💬｜general-chat

[rapid atlas]

Mobile, but I'll restart my phone lol

[wild steppe]

It should be available now! 👍

[fervent thunder]

I can feel SD GPU's going brrr under all this bot pressure 🫡

[deft tiger]

tanks!

[karmic brook]

[sleek osprey]

The finetunes for xl might end up better than midjourney bruh

[clever stone]

how many images does each person get

[wild steppe]

no limit atm but please vote!

[clever stone]

vote where and for what

[wild steppe]

🅰️ or 🅱️ underneath the generated images & #1100484581037195384

[clever stone]

ok

but why

[wild steppe]

By participating in the A vs B voting, you're helping to train the model by providing valuable feedback on which images are preferred by humans. So not only is it a fun activity, but it also helps improve the quality of the images generated by the bot in the future!

the highest voted images will be iteratively showcased in #🏅｜pantheon

[spice plank]

Lol how many rooms this server got now?

[clever stone]

okkk ty

[twin pasture]

do i have commercial usage permissions for the images generated on discord?

[wild steppe]

A TOS is being updated and will be available soon for our discord bot soon to answer questions on usage!

[twin pasture]

may i ask for some of the info now, or is it prohibited to leak that now?

[wild steppe]

It's still being finalized so I'd rather not give you any inaccurate info

[twin pasture]

alright then

thanks

[wild steppe]

Thanks for your patience!

[kindred idol]

Hi ! Is there any possibility to get the sdxl model for research/testing purposes ?

[broken gull]

can I convert diffusers to ckpts?

[hidden bridge]

How do you get more realistic results because what I type is not what I get usually

[stark hornet]

is the two point one channel gone ?

[wild steppe]

2.1 channel is renamed to #1045349359044280360 as we've launched our SDXL Beta bot! #1100170312106127410

[stark hornet]

oh

will sdxl run locally ?

[wild steppe]

eventually yes it'll be open sourced

it's in testing/research preview atm

[stark hornet]

will it run on a 4090 ?

[deep stone]

I have a very quick question

I got the python scripts installed but I forgot to take the local URL

is there a way to find that?

nevermind

I found it

[hallow nymph]

does anyone use Draw Things app on mac? is the latest SD model always the better one or are older versions better at some things?

[shrewd perch]

O

[spice plank]

I updated webui, and controlnet and I don't have the reference only mode

[fervent thunder]

privet pupsik

[waxen pasture]

Are we supposed to be able to write in #1103141787960344576 ?

[hasty bane]

hello, I'm uploading some images for sale on image banks, and reading the site's guidelines, it says you need image rights permission. I don't understand much of this, so here's the question, is it possible to sell the images in image banks without problems?

[jagged remnant]

Ayo, why is my ETA timer increasing? 💀

[upbeat hill]

gm

[fringe current]

how do you guys add the negative prompt in here?

[karmic vigil]

Anybody have a landscape prompt they wanna give me to try?

[graceful spade]

hello

[pine fulcrum]

Any good models for generating dnd city/world maps?

At least like the base for one

[vital wadi]

I tried to use your style csv file shared from some yt AI Master but I have problem. I do not have originally in my root folder style.csv file. When I coppied there file SD didn't read it. I created my own style, and save it. It somewhere store, becasuse I closed and open SD and styles are save but not in styles.csv file. Any Idea what is wrong?

[radiant nacelle]

Hey this new model looks cool

[dim latch]

[spice ledge]

Hello, did we have models more specialized for no humans creation ? I need to design a card but not for illustration purpose more to add text on it after and don't see anything on CivitAI (and to be fair, i not launched SD since months , impressed by the evolution since that time but kind of lost about the possibilities we have now)

[fervent thunder]

why I do not get the option copy as path for Python to paste it on the notebook and I only have just copy? Any ideas?

[errant plover]

worst discord server ever

[alpine bear]

Ikr? Free Guiz!

