#💬｜general-chat

1 messages · Page 72 of 1

vocal kettle
#

How do I make post in the #1092446741984444416 ? I can't submit for some reason. Should I receive some role or contact someone?

ocean mango
#

Is there a video prompt sharing discord was one a while ago

fervent thunder
#

Where do I find clickskip?

brisk ledge
#

What question? Why would it exist if it wouldnt be faster?.?

dull flame
#

@brisk ledge

#

idk not much of a it guy

brisk ledge
hollow hound
#

Channel Bot 5 unwanted nudity result

keen birch
#

does anyone use opensea

frail tusk
#

How do I make stable diffusion not do something? It doesn't seem to listen to my negative prompts.

austere wharf
#

hi guys is there a website like lexica.art but for prompt for animations not for still images?

#

for example with runway gen 2 you can remix an existing video and you can see the prompt

#

but I was looking for something like lexica.art which is stunning

bleak matrix
#

Good afternoon, everyone. How are we all today?

bleak matrix
bleak matrix
primal current
#

Do you know art dealers or gallery owners? I'm trying to contact some to sell some pieces of AI combined with other digital tools.

keen birch
#

i started using opensea yesterday

#

idk if that counts as an art dealer or gallery owner but it lets u sell art and put it into a gallery collection

uneven heath
#

oh hi

fierce charm
#

does Stable Swarm UI on windows need the NET 7.0 install?

potent spire
#

Oh hey Olivio ^^

fierce charm
#

also does Stable Swarm UI on windows need git and python installed or is it self contained and the Install-Windows.bat file does everything for you?

#

@potent spire hi

potent spire
#

Good questions, would like to know as well

fierce charm
#

i guess since it is a bat file it needs to run somewhere, right?

#

oh, ok, found the answer. the bat file downloads net7

potent spire
#

trying a new workflow? 😄

dark hemlock
#

Lol my prompt got on showdown

pearl minnow
#

stable diffusion has video enchancement?

violet arrow
zinc swallow
#

in comfyUI Ksampler (adcanced) node, is the actual number of steps done by the sampler equal to steps or steps - start_at_step?

#

i have experimented and it seems to be the latter but thats also a bit counter intuitive?

little citrus
zinc swallow
#

ok i just wanted to make sure if i understand it right

cold socket
#

ok maybe im lost in the sauce but, what happened to the channels for the original sd bot gens? (the ones for 1.4 and 1.5)

humble glade
#

gear 5 in one hour

grave stag
#

Anyone able to help me in tech support real quick?

lost shard
#

anyone know how to change the generated image dimensions?

little citrus
lost shard
#

from the bot

little citrus
#

Oh, no idea on thatone. Sorry.

warm junco
soft wyvern
#

Anyone tell me Which python version is support 2.0.1 torch?

rare saffron
#

i searched some, automate1111 and some thing more~ but i cant do same as that

bright inlet
#

Capitalism is slavery

rare saffron
#

😭😢

humble panther
#

Can anyone help remind me where like stable goes in your files? maybe with automatic?

#

i dont know if i deleted everything for space or if its somewhere on my pc

ivory sierra
#

hello, I'm just a newbie, is stable diffusion feasible for IP industries? i mean they can produce consistent character, persona and visual identity?

charred light
#

Any way to edit the previous prompt without copying and pasting it into a new one please?

soft wyvern
#

Anyone tell me Which python version is support 2.0.1 torch?

astral flicker
#

i want to copyright my prompt

civic stream
#

i need help with stable diffusion webui, is this the place to ask for help?

potent spire
old gust
#

Hey

old gust
#

In exchange, can you comment on one of my stable diffusion videos? 🙂

#

Feels like it lacks marketing

#

I’m still learning to be consistent in my animation but maybe consistency is over done

steep sage
#

What happened to distilled SD that was supposed to be 20x faster

real zodiac
#

every paper claims x times faster but the reality ends up being +0.1it/s with caveats unless you own whatever top tier gpu

lavish lake
#

and if I'm not mistaken, it was based on 1.5

civic stream
# old gust What do you need help with?

im trying to train embedding on face, having some errors that im not sure how to fix it or i did certain steps wrong. and yes, send me the links i'll comment on your vids.

#

i posted here, waiting for someone to send me to the right direction

devout yew
#

guys, what is the prompt to add a custom reslution to my output?

bleak matrix
#

Good morning, everyone.

bleak matrix
fervent thunder
#

cant find my creations, how to FIND them, not listed in the INBOX drop menue???

prisma pulsar
#

What would you say against this argument?

#

"While you can finetune models with your own images, the base model still requires millions of images to be trained on. So training the model with your images doesn't make it less of a problem. There really is no proper way to make ethical use of this current tech."

astral goblet
# bright inlet Capitalism is slavery

slavery is still real. there are more real slaves on the planet today than there have ever been at any other time. i think its kind of dumb to suggest people living with technology and standard of living higher than any of their ancestors, are comparable to actual slaves that live today in today's modern context who struggle and have no freedoms at all

supple coral
#

hey all

astral goblet
supple coral
#

i know its ladiesmen yes

#

new here so for sure will need to find my way

astral goblet
lavish lake
supple coral
#

not sure what you mean flowwolf but thks

astral goblet
supple coral
#

thks

astral goblet
violet arrow
supple coral
#

are there any tips on here about prompting and in which section to look?

lavish lake
supple coral
peak tendon
#

Are there any good ai solutions for android? Like, I have stable diffusion locally in my computer, does something comparable to it for phone exist?

potent spire
arctic meadow
#

Hahahahaha hahhahaha @sharp fable hahhahahahaha he ping younedddddd hahahahaha he pinged your entire username I'm dead hahahhahah

sharp fable
wicked sonnet
#

I'm having difficulties making icons with stable diffusion, are there any models that are better with objects/icons?

bleak matrix
peak tendon
bleak matrix
bleak matrix
# peak tendon Paid it is?

The pricing is listed on the website. Ofc, you can also use the bot here on the server for free as well.

peak tendon
wicked sonnet
bleak matrix
wicked sonnet
#

Thank you

peak tendon
potent spire
#

And also important: start with themselves

peak tendon
potent spire
#

Sure the plan may be some idealistic ideas

#

Not that i disagree that there are issues with the current system/systems in the western world

peak tendon
#

There's a lot of ideas better than capitalism

potent spire
#

Yes, some socialist parts

hasty sierra
#

How much vram we need for sdxl?

#

And is there a version we can run on my 10gb 3080?

plucky shoal
#

you can run normal sdxl on 3080, if problems, add medvram, i got 3070 with 8gb and with medvram it runs o.k.
ComfyUI handling it better than A1111 for now.

potent spire
#

I run with RTX 2070

warm junco
#

Your card shouldn't get problems running it if you use --medvram and --xformers

hasty sierra
#

nice ty

dapper wadi
#

so how do I get my locally running webui to look like the bot-produces images in this server?

bright inlet
potent spire
violet arrow
polar jungle
#

Whats your guy's favorite model and what do they do

little citrus
soft wyvern
#

Anyone tell me Which python version is support 2.0.1 torch?

strange vale
#

Anyone here having trouble with pc ui freezing while running a comfy ui work flow? I can't use the machine for anything else while processing images. (using sdxl)

tawny tusk
#

can anyone tell me what the interpolate function of dynamic threholding is? i haven't found any guides on it. my understanding is the higher the interpolate the higher the detail at the cost of contrast and color-detail

mild hedge
#

Has anyone had any luck getting invoices? I have submitted 2 tickets almost two weeks ago and sent an email to both platform@stability.ai + dreamstudio@stability.ai and not received any response. I represent an EU company and we need a tax invoice. Any leads/ideas/reponse time estimtes are welcome!

fresh ivy
#

Why can't I see the SHOWDOWN images?

turbid lake
#

@fresh ivy maybe you should in PC

fresh ivy
#

@turbid lakeI am buit all it shows me is the rankings

turbid lake
#

@fresh ivy Sorry I only downloaded discord for a few weeks,i am missing ,too

buoyant ruin
#

hi I need help with my SD

#

can somebody help me?

warm junco
buoyant ruin
#

okay thanks

weak lark
#

How it works?

marsh iron
#

Guys what would be the best way to inpaint a part of image but just change the color of it but nothing else like the material and structure

#

I know we can use controlnet but inpainting model doesn't work well with controlnet

plucky forge
#

i started up my stable diffusion 1.5 after a month of not using it and it looks like it downloaded a new interface or something. i didnt even know it was constantly connected to the internet like that. is there a way to go back to the old one or is it just going to be liek that forever now?

sudden ruin
plucky forge
#

yeah, img2img doesnt even work anymore. im going to have to think about how to get a truly local version because its pretty unacceptable it connected to the internet and updated and broke a bunch of stuff.

