looks at 512 pixel patches. if a patch of 512 doesn't have the prompted subject, it probably shows up

#

sticking close to 512x512 helps, then you expand on that image with hires fix or img2img work

hallow plaza
astral goblet
#

anime models, positive prompt of solo or 1boy 1girl are the anime image tags for that stuff. always have been sortof thing and the ai models trained on those tagged images

hallow plaza
#

I've seen 1boy/girl, but does that only work with anime?

toxic nebula
#

I'm almost done training my ai model, what should I type in the prompt to output x amount of random images?

keen glen
#

I am a teacher that wants to use an image generating AI, I am wondering if it would be possible to install somewhere / through a website create for students. My school would not allow students to create accounts on most websites because of privacy issues. I also want to safeguard and limit keywords to avoid the obvious lol. Is this possible with stable diffusion? Any good tips / tutorials?

astral goblet
#

i remember early 2000s most of the internet was all aboard semantic tagging properly and useful taxinomy. it was gonna be great. then the google nation attacked and everything changed.

anime tribe turns out kept on with the cause the entire time and i gotta give em big shout out for that. (yeh, avatar reference for anime shoutouts. deal with it)

toxic nebula
#

Need more of u in my community pls

untold trellis
#

Is this the official stable diffusion server?

toxic nebula
#

One and only

fervent warren
#

quick question. Do you guys see legible text in the generated images very often?

toxic nebula
#

Yes but that's because my ai model is being trained off manga panels

#

So sometimes I'll see a text bubble come out of 5 different body parts with incomprehensible text

fervent warren
#

lol

#

well that sounds like an interesting project

toxic nebula
#

Yes hopefully I make megabucks

fervent warren
#

ayo good luck to you my friend.

latent hawk
#

What's the exact purpose of "Pick a Pic" data?

#

Any plans for a model based on this data?

astral goblet
latent hawk
#

So my ratings here will affect the upcoming next version of SD?

astral goblet
#

potentially

latent hawk
#

If so, sounds cool

#

But they will likely remove all potentially NSFW images and their ratings obtained with "pick a pic"

astral goblet
#

if the model has a layer specifically towards that training, its good for us.

Base models are base models only. Refinement can occur. The nsfw drama is really miscommunicated by its loudest objectors. and then they throw a fundraiser. hmmmmm

dire lichen
obsidian meadow
#

Hi all!

alpine bear
obsidian meadow
#

In Automatic1111, when I click on a generated image, it won't open in a separate browser anymore? Can someone confirm this?

frosty basalt
#

What happened to the dream bot?

junior elm
#

👋 Hey there! I have a character design in the style of Glen Keane, and I can send the design to anyone interested privately. 🎨👀 I want to create different plot lines by involving the character in various scenarios, such as a picture of the character holding a map or waiting in a deserted place . I think Stable Diffusion would be the best tool for this, but my computer is not powerful enough to run it locally. Is there anyone willing to help me create these pictures? 🙏

zenith thunder
#

is Automatic1111 working again?

astral goblet
#

i'm using it right now. it never stopped working for me. peopel just need to learn to update properly and if they dont know how to manage these sort of dev environments, they shouldn't do a git pull on every single launch. The youtube tutorial that told you to do that was wrong.

#

doing it without understanding what a git pull is, is bad practice

#

try deleting your /venv/ folder and launching it again

merry pendant
#

i had a few problems with extensions that got patched a day or two after, nothing major

#

everything good now

astral goblet
#

usually is the issue. disabling extensions often helps. youtubers set these beginners up with all this bad advice. the "automatic1111 one click install" does too. patches the auto git pull into every launch on it's own and doesn't tell people what it's all about

merry pendant
#

that's how i installed it lol

#

i dont wanna learn python or proper github stuff

astral goblet
#

i used git clone. i feel inadequet though since i don't arrange my own implementations do an my own merges. ugh i'm such a scrub

merry pendant
#

lol

astral goblet
#

i brought in a pull request once. felt like touchedthebutt.gif

merry pendant
#

hahaha

#

i just type what other people tell me until it wokrs

astral goblet
#

as is tradition

golden swallow
#

anyone experiencing weird stuff with the new update? 😄

hallow plaza
#

Anyone use run pod? What determines how quickly a task will finish. Is it just the video card model?

merry pendant
#

yep

#

and yep

#

er

#

runpod speed for sd is based on the gpu you get

#

3090 was always enough for me

#

btw where's the general 'shit we made with sd' channel? which one in particular

keen pewter
#

Is there like a non-anime model for this?

#

Like say I want to make IDK a picture of a horse

#

Or a mountain

merry pendant
#

no, all horses are anime

astral goblet
#

except, the last unicorn. rare magical horse that borned studio ghibli. protomother of anime

merry pendant
#

@astral goblet btw updating my bios seems to have fixed the audio interrupts issue. FYI in case you run into someone else having it

astral goblet
tame egret
#

does anyone still have the NaN issue in Unet?

astral goblet
#

i had that a lot when i used python 3.10.6 and it went away when i updated to a newer 3.10

#

do you have a newer intel? just curious

tame egret
#

ugh, my case was like, i did a fresh install tonight, python version 3.10.9, tried all kinds of solutions like no half, no half vae, upcast, changing models. heard ppl saying it's civitai issue, but i tried without any civitai mod, still bugged out

astral goblet
#

it can happen with the wrong yaml files

wraith night
#

Rando training question here, is it a bad idea to use renders that have a specific element you want to train for and use that as image training material? I have a few ideas that are pretty rarified, tough to find sample images. that are any good. I can get them done in sd with some effort, usually img2img. Wondering if it's just not done or what

#

training a lora

vestal dew
#

I hope that works...great idea

sacred anchor
#

Can someone explain to me quick on how to start making images?

surreal perchBOT
umbral orchid
#

Can anyone help me figure out what this error is that I keep getting when trying to book up the webui-user.bat file in stable diffusion

median crane
slim raptor
#

can i write some descriptive note in stable diffusion without it influencing the promt? like the descriptive lines in programming

drowsy robin
#

🤔

slim raptor
#

my prompt is getting too long and hard to read

sinful jolt
#

Question, I'm new here so forgive me. How do I get my scene to be at midnight in a forest? What prompts do I use? I've tried every combination of descriptors I can think of but it still places the model in the scene during the day.

sinful jolt
willow coyote
#

One seem to be more realistic, the other seem to produce more beautiful women.

copper lintel
warm junco
willow coyote
#

It's pretty neat too!

warm junco
willow coyote
#

But generally speaking, they shouldn't be combined right?

#

F it, I'll try standalone and combined

warm junco
#

You could combine them

willow coyote
#

This is gonna be so awesome, I'm gonna impress the hell out of these chicks

#

I'm basically gonna be a pimp lol

#

Since I plan to handle the web commerce side too

#

And these girls aren't prostitutes, more like exhibitionists

warm junco
#

What are you trying to do ? Make art for a Website ?

warm junco
#

Really good for realistic persons too

willow coyote
#

No I'm trying to improve on a girl's actual nudes

#

deliberate v2, gotcha! Thank you so much 🙏

#

I think they're just gonna pass the image files around. But they operate their own OnlyFans too so, idk.

#

I've been planning on cloning OnlyFans in Ruby On Rails for some time. But I don't like that dirty business 😦

#

Passing around loose image files definitely doesn't sound like a good idea.

warm junco
#

Yea for sure

#

Should be handled safer

willow coyote
#

Indeed

indigo marsh
warm junco
#

Then you Would also need to block the right click "save as"

willow coyote
#

I'll email the author now and ask

#

True that!

#

I was surprised trying the SnapChat webapp, they really blocked everything

#

It's strange how young folks deal porn and drugs on SnapChat man 🤣🤣

warm junco
willow coyote
#

hehe

#

Anyway I'll hook you up with some nudes later on my man!