[radiant nacelle]

is there a website for all kinds of art styles and artists?

does anybody know how can i do that

[thorn pilot]

https://proximacentaurib.notion.site/e28a4f8d97724f14a784a538b8589e7d?v=ab624266c6a44413b42a6c57a41d828c

[radiant nacelle]

thank you

[tribal root]

what happen too the auoto1111111 web ui?

[fervent thunder]

why I do not get the option copy as path for Python to paste it on the notebook and I only have just copy? Any ideas?

[balmy jungle]

i am now discovering it's very difficult to get stable diffusion to make a cow as an armoured tank.

[near silo]

I have a quick question for you all, not sure if anybody knows the answer

if I am using a model that has a tag that is like

character (movie they are from)

how would I go about triggering that tag?

Cause the example I have just triggers the character name and movie separate

[copper crystal]

if tags have brackets you need to escape them so they're not read as emphasis brackets

\( moviename \)

the actual prompt there would be (moviename ) and won't be ignored for emphasis syntax

[twin dirge]

is anybody here that works with Sad Talker or Spin thin motion model?

[near silo]

See, I tried that and it didn't seem to produce any better results

For example, lets say its like

Toothless (How To Train Your Dragon)

it would then be

Toothless \ (How To Train Your Dragon\ ), right?

or are the spaces important in this case?

I appreciate your help BTW!

[copper crystal]

\( the slash immediately before the ( to escape it

[near silo]

ok so

[silver flare]

what's the difference between stable and unstable?

[near silo]

immediately before the first and second?

[copper crystal]

the tag may use underscore where the space is. i know that's how booru tags work

[near silo]

so like |( |)

but with \

[near silo]

Luckily, I have a list of all 81.3K tags in this model 😅

[copper crystal]

yeah haha. to put \ on discord you need to do \\ and to show that i had to do \\\\

[near silo]

alright, I really hope this works

Thank you again BTW!

[copper crystal]

hope that works for you. it'll prevent the brackets emphasizing the prompt at least

[near silo]

I should still be able to put parenthesis around the whole thing tho, right?

like (Toothless \(How To Train Your Dragon\))

[copper crystal]

yup

[near silo]

alright, thank you! It doesn't seem to be triggering right, but I can only assume thats cause part of the token is much stronger than the part I am trying to trigger

[copper crystal]

try the _ where spaces go. booru tags are listed with no spaces too, but the training adds a _ so it is a consistently understood token

oops i mean _

[near silo]

I tried that as well and it seemed to make no difference unfortunately

I asked on the civit page if anybody knows. This is the most insane model I have ever used, and if I can get these to work, I will be unstoppable lmao

[calm zinc]

how to create private channel in this

[copper crystal]

whats the model

[near silo]

I can't share it here unfortunately, but just know its hands down the best thing I have ever used. 1900+ artists, 850+ species, 71k+ exact tags, 50 hue sorted colors, full NSFW (outrageously good), 35 different baked in framing presets, and reliable multi character/subject support

[floral timber]

After using SD for about a week now, I can see my interest in it fading fast. To be exact, its inability to use the same characters and environments from simply different angles or situations makes it sort of a one-off novelty creation tool. Yes, you can sort of train it to create a certain person, but it's always going to be a different generated outcome. Now, if there was an extension called "Same Scene, Different Angle" and "Same Character, Different Pose" then I'd be on this train and hollering. Are there such extensions that I'm just not seeing?

I'll stick with DAZ3D.

[near silo]

there are several ways to do it now. There are additional networks for multiple views of the same character

multi layer diffusion where you can generate the background, character, and lighting separately

And now we have reference only controlnet, which can basically look at any image you feed it and transfer the features and style of the image you provided, and it works damn good

[floral timber]

OK, so hip me to what and where to find this multi-layer diffusion thingy.