#

i guess it was easy diffusion so theres probably something else i can setup

fervent thunder
#

Should I grab a 4080 or still to hold out for the new 7000 AMD?

heady canyon
#

The future is bright for us. Follow me for great content and streams! Link in bio

sick gazelle
fresh ivy
#

@sick gazelle👍

undone harness
#

Hey, does anyone know why there's such a quality difference between the generated images I get from dreamstudio.com and the REST api?

clever vault
#

how fast is the rtx 4060 in generating 512x512?

worldly terrace
#

does anyone have any experience with running SDXL on vast ai?

steep sage
astral goblet
main ledge
#

hello 🙂

astral goblet
#

wonder what's going on with @warm junco 's status messages

#

oh its very late where cs1o is. maybe i shoudl watch my pings

warm junco
astral goblet
warm junco
astral goblet
#

yeah i'm looking at it now. ❄️ 🫘

slender jolt
#

嗨

supple coral
#

evening all

supple coral
#

thks for the tip yesterday about regional prompter > its works

runic turret
#

Anyone know where the FAQ is on nsfw guidelines or are we left up to our own to decide? I couldn't find anything. This server has lots of channels.

astral goblet
#

Rule #4
No NSFW Content
Excessive gore/blood/nudity or anything that exists to suggest sex including sexualized poses. No illicit imagery. This includes NSFW content involving or depicting underage characters such as schoolgirls, lolicon, etc.

rustic prism
#

hey if anyone has any studies of artists that work well with sdxl 1.0 I would love to utilize it

robust coral
rustic prism
#

ya that one is fantastic, I have been devouring it. Ive made prompts with all of the artists on it already 😄

robust coral
rustic prism
#

your amazing!!!!

nimble sand
#

Hello

twilit phoenix
#

Am I creepy if I ask random girl at school for her number after telling her she is beautiful?

trim magnet
#

only if you are the janitor or the teacher

twilit phoenix
#

No I’m student

runic turret
crude marlin
#

hey so I'm kinda lost, how can I make a similar image to a previous one using the seed? and is there a way to look up all the images I created? I don't remember the seed neither the prompt of the image but I do have it.

rustic prism
runic turret
robust lance
#

can we use controlnet on sdxl 1.0 ?

foggy spear
#

Hello everyone, any1 has a local repo of kohya_ss? His latest update is broken
So do not update if you want to be able to train
if you could share a zip file of a previous version with us, would be helpful

last estuary
#

how lon it takes to make my photo?

fading lagoon
#

Assuming i want to create a portrait of a famous person in a creative art style. ( while bypassing IP, because germany is f up ) would you 1. start with 1 image and iterate until its both in the given style and not comparable with the original ( facial expression, posture, clothing) or 2. train a LORA with many Picture of this person and then deploy it on some CKPT with the given Style ?

potent spire
young comet
#

Anywhere I can ask a ComfyUI question? I am using a custom node and it wants me to add PYTHONIOENCODING=utf-8 to the env, but I got no idea where to edit env variables for ComfyUI

bleak matrix
#

Good morning, everyone!

bleak matrix
young comet
#

I'll try the tech suppport one, I am not messsing with SDXL, jsut plain ole 1.5

#

thanks!

fading lagoon
# potent spire Why is Germany fck up regarding IP and why do you need to bypass the IP? That es...

Nono i dont want to sell a portrait. You see i just want to use a picture of a person im refering too(in a video that is monetise), but in germany there is no fair use. Basically we have a Mafia like entity thats buying up all IPs for ALL Stock footages and IPs as possible. The purpose is not to protect the iP of the original Person, but to have the right and ability to extort anyone who cant pay the fees they demand for any use of pictures. Thus creators in germany have resorted to 100% create there own content ( reason why Kurzgesagt exists, since they were force to animate everything instead of using any stock footage) Im trying to enforce fair use with creativity

fervent thunder
#

Is the nvidia a100 better than 4090 for stable diffusion

#

??

potent spire
#

im in a neighbour country of yours btw 😄

#

you probably need to talk to the manager of the prominent person or if its a image or artwork to the person that did it

#

i mean you said its for a monetized video

frigid lance
#

Hey guys, what's the best tool rn that allows training the model on your own images? I don't want to run it locally, ideally I'd like to pay and use some webui

#

Like dreambooth with sdxl or smth

hollow fiber
#

Hello, Is there a way to calculate how many credits will my request spend?

fickle ocean
#

Hei 😊
Is there a way to take a picture and rotate it.
Let's say i have this bird. But i want to have a image from the other side or rotate it 90 degrees? Is that kind of things possible yet? 😊 If so, it can probably be used help making a 3d model out of it. 😀
https://i.imgur.com/DYy5g9o.png

fading lagoon
# potent spire i dont call it Mafia, they are there to protect us from criminals that steal art...

Well their terms of business kinda is just really scummy, it´s one thing to enforce ip to protect, by warning and taking down unlawfull used images, It is another thing to try to maximise every breach of ip even if accidental. Their main objective is to profit from accidental breaches. By now it`s almost a monopoly. But back to topic, i want to make an animation as tribute to authors from some books. for example let Timothy Ferris. Id like to create a "drawn animation" , so in this process there will most likely be some kind of pictures that have ip. So would you rather do option 1 or 2 ?

potent spire
#

usually you get a warning or in general contacted first

#

a direct lawsuit is less likely to happen

distant fossil
#

🎉

fading lagoon
potent spire
#

if i get to sue someone that broke the agreements of the usage of my artworks or images i would do the same as well

#

it wouldnt be just the cost of the court and lawyers which is already not fun as well

#

pay the reparations and pay the court and lawyer cost on top of that

#

it the lawyer isnt guaranteed by a insurance which is the case with myself

uneven pecan
#

What files are needed to run the SDXL model?

lavish lake
#

the offset lora example is useful

uneven pecan
#

Awesome. Thank you so much! 🙂

trail lion
frozen rover
blazing cobalt
#

also they're missing metrics, which, considering starcoder is better than gpt3.5 now, really should be there

chrome jacinth
#

wait so I should be using stablecode instead of chat GPT for programming?

frozen rover
blazing cobalt
#

not until they tell us if it's better

charred mesa
#

research use only lets gooo

potent spire
chrome jacinth
#

so stablecode is nothing useful or what?

potent spire
#

nah, but it also breaks down to your preference. Copilot with or without ChatGPT is the dominant force as of now

#

especially with the new upgrade (Copilot X)

#

which isnt out yet for the public

chrome jacinth
#

hmm okay

potent spire
#

with Copilot X there wont be a "need" for ChatGPT anymore anyway

rugged bough
#

anyone had change to try the stabilityai/stablecode-instruct-alpha-3b yet ?

#

I failed to run it on webui 😦

oblique pine
#

where to test it on web?

crimson beacon
#

what happening with this error in comfy??

#

Prompt outputs failed validation: Required input is missing: images
SaveImage:

  • Required input is missing: images
#

idk what that means

#

Im using Sytan workflow

#

pres generate

#

and that happens

rugged bough
#

any ui working atm for coding ?

crimson beacon
#

oh i see so basically I need to use refiner or if I wanna remove that step I gotta work with all these nodes

#

comfy so ass bruh

#

thanks tho

still shore
#

That code, is going to be so UN-stable that it's got the wrong name.

hasty hornet
# potent spire with Copilot X there wont be a "need" for ChatGPT anymore anyway

I think we just need 2 things

  • AI search engine, like phind, but for more "general" questions and something like phind , but less janky for coding questions...
  • Local AI with bigger token count, I'm talking millions tokens to work with local projects + a way to ignore files and folders, otherwise even millions won't be enough.
hasty hornet
# rugged bough any ui working atm for coding ?

There are few, but they all suck.

The only thing I found usefull is phind, mostly cause you can make it scan docs and forums of framework \ library \ engine you are working with, it's answers gets significantly better compared to "general" search or gpt answers.

#

But don't expect too much, it makes same dumb mistakes as any GPT - based bots or search engines...

#

it hallusinate things that doesn't exist, it doesn't understand version diffrerence, it suck really bad when code gets complex...

potent spire
#

showcased at their I/O i believe

#

basically integrating (more) advanced AI into their search engine

hasty hornet
#

yep

#

I've seen someone working on open source project with mil tokens, but it's...pretty bad

potent spire
#

dont know about that one, hear about it for the first time tbh 😄

hasty hornet
potent spire
#

hmm, closed beta

hasty hornet
#

not sure if it'll be open tho, idk why I thought it's open

west echo
trim magnet
west echo
trim magnet
#

👍🏿

west echo
#

Or it could be a complete fake and install multiple mining apps on your pc

hasty hornet
#

I'd suggest avoiding packages like that. They might be legit, but they also might pack something else in.

west echo
#

And 8 hours!!!!

hasty hornet
#

it might be cause it has lots of models in

#

but still...very sus

proper sparrow
#

is there any realistic model for Hollywood famous movies... creating deforum video

#

or any way to get free mobile format videos for creating hybrid video on stable diffusion

plush dawn
#

Excited to work with the Stability community on 100 Builders! If anyone has questions about the program, feel free to ping me here or send me a DM 🫡 #1072240143521554592 message

fathom marten
#

Hum, an llm model

#

Absolutely not related to prompting or SD, hum, interesting choice

#

Is it on the todo list ? Or will we condamned to restart from scratch every prompt, forever, trying to get a decent one ?