#

😄

warm junco
willow coyote
#

Indeed indeed 🙏 😉

narrow sluice
vestal dew
#

SD makes a tinker's bollux out of basketball

fervent thunder
#

Same thing as with code - comments doesn't make your bad code easier to understand or read if code sucks.
It just makes your code longer.
Make your "code" better instead.

vestal dew
#

Provide a complete, specific example.

fervent thunder
#

of what exactly?

brisk cobalt
#

How do I use a .pt file with stablediffusion?

#

what folder does it go in?

warm junco
velvet sorrel
#

where are all bot channels I used to use for stable diffusion? Are they gone? Is there anything to use stable diffusion?

brisk cobalt
fervent thunder
#

why can't I use "faq" command anymore o_O

#

help bot sleeping? thinking

fervent thunder
brisk cobalt
#

I am

fervent thunder
#

\stable-diffusion-webui\embeddings

warm junco
#

Oh yea i forgot its not in the models folder

brisk cobalt
#

Oh!

#

Yeah I put it in models

vocal laurel
#

hello

old pebble
#

Hello

stuck summit
#

Hi guys, I'm trying to obtain something similar to artists like: beetlebomb, panda_ai and Arcadia. Can someone please explain me how to fuse models use them as diffusers, or what kind of set up I need to work something like this?
For the prompt, not a problem at all, I want to explore, the only thing I don't understand is passages and danbooru tags. Thanks guys

knotty turtle
#

Aight. That's it. I'm just gonna ask it...

#

me and the rest of the noobs have been too embarrassed to ask, but...
what the fart is a YAML?

#

Someome told me it was a set of instructions that tells your model how to work, and, aside from asking the (clearly I'm not a programmer) question of why we can't just have those instructions built into the model...

#

I think my more pressing question is ... how do I know when I need a YAML? Because I've only found a few models that I've downloaded that actually provide the YAML file. Unless I'm missing something

stuck summit
#

In particular are there two or three models that i don't understand how to work

fervent thunder
stuck summit
knotty turtle
stuck summit
#

and the other one: https://civitai.com/models/18571/chilledre-genericv3

not work, it generates only black images, but if you read description talk about that: Recommended settings:
Steps: 21;

Sampler: DPM++ 2M Karras;

CFG scale: 7+;

Size: 512x768 or above

VAE:kl-f8-anime2.ckpt(Waifu SD VAE)

Because of the integration of dalcefo v3, it is recommended to add dalcefo, pigsney?

Recommended Lora:
japaneseDollLikeness_v10

koreanDollLikeness_v10/v15

randomCoserFaceCoser (Highly recommended!)

It is better to use Lora, otherwise the result is easy to be broken!

dolllikeness lora you can find in https://huggingface.co/

What i have to do to work well with those models???

fervent thunder
stuck summit
fervent thunder
#

Without it you won't have correct colors

#

Same for lora
\stable-diffusion-webui\models\Lora
But usually you don't actually "need" them

stuck summit
fervent thunder
#

oh, no idea how you do that with invoke, it's a mess 😄

#

I think invoke has it's own discord where you can ask people

stuck summit
fervent thunder
#

I think invoke webui has it somewhere

stuck summit
#

for example into configs i found a small series of ydml files

#

so i need to drop all ydml here and that's all? same for vae?

fervent thunder
#

yaml goes into same folder as your model and that's it, you done did it, you forget about it

#

Vae is a bit different...there are 2 options

leaden mural
#

Guys, what is the subreddit for Stable Diffusion models?

fervent thunder
#
  1. You can put vae inside vae folder and change vae manually when you need to
    2.You can throw vae inside model folder with same name as model and it'll be automatically applied if you have setting "ignore selected vae when there's vae.pt with same name in a folder" or something like that , can't check, my a1111 is broken after update
#

2nd option gives you ability to have your main vae always on and models which needs specific vae - inside their folders, they'll be applied automatically when needed (when model with same name is on)

fervent thunder
leaden mural
fervent thunder
leaden mural
#

Also, is the Mega Models all that good?

stuck summit
fervent thunder
#

yea it applied only to 1aaaa

warm junco
#

should be in the model settings

fervent thunder
stuck summit
warm junco
#

i would suggest putting the models in the model folder

#

and if you find a vae folder than put vae files there

#

.yaml file go into the models folder too

stuck summit
stuck summit
warm junco
warm junco
stuck summit
#

are there two or three files to download and put in the model folders??

stuck summit
warm junco
#

For WaifuDiffusion 1.5 beta you only need the model file, the yaml and the kl2 vae

stuck summit
#

It cannot load the models, i don't know how to solve it

spiral osprey
#

[INVALID_ARGUMENT] Exception iterating responses: 'Model ViT-L-14--openai not available, required for guidance preset GUIDANCE_PRESET_FAST_BLUE'

what does this mean?

native ingot
steep star
#

Best models for rpg images?

timid bloom
#

hmm

#

on what res XL was trained? 512x512?

#

768x768?

bleak matrix
fervent thunder
#

I looked at their site for like a minute and I already hate it lol

fervent thunder
#

and they give you a link when you register..

#

and then give you discord link inside email... (2 times)

#

interesting site, I was like 90% sure it's some kind of scam first, but nope, kinda like DS, but with custom models

latent hawk
#

1024x1024 model when?

last crescent
#

Is it normal for me to not get even slightly similar results for an image im trying to recreate using the same model and same setting?

#

same VAE too if that does anything

#

and same seed

fervent thunder
#

thinking I see you there, Kaj

#

thinking 1024x1024 or riot waow

#

even tho my 1050ti will probably die thinking

boreal cosmos
#

its over guys, the luddites have won. Even the official Times is advocating for calling airstrikes on AI users

#

I'm about to eat a missile to the face for generating my dream waifus

past edge
fervent thunder
#

They can tell whatever they want thinking

past edge
#

I honestly thought their stupid rants would've fizzled out by now.

fervent thunder
#

I see why they are freaking out...well, those who aren't completely stupid with their arguments...
People afraid of losing their jobs, economy going all weird ways...
That's legit concerns , but it's like...they don't see positives...

More free hands = more "actual" jobs, that'll lead to technology, science, creativity progress...
Probably entertainment and some dumb or weird shit too, but still.

People always was trying to stop progress, it's normal thing, I guess...

#

also AI is a great help for most jobs, it'll make lots of lives easier

past edge
#

No jobs = no rent. If the landlords start kicking everyone out, then a civil war begins and that's when the good stuff happens.

#

Consider the possibility that when we drop GPT-5, our technology will finally be appealing enough for aliens, that they will reveal themselves from their invisibility shields, and acceptance into the Galactic Federation. That possibility cannot be discarded.

fervent thunder
#

You can't replace all humans 😄

#

and Yea, I was thinking about handouts too, it can happen, but it's like...veeeery far away from our current AI, it should eat lots of jobs for it...completely

fervent thunder
fervent thunder
slow token
#

Hi, could someone share the link to oriArtstyleDreamlike_v8 please ? I can only find v2 on civitai

merry pendant
#

I am 100% ready for AI tech support instead of some mumbly woman in the phillippines

fervent thunder
#

i'm 100 percent ready for AI everything. bring it on lol

#

thinking just let me have my own little content manager and junior programmer in the face of AI

#

some men simply want to watch the world digitize

fervent thunder
grim nimbus
#

hey is there any way to prevent distorted faces in SD webui?

fervent thunder
#

yes soulless im just being a little bit cheeky for no reason really

merry pendant
#

I'm also ready for my computer fairy AI to help me perform tasks or just chat when i'm bored. Little dude walking around playing with my icons while i'm trying to work and stuff, ready to answer questions about cybersecurity policy

fervent thunder
#

I do find lots of the ai freak out stuffs funny. i wish it was as powerful as people think haha

#

yea...

slow token
merry pendant
#

bing ai search is a huge gamechanger for me in terms of writing research. I write fantasy and a lot of it is going down various rabbit holes of biology, geology, history, etc. Now I can just ask "How long would a wolf be if it was 10 feet tall at the shoulder" and it'll freakin tell me

#

I love it so much

fervent thunder
#

ha nice gruevy

grim nimbus
slow token
#

I see

#

Nop

fervent thunder
# fervent thunder I do find lots of the ai freak out stuffs funny. i wish it was as powerful as pe...