[near silo]

don't personally know off the top of my head, haven't used it in months

this is what reference only can do

or wait

move over to #🏞｜general-with-images

[warm junco]

Same person can be done with using a trained Lora

Then you can easily adjust the Pose or Setting

[copper crystal]

if it doesn't fit your established workflow, then i wouldn't try shoe horning it in either. like cgi in movies where practical effects would've worked better and cheaper. sometimes it's just not needed.

long established workflows (like daz3d for you) still have a huge place in creativity. It's good to learn this technology while i'ts new and young. I think you'll end up comign back to this though still.

there's a new "reference only" mode in controlnet that is quite remarkable. able to do same scene and character pretty well!

[copper crystal]

so its not on civit because it's a celebrity deep fake model.go figure.
check yo ethics bro

[indigo ermine]

Anyone have a good recommended flow or method for taking a photograph and turning it into a painting via img2img? In my experiments thus far, when I push it toward the prompt for the style, I end up with a totally different subject.

[copper crystal]

controlnet

or loopback script with a mild denoise setting too

[indigo ermine]

yeah, I had been using Controlnet but mostly just "trying settings at random" with minimal success 🙂

haven't tested loopback at all... I'll look into it

[copper crystal]

it's an img2img script. sends the image back in and runs it again

[indigo ermine]

yeah, I've seen it mentioned and such but didn't think to test it out for this project... gonna check it out

I'm helping with a theater production that calls for a painting of one of the actors as part of the scene, so I figured I'd throw some AI at it!

[near silo]

uhhh, it has nothing to do with celebs, but ok

it is on civit, its name just have lewd terms in it lmao

[near silo]

Not even sure where you randomly assumed that, as I mention literally nothing about celebs lol

[gloomy cedar]

where you reading about celebs on there dude?

[copper crystal]

lol y'all acting like i'm making shit up and deep fakes don't exist. okay. carry on.

not very many reasons to be shady about sharing unless you're out in that clearly unethical zone

you do you. i'm just saying. it's obvious.

[gloomy cedar]

he literally is sharing it

its on civit

it just cant be shared here

because this server doesnt like nsfw...

[copper crystal]

i bet dude

[gloomy cedar]

🤦‍♂️

i hate people like you

[copper crystal]

thats so nice. cool story.

[karmic brook]

Please be respectful to other users @copper crystal
Uncalled for attitude

[copper crystal]

why is it whenever deep fakes get brought up, people get gaslighted? mods gotta understand that the people who want to do deep fakes will rush to you and make stuff up immediately. don't fall for it. lets move on

[karmic brook]

Im not here to fight you about deepfakes
I came here because of your outright attitude for your own misunderstanding

So, yes. Let's move on

[copper crystal]

i still don't believe anything was misunderstood. obvious reasons are obvious. i'm allowed to believe what i wnat and i'm not imposing that on anyone else. lets move on.

[karmic brook]

Okay. Come back when you actually want to move on.

[gloomy cedar]

bro he literally never said anything about deepfakes... at all you are just guessing very hard for nothing

[karmic brook]

Yeah, I had to reread a couple times myself

[near silo]

Its a furry model man, I am using a furry model with a name that is too NSFW for this server. @copper crystal

https://cdn.discordapp.com/attachments/978453842733719632/1108067218228514996/01419-1375993588.png

just making portraits, stop thinking I am some unethical perv. I'm not Sen lol

[gloomy cedar]

link me that model dude

[near silo]

I'll link in DM's

[swift roost]

Welp, HF chat is now log-in only

About time I get my servers ready for self-hosted AI

[uneven moth]

Can anyone explain why using Automatic1111 I cant get SD to create a scyth. Every attempt just gives me random looking characters

[fleet kayak]

hi yall

[bleak silo]

what is the model behind sdxl bot?

[near silo]

Haha. A Scyth (that used to be my old brand name before Sytan)

If you mean the weapon, its spelled Scythe, with an e at the end.