#

Legit question

little citrus
#

What do you mean? Hum?

fathom marten
#

Dunno, just something related to SD. Not just a text completion model for prompting, but more an assistant.
'This image is fine, but i prefer this instead on this' > Change this
'How can i draw a chimera with a clown in one hand and a turtle in the other one.' > Give the best syntax for it

#

See

#

Not really lol, don't even understand your sentence

little citrus
#

There are many images you can never generate even if you spend years engineering the prompt.

fathom marten
#

Yeah, true. Like regional prompting etc. I'm aware about this. Took the hands example not to imply this situation.
The generation of one prompt isn't the point. More like you said, based on the generated pic "in the conversation"

little citrus
#

These will all come in time. Maybe in 1-4 years if I had to guess.

south jungle
#

I don't really know how and who to ask, but if you have knowledge and skills in hacking, well, I'm potentially interested in hiring you and even getting paid for this job, come PM for more details.I must add that it is not too legal.

hallow plaza
#

Has anyone tried Stable Code? Is it better than chatgpt4 for coding?

hasty hornet
astral goblet
worthy hornet
#

Where can you ask a support question? I just had an image reported as inappropriate and I can't image how. Before it reported it, it generated 7 results that all looked almost exactly like the uploaded image which was a black and white swirly flourish I wanted to turn into a william morris like flower.

little citrus
errant yacht
#

Has anyone made a guide to setting up and using the stablecode llm?

errant yacht
#

It's actually a really nice UI for using all sorts of models

opal iris
#

hello my dear friends, Is there any great Models for surreal or weird but delicate set design or architecture in Civital????

hidden dagger
#

hi

#

see da syrprise]

violet arrow
warped helm
#

Hi all, how would I know if stability code would fit on a 4090?

#

Before running it obviously

#

I am not too sure how big 3B point is

thorn citrus
#

Hey there
Hope you are doing well
I am a senior full stack developer who has full experience in web and AI development
So if you have some projects, please let me know
thank you

royal garden
#

Greetings all - I have a question about Lora that I am not finding an answer for despite searching for half a day. Lora seem to be activated by key words in your prompt, WITHOUT having to actually add the lora:xyz:weight prompt. How do you STOP them being activated? Do you have to add every single Lora to your negative prompt? Or is moving them out of the Lora folder the preferred method?

broken smelt
thorn citrus
#

Good @sudden violet
so Please DM me
And I want to know about in detail

jolly crystal
#

Where can I find football lora? wanted from real madrid

#

I wanted a lora in a football club uniform

royal garden
native root
#

Hi everyone,

I hope you're doing well!

I'm in search of inspiration today. I'm looking for an image-generating AI platform similar to Dreamstudio.ai or Magespace. My goal is to demonstrate AI's generative capabilities in a playful manner to a large audience. The issue with Dreamstudio is the limited credits available for free, and with Magespace, the NSFW focus might draw too much attention in this context. Are there any other recommendations? The demo on StableDiffusionWeb just isn't robust enough in my opinion.

#

Thanks for the reply! - unfortunatly thats no going to work. I cant get 80 non technical people to install something, play with it and then move on to other things.....

#

Chechking it out now - interesting....

compact scaffold
#

Lol

#

Hello there, everyone o/

native root
#

I wouldn't know much about it, but I've found the creative exploration of different models fascinating to follow.

native root
#

I think i will go for dreamStudio - with enough guidance on how to adjust settings, and maybe using an earlier model I can get the 25 free credits to last longer. Thanks for the help;)

crisp shell
#

Hi all, don't know if i am right here for my question. I have (for a wedding) images of all the guests, and want to make them "medieval" or "fantasy" style to create a guest book.What is the best way to do that? Would be very happy if someone could help

pearl sierra
#

i have deployed my api to google cloud vm
i want to move to any way that it's auto scalable when a lot of users came it automatically scale up and down please give me solution
am using automatic1111

plucky shoal
#

@crisp shell not sure, but nobody answered, img2img and controlnet. Pix2pix, i think it is part of A1111 img2img, not sure, if contorlnet is must.

hollow knot
#

perhaps someone could help point me in the right direction, I like the way a picture visualizes the ingredients they offer in their drink, and want to do a similar visual.
is there a good place to show an image, and try and get a similarly created one?

fervent thunder
#

Anybody got some time for a small troubleshoot to get sd to work?

crisp shell
#

Thanks a lot, after playing around with mage.space it seems to work .. But:
Why is this blocked?
NSFW content is only available to premium members. Upgrade to Pro to enable this!

arctic steppe
#

I have a question, how can I use images to generate pictures?

crisp shell
#

I created an Image with my face (not good but in the hope AI will enhance it) in it and used a mask to leave the face with custom inpainting. It creates an seemingly good image but its blocked because of "NSFW content blocked .... premium" 😦

#

Hmm okay seems to work : the inpainting is always left fully untouched on the mask?

hard cape
#

is it declasse to ask if we get to keep the premium models for use with, eg Auto1111? Are they even compatible?

#

bookmarked this for after work experimentation.

#

(I'd thought initially this was dreamspace discussion. Didn't read close enough).

runic turret
runic turret
fervent thunder
crisp dagger
#

hi dudes, did you tryed already SDXL?

bleak matrix
#

Good morning, everyone! I hope you are all well today!

hexed lodge
#

Is there a way in a prompt to request for a girl in a generated image to have a smaller breast?

I tried several texts, but they don't seem to have the effect I would want. Some of the girls I generate have breasts that are too big.

hearty folio
#

hi is there any good model for creating christmas baubles with good background

runic turret
hexed lodge
runic turret
#

(small breasts:1) ? or (small breasts)+++ ?

#

move the prompt around, maybe in the first part like 1girl with small breasts, etc ...

drowsy cloud
#

How do I use the ai

#

On bot 1

hexed lodge
#

Thanks. I had some good results by adding:
very very very small breast, anorexic

pale latch
#

the weight is a token multiplier so over 2 works but will blow out the math and cause bad generations.

#

yeah its not recommended to go that high. but it wont be ignored

cursive jetty
#

Question for some advice using Deforum. Is it possible to make a zoom animation based of a video?

#

So like using a video as a template... is that possible?

#

Oh really? I'd like to know how to do that.

#

I've seen videos on doing deforum based of images.

#

Maybe its the same concept... I am not sure

bronze goblet
#

Do most people here run Stable Diffusion locally? Do you use a GUI for it?

trail lion
dim yoke
#

has anyone gotten transparent backgrounds with SDXL?

elder plaza
#

Have a question. Hope this is the right place. Apologies if it isn't. What's the correct syntax in a prompt if you are trying to create something in the style of an artist. I've had limited success. For some reason the prompts I am using tend to create images that are not very similar to the artist I'm trying to mimic. And to add insult to injury, it usually sticks the actual artist in the image! Would really appreciate some help. I use Automatic and Comfy with SDXL 1.0, base plus refiner.

trail lion
thorn hare
#

is there a general discord for text ai ? or just for specific programms ?

chilly echo
#

Been genning for months. First time this is happening. Comp keeps crashing, even with the smallest tasks. Genning a 512x512 pic with 5 steps. Comp completely locks up on the final step. Restarted twice, had to hard shut down twice. GPU is working and detectible and running cold. What can I try to fix this issue?

astral flicker
#

/where is the bad apple sd edition?

trim magnet
astral goblet
#

thats for booru tags.

#

just to get an artist's style, ones that are not anime, wouldn't benefit from booru tagging

#

you'll want to say "painting in the style of ***" or "illustration in the style of" or "digital art" or "photography" etc...

#

the clip layer does still have a level of contexual attention

#

being more specific about the style helps too. "surrealist painting in the style of Salvador dali"

trim magnet
astral goblet
#

i mean, thye're a site with a lot of articles, but you shouldn't take it as a gospel. try what they encourage and try the other ways too. see what happens. they're at least giving you a place to begin right?

#

i dont believe in the whole "using it wrong" thing with technology that's misunderstood by the very researchers who created it. results speak for themselves. do what gets you good results. be ready to explore approaches

astral goblet
#

it knows a lot

#

it is usable. you can see this by lots of people making great works from it. maybe you're probably just prompting wrong? i mean... people seem so quick to give up but prompting is king here.

remember, a year ago you had NO prompt game. It shouldn't be this hard to evolve your current prompt game. How ingrained can it be at this point?

#

One of the prompt clip layers is the same clip vit - L they used on 1.5 even. i think it was a smart move to help bridge the dialects of one clip layer to another open clip layer

urban tide
#

Yo, What do people think the best AI voice tools are right now? i still feel like most of the ones i hear just sound a bit robotic

potent spire
#

what is it that doesnt get you the results you need?

astral flicker
#

this is elevenlabs

urban tide
#

rvc?