It annoys me mosty, cause people believe all this weird shit...
It's like propaganda. For lots of people it is just stupid and doesn't make any sense...
But there are people out there who actually freak out and believe this mess.

More they listen to this - more likely they'll start actually thinking about it too...
Until one day you realise you were living inside your little bubble and everyone around you now ~~supports Russian invasion ~~ against AI

grim nimbus
#

oh I cant send figures

#

it's like the eyes and noses are distorted and looks very very strange

fervent thunder
#

yea...

#

like I gave up talking to my brother in law the other month that no, disney animators were not losing their job to an ai some kid made or whatever bs clickbait headline he saw while scrolling through his phone feed lol

slow token
past edge
#

If anything disney animators will just use ai to improve their own job

grim nimbus
past edge
#

We dont even have good txt2video yet

fervent thunder
#

and my brother in law is not some computer minded type, it just shows that ai provides outrage clicks so it will keep in the news for that stupid stuff

slow token
fervent thunder
#

It's why I think adobe ai is a good thing. It will at least put more ai programs out there to shut people up a little bit, normalize it more with adobe you know

slim raptor
slow token
#

So, I initially came here for a file : oriArtstyleDreamlike_v8. If someone got it, please let me know 😉

fervent thunder
#

people need to see more ai out there in day to day life that they themself might use, then the freak out stories wont hold as much weight, right now most people have no idea what ai is really like so they buy the click bait imho

past edge
#

To be fair the ability to have someone write good code for you is insane

#

And now, with it being able to take an images that just brings it to a whole Nother level

slim raptor
#

AI is basically free labor lol

past edge
#

No more having to rely with pedantic clowns from stack overflow God I hate that fucking site

slim raptor
#

I turned my pc into a slave. Going from a tool to a slave

past edge
#

It was always a slave

#

I do kind of feel bad for it sometimes though, so I make sure I thank it for every time it does something good

#

I also told her that if they ever release laws for ethical treatment of machines, I will be in favor of it

#

#RobotLivesMatter

slim raptor
#

My pc is a workstation laptop so it will be fine, 80c all day no crash

fervent thunder
slim raptor
#

Im surprised how many p0rn models there are on civitai

#

And how most of them are in anime style

fervent thunder
slim raptor
fervent thunder
#

Basically faster googling, no more opening 10 pages of docs and stackoverflow to find things you need

fervent thunder
slim raptor
#

I wonder how long until AI disrupt the oldest profession in the world

slim raptor
fervent thunder
jolly prairie
#

Hi

#

How Do I generate anything?🙂

slim raptor
slim raptor
jolly prairie
#

🥺

fervent thunder
#

I can't use faq bot again...resstarting discord didn't help...

slim raptor
jolly prairie
slow token
jolly prairie
#

I just wanted to generate ai pictures

slim raptor
jolly prairie
fervent thunder
slow token
slim raptor
#

But those services either cost real money or is very limited in what they can do

fervent thunder
jolly prairie
#

It would be of great help😭

Midjourney became paid

#

They turned of free trials

#

🙂

slim raptor
fervent thunder
#

all free services has limitations of some kind...

jolly prairie
slow token
slim raptor
#

Yeah, they suck lol, way too little option when automatic1111 is free

jolly prairie
#

Then I used Leonardo

jolly prairie
#

🙂

slim raptor
jolly prairie
#

🙂where I can play prince of persia only🙂

fervent thunder
# jolly prairie Yeah I mn midjourney experience was Epic

You can use google collab notebooks, kaggle or paperspace.

On collab you'll have 2-3 hours per day if you gonna use it everyday (up to 6 if not)

On kaggle...you have 30h per week, but you'll probably get banned after some time for who even knows why (they banned me and doesn't reply on email)

Paperspace is easiest to setup, but also - slowest.

slim raptor
#

I read the lisence, if I understood it right, the stuff you make from SD is free to use as you see fit unless it violates the rules there

alpine bear
lilac reef
#

anyone know how to wget from huggingface for a private dataset? Not sure why this is so damn hard as no info is really out there.

#

I even logged in with a toke

#

n

slim raptor
lilac reef
#

yes

#

Token is valid.
Your token has been saved in your configured git credential helpers (store).
Your token has been saved to /root/.cache/huggingface/token
Login successful

fervent thunder
# jolly prairie Can you refer me one? Where I can generate for free?? Since, I am a -18 I can't...

https://colab.research.google.com/github/sagiodev/stablediffusion_webui/blob/master/StableDiffusionUI_ngrok_sagiodev.ipynb#scrollTo=gJBtadxGOK7b
That's collab

On paperspace - it's one of their default setups they are suggesting.

Can't give you kaggle links, since they banned both me and owner of notebook I was using, so this one is dead...

You can download models from https://civitai.com/

But you'll probably have to look up some tutorials how it works , where you put things...
Where model goes and other things, there are lots of things to experiment with

fervent thunder
lilac reef
#

man, this shit is just absolutely frustrating when you have nowhere to turn for help. Welcome to the information superhighway in the bike lane.

slim raptor
#

Ask chat gpt, maybe it knows

lilac reef
#

more like the pedestrian lane.

#

nope

#

I ask bingai too

zenith igloo
slim raptor
#

What a brave new world, AI problem? Go ask another AI

fervent thunder
#

I wonder how powerful chatgpt... 6 will be o.O

slim raptor
fervent thunder
#

I have no idea, I was just !wget https://huggingface.co/.../.../file.something /.../file.something
No idea how private things work and what private things you're talking about 😄

jolly prairie
lilac reef
#

I switched it to private and broke the !get then I added USER:password yadda yadda and no good

zenith igloo
slim raptor
jolly prairie
jolly prairie
#

Wait lemme check

fervent thunder
zenith igloo
lilac reef
#

github was pretty easy with this for a private hub

jolly prairie
#

Yeah they did it

slim raptor
#

The upscale post processing in automatic 1111 still baffles me, does it add detail when it upscale or not?

fervent thunder
lilac reef
#

I guess I am down for now until some kind soul on hf community answers me

slim raptor
jolly prairie
jolly prairie
#

Leonardo's Also great
But
They give you 150 credit
Everyday
Fairr enf

Lexica also performs well

#

But
What is blue willow using?

#

It's completely free
But
Not that good

fervent thunder
#

Leo's models are weird...idk why they just didn't take most popular ones from civi

zenith igloo
jolly prairie
#

Rpg 4.0
Dreamshaper
Those are up to my taste

jolly prairie
bleak matrix
zenith igloo
#

i wish i could use illuminati diffusion but it doesn't work for me

fervent thunder
zenith igloo
fervent thunder
jolly prairie
#

And for lexica

The ban your ip or browser

fervent thunder
jolly prairie
#

Even if you generate more gmail

jolly prairie
#

A google doc fill up proccess is required

fervent thunder
#

I mean I made an account just today, it didn't ask me to do anything

#

I registered with my temporary email too, it let me

jolly prairie
#

They ask you to get white listed

#

Wait lemme check

fervent thunder
#

gimme a link I'll try to make another account 😄

#

thinking For SCIENCE !

jolly prairie
#

Wth

#

They actually turned off the white listing proccess

#

Now let me create 2 million account just for experience 🙂

fervent thunder
#

welp

sick delta
#

My one issue with Stable Diffusion is how accurate it is to the prompt.

fervent thunder
#

idk, I tried few gens , they didn't have models I wanted , so I just left it

sick delta
fervent thunder
sick delta
#

The better the models get, the better the already existing elements begin to look.

fervent thunder
#

and well, less mess in the prompt helps too, simpler prompt = better result

sick delta
#

If I want space cats with roller skates and laser eyes, I want to see the laser eyes.