It could also be because stable diffusion decided to go out of their way to prune all weapons for whatever reason

[fervent thunder]

https://skateparkhero.org/2023/brandon-34

[next tulip]

is there a model for SD that can turn pixel art into other styles? more realistic etc. My assumption is try to use something like pix2pix perhaps

[tall blaze]

You got problems

[near silo]

you can gladly keep your opinions to yourself, thanks

[drowsy agate]

What does it mean to "Ensure" XPU availability?

[silent dome]

Hi all, I have a question what's the best channel to use for questions?

[cinder herald]

anyone interested in making some money by making a bot for me with stable diffusion

[silent dome]

I guess I'll just ask on here unless told otherwise. I have stable diffusion set up locally with auto1111 and installed checkpoints, Lora's everything. It works but my pc sucks lol. Is there a way to use a cloud service and just upload my stable diffusion folder with all my chkpoints and Lora stuff directly and just run it or what exactly is the process?

[stray nymph]

I saw a comment on reddit that in SD 2.1, 18+ stuff is gonna be banned. I just wanna know if this is true or not by just a yes or no, no further explaination or discussion needed.

[waxen pasture]

2.1 has been around for quite a long time, and NSFW was removed from the dataset, most speculate it was due to press and legal issues related to CP and stuff. There are probably finetunes of NSFW for 2.1 but I am not gonna go looking for them.

[stray nymph]

thanks for the answer, all i needed

[placid token]

Hi everyone. Quick question. What is the difference between prompting here and prompting on the website Dream Studio?

[mortal ocean]

Hello everyone. Quick question, does Stable Diffusion have a tool like MidJourney's /describe feature?

[muted canyon]

Where’s the porn

I want porn ai

[waxen pasture]

wont find that here... try like unstable diffusion or smthn

[tranquil bough]

Anyone know how to get video like this?
https://www.instagram.com/reel/CsOsJcRIZOC/?igshid=MzRlODBiNWFlZA==

[spice plank]

So the various versions of SD, I'm a little confused. If I have stable diffusion on 2.1 and I use a checkpoint from 1.5 would the nsfw stuff not show?

I'm very new to this stuff, sorry if it's a dumb question

[slender cave]

whats happening

men at work who could it be now

just when you thought it was safe to go back into the water LOL

[fervent thunder]

i found this extension in automatics webui extension list https://github.com/Physton/sd-webui-prompt-all-in-one
and apparently it has some tencent api stuff built into it and it bricked my venv

reinstalling now

[outer crescent]

for the love of god don't send things you've written in a prompt box through an api to other countries. I'm iffy about even using gfpgan and t2i-adapter without outright sending actual text information abroad. every country has different privacy and ai laws.

no point going all davinci code on it, but avoid.

[fervent thunder]

yeah

fastest uninstall and reinstall

its some kind of auto translation

def translate_baidu(text, from_lang, to_lang, api_config):
    url = "https://fanyi-api.baidu.com/api/trans/vip/translate"
    app_id = api_config.get('app_id', '')
    app_secret = api_config.get('app_secret', '')
    if not app_id:
        raise Exception("app_id is required")
    if not app_secret:
        raise Exception("app_secret is required")
    salt = random.randint(32768, 65536)
    sign = app_id + text + str(salt) + app_secret
    sign = hashlib.md5(sign.encode()).hexdigest()
    params = {
        'q': text,
        'from': from_lang,
        'to': to_lang,
        'appid': app_id,
        'salt': salt,
        'sign': sign
    }
    response = requests.get(url, params=params, timeout=10)
    result = response.json()
    if 'error_code' in result:
        raise Exception(result['error_msg'])
    if 'trans_result' not in result:
        raise Exception("No response from Baidu")
    translated_text = ''
    for item in result['trans_result']:
        translated_text += "\n" + item['dst']
    return translated_text
    # return result['trans_result'][0]['dst']```

has things like this

good for people who dont use english i guess

[outer crescent]

yeah i can tell from the gifs. the bottom of that page is covered in crypto donate links. It mightn't be intentionally sus. And it is tagged as online. I suggest you disable online extensions from the menu. for one thing gradio is just a demo framework so better kept isolated from the internet as much as possible

[fervent thunder]

yeah it did what it should do but its bloated with all these things for just a text formatter

[hallow plaza]

Are there any blip2 git repos that I can run remotely like on runpod?