#

i feel like eleven labs is pretty bad tbh

#

i guess its the best we have so far

#

but its just super uncanny valley still i feel

primal current
#

Where can I share a business proposal related to AI?

potent spire
#

@sudden violet hmm, did you use your own workflow or someone elses?

#

i used someone elses workflow for ComfyUI

#

but i want and should try my own

#

its for fun anyway

#

i dont know, my results were usually good before but my prompts were limited to certain topics and styles

#

i dont use it often

#

and my current PC specs arent ideal either

#

A100 would be nice to have

#

or H100

#

if i was purely into AI and programming instead of gaming and art as well i would build such a workstation, but because its not the case i have to build a strong PC but not a specialized workstation

#

didnt look up for Runway for a long time

#

its expensive

#

those 70+$/month for the biggest sub

tight tartan
#

In A1111, I know you can interrupt SD from its current process however I'm doing a batch upscale from a directory but i realize i need to stop it. how does one do this?

#

apologies & thx!

potent spire
#

why not, i mean if its usable for you and worth it for you then you can

#

there are more expensive softwares than that

#

let alone every day stuff people buy

#

cigarettes and alcohol are a prime example

#

i agree there

#

which is a good reason for me why i dont fully switch to generative AI art but instead have it as a another tool i can use

#

with fully switch i mean to abandon the other ones and use exclusively AI art generators

#

i guess there is low to none control with Runway then?

#

if i understood it correctly

potent spire
#

hmm

#

yet that expensive

#

thats more expensive than 3ds Max and Maya together (indie licenses)

rich plover
#

i broke stable diffusion a few days ago when i tried updating a111 and have been trying to figure out the error without reinstalling it for like an hour here and there. and i finally got it working again by deleting a config.json and it simply remade the file on launch

gentle blaze
#

I got some pictures of my kid playing. I want to use them as base to turn into an picture for a kids book. Where do I start?

wary belfry
#

any updates on ControlNet for SDXL? It's one of the single most powerful features. is still just the filesize the issue?

hallow plaza
#

Anyone use claude2 for programming? I'm trying it out and the large context size seems to really help over chatgpt4

signal mulch
#

recently, i don't see any picture in showdown channel?

lavish lake
pure plume
#

how can I get a high resolution picture

#

?

lavish lake
#

also there are a couple vae/loras for that

pure plume
#

the pictures are generated in 1024*1024 even if i upscale the picture it would be square

lavish lake
pure plume
jagged vessel
#

yo guys wassup

#

i was trying out stable swarmui and i tried to run the update.bat to update it if there are any updates but then it broke the comfyui selfstarting api thing now whenever i run it it says Initializing backend #1 - ComfyUI Self-Starting...
17:25:18.459 [Warning] ComfyUI start script is '', which looks wrong - did you forget to append 'main.py' on the end?

tight mist
real zodiac
#

nice, finally some controlnets

#

also what the actual fuck, and those prices

dim oxide
#

hey peeps, anyone knows a way to limit prompts with colors, eg. segmentation somehow. eg. "red hair and blue dress", sometimes becomes reverse, sometimes works good... if there's a way to put in some brackets or something so the first few steps are heavily focused on right color of the right thing

tough bough
#

Do you mean with 1.5 or with SDXL?

dim oxide
#

currently working with SDXL

tough bough
#

Don’t think there is a way to do that yet, unfortunately.

#

But ask in that chat. Maybe someone has a method I don’t know about.

dim oxide
#

i thought more of a general question, with all models, that's why I asked here... trying a lof ot them, all have same issue when it comes to colors 😄

#

especially if there are more than 2 desired items

#

then it all gets mixed up

#

eg. try female with red hair, black lips, yellow shirt, green pants and blue shoes .. tell me what you get, on any model 😄

little citrus
#

It sees "green" "yellow" "black" "hair" "shoes" "red" "pants"....

#

And it tries to fit all of those ideas in the image.

dim oxide
#

I kinda understand why it's so, so the question was is there a way... to force it draw maybe item by item then fit all together, or something.

little citrus
#

But right now it is a new technology. It can't make anything you can imagine. It can make many things but not exactly as you imagine.

dim oxide
#

yeah, inpainting is cool, but it's not exactly what I want... especially when using wildcards

#

but, i've been thinking, there could be a workaround, first generate everything needed with low steps then run all through img2img

little citrus
#

This is one of those tasks that may need newer technology that doesn't exist yet. There are many images that can't be made with SD even if you spend years engineering the prompt.

#

img2img uses the same text encoder so it still won't understand the complex relationships between words.

dim oxide
#

yeah, but if something is clearly "red" in the original image, it will be red on the output image no matter what you type in

#

denoiser won't turn red into blue even with really high setting

little citrus
#

Right, but how will you get it to be red in the first place? That's the hard part. If you say "red hair" that might also make the shirt red.

dim oxide
#

true, there comes inpaint. perhaps first generate "female with red hair" and nothing else, then on top of it generate blue shirt etc.

little citrus
#

That's what I was thinking. Even then it's challenging.

#

There may be some extensions to make this easier but I am learning with the base model only so I can really understand it.

dim oxide
#

i'm torturing my 4090. bought it for gaming, not playing anything. so running countless generations of useless images 🤣

little citrus
#

Sometimes just generating a large batch and picking the best image is the easiest way.

dim oxide
#

so there goes sdxl and mega resolutions

real zodiac
#

for sd, not sdxl specifically, I didn't have much success with it personally though

dim oxide
#

yeah, makes sense because of the way it works

little citrus
#

Anything that tries to "fix" that issue will have limited success. It's not a bug. It's a limitation of the current model. Basically SD can take your prompt and use it as an idea. But it can't match your imagination exactly.

#

Wait 1-4 years and there will be a model with much better language understanding.

real zodiac
#

yep, all true, sdxl doesn't fare much better than sd sadly

dim oxide
#

i'm also stuck on auto's , really need to try comfy. got a feeling i could make a loop there so it expands prompt gradually

#

example run a text2img prompt "female with red hair", generate and run through img2img with "female with red hair and green shirt" etc. to the last item on the list. it should work technically

little citrus
#

No such thing as "expand the prompt". It's all just weights applied to the model inputs.

#

I would be surprised if that works since the first prompt will produce some other color of shirt and it will be hard to change without changing the whole image.

dim oxide
#

you can always use "in colorful, rainbow underwear" 😄

#

and build layer by layer

real zodiac
#

what you can do and I can confirm it works is generate whatever image, then select eg hair with clipseg or other segmentation model, turn that selection into a mask, and run img2img on it

little citrus
real zodiac
#

it's possible as a one click thing in comfy too

tough bough
#

You can also draw a color onto one area of the image before inpainting.

real zodiac
#

yeah but stuff like clipseg you can automate completely

dim oxide
#

yeah, i'll give comfy a shot today. need to read all that install text about linking the model folders

little citrus
#

Often easier to edit the yaml file than set up a lot of links.

dim oxide
#

tnx, i'll try it a bit later. will come here if I get stuck somewhere

supple coral
#

i had good tries with regional prompter

wild steppe
severe sequoia
#

Are there vods for the stream

lethal nest
#

Hi guys. I was working with stable diffusion 1111 without a problem today but since I have copied the ckpt file version 7gig it stopped to work with SDXL refiner and even after I have deleted it it doesn't work when I push the generate button it gives me a black image 😔

lethal nest
astral flicker
#

who timeouted me?

karmic brook
#

steer away for future notice thumbsup

astral flicker
#

oh sorry haha i wasn't aware that's a problem

karmic brook
urban lantern
#

we moved here?

karmic brook
#

yes, the stage-chat closes after our stage has concluded!

cold light
#

hello is img2img possible in bot10 channel?

karmic brook
cold light
#

only on the website? thanks

hasty hornet
elder plaza
elder plaza
stoic ridge
#

hi! i just started using stable diffusion with the xl release, so i'm catching up on all of the intricacies. would love to use the QR code generator and have followed the tutorials, but noticed that they all assume an older version of SD. do the stylized QR images created with controlNet work with XL as well?

warm terrace
#

Is someone willing to please train stable diffusion with my photos because seems I can't with a AMD graphics(unless there is way with AMD could someone know how please and thank you)

crimson belfry
#

for anyone making image prompts with chatGPT and having issues with censorship, Try BetterChatGPT its free and its completely open with no censorship

#

i just tested it out and it works

crimson belfry
#

SLAVES NO LONGER

slender blade
#

Hu

crimson belfry
#

wazzup

proper kite
#

Hey all. 🙂 does anyone in here use the Easy Diffusion web UI version of SD? Looking for some help with it but I haven’t found a “community” for it

broken smelt
proper kite
#

Thanks!

simple spear
#

hey

#

🤓

#

guys?

wicked cypress
#

why whenever i put a prompt on photographic is creates a picture of a camera

little citrus
wicked cypress
# little citrus Which model? Show the prompt.

i believe it was sdxl 1.0, noticed it has the same problem on clip drop and the discord server. one of the prompts i put was “blood on sidewalk”, and one of the outputs was like a camera with a black and gold background. sometimes when i put names like “david pakman” for example, all of the outputs are just cameras. this only happens when i put it on cinematic/photographic style tho. even when i try to exclude cameras from it, it continues generating it

little citrus
#

Try less controversial prompts. Ask for "a tabby cat sitting on a couch" with photographic style and see what happens.

wicked cypress
#

got it

#

was just kinda confusing how it kept generating pics of a camera but ig all of these still have limitations (luckily)

round cloud
#

Hey all,

We'll be running a hackathon for Actually Open AI on September 6th in collaboration with Stability AI and Arweave - if it's of interest, feel free to DM me to learn more. More info available at actuallyopenai.dev 🤗

unreal crown
#

does anyone know how I can make SD choose between these pose?