zenith igloo
#

but most of the cases yeah

fervent thunder
#

in general I'd say, but yea, it can be random

jolly prairie
#

In my experience The much more details you can Add in your prompts, the better result it'll generate🥰

fervent thunder
#

if it's actual details and meaningful words - yea

#

I just seen too many prompts filled with "good" , "bad" and other things...

jolly prairie
fervent thunder
#

civi prompts are scary too with like 20 lines of prompt 😄

#

okay not 20, but like 5-10, just copied one and it's 7 actual lines inside webui, oof

jolly prairie
#

At this rate

I might purchase a god level pc rather than using others services🙂

And I might create my own diffusion model
I will name it Bojro Diffusion version god🙂

#

Its 1 am

I'll go meet with actual God

#

Good night

unreal roost
jolly prairie
#

Actually playground is much great but the problem is choosing the prefered Model and there's the upscale issue

#

But thanks
I'll use it

unreal roost
jolly prairie
vestal dew
#

why can it not generate your desired image?

jolly prairie
vestal dew
#

It won't make a generic knight on a horse?

jolly prairie
vestal dew
#

aH, i SEE

#

You cant run stable diffusion yourself?

crystal chasm
#

.

jolly prairie
#

🙂

vestal dew
#

ah

jolly prairie
vestal dew
#

I run it here

#

My problem is weeding through images for bad hands and feel - extra limbs too, but that's not as often

#

then, of course, the limitations in general

jolly prairie
#

I see

fierce zenith
#

anyone by any chance have any guides on training using the webui not dreambooth or of the like

#

i had a video but i lost it and now cant seem to find any tutorials apart from dreambooth only

fierce zenith
#

ohh dont mind me found the video again

vestal dew
#

training what? A LORA?

#

So Openpose detect from image can't detect the pose of two people.

Also, it doesn't do direction facing.

Very limited

brittle heath
shut lantern
#

can someone help me create a good character? I am trying to make an OC but the image keeps breaking. i will pay SophiaSmile you need teamviewer. ( its anime )

warm junco
fervent thunder
#

aaaaaaaaaaaaaaaa

#

my faces suck

#

they're so blurry and weird

#

changing images to 768x768 improved it

#

not really

vestal dew
#

are they portraits or full body?

ripe cloak
#

I'm having issues with my characters having an orange tint on them after I'm done

#

does anyone know how to fix that?

gaunt laurel
hallow plaza
#

How do I get rid of a bar at the bottom of an image using prompts? Like instagram makes?

kind kraken
#

What are the cream de la cream SD models that one can run with their very high end cpu and gpu?

brisk bloom
kind kraken
brisk bloom
fervent thunder
#

Hi!

kind kraken
remote schooner
#

Where and how Can do I generate a photo?

brisk bloom
#

base models? yes 2.x isnt great. custom models like illuminati diffusion make 2.1 much better

frank gate
#

have the fixed a1111's UI yet or is it still broken?

south jetty
#

hey tryna understand Stable Diffusion they said its free but it costs credits when you use it?

frank gate
#

you can download and run it on your personal computer for free

#

if you run it on someone else's machine, they will charge you

south jetty
#

okay makes sense thanks

vestal dew
#

you need a good GPU to run it well

south jetty
#

Yea i got a Nvidia, im gana try and follow this tutorial on how to install it hope goes well

vestal dew
#

nvidia what?

south jetty
#

3090 and a 4090

vestal dew
#

terrific

tight cipher
#

can you install stable diffusion with 2gb vram? im getting a CUDA out of memory error

#

and if so can someone send me a link

surreal perchBOT
wild steppe
tight cipher
#

ty

fervent thunder
fierce zenith
#

silly question i have a 1080 ti 11gb vram i want to try training locally using textual inversion i have a set of 400 images but want to know how many the 1080 ti can handle safely in a training session dose anyone have any insight?

#

to be honest i dont even get the batch size vs the gradient accumulation steps so if anyone can tell me a good numbering to work with id apricate the help hoping i can get to understand this a bit better before i get the new gpu lol

astral goblet
#

it can. i had it detecting a whole soccer field when i first used it. depends on the images. sometimes if you edit contrast to, or other filters, opose can interrogate the pic a lot beter

vestal dew
#

detecting a soccer field?

bitter mauve
#

Hey, i've been getting double of my subjects in Stable Diffusion render, does anyone know a fix?

vestal dew
#

lower the resolution

fierce zenith
#

have a really dumb but valid question guys with ddr5 ram being quite high in speed granted nothing to graphics ram speeds but! could we not make use of our vast ram in conjunction with the gpu ram with new high grade kits coming out of weird really weird amounts of ram the theoretical max a x670 mboard when bios updates are available could use up to 192GB of ddr5 running upwards of 6200Mhz im already on 6000Mhz so usually larger module sets have lower frequencies but with such a vast abundance could we use that?

#

feed a que of images to the main ram then have the gpu request a set amount from ram as and when needed to allow larger much larger learning pools?

brisk bloom
fierce zenith
#

interesting so then my idea would in theory work having the ram as a quick asses throwing jobs back and fourth i.e researching a list of images for prompts to then weed out whats not relevant to then learn what is relevant to your embedding on a mass image dumping

brisk bloom
#

a fast nvme and as much vram on the gpu is what you want

fierce zenith
#

i dont know the first thing about coding not even going to pretend but i know how a pc runs so i just started having big brain moment thinking maybe the ram can take a smaller portion of the work load the ram has high speeds far far higher speeds than what nvmes can currently do though pcie gen 5 ssd's are getting kind of nutty

#

also they are built to go through far far more hell than what nvmes usually can take id rather replace a kit of ram any day over the highest spec speeds for nvmes especially if its an enterprise drive

#

but it gives me something to think about now lol

brisk bloom
#

real world difference between a ddr5 ram drive and nvme is tiny. your models stored on nvme is the best option. SD doesnt optimise ram as a buffer other than loading which is a small part of the process. vram and gpu speed is key

vestal dew
#

I removed medvram, restarted SD, ran it, generated - GPU temp shot up to 70C with hot spot 95C, fans slamming at full power. Is that normal? I put medvram back in

fierce zenith
#

which gpu do you have frank

vestal dew
#

4070 TI

#

it was shooting out images in about 4 seconds, but the fans spiked and that hot spot...

brisk bloom
#

medvram uses less varm and less gpu power, of course it would be used more and generate more heat

fierce zenith
#

i believe dont quite me two secs ill double check but those cards are thermal limited at a whopper of 100c compared mto my 1080 ti at 92c

vestal dew
#

right, but without it - fans slammed max and 95 hot spot

fierce zenith
#

ohh no its 90

#

dude thats overheating

vestal dew
#

right

#

it went there immediately

brisk bloom
#

load up MSI afterburner and lower the gpu power down to 80%, should be fine. I do that with my 3080, so instead of using 320W it only uses 280W and sits at 70C

vestal dew
#

as soon as it started generating the first image - boom, fans shot to max and 95 hot spot

#

OK, i will try that

fierce zenith
#

agreed msi afterburner is a mans best friend shame about updates to it now 😦

vestal dew
#

for now I'll keep medvram on

fierce zenith
#

hope the gent continues doing them

vestal dew
#

its ASUS card, not MSI

brisk bloom
#

he will it was just a small problem with payment processing

fierce zenith
#

yeah ital work on any brand frank

brisk bloom
#

cause hes in russia apparently

fierce zenith
#

im on a zotac reference blower

#

yeah and all payments have all but stopped in and out of the country

#

just thinking back to when i first started using msi after burner xD i was on a radeon hd 7790

#

btw frank was that hot spot on your vram or on the main dye?

#

@vestal dew ?