[thorn dove]

Hi guys

Im new to stable diffusion

I just wanted to ask one simple question:
https://pbs.twimg.com/media/D6ylAv6UcAAYRK-.jpg

Which model should I use to expect such result? Cinematic color grading + photorealistic model

Most of the photorealistic models i saw shows results of potraits. But I want a full body result of a man sitting Down

[karmic berry]

none

at least not with that amount of detail, if that image is AI, then it is a multi-step process

[radiant nacelle]

I think its gonna evolve some day to that quality

I do not think im sure

It's developing very fast

[little sierra]

Hi guys, I'm new to SD. For some reason I am getting very similar results even after changing prompts completely. Seed is set at - 1 and I am using control net. Am I missing something?

[karmic berry]

what model are you using

[little sierra]

Realistic vision v20

2*

[velvet owl]

Just joined but damn super disappointed to see there’s no monthly subscription option 😞 and no bot

[torpid raptor]

Dumb question, but is deep-floyd available to use by stability.ai's API? If so, what engine_id is it? Or do I need to spin it up as my own api?

[plucky shoal]

there is dream studio. @velvet owl

[surreal perch]

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

Currently, there is no public bot on the server that generates images. There is an experimental bot available for early server members & Stable Society. You can obtain the Stable Society role by winning our weekly events #⭐｜pow-info #1087493421209485393! 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 owl]

Just a shame there’s no discord bot, especially one you can dm. Would make it so much easier.

[plucky shoal]

but there is subscription. Cheap as your OP.

[velvet owl]

What do you mean?

[warm junco]

I thought Dreamstudio has a monthly subscription. Its the official SD Website from stability-ai.
https://beta.dreamstudio.ai/generate

[velvet owl]

I don’t see a subscription option, just the option to buy credits

[warm junco]

Oh your right, i thought its there. Will give them feedback to implement it

If you have a good PC you can also run it for Free localy

[plucky shoal]

yes @velvet owl but it is very cheap. I think it is easy thinking about it as about monthly subscription

[velvet owl]

Thanks 😊 could you also give feedback about having a discord bot please? I’ve used MJ for so long that using anything but DMing a discord bot seems silly lol

[karmic berry]

this isnt mj

most people here diffuse locally

[velvet owl]

Thanks genius.

[karmic berry]

why would SD aim to be a MJ clone

[warm junco]

Can you access https://discord.com/channels/1002292111942635562/1100170312106127410 ?

[velvet owl]

Is it more expensive to buy the tokens or do monthly subscription?

[velvet owl]

I can 👍🏼

[warm junco]

Nice there and in the other bot1-10 channels you can try the new SDXL model.

[plucky shoal]

i think there isnt monthly subs. But i read it is so cheap to just cover electricity and etc.

[velvet owl]

I meant one I can dm with so only I can see the results and I’m not having to scroll through everyone else’s lol. Do you know if that is/will be an option?

[velvet owl]

I just used the 25 starter credits in 3 generations lol maybe I’m doing something wrong

[warm junco]

I dont think that will be an option for now. As the SDXL is launched here to get User feedback and to test it.
But i dont know if it will be an option in the near future.

[velvet owl]

Fingers crossed, it’s a feature I love about using MJ.

How did I burn through 25 credits on 3 generations though? I didn’t modify any settings except the aspect ratio. Steps was automatically set to 50.

[plucky shoal]

i think it is for tagging images that feedback.

[velvet owl]

I don’t understand, sorry?

[plucky shoal]

I mean you are helping SD to vote for picture that fits more positive prompt or if both picture do, which is better.

[velvet owl]

Ahh okay I get you now.

[warm junco]

If you use the bots here yes. Not in Dreamstudio

[velvet owl]

@warm junco is this a question you’re able to answer?