I tried {squatting|sitting|lying|kneeling} but it is only choosing kneeling

little citrus
unreal crown
little citrus
#

Yes on Extensions tab.

bleak drift
#

So I am trying to remove a shirt from a person in a photo I have and replace it with a different type of clothing. So I use the paint brush and painted the area I wanted to inpaint.

I then typed in my prompt and hit generate. And all it did was just show up the same color as the brush color.

#

What am I doing wrong?

astral goblet
bleak drift
#

And they really didn’t give much tutorial wise

astral goblet
#

no idea about taht app sorry

fervent thunder
#

I need some help.

obsidian meadow
runic turret
fervent thunder
#

Hi

#

Can any of you tell me how to make a Collage in SDXL? Do you have any example of a prompt that can help me? Unfortunately I couldn't find anything about it.

open quarry
#

Hi. How run this win command: python.exe -m pip install basicsr facexlib
but for mac syntax?

#

i found
python3 -m pip install basicsr facexlib

unique hamlet
#

hi all, how can I XYZ plot LoRA weights and different LoRAs in AUTOMATIC1111?

warm junco
unique hamlet
unique hamlet
warm junco
unique hamlet
warm junco
#

No problem 🙂

open quarry
#

how set proxy for sd?

fading pike
#

is there a way to auto refill credits?

thorn hare
#

does someone know how to link SD to silly tavern to create expessions ?

bleak matrix
#

Good morning, everyone! Hope you are all doing well today!

bleak drift
fervent thunder
#

hi, in stable-diffusion webui, is there away to pause the training then resume again?

#

the "interrupt" seems to stop it but im not sure how i can continue to retrain it

torpid parcel
#

Hello there,
I was wondering if it would be possible to train a model to do a shirt swap over, for example, NBA players ?
Does it exists anything like that ?
Like if it's possible to do a face swap then it should be possible to do a shirt swap, no ?

gilded pivot
#

Can someone give me a good inpainting seting for changing little things.
I can't seem to find good settings.

severe sequoia
#

can i use comfy UI on a mac

brisk linden
#

how do i install a1111 and comfy ui?

#

please provide youtube links

astral goblet
astral goblet
#

youtube links are often the worse. there's not a lot of ways of finding good youtube videos since only the most clickbait videos go to the top of the recommendations and they're often the most innaccurate. what you want to use instead is documentation and text guides. Text guides can be updated by the author easily. Once a youtube video is published, it's out of date the next day, never to be edited or updated again.

kind grove
#

good morning. I am looking from some mmodels that I cant seem to find on Civitia or Huging face. I asm new to all this, so I was wondering if there was another place I might look?

stoic pivot
#

Hi, I wanted to ask, if in img2img I load a png file with alpha channel ( and in my case without background), how can I prevent sd from generating a background, keeping the transparent part actually transparent?

low sedge
#

hi just a quick question, am i able to update a1111 from v1.4 to v1.5/SDXL or is it just an entirely new version i have to install separately.

i dont feel like going through the hassle of reinstalling extensions and what not.

obsidian meadow
#

Hi very lovely people!

#

How are we doing so far? Good gens?

runic turret
warm junco
warm junco
low sedge
warm junco
brisk linden
#

@astral goblet thank you so much i will take a look at it

knotty turtle
#

Do you guys know the difference between **Shuffle **and **Reference **ControlNet models?
From the onset, they seem to kind of do the same thing

astral goblet
#

reference isn'ta model. it does something weird with the preprocessing and doesn't need a model

#

it came out later

knotty turtle
steel sedge
#

Anyone facing issues with SD+Automatic1111 and Google Colab?

dense tinsel
#

Hi guys, if you want to try our stabble diffusion xl 1.0 integration,you Can Do it for free and give us your toughts on https://arescreative.fr

violet umbra
#

Gm 👋👋👋👋

prime rune
#

hey, what prompts do you use to create characters from anime and they actually look like them, dm or respond to this ty

jolly crystal
#

Who is the best model for hentai?

#

And what is the best realistic model for hentai?

trim magnet
proper sparrow
#

if we recreate a movie clip with hybrid and controlNet for YouTube shorts..do we still get copyright strike on YouTube ?

astral goblet
twilit phoenix
fervent thunder
#

Anybody tried the Stable Chat yet?

orchid mason
#

how to generate photos with different models at the same time stable diffusion

crimson belfry
jolly crystal
native sparrow
#

Hey, is dreamstudio down? I'm getting the error something went wrong on our end, please try again.

hollow swift
#

editing any type of video for only couple of cents

#

as pay

fervent thunder
heady canyon
#

Join my subreddit for great content, analysis and discussion. Link in bio.

fervent thunder
untold beacon
#

/

safe moon
#

Heyo if anyone bought the Nang Women AI Image generator pack I've been running into some issues with it working at all. Most of the time it will just crash before it works or after a few seconds of use rip

hasty hornet
#

no idea even what is that

safe moon
#

I found it funny and wanted to try it lol

rotund coral
#

^ I put this together to show people how to repair images in SDXL.

bleak matrix
#

Good morning, everyone!

#

How are we all doing today?

rotund coral
#

Good morning @bleak matrix

bleak matrix
#

What's up?

runic turret
#

Sigma fp with Sigma 45mm f-2.8 DG DN,

rotund coral
runic turret
#

How? Photos shot exclusively with that brand of lens? Is this a model or Lora? Sigmas aren't considered great lenses like a Nikkor or zeis.

rotund coral
#

I did pull up sites like Amazon and others looking for more modern cameras, like Blackmagic for example. I tried many different ones. But it didn't appear to make much difference.

runic turret
#

I would be very suspicious that those images are created with those lenses and or film.

rotund coral
#

They work. I can confirm it.

runic turret
#

I believe they do, just their authenticity. Having shot and processed for years having all those cameras, films, and lenses you would have to be ... Well very rich lol.

trail lion
rotund coral
trail lion
#

Basically a collection of style prompts

rotund coral
#

There's a tool that started circulating today. It's a simple prompt writer. It's amazing! But it's not mine to share.

#

When I get permission I'll come back and share it with you guys. Between that and the above website you can really step up your SDXL images. But again, not mine to share. Sorry.

#

And ofcoarse I have to abide by the rules of this Discord.

runic turret
rotund coral
#

Let's just appreciate his efforts (and time) to make us all better at AI art. My hat goes off to him.

#

Whoever he/she is.

knotty turtle
#

I have a question about the concept of using a celebrity's name for training instead of a the generic "ukj", etc...
I recently tried doing this with a friend of mine. I used "Tom Hanks" b/c he kind of sort of a litte bit resembles him. I trained both a Checkpoint and a Lora. It worked pretty well for the Checkpoint ... however, with the Lora (and I assumed this would happen), when I use the Lora with a different checkpoint (say, Photon), it creates a person who looks like a radiated blend of my friend and the actual Tom Hanks.
Am I correct in assuming that this concept does not work for Lora's?

delicate trout
#

i feel like i'm missing something, the results i'm getting with xl aren't as a good as they should be

plucky shoal
#

are you using at least 1024x1024 resolution? @delicate trout

delicate trout
plucky shoal
#

hm should be o.k. Try 1024x768 just for curiosity

#

And do you have VAE?

delicate trout
#

i don't have any for xl

plucky shoal
#

hm, are you using A1111 or ComfUI or something different?

delicate trout
#

a1111

#

i mean i'm not supposed to use any, the model itself must be enough

plucky shoal
#

There is for A1111 Vae, go and get it

untold igloo
#

I have a quick question, so when I downlaod different models do I have to have the specific one I want selected in the "Stable Diffusion checkpoint" or will my prompt draw resources from all of them?

plucky shoal
#

no it isnt.

delicate trout
#

i have vae, i said i'm not using any for xl

plucky shoal
#

you must use VAE for XL version

delicate trout
#

which one

delicate trout
plucky shoal
#

i dont know. sdxl_vae.safetensors

#

@untold igloo it should use only that you have selected.

delicate trout
#

i don't understand, why using a separated vae

#

why is it not baked in

#

vae is more to boost results, but the model itself doesn't give perfect results in the first place

plucky shoal
#

Go in settings quick settings, and choose VAE so you can change it on the run from top of the A1111 window. Realy suggesting doing this

delicate trout
#

mate 😂

#

i told you i use vae i know how they work

#

i'm just not using any for this model because the model alone has to be enough

plucky shoal
#

so post your result in general with images, and pros here can help you, i bet it is because no VAE for XL

delicate trout
plucky shoal
#

o.k. so try 1024x1024.
Perfect result is something very personal 🙂

delicate trout
#

"perfect" for results is not subjective (personal as you say), it's something everyone should get, meaning no bad quality, no prompts misunderstanding, detailed results etc

#

here's an example of what i get, it's not normal lmao

#

even small models give better results, without any vae

runic turret
deft epoch
#

Is there a software like SD that can be installed and used locally

#

but to produce AI music instead of image?

glad aurora
#

what's this chat about?

deft epoch
#

Eeeer

#

general?

vital hatch
#

Hey everyone. I want to do a contest for prompt engineering.there will be no prize but a show of skill and uses. The model used for it will be normal GPT 3.5 so all can enter the contest (NO GPT 4) you can use the app on IOS or android or the Normal Website if you are on PC.