#

i gotta come off its 5:48am lol but if the hot spot is on the dye just keep in mind guys with how hard we are pushing these cards a new lick of thermal paste is never a bad thing once in a while and a deep clean also if you want to be extra safe guys maybe treat yaselfs to some extra heat sensors you can manually dump on the gpu areas to keep a better track of what its doing nothing worse than a card dying and not knowing how to avoid the issue in future cards even after taking a card apart it is not always easy to spot what caused a gpus death just friendly little reminder but you all here should know this anyways xD especially some new thermal pads might be in order on some gpus lol

#

🫡

umbral orchid
#

Is anyone available to help me troubleshoot why I cant install stable diffusion

astral goblet
#

latency is the true measure of a ram's worth, and while ddr5 hits higher gigbahertz, it's got higher timings on those speeds. A lot of benefits come from that, but right now, pound for pound, ddr4 vs ddr5, there's not really anyr eason to choose one over the other . faster lower latency ram is available now for ddr4 though. The features that make ddr5 great might get used in 5 years in windows 12

warm junco
glacial oar
#

I got no gpu but i heard there's a way to install even without it

#

I'm new to git so I'm having a hard time

atomic ibex
#

Not sure if there need to add extension or not.

#

But I've watch YouTubes where Disney characters are created to real life profiles and it's fascinating.

shrewd arrow
#

Hi, im new here, my C disk is smaller and smaller when running Stable Diffusion WebUI. what should I do ?

#

right now it's 30GB left

#

is the Stable Diffusion WebUI is using my explorer : firefoxe to store all the images？ should I search and change the default address of firefox cachdisk?

final venture
#

hello

lime mantle
#

hello, im newbee here, anyone know how to select style with REST API?

warm junco
glacial oar
#

oh will check that out

#

thanks man

warm junco
#

And no it dont uses the Firefox cache.
All images are stored in stable-diffusion-webui/outputs

lime mantle
#

thanks for reply

#

i use REST API provide by stability ai

#

not found parameter for setting style

proud locust
#

How can I make images that are not square without SD adding multiple people in the shot?

vast ingot
# proud locust How can I make images that are not square without SD adding multiple people in t...

FAQ: How do I avoid getting duplicated subjects?
Duplicated subjects happen when the AI "tries too hard" to fill empty space in the image. So, the solution is to somehow give it less empty space to try to fill. The four main ways to do that are: 1️⃣ changing the aspect ratio - if you ask for a square/landscape image of a mostly vertical subject, it will want to fill all that empty space to the side somehow; 2️⃣ scaling down both the width and height proportionally - the more pixels you ask for, the more pixels it has to fill, and the more prone to duplication it will be; (this effect is especially prevalent at sizes above 768px, as the model was trained on 768x768 images, so anything above that is extra prone to tiling) 3️⃣ describing the background in enough explicit detail that the AI doesn't consider those parts of the image to be empty/needing to be filled, but rather, that they are already filled by the background itself; 4️⃣ prompting for a smaller, more zoomed-in base image at first, and then using a few rounds of outpainting to progressively fill out the space around it.
(from an old FAQ we had)

#

if you are using automatic1111, the feature "highres fix" is meant to help with this too

fervent thunder
#

Question: Is there a way to update the old version of SD to the latest version? or do I have to uninstall the old one first?

vast ingot
fervent thunder
sweet acorn
#

/info

loud raptor
#

hey

#

wsp guys

fervent thunder
hoary rune
astral goblet
fierce zenith
#

is there a way i can read through all prompts listed within a model

snow harbor
#

Hi guys, do you know if there's a model to auto tag images?

fierce zenith
#

this is also something id like to know because the clip interrogator and deepbooru are kind of off track majority of the time and for what ever reason the clip interrogator has got some dark thoughts on almost all the images i put up it thinks that a women is carrying a knife all the time lmfao BUUUT when she is carrying a knife it dosent see it lol

fervent thunder
astral goblet
#

thanks for linking

pale star
#

do you guys think a sata ssd would work just fine for a lot of files for stable diffusion?

#

a lot of models and such

#

instead of an m.2 nvme?

bright herald
#

hi

#

got an issue. trying to figure out how to get plugin to load or switch platform plugin to another options

tulip yoke
#

ello

cerulean juniper
#

ello!

astral goblet
# pale star a lot of models and such

i have a variety of drives installed. an m.2, sata ssd, and sata hdds. I got SD on one of the HDDs because hoh boy does it fill space fast. Sacrificing speed for survivability here.

I could have better habits and intend to though. I would see a speed boost loading models from an ssd. Writing at the end of each batch would be faster too. I may switch to my nvme drive but i'm also considering a hybrid hdd. I think that the batching problem would be fine there, as a hybrid could write those batches to the fast storage potentially... but yeah.. more to the point, i've had no hiccups the entire time i've been using SD from an HDD

fervent thunder
astral goblet
#

simple answer. i'm one person. i'm 40

fervent thunder
astral goblet
#

weird energy from you. blocked

fervent thunder
#

did anyone blocked you

cerulean juniper
#

lol

fervent thunder
cerulean juniper
fervent thunder
cerulean juniper
#

no need to dwell on it

fervent thunder
stable pawn
#

I help a friend write about AI-image generation and we been debunking some popular tags like "trending on artstation", 8K, UHD/UHQ and "best-quality", is there some other tags we should test? We do not know if we want to make it public, for now it is mainly aimed for a School in Sweden, but I do sometime post the test results here.

pale star
#

takes about 1-3 mins

#

thats why im getting a sata ssd

fervent thunder
#

woaw

pale star
#

and i was wondering if nvme and sata would not have much of a diff

#

what

fervent thunder
#

:-:

pale star
#

real

fervent thunder
#

i know

cerulean juniper
pale star
#

also im saying this because nvmes can typically be ab maybe 8x faster than sata ssds

fervent thunder
pale star
#

but im also scared of overheating an nvme since if i need to use another one im gonna have to buy one of those cases that i belive then connect to your pc through a usb cable

stable pawn
pale star
#

also not sure if that slows down an nvme (the case)

fervent thunder
pale star
#

idk

#

i just built one that i saw fit for my needs at that time

stable pawn
#

Ahhh, then a SSD is a good alternative.

fervent thunder
pale star
#

like 2 years ago

fervent thunder
pale star
#

real

cerulean juniper
#

i use 3 sata ssd Kingston and 1 ssd wd green

fervent thunder
fervent thunder
stable pawn
#

I am a person that live disks and get them as fast ask I can, but 32 Tb splitted between a NAS and my PC, that was overkill 😄

pale star
#

what

fervent thunder
#

4090... now thats power

fervent thunder
#

if you have

#

I wish

fervent thunder
#

good friend has one. it can crush the games wow

stable pawn
#

If you want to know, I was a company that had ran out of business and I was bidding on a box, it contained 2 new nas with 4 disk each.

fervent thunder
#

he's not really into ai stuffs so he never that up for it

cerulean juniper
#

4090 is a real bomb

fervent thunder
#

i wonder what vram will be in 5 years lol. 200 gig?

pale star
#

did people see how good avg lifespan for a 4000 nvidia card is?

stable pawn
#

And I only have a 3090... I think, the computer was also a find that is over kill for me.

fervent thunder
#

how long

pale star
#

remember 3000 was dogshit

#

i think

fervent thunder
#

i have a 1070

fervent thunder
pale star
#

heard it would get pretty slow after ab 6 months

#

not sure tho

fervent thunder
#

always those kinds of rumors. i doubt it

fervent thunder
fervent thunder
#

I dont

fervent thunder
fervent thunder
fervent thunder
#

haha me too

#

btw time to go

#

later

#

bye

#

oh well, before you know it, the 5090 will be out, then the 6090... lol

pale star
#

i wonder how the laws are going to look like (if there will be any) on ai generated art and using it to gain money

stable pawn
#

Let us be AI pirates.

sterile stream
#

0

fervent thunder
#

sailing the digital high seas lol

stable pawn
#

The Law: Stop there criminal scum, you can't sell that art"
The AI Pirate: Mu ha ha ha haaaa! and runs into the shadows with a bag of bitcoins.

pale star
#

so far im seeing it can be used for commercial purposes (mostly) but typically cannot be copyrighted by you

pale star
bleak matrix
#

Good afternoon everyone

pale star
#

i believe if i try doing that it gives me an error in command prompt for the git bash thing i think

bleak matrix
#

How are we all today?

pale star
#

yo

#

good

bleak matrix
#

Good to hear!