[plucky shoal]

there is man used showing how cheap it is using max 30 steps. Depends on resolution and steps very. 50 is way too much.

[warm junco]

I think the models and resolution impact the credits used

Also check your dc dm

[sudden osprey]

Hi guys I just wonder if I have architectural sketch in b&w. How I can make beautiful render picture from stable diffusion?

[sand adder]

Where would i share Videos?

or would i just post a Yt link

https://youtu.be/yLji9njYBag

I created this using stable diffusion and it took around 20 hours or so

or maybe more like 17

[sand adder]

im new

lol

you have any art that you would like to showcase?

[lone geyser]

yo

[sand adder]

Yo

[lone geyser]

im having some problem installing pip

[sand adder]

Idek know what that is

Something to do with Stable diffusion?

[lone geyser]

ye

when I click on the

webui-user-bat

[sand adder]

Open the launch file first

That’s what I’d do

I did*

[lone geyser]

what launch file

can we chat in IB

the image doesnt support here

[sand adder]

You got stable diffusion for the automatic1111 GitHub thing?

[lone geyser]

ye

[sand adder]

There should be a file named launch on true same folder as the stable diffusion file that you run every time you want to run SD

I don’t recall everything perfectly as I am on my phone rn

[lone geyser]

can u like guid me

guide*

come to the inbox

[sand adder]

@me in it if you can

[lone geyser]

i have

[amber dust]

Can anyone help me with a matter

[soft sparrow]

is IF open source yet? Can i use it to inpaint?

[torpid zephyr]

With this new version of the DreamStudio GUI, did they get rid of the slider option that lets you decide how close to the source text the image will be?

I can't find it anywhere. If so, that seems like a pretty big omission.

[spark kite]

New state of the AI Horde is out! https://dbzer0.com/blog/state-of-the-ai-horde-may-2023/

[mortal ocean]

Hello everyone. Quick question, does Stable Diffusion have a tool like MidJourney's /describe feature?

[spark kite]

what does that feature do?

[mortal ocean]

It allows you to upload a image and it creates prompts for that image.

[spark kite]

something like this? https://tinybots.net/artbot/interrogate

[mortal ocean]

Thanks I will give that a try.

[ashen hound]

Hello everyone , How can we run stable diffusion models in a low end pc , whenever i install it my laptop hangs ,therefore i have to delete it . Is there a way to run stable diffusion models on a low end pc?

[steep sage]

https://www.nasdaq.com/articles/openai-chief-goes-before-us-congress-to-propose-licenses-for-building-ai

[ancient meteor]

Hey! I'm trying to get the webui to work, but I'm getting "exit code: 3 stderr: The system cannot find the path specified"
Does anyone know a fix to this? Is this even the right channel to ask?

[plucky shoal]

#🤝｜tech-support better ask there

[latent depot]

Can you help me to create images in landscape mode which text should I use? so that the images appear horizontally

[fervent thunder]

Hello, I need the best model for realistic cars

[unreal oriole]

Get a good realistic model + car Loras lol

Also civitAI is down

[copper crystal]

yeah civit is 503 right now

can look on hugging face for models but, its no civit

[unreal oriole]

huggingface has like 3 models

CivitAI is back

[wild steppe]

starting shortly! see you in #1029055412764422214

https://discord.gg/cyP4Vkse?event=1107712581495439431

[ocean venture]

are stable stuff just discord admins or stability ai staff

[wild steppe]

Stable Staff are all Stability AI staff

[ocean venture]

when you finish this if you dont mind can we go dms? I need a question answered

[wild steppe]

We don't answer questions in DMs but feel free to open a private ticket in #1010934719455707218 if you have any questions for the team!

[ocean venture]

its a really dumb but important question i dont think i would get an answer

i will try anyways

[twin dirge]

Hello, i dont knwo if this is the right chat but im posting it, feel free to delete if im in the wrong chat for such things. im getting into stable Diffusion and planing to work with Sadtalker, thin plate spine motion model,....