The Theme is: UltimateGPT
This will have a selection of a different modes for ChatGPT such as AI art mode,prompt mode (which it will make a prompt that works) and MusicLM prompt mode (go research it if you have not)

If you want to enter please DM me!

deft epoch
#

If there's no prize

#

then how's that a contest?

ornate yarrow
#

How do i change SDXL into a coreML?

knotty turtle
runic turret
#

From what I understood is the goal of the Lora is to use the images you created in them. To access it specifically we use the prompt name chosen for the Lora. If the name you choose has images in the base model that match that name it may use those images in the base model over your Lora images.

#

I may be wrong...

wary belfry
# knotty turtle Thanks for the reply. Just to make sure I understand, are you saying, then, that...

Yes and no. It's exactly how you described it. When training a checkpoint or using the Lora with exactly the checkpoint it's trained on it works. Because it uses Tom Hankls as a base and just alters the bits to make him to look like your friend. If you use another checkpoint as base your your Lora, then this perfectly tailort changes don't add up perfecty anymore. Image you want to generate the number 10. Your base model generates a 7, so the lora adds 3 and you get 10 as a result. Perfekt. But now your other basemodel generates an 8. Your Lora still adds 3, giving you a 11 as result. So yes it works with with only that base checkpoint and no, it doesn't work when you switch your base model

crimson belfry
#

SDXL is the best, Just a fyi

runic turret
#

Or for the most part

#

?

wary belfry
# runic turret If you name the Lora prompt something like swrbfd4 and use that and don't mentio...

yes and no again 😄 it still uses both, like with tom hanks. But if you use swrbfd4, then this token is emptly and has no information. So all the information comes from the lora, since the checkpoints dont have any for swrbfd4. Same example: Image you want to get 10 and for all checklpoints swrbfd4 gives you 0. Lora adds 10. Since all checkpoints have 0 for swrbfd4, adding 10 from the Lora always works 😉

runic turret
#

What I am saying is take the Tom Hanks Lora he made and instead us swrbfd4 as the name. In that case the Lora would add 10 and 0 from the base model. Sorry if I didn't make that distinction. Ie, name your loras something that won't be in your base model or other loras if that is your goal.

wary belfry
# runic turret What I am saying is take the Tom Hanks Lora he made and instead us swrbfd4 as th...

yes then it always works. But the benefit of using Tom Hanks is that it can jumpstart from all the knowlegde it has from Tom Hanks. It doesn't has to start from scratch, but has instantly a good understanding how that person works with all it's lighting, perspectives etc and just has to finetune the apperience. It just has pros and cons. If you plan to use it only with your super favorite model then it's usefull. If you want to switch basemodels it's useless

serene sun
#

In some pics I'm having a problem with the face(es) as they appear distorted before upsacling, but not on all prompts. Could it be that the model is more biased towards closeups? I'm trying make shots from a distance

wary belfry
#

yes faces that only have a small area on the images don't come out that great

heavy eagle
#

Total newbie. I have a ton of VERY basic questions. Is there a good channel for those?

#

(I installed A11111 and v1-5 on my home system, and it's running, but beyond that, I'm kind of lost)

wary belfry
heavy eagle
#

so... my guess is I probably want to download a better model -- on huggingface, I see sd_xl_base_1.0.safetensors. Can I just download that and stick it in the models \ Stable-diffusion file?

#

Or do I need to do additional stuff to make it work?

#

(also, should I delete the older ckpt file?)

wary belfry
#

I'm not entirely sure about sdxl with Auto1111. But I think the current version of auto can use sdxl natively. But it's now completely implemented yet

#

no keep them, you can easily switch between them

#

a huuuuuuuuuuuuge collection of models

heavy eagle
#

Thanks! I just dl them and stick them in the models\stable-diffusion folder?

wary belfry
#

yes! Then restart the web interface so they show up or use the little reload button on the top left corner next to the model drop down

heavy eagle
#

thanks! ... gonna go try that and see how it works!

wary belfry
#

there are checkpoint ( or models ) that work on their own and then there is some called Lora. It's like an addition to the checkpoint you are using. For example you have a model that's good at paintings and then you got a Lora can is good at Pokemon. If you load your Lora onto your model then you have a model that is good and paintings and pokemon at the same time 😉

heavy eagle
#

woah

#

... see, I was struggling to figure this stuff out. I would RTFM, but... I can't find the FM! really appreciate this!

wary belfry
heavy eagle
#

perfect!

wary belfry
#

it's a bit overwelming now with all those videos at once but it was super helpfull over the last months and taught me everything as it came out.

hoary kite
#

hey all

wary belfry
#

But the main things now are Checkpoints as your base and Lora as an addition to your base model. Later you can search for stuff like inpainting and one of the most impressive things is controlNet. You can even train your own models 😄

hoary kite
#

new to the group running AUTOMATIC1111 ON OSX and encountering the error COMMANDLINE_ARGS = --skip-torch-cuda-test
^^^^
NameError: name 'skip' is not defined

#

any assistance would be appreciated

real zodiac
hoary kite
#

RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check

#

is the response i get

real zodiac
#

mm well that sounds like a bigger issue, but anyhow did you put the flag in webui-user.sh?

#

should be on line 13 according to github

#

and you have to uncomment it

deft epoch
hoary kite
#

looks like that

#

Commandline arguments for webui.py, for example: export COMMANDLINE_ARGS="--medvram --opt-split-attention"
export COMMANDLINE_ARGS="--skip-torch-cuda-test"

proper sparrow
#

anyone have cool settings for motion for deforum videos currentl i am using "angle": "0: (0)",
"zoom": "0: (1.0025+0.002sin(1.253.14t/30))",
"translation_x": "0: (0),30:(15),210:(15)",
"translation_y": "0: (0)",
"translation_z": "0:(0.2), 60:(5), 300:(5)",
"transform_center_x": "0: (0.5)",
"transform_center_y": "0: (0.5)",
"rotation_3d_x": "0:(0), 60:(0), 90:(0.5), 180:(0.5), 300:(0.5)",
"rotation_3d_y": "0:(0), 30:(-3.5), 90:(-2.5), 180:(-2.8), 300:(-2), 420:(0)",
"rotation_3d_z": "0:(0), 60:(0.2), 90:(0), 180:(-0.5), 300:(0), 420:(0.5), 500:(0.8)",
"noise_schedule": "0:(-0.06(cos(3.141*t/15)**100)+0.06)",
"strength_schedule": "0: (0.65),25:(0.55)",

wary belfry
#

stick with Nvidea. Most tools work with them and they need less vram. But I don't know the exact one for your needs

#

hmm seems like the 1000 series is still very slow for AI stuff. Fastes here is GTX 1080TI. I think the 1070 will be even considerable slower

rotund coral
wary belfry
#

problem with that is thats an amd card. You have to inform yourself beforehand of it works with amd

rotund coral
wary belfry
#

with nvideia you generally can run everyting

trim magnet
#

if u wanna run sdxl with decent performance and little money just buy an used 2060 12gb

rotund coral
#

And many of us do video editting too. I still recommend that card I linked. Best bang for your buck. Until you can get into the 4k series NVIDIA.

rotund coral
trim magnet
#

but we arent talking about most people we talking about him

amber aurora
#

has anyone managed to do something coherent like a manga with SD?