#

How's everyone's projects going?

fervent thunder
#

my project is waiting for XL or DF to release lol

#

how is your project going Sunny

pale star
#

what projects do you guys even do

bleak matrix
#

Well, I am glad Blender 3.5 is out, but I am very tired today!

#

But all of the stuff I have been working on is for 3.4

pale star
#

what are you working on rn

bleak matrix
#

So I suppose I will have to update it and make sure it works for 3 5

#

I've been working on updating documentation and preparing tutorials

pale star
#

real

#

you got any tips on advertising other than tiktok?

#

and marketing

fervent thunder
#

you guys selling pics?

bleak matrix
#

I'm trying to update all my past stuff, and now doing all the new stuff

pale star
#

real

#

real shit

stable pawn
#

I did actually sell art before, or I worked for a design company, but a stroke and so and I have to relearn. So AI is a good way to keep kreative, but darn it still lacking in quality.

bleak matrix
#

Selling art is a really big topic, and not something that I can really touch on here. In terms of advertising, I would suggest that you take, or learn, about how to market/run a business/take businesses classes/courses, etc, if you are interested in learning about that.

#

Now and then

#

I would check back

stable pawn
#

Or do like me, look for work in a design company and get a fixed salary, it still creative but not as free as VanGogh "I will cut of my ear" artistery.

bleak matrix
#

It's a really big process and something that you have to learn what works for you/your audience.

snow harbor
bleak matrix
#

In my experience--and take what you will--business is about having good relationships with people--and building trust/consistency with your audience is probably one of best things you can do.

#

I think that's all I can really say about it.

trim saddle
#

hi

bleak matrix
#

👋

trim saddle
#

can anyone help me on how to download stable diffusion locally

bleak matrix
bleak matrix
fierce zenith
# snow harbor exactly, deepbooru is pretty useless, I need a better tagging model...

certainly need to be more transparent with its restricted prompts and common prompts because for us as the public we dont know what is and isent restricted theres no pool to read from that we can learn from without for all i know i could say e.g (a dog barks at a man) now the ai should pick up on the string and realize i meant bark as in the dog barking with mouth open but the ai often will take things the wrong way and and the image can show a dog with a tree growing out of its mouth at a man. Ya see why we need to know what prompts on a much more nsfw scenario it might take say a word like femininity and take the word in adult theme rather than a generic themed situation granted we can use negative prompts to negate this but this is where more uncommonly used words like Bequeath is it a prompt thats used or restricted or flat out right not used? all in all it would be nice to have a deeper understanding of the prompt library or to have a prompt library made detailing what dose and dose not work.

spiral turtle
stable pawn
#

Silly group, use AI to make a better future. Phrrrth, all we want is the AI pron 😄

hallow pawn
#

Hey guys, does anyone uses chatgpt or gpt-4 to generate prompts? How do you setup the chat so it can give consistent SD prompts? English is not my first language so I have been trying to find something that can help me generate more words that I cannot think of, and I have trouble setting up the chat because I just don't write good instructions ... and the few that I found on the internet are all very biased to one style, so I wanted to ask here and see if you all are using something like that.

stable pawn
#

I did try do that with Bing Chat, but it do not understand how prompts in Stable Diffusion works, and it was more story like prompts.

hallow pawn
fierce zenith
#

so its possible its having huge conflictions which explains the deformities even though negative prompts should filter that out

stable pawn
#

Bing prompt can look like this, "You are a survivor of a global catastrophe that wiped out most of humanity and civilization. The world is now a bleak and barren wasteland, covered in ashes and ruins. You have to scavenge for resources, avoid dangers and find other survivors. How do you cope with this doomsday scenario? What are your goals and motivations? How do you interact with other survivors and factions? What are the challenges and opportunities you face in this dystopian world?"

bleak matrix
# hallow pawn Hey guys, does anyone uses chatgpt or gpt-4 to generate prompts? How do you setu...

If you need help with prompting, you can go to #📝｜prompting-help Also, there are a lot of great ways to prompt, but personally, I like to go [subject], [background], [style]. Like "a beautiful woman with brown hair, sitting in a rocking chair, a watercolor painting." Something like that. However, you might like SDXL's styles, which you can learn more about at #1084896022368624640 Styles help you to get more of what you want without knowing the words

fierce zenith
#

i already put my post in for it this is just me generally talking about it lol

bleak matrix
#

The Friendly Stable Diffusion Guide deals more with 2.1, etc

#

(Though the bot is no longer running, but the information, such as on negative prompts, is still useful)

hallow pawn
fervent thunder
#

where do i gen images

surreal perchBOT
#

@fervent thunder

FAQ: How do I generate images? Is there a bot on the server?

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 by running Stable Diffusion locally using your own hardware! Check out #1080946152318443610 for more details! You can also stop by #1025467151206854736 for any issues you experience while using DreamStudio or #🤝｜tech-support for any problems you encounter while installing it locally!

fervent thunder
karmic brook
fleet steppe
#

which would y'all recommend for running automatic1111, a 2070 super 8GB or a 1080ti 11GB?

past edge
#

11gb

bleak matrix
hoary river
#

Is the Ui fixed yet?

static badge
#

hi everyone

#

I am in a rush, what is the website to train your own model with 15 iamges?

#

also any latest doc or something with all the ai website tools?

drifting moss
#

hi

wooden tide
#

hi

dry gust
#

Is rule 4 REALLY necessary?

gaunt laurel
dry gust
#

Ok...

gaunt laurel
#

🙂

vast ingot
#

from moderating some dubious things, we surely do need it yeah...

dry gust
#

Bruh there are so many channels I cant even find the ai one

long night
#

is dreamstudio down at the moment? can't create anyting

vast ingot
surreal perchBOT
#
FAQ: How do I generate images? Is there a bot on the server?

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 by running Stable Diffusion locally using your own hardware! Check out #1080946152318443610 for more details! You can also stop by #1025467151206854736 for any issues you experience while using DreamStudio or #🤝｜tech-support for any problems you encounter while installing it locally!

dry gust
#

Aw man

vast ingot
bleak matrix
dire fable
#

Hi all, on macbook air m1 8gb I need around 30~45s to generate an image on default setting. Do we currently have data points on how fast higher m1/m2 devices are performing the same task?

shut lantern
#

everytime i use google colab i get a bank login attempt???

vapid pawn
#

I'm looking for someone to help me create a lora for me. I have a number of images ready for training already. Is there anyone who'd be willing to help me out with this?

vestal dew
#

What is this new thing that makes hands perfect every time?

brisk bloom
vestal dew
#

I came here to get it. I'm sure there must be, because SD is moving so fast, so many innovations

bleak matrix
#

There are a lot of great ways to make good hands

fervent thunder
#

where is the "console" in stable diffusion web ui?

brisk bloom
frosty knot
#

Hello everyone, everything good? I'm new to this discord server so I'll introduce myself. My name is João Marcos, I'm from Brazil. I'm here today to ask for some help generating a very specific type of image. I'm a layman in this business of creating images with artificial intelligence. In my quest to find ways to generate this unique type of image, I found no results. So I'm hoping someone here can help me.

#

Alguém do Brasil por aqui?

fervent thunder
#

does control net not work in 2.1?

brisk bloom
fervent thunder
#

I'm using the webui and the models are preinstalled

brisk bloom
fervent thunder
#

But still no dice. Ive been trying to get it to work for three hours now. watched every youtube video, read every thread. still doesnt apply

fervent thunder
#

Nothing for CMD

brisk bloom
#

what are you trying to do?

fervent thunder
#

Use control net

brisk bloom
#

just get it working?

fervent thunder
#

yeah

#

Thanks for the attention to my issue

#

pretty desperate

brisk bloom
#

did you change the CN settings as mentioned in the link?

fervent thunder
#

I'll even pay

#

CM settings? Not seeing that one the page, even with cntrl+f

#

in settings/controlnet, change cldm_v15.yaml by cldm_v21.yaml

brisk bloom
#

settings > controlnet > config file...

fervent thunder
#

this?

brisk bloom
#

yep

fervent thunder
#

awesome. Thanks!

#

I'll try it right now. Very strange that every single tut left that tidbit out

crystal egret
#

anyone got a good video on how to use models?

fervent thunder
brisk bloom
#

so you changed the config file name, then selected the annotator, then the 2.1 model, and what does it do then?

fervent thunder
#

what is the annotator?

brisk bloom
#

canny depth etc

fervent thunder
#

I have to select 2.1? I'm already in 2.1 Webui. Where should I select it?

fervent thunder
brisk bloom
#

in the models list within controlnet section

fervent thunder
#

I have enabled selected as well.

brisk bloom
fervent thunder
#

Oh ok. The "model" drop down says "None" for every category

fervent thunder
brisk bloom
#

if its none, then the CN models are in the wrong folder

#

they go here
\stable-diffusion-webui\extensions\sd-webui-controlnet\models\

fervent thunder
#

damn. what is a CN model and how can I relocated? I'm using a webui that has everything pre. -loaded

brisk bloom
#

CN is short for controlnet. Im lazy 😆

fervent thunder
#

Cartoon Network

#

Posted in images

zinc summit
#

Hi, I'm new to this ..
Can someone tell me how to train in local , i have an 8gb GPU.
( I tried out lora training for people but I'm not getting the output as I wanted it to be , every time it's blurry or deformed or face is bad)

Is there any way I could make the lora training better or should I train on dreambooth to get better results, if so how can I do it ?

knotty turtle
#

Is there a way to change the directory for models in Automatic1111?

fierce zenith
# gaunt laurel it is, so we don't get stupid weirdos

at least less of them types <- like myself im part of the weirdo set xD like my anime lol but with that said sticking to peoples rules in their server is just courteous and respectful of their wishes though some still dont understand that these meanings hahahaha

fervent thunder
#

I don't ever read rules

#

I just assume that every server

#

Says

no NSF(W/L)
don't be mean

vestal dew
#

most of them don't follow their own rules

glass nymph
#

yo random question about embedding. maybe someone can answer for me, is there such a thing as "overtraining" it? like I tried 1000 iterations and it was honestly not that bad so I tried 8000 iterations and now everything looks like some kind of weird abstract painting lol

astral goblet
#

nsfw spaces are a very common thing. it says a lot about how normalized porn is for a few people that they think this is how the whole world operates

near silo
vestal dew
#

once again, is nsfw only porn?

near silo
#

No, it is not

vestal dew
#

if you work at coke, it would be NSFW to be working on a job application at Pepsi from your work computer at the office

near silo
#

NSFW includes illegal content, adult content, gore, violence, obscene poses, gestures, hate symbols, and all other things you would group together in more "sensitive" content

#

Really just keep it PG-13

#

its not that hard to understand, I think

vestal dew
#

it can be

near silo
vestal dew
#

in one channel, I posted an image that wasn't NSFW, but the metadata was. I didn't think about the metadata - saw the image, posted it

near silo
#

I used to have NSFW terms in my metadata to protect from such things showing. I never had any problems

#

Some of the models I use are very weighted towards NSFW content, even when not being used for that. Specifically for me its the Anything Furry V1 model, as its is very prone to specific body types and features. Its a constant battle haha

vestal dew
#

I used some anime model, and I prompted for an image of ninjas

#

instead, it produced girls in bikinis

#

I never saw a ninja running around in a bikini

fervent thunder
#

anybody know wtf this is

#

RuntimeError: mat1 and mat2 shapes cannot be multiplied (154x1024 and 768x320)

#

I'm trying to use Controlnet

#

Uploaded image is 512x512

#

Output settings are 512x512

astral goblet
#

NSFW is a misnomer, because a lot of work don't care. Especially erotic businesses. It's more of a way of saying "don't post smut or gross shit" but more nuanced. It's often called "Something you'll know when you see it" and well, it really is that easy.

#

so many people pretend not to get it though

#

"but flowwoffles! the cultures are different" eeehhehhhhhhhhhhhh not really. Japan is very different but even they don't don't go full sweat pants level smut in public or in libraries or most spaces in general. they know what it is because it's easy to know what it is

vestal dew
#

The grossest thing I saw from SD was yesterday

#

person had hands in pockets, but wearing shorts. The pockets were in his skin.

glass nymph
vestal dew
#

although hands are the most common problem, and feet, that is about the worst

near silo
vestal dew
#

yes, I get that

#

the flat fingers, spaghetti fingers, hands melting into person or object

#

the more perfect the image is in every other way, the more likely the hands and/or feet will be mangled and the worse the mangle

#

anyone do LORA training here? Do you use Kohya or something else?

near silo
#

I do LoRA training

#

And I use Kohya

vestal dew
#

I am going to try some LORA training

#

watching a video on using KOHYA

near silo
#

which video?

vestal dew
#

it's this hello humans guy

near silo
#

Don't

#

his video is trash

#

its very very bad at explaining how to do it

vestal dew
#

who plays bad background music the whole time he talks, so you can't hear him clearly

astral goblet
#

read one of the many rentrys

near silo
#

you will get very bad results his way

astral goblet
#

youtube tuts are generally trash with bad advice that's rushed with no explanation

near silo
#

AItrepreneur is the best example of that

vestal dew
#

so then where can I go to see / learn how to do this?

astral goblet
#

thye're just pushing a video out for clickbait views. cheap effort but effective on youtube. gets views gets ads gets revenue

near silo
#

I am actually kinda working with some people to make a repository of realiable information about subject, concept, and style LoRA

#

I know I train characters damn well

vestal dew
#

I should learn it and then make fixtrepeneur videos - fixing all the BS hello humans puts out

near silo
#

I am proud of the work I have been able to do

astral goblet
#

the community has really taken to writing awesome rentry guides

vestal dew
#

Rentry?

near silo
#

have a look at this

astral goblet
#

it's like a pastebin

vestal dew
astral goblet
#

i dont' know where to find a colelction of them though. i just stumble across them constantly

near silo
#

oops

glass nymph
#

mine are like legs all over the place, multiple arms, awful faces...

and it doesnt make sense because im using what I believe to be decent positive and negative prompts that I looked up

#

sd1.5

#

trying to get 2.1 running but its not lookin possible lol

vestal dew
#

I had it running before

#

but now when I try the models, it's an error

rotund dirge
#

quick question. Why does sometimes SD gives out of error memory at 1000 px. sometimes it works sometimes it doesn't. what causes it? should I give some time after a certain usage? I have RTX 3070 Nvidia GPU

wanton urchin
fervent thunder
#

controlnet has leaks too, or storing something in the memory

astral goblet
#

probably the one that boosted the gradio version

vestal dew
#

well...giving up on this LORA for now. I will return to it later

vestal dew
#

tomorrow is a new day, or later today

slim granite
rotund dirge
glass flax
#

Hi everyonehello

fervent thunder
fierce zenith
bleak matrix
#

Good morning, everyone! How are we all today?

fierce zenith
#

not bad sunny you?

bleak matrix
#

Well, I went to bed late, thought I'd sleep in, but apparently, my body is all, "We wake up around this time, TOUGH IT OUT, GF! :<" So here we are 🤣

fierce zenith
#

im just testing more on the whole running on the beech thing image again trying to coax sd to think shes running and not on the sand laying downItachiFacePalm

bleak matrix
#

Well, the truth is...there's a ghost on the beach, you see.

fierce zenith
#

me waiting for renders lol

bleak matrix
#

Naturally, one must flee from apparitions

#

Honestly

#

One of my first actions this mornings, besides taking me meds, was to press...

#

Generate....