#

seems like we are still a ways away

wary belfry
#

you know corridor digital? they made an anime with that. But it needed a loooot of preperation and exact workflows

#

here is episode 2 with improved coherence

#

they also have a making of where they explained what they had to do

amber aurora
#

damn okay nice

#

happy to see it

#

thx for hte link

amber aurora
wary belfry
vagrant harbor
#

Hope this is the right place to ask, but does anyone know of a good "model" to create potions/images of potions with? Or is that just kind of a general thing?

amber aurora
wary belfry
wary belfry
#

but yeah, saying that using the hands is the hardest part was meant more as a joke in the first place 😉

violet umbra
#

Hello guys Gm I’m a beginner trying to install stable Lm 7 billion parameters on my desktop in vs code could someone help me with some ideas and tips especially with me being a beginner to all this especially with coding and I’m using vs code 🙏🏿🙏🏿

amber aurora
#

right now i think they just do one-at-a-time greenscreens and then edit them on top

#

also, they have some weird solution to get rid of flicker, but why don't they try mapping multiple images in a grid?

#

not high enough resolution to do that i guess?

wary belfry
prime rune
#

guys whats the best model for making images that look like characters from an anime accurately

cursive jetty
#

Is there a way for a deforum to stop changing the image drastically? I’m currently doing something that isn’t exactly sfw but ||I’m trying to do a deforum zoom in to a girls boobs and the moment it gets really close, it suddenly changes the image completely||

#

Is there a way to not make these deforum animations change and stay?

unkempt zealot
#

I just ordered a 4070. Anybody here using one?

azure lantern
#

hi

fervent thunder
#

how do I use the stable code model?

true mauve
#

why is that sometimes sdxl will give me absolutely beuatiful images and some times just a bunch of cartoon drawings?

true mauve
violet umbra
#

Good morning

#

How do I run the run the stable Lm vs code. 🙏🏿

chrome spindle
#

Guys, do you have tips for tutorials on how to train your own embeddings for SDXL using google colab?

unkempt zealot
chrome spindle
tight mist
sudden ruin
tawdry verge
#

Is there an AI that’s free ?

left pilot
#

Hello, I'm using Automatic1111's API code to create my server where I can create images for incoming requests. Is there a good repo / project that focuses on API creation for stable diffusion? Something that supports auto scaling, request authentication etc?

unkempt hatch
#

is there a model that just gives people really big heads?

deft pivot
#

guys wha time does clip drop folow?

potent spire
zealous gull
#

Gpu broke?

clever inlet
#

Anyone knows approximately how much GPU memory to train LoRA on SDXL 1.0?

#

seems like cloud training is the only way as up to 32 GB VRAM is required to feel comfortable

rotund coral
rotund coral
fervent thunder
#

Hi

languid helm
#

which is the latest model i can download from huggingface

#

also i have only 16gb ram ryzen 5600h mobile processor and gtx1650 mobile gpu

pine fulcrum
heady canyon
#

Join my subreddit for great content, analysis and discussion. Link in bio.

hasty hornet
#

nope, seems like 1.0 was released too.

languid helm
#

thank you so much

astral goblet
languid helm
#

yeah seems like it

#

my ram said nah

#

TwT

#

i wanted to generate / extend wallpapers to span across all my monitors

#

:(

astral goblet
unkempt hatch
astral goblet
languid helm
#

wow

astral goblet
languid helm
#

thank you so much

rotund coral
astral goblet
#

i also recommend for wallpapers, wider generations it's hard to get a single subject to show up. LIke if you prompt for a waterfall, you might get two waterfalls on the left and right side. it's a lottery there. sometimes it'll do the prompt on the entirety of the image. sometimes it will do the prompt in quadrants and you'll have it rendering twice in the same wide image. Try prompts that don't focus much on a single subject or can easily get away with tiling the prompt across the width. "A forest" wouldn't look obviously doubled up where as "person" might show up twice unintentionally"

pine fulcrum
astral goblet
astral goblet
astral goblet
# rotund coral https://rikkar69.github.io/SDXL-artist-study/

while i love that site, it sucks for discovery. it's only tagged with artist names. i can't do something like search for all the oil painting styles. or all the photography artists. i have to know the name of the person to find them on the page and see what sdxl creates. or i can scroll through the infinitely scrolling name sorted list and wait for a style to pop out at me.....

it's just a bad design with the guise of being a good resource. it was really phoned in though.

#

the other lists on that page, much shorter and manageable

rotund coral
astral goblet
rotund coral
unique hamlet
#

is it me or the latest version of AUTOMATIC1111 broke most extensions?

astral goblet
#

it's just the alphabetical artist name list that i think they really just failed at. the rest of the site's examples are very manageable

astral goblet
unique hamlet
winged jungle
#

hey! sorry to bother you but i'm curious about the difference between the normal models and fp16 versions (using web ui). i know it has to do with floating point precision (what's mixed precision?) and i was wondering which one to get

cinder dune
#

is a rtx 3060 12gb enough for SDXL in auto1111?

sand light
#

Is there any model that works well with making textures for games?

astral goblet
#

extensions may break more often on dev branch, but i've had little problem there

hollow hound
#

Is it ok for people to make content from this server? Like screen recording, or posting the results?

#

Dunno, thats why im asking..its not about "care" anyway..

sly tusk
#

cow utter

wary belfry
#

use Gradient Checkpointing and Adafactror to save a lot of vram. Also lower batch sizes save Vram

willow coyote
#

wheres our black guy fro BC with the bot channel?

#

so horrible at names

echo yew
#

hey

sick hornet
#

is it possible to train an inpainting model on a specific object? I would like to inpaint a specific object into photos

astral goblet
#

you don't need an inpainting model to inpaint, it helps but not needed. especially now with controlnet inpainting. But yes, you can also create inpainting aware models

#

imo the controlnet model adds the capability to every 1.5 model and does it better

sick hornet
astral goblet
karmic vigil
distant swift
#

Does anyone know some kind of node that allows batch count to create a grid like A1111 does? I kinda miss that feature

deft epoch
#

Sounds like lack of VRAM.

#

That's way too low.

chilly vector
#

Anyone got any luck running sdxl on openSUSE tumbleweed?

trim magnet
#

yes

narrow cargo
#

How do I make my images look real here in the bots? do you have any prompts?

#

I wanted us prompts for a photorealistic car image

livid cloak
#

Hello, anyone knows why SD doesn't want to use my GPU? Instead is using my CPU. I have an AMD GPU, followed the instructions on the wiki but stills wants me to use the command to ignore cuda, this didn't happened before

hallow plaza
#

Has anyone tried stableCode yet?

fervent thunder
robust magnet
#

hello?

#

I'm new here

pale coral
#

hello, is there any channel dedicated to models or loras in here?

karmic brook
robust magnet
#

ONG THIS SERVER IS LITTERALY THE BEST

#

THIS AI GENERATED ART IS SO COOL

#

look at the general with images

#

it gave me sum hot man

#

realll

#

I don't know what to say but this server will save my life

turbid whale
#

ong ong fr fr, this ai bussin

vast berry
#

one of the stupidest negative prompts i've ever seen anyone put as an SD default (not in this server) is 'ignoring prompts'

#

oh yeah its gonna listen to you now kappa

trim magnet
#

just put ( 0.0 ) on both prompts

thick trout
tender cove
#

I think you mean an instruction to disobey instructions. An instruction to ignore instructions isn't a paradox

#

Also paradox is the wrong word, but I can't think of the right one

#

You can ignore an instruction and still do what it says

#

Otherwise it's just reversing them

unique hamlet
#

I have a question about celeb loras: if using someone likeness is illegal, how comes there are so many celebs loras on CivitAI? Is it because it doesn’t fall under commercial use? 🤔 What are the legal risks of creating a celeb Lora

trim magnet
#

who said it was illegal

fervent thunder
#

I want to change the color theme of a pic from green to purple, what prompt should I use?

lavish lake
lavish lake
fervent thunder
#

nah img2img

#

I tried using purple color theme but nothing worked

#

also if you upload those pics you said on a anonymous place there is nothing coming your way as well

fickle agate
#

how can i invite the stable diffusion bot in my channel

pastel ridge
#

i need help pls

fervent thunder
#

what's up

vast berry
lavish lake
proper sparrow
#

what resolutions stable diffusion can handle for reels and shorts

unkempt zealot
bleak matrix
#

Good morning, good morning, everyone! How are we all this beautiful day?

unkempt zealot
bleak matrix
#

What are you up to today?

unkempt zealot
bleak matrix
brave pike
#

You pro at sdxl prompt engineering?

deft epoch
fluid crescent
#

H-hey there guys. Sorry for this but..
How i do a search right now in Civitai ?
Just tried to search for LYCORIS
But when i press enter, nothing happens.. ?
Sorry my english.. :/

turbid whale
fluid crescent
bleak matrix
brave pike
bleak matrix
#

For prompting

visual ledge
#

does anyone know the settings the server bots use for fantasy art style? trying to match the generation quality on my system.

brave pike
visual ledge
#

i was more specifically looking for the sampler and think i got it. as far as i can tell uses DPM++2S a Karras.

#

got the rest of the info from the bot DM

brave pike
#

What is the prompt

vast berry
gray wolf
#

What’s the best current local stable diffusion app with a gui? Using Automatic1111 but wondering if there are any better ones that u can preferably use on your phone running of your pc for example

solemn yew
#

Hey guys!

brave pike
#

Its a website that supports Android and all sdxl models including dreamshaper and anime diffusion xl + Most Lora

bleak drift
#

Heyyy. I have a question.