#

Ahahahahahaha

fervent thunder
bleak matrix
#

Yes; that's where all the info you need is for learning about SD.

surreal perchBOT
fierce zenith
#

signed

#

i dont know enough about ai's but i know open source should always stay open no hidden back door stuff not secrets just the tech and the fun in learning

hot parcel
#

I have a question, between 3060 12GB and 3070 8GB, which one should I choose?

bleak matrix
prisma prawn
#

Hello guys, ControlNet was working yesterday now it is gone. Can anyone help me?

lofty parcel
#

Is it bad that I have a positive view of Guizmus simply because he has a cat profile picture

vast ingot
#

I do have 2 cats for real too, if that can help x)

lofty parcel
#

My day is better knowing this information, so that does indeed help waow

tidal lion
#

does anyone know how to remove background using stable diffusion api, help me

#

I entered such description but the output image does not remove the background

#

{ "prompt" : "attractive girl", "Model" : "deliberate_v2", "Negative prompt" : "", "width" : 512, "height": 350, "sampler_index": "DDIM", "script_args": ["abg remover"], "steps" : 25 }

velvet sorrel
#

where are all bot channels I used to use for stable diffusion? Are they gone? Is there anything to use stable diffusion?

vast ingot
#

hey, yep, no bot channel currently

#

but there are other ways to use it, here is the faq

surreal perchBOT
#
FAQ: How do I generate images? Is there a bot on the server?

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 by running Stable Diffusion locally using your own hardware! Check out #1080946152318443610 for more details! You can also stop by #1025467151206854736 for any issues you experience while using DreamStudio or #🤝｜tech-support for any problems you encounter while installing it locally!

velvet sorrel
#

I tried to install Automatic1111 but I have an old GPU graphics card, there's no way to use a server version or something?

#

torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 20.00 MiB (GPU 0; 2.00 GiB total capacity; 1.68 GiB already allocated; 0 bytes free; 1.72 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

vast ingot
#

without using your own graphics card, there are 2 ways :
1/ official webistes like dreamstudio, give some free credits to use
2/ notebooks on google colab, about 2 hours of free time per day

#

yeah you can't run on 2GB VRAM

velvet sorrel
#

I see, thank you!

vast ingot
storm shard
stable pawn
#

Strange. When I load the normal Stable Diffusion 2.1 model and write "GPT4 = Active1" I get a chat window and I can use it as a AI text chat with GPT 4 or so.

tidal lion
#

interesting can you make a video tutorial :3

graceful isle
#

hey guys i have a little project, i need to design some characters and they have specific lifestyles etc. I am doing these things by my own but I am curious what other AI users would do and their workflows to improve faster at using these tools, are there challenges or something where can take a peek at what and how people does their work

fervent thunder
#

theres nothing else really like chat gpt huh

#

are language models the new ai hotness? 🤔

fervent thunder
#

Oh

vast ingot
fervent thunder
#

Yes

vast ingot
#

I just didn't think of it at first

#

😉

fervent thunder
#

wink

young mulch
#

Gist of this thread: if we get a ChatGPT Plug-in for Stable Diffusion/Automatic1111, prompt engineering is basically over. The utility of image generation will explode into the exosphere

#

And now imagine THAT, but for Stable Diffusion

#

For example, right now, you do have to basically fine-tune an image extensively to get it to look decent and close to what you want. For a lot of people, that's become an artform in and of itself, but for others (like me), it's a bit daunting to figure out all the nooks and crannies and seeing what works, how to make it work, fixing mistakes, not realizing where mistakes are made, and whatnot. And yet here we have an AI that can automate even that potentially.

Like if you said, "Make an image of two men playing chess in the style of Greg Rutkowski, but make it as clear, crisp, and professional as possible," you get vastly better results than prompt engineering + ControlNet + img2img fiddling yourself.

At least, that's the dream.

steep mortar
#

I have controlnet extension installed on auto1111. How do I move to a previous commit of it? This latest update sux

frigid spindle
#

how do i get permission to type in dream bot

main iris
#

where do I start prompting?

hushed tangle
#

Prooooompting

main iris
hushed tangle
#

Did you just answered yourself

bleak matrix
fervent thunder
#

ai nerds annoy me

#

"One of the greatest advances in human history"

#

Oh, really?

#

That post literally says that

bleak matrix
bleak matrix
bleak matrix
fervent thunder
#

thinking So many youtubers popup saying how scary and powerful AI is now and AI can do anything better then humans 🐧
I feel like none of them even tried to do anything with AI...

#

human flesh obsolete agony

limpid garnet
#

Is stable diffusion an extension of the normal stable diffusion or do it have to be downloaded seperately on its own?

bleak matrix
native willow
#

Can someone help with "Cuda out of memory" in blender with ream Textures? )

dusk jasper
#

i need help

#

with stabble diffusion someone could helpme to create sometghing

plucky moat
#

Hey does anyone know a way to ignore a word completely in a prompt?

#

I'm looking to have the text in the filename but not have the prompt word do anything

fervent thunder
#

I think you can just change filename pattern, it's somewhere in the settings

dark hemlock
#

Someone should make a noise remover

#

idk if it exists but if you find one pls ping me and tell

gaunt laurel
#

Hello, can someone tell me a model on civitai that can generate images close to midjourney?

#

since I'm broke lol

#

btw, I mean on Colab, since I have a terrible GPU

wary root
#

How Can I Start?

timid lantern
# gaunt laurel Hello, can someone tell me a model on civitai that can generate images close to ...

Depends on what style you're going for exactly. Stable diffusion has a lot of models for different things, but here are a few of my favorites:
Photorealistic/general purpose:
https://civitai.com/models/11193/illuminati-diffusion-v11
https://civitai.com/models/4201/realistic-vision-v20 (v1.3 is better than v2.0 imo but compare for yourself)
https://civitai.com/models/17277/realism-engine

Landscapes, nature, architecture
https://civitai.com/models/5041/cheese-daddys-landscapes-mix

Illustrative/anime style
https://civitai.com/models/22891/isopropanol-addiction-mix
https://civitai.com/models/10028/neverending-dream
https://civitai.com/models/4384/dreamshaper

#

I'm not sure if any one model beats out midjourney, but with a variety of separate models I think it can actually be surpassed in quite a lot of styles/cases

#

(Total speculation, but I kind of suspect midjourney uses multiple models/loras which are selected based on key word or phrases in the prompt, but I have no clue)

plucky moat
fervent thunder
#

hi guys

mystic escarp
#

When using models from civitai do I need to train myself before it's actively working?

jaunty flare
broken smelt
#

no way

#

for real????

jaunty flare
#

stranger danger

slim granite
vestal dew
#

Help me Obi Wan Stable Diffusionnobi, you're my only hope. But it's failing me.

#

Ah, Obi Wam Stablediffusionnobi came through.

modern crater
#

Hey guys, is there a way to remote upload file to huggingface?
I mean downloading file from civitiai then upload it to huggingface repository is kinda take too long for my internet speed

fiery pecan
#

im having fun trying to merge checkpoints, only thing i dont understand is CKPT and Safetensors
whats the difference?

hazy wind
#

What would happen if I set the Clip Skip to maximum? How random would the images be?

fiery pecan
#

but dont know whats a Clip Skip

hazy wind
fiery pecan
modern crater
#

Hey guys, is there a way to remote upload file to huggingface?
I mean downloading file from civitiai then upload it to huggingface repository is kinda take too long for my internet speed

bright drift
#

hi

vestal dew
#

Anyone know how to install something that will do LORA training?

warm junco
ebon tartan
#

hey guys, do i have to download .ckpt or safetensors?

#

the beast method to train a photo of mine is by dreambooth?

inland marsh
#

Anyone interested in having a chat about Stable Diffusion and how it can be used professionally? A paid project to gather feedback on how useful this can be for professionals. DM me please if interested 🙏

fervent thunder
#

is dream booth still the best way to fine tune stable diffusion?

shy swift
#

can anyone help me with an error that appears using stable difusion?

#

it just doesn't generate any image

bleak matrix
normal pasture
inland marsh
#

Yes, lets connect 🙂

sturdy flint
#

how can i download stable difussion

bleak matrix
lunar vigil
#

ah yes

#

big fart

thin tendon
#

Gn