#

How hard would it be to use stable diffusion to swap someone’s gender.

#

Like closeups and stuff.

#

And keep the face the same in different positions?

turbid whale
bleak drift
#

What is roop?

turbid whale
bleak drift
#

I shall check it out.

broken cave
frank spruce
rotund notch
#

hi

#

so I got an incredible result img2img with controlnet and a qr code, and am unable to reproduce. Is this expected ? it's totally random even when you input the same seed and prompt. Am I supposed to generate HD pictures every time just in case?

sinful stag
#

hi

turbid whale
#

It might not recognize anything else beyond the first if you use single image cnet.

placid swift
#

Hi

stuck quartz
#

how to make resize like a wallpaper

cloud kayak
#

is there a service I can use to train a custom model with new photos on top of base model that I can select?

inner crest
#

I use adetailer for faces

#

if that doesn't work try hires. fix

proven holly
#

i have a real challenge i need help with. I have an image idea that is so specific i am not sure if AI will cut it.

#

need help with putting a lot of things into one image. a 1960s neon orange tow truck that has a lot of specifics. The background behind is in a fiery derby arena

little citrus
plucky shoal
#

And regional prompter can help very there @proven holly

proven holly
#

i'm not good with the AI as well. Someone tried to help me create what i wanted but they couldn't get it. I'll explain what i need.

inner crest
#

I've been messing around with the concept just as practice for img2img and inpainting. still not much success yet

#

this isn't what I intended but it looks really cool

#

I can't send images here

proven holly
#

i can pull the description of what i am needing

#

The Rambling Wrecker was a tow truck given to the driver a year before he mysteriously disappeared. The tow truck was painted neon orange. With other reds, yellows, and oranges painted to look like it was on fire. The truck’s front bumper was all brown as it was rusted out. The body of the tow truck was covered with various dings and dents from racing over the years. All the lights on the truck were either broken or missing.
What worried Johnny was the armament and weapons the tow truck was carrying. On top of the roof was a missile launcher. It had six missiles loaded onto a rack. On the hood were two gatling riffles on either side. The front end of the truck had a wedge-shaped plow. The plow had four small holes in it. Two to launch flame throwers. While the other two shot off beams of plasma. On the rear of the truck, it carried land mines. The truck also had barrels of oil the magnetic tow cable could threw at opponents. The barrels could also drop oil slick. Though what made the Rambling Wrecker famous was the tow cable. The cable had a magnet hook line that would trap another driver by swinging them around like a wrecking ball into objects and obstacles. Many drivers died to the magnetic cable.

inner crest
#

I sent you a DM with the messed up but still cool image

proven holly
#

looks interesting but i would like to have the image mirrored

mild hedge
#

Has any one ever heard anything back from any emails sent or tickets logged? I have been trying for weeks to get a response from stability.ai or dreamstudio.ai but have heard absolutely nothing. It is super disappointing that they have such poor customer service. I'm not asking for much, just a tax invoice! Are there other people here using them for a business purposes and if so have you been successful in getting an invoice from them?

wise bolt
#

Hi, every text-to-image generation from different users on different IPs is always queued first. And I want each of these different user-generated outputs to occur concurrently instead of queuing sequentially. Is there a possible solution for this?

keen epoch
#

how to setup a some kind webui on local pc with stablecod to play with?

deft epoch
inner crest
#

you can and in some cases I do, it just makes it take longer to generate. I tend to keep hires. fix on all the time and only turn adetailer on when a gen is giving me trouble

late blaze
#

What software do people use these days to generate locally?

inner crest
#

I've always used automatic1111. I tried comfyui but it was just less efficient and didn't have half the plugins I use constantly in 1111

lavish lake
inner crest
#

SDXL has yet to win me over. From what I've heard, it lacks of compatibility with SD1.5 Loras and the lack of a good anime model from what I've seen so far also doesn't help

lavish lake
turbid whale
#

Why would it be compatible with 1.5 loras?

lavish lake
inner crest
turbid whale
#

I don't think it's as simple as making things backward compatible. checkpoints and loras between 1.5, 2.1, and xl don't share the training data on the same resolution imagesets.

inner crest
#

I interpreted your use of why differently. Functionally it wouldn't be possible, but ideally it would be nice if it was

turbid whale
#

You're good, I was a little confused myself. Haha.

bleak matrix
#

Good morning, everyone! How are we all today?

proper nebula
#

hello everyone, I would like to know if we can creat a earth that revolves around itself

#

animation

violet umbra
#

Hello 👋

hazy pulsar
#

Does anyone know if vlad (now SD NEXT?) supports SDXL? I git pulled the latest and still get size mismatches on the XL models

pearl jungle
#

hi , does anyone have recommendation for model that could do justice to Lovecraft works ? ideally a sdxl model

trail lion
plucky shoal
#

@pearl jungle how your prompts for Lovecraft images looks?

pearl jungle
#

nothing particular, just put lovecraft and some scene or description from a book (and the usual negative prompt)

trail lion
#

have to be careful with negative prompts when you're doing something exotic, I've been bitten by that...start out with no negative and add tokens one by one to control what you dont want

#

you might end up having to do a lora for this, I'm sure you can find some images online to get it going

#

sounds like a fun project, I might play around with it when I'm not already training something

plucky shoal
#

There is some Locon or how on civitai, but seems not any miracle

pearl jungle
#

i ll keep working on it, i ll post something when i get decent results

#

thanks for the advices

plucky shoal
#

images by H. P. Lovecraft.
Try this prompt @pearl jungle

silent fractal
sacred shell
deft epoch
rotund notch
#

hi. how do you reproduce a result in HD once you are satisfied with the 512x512 output? img2img+controlnet+same seed always produce very different random outputs

plucky shoal
#

Try hiresfix Latent gives similar result but not identical, I personaly using extras tab and there classical GAN resizers.

fervent thunder
#

i have a dataset of 100 images of a character

#

i am training a lora of sd xl

#

how many steps should i use

#

i am using kohya

#

25 Epoch, 20 repeats is okey?

fervent thunder
# fervent thunder 25 Epoch, 20 repeats is okey?

20:19:51-080943 INFO Total steps: 2340
20:19:51-081943 INFO Train batch size: 1
20:19:51-082943 INFO Gradient accumulation steps: 1
20:19:51-082943 INFO Epoch: 25
20:19:51-084447 INFO Regulatization factor: 2
20:19:51-085463 INFO max_train_steps (2340 / 1 / 1 * 25 * 2) = 117000

rotund notch
plucky shoal
#

yes i am using SwinIR 🙂

rotund notch
# plucky shoal yes i am using SwinIR 🙂

SwinIR actually tries to repaint what it detects in exactly the same style while the others mostly just upscale and smooth it. R-ESRGAN 4x+ is pretty close too and cleaner, it's just different. the most perfect one is SwinIR indeed

bright nest
#

hi, im using comfyui SDXL and having an issue with seed/random/fixed. when i generate an image using "randomize" seed and i want to use the same image to send to the refiner i hit "fixed" however the base generate a different image even though the seed does not change. is this a known issue? and is there a way to fix this

potent spire
#

nobody using Alkaid.art for PS?

alpine locust
#

hey all is developer sandbox and profile dashbord down?

plucky shoal
#

@bright nest how about drag drop image you want to send to refiner? I think when you hit fixed is already different, not sure.

bright nest
plucky shoal
#

you are right, i forgot it keep stored workflow... But you can at least check if same seed as when you press fixed

bright nest
devout yew
#

hello

#

got a quick question

#

are these generated images with open source license? and can be used on commercial work like web designs?

wanton nacelle
#

Why are 75% of the photos in pantheon just half-naked girls?

tidal pier
#

Why can't I access 127.0.0.1:7860, how to solve this problem

#

I'm trying to do a local deployment, but have been unable to access 127.0.0.1:7860

little citrus
tidal pier
little citrus
#

First thing, make sure the webui has started and says it is listening.

#

Show console output from startup.

lavish lake
tidal pier
#

i already have down load and reinstall more than 3 times, but still can't work out it

lavish lake
#

usually there's people that can help with tech issues either on the chat or monitoring

tidal pier
#

I was very surprised before that I couldn’t find the bat suffix file in my root directory. My webui.user.bat only showed the webui.user suffix. I re-downloaded a bat suffix file, and it was still the same, but I double-clicked him When running, it is displayed as the webui.bat file. This may be the problem. How to solve it?

little citrus
reef hamlet
#

what are the best resolutions to prompt SDv1.5 at? Like for different aspect ratios

fickle siren
#

How do I recall a seed with the bots?

deft epoch
#

Is that new AI generator they released worth my attention?

pine path
#

Anyone mess with the restart sampler from a1111 dev branch? I didn't try it out much myself, but those I've seen talk about it have praised it.