I found that my AMD GPU is not the right choice for it and i got told i should choose the INNO3D GeForce RTX 3060 TWIN X2. Thats like 340€ here in Germany and its my absolute max investment i can do.

What do you guys think?

is this the best choice for this Price?

Should i choose exacty this one or one of the many other 3060s?

I got told that i definitly should choose the 12gb version.

I could take some advise. Thank you

[spice plank]

If there is a cheaper 3060 12gb and you can't go Higher in price to a 3070/etc then yes. Brand is up to personal preference, never used inno3d and haven't really heard of them so I can't say.

[copper tangle]

I'm a 3d artist and I'm looking to use stable diffusion to create clothing variation for my characters, is there a recommended model to be able to do img2img and add clothes/armor to a bare human body?

I'd like to stay in the realistic realm so no anime models or anything, not sure where to start first

[hazy wind]

If the latest open source version of SD turns out to be much better than 1.5, will it mean that future models will also be much better than those created with previous versions? What would happen if all the dataset that was used to train a model was used in the new version of SD?

[polar pier]

Question RE: ControlNet: Should I update to the v1.1 models for canny, depth, and pose?

[maiden eagle]

anyone attending the nyc meet up on Friday?

[glad skiff]

Where can I find info on this?

[maiden eagle]

https://partiful.com/e/AqsGTfRFmbCgIlMMnbEk

[eternal geode]

The channel icon disappeared for a second

Or i didn't saw it

Weird

I didn't saw it most likely

[spice plank]

Pe?

[woeful sun]

._.

[steep sage]

Fuck openai

They were at congress trying to kill opensource ai

[balmy jungle]

we need an AI that scores pictures on originality. i am so sick of seeing batman in dramatic poses.

[balmy jungle]

EU parliament is drafting legislation to prohibit opensource / small company LLM's too.

kabbal doesn't want any tools in our hands...

[crude pulsar]

hey now

[potent spire]

@balmy jungle fake news

[potent spire]

Also fake news

[forest lily]

For a company called open ai they sure seem to hate things being open

[hexed cliff]

What happened to chat 2? Did it disappear? Is this a bug in my discord? Good morning.

[high falcon]

https://www.reddit.com/r/StableDiffusion/comments/13l0pf0/do_you_even_link_bro/

[wild steppe]

Not sure what you mean by chat 2? Good morning!

[rapid yarrow]

Is there a way to enable a safety filter for self hosted SD?

[high falcon]

Anyone have a Luigi (SMB) LoRA laying around?

[gray wolf]

How do I get automatic1111 to auto update?

[peak portal]

that's the model publisher's fault

i'm constantly converting and publishing good models from CivitAI to HF Hub

especially if they're 2.1 based

i do not really care about 1.5 anymore.

[mighty stone]

hello

[copper crystal]

new 4060ti with 16gb is announced for this summer. $500 msrp. save up for that

https://www.nvidia.com/en-us/geforce/news/geforce-rtx-4060-4060ti/

[ebon brook]

Since I don't have a GPU and need multiple subjects in one image, I think I need controlnet? Any recommendations for an online GUI for that?

[copper crystal]

i dont think many reliable services offer controlnet. mightneed something like runpod where you just rent the hardware and set the software up yourself

[unreal oriole]

bro it's been 24 hours

💀

[copper crystal]

attention spans! like, how even do they work??

[peak portal]

i publish on both concurrently

[ebon brook]

What is controlnet / pose library called on other platforms like midjourney?

[unreal oriole]

a paid feature

[potent spire]

They better reduce the price of RTX 4090 with that

[peak portal]

lol why

different audience

[plucky shoal]

but it will be tricky, nobody can say it will work on pci-e 3 which is majority o.k. pci-e 4 should be o.k. but 3 nobody knows

[peak portal]

if anything the price of the 4090 will go up if they reallocate manufacturing capacity to the new GPU and have fewer 4090s on hand

supply and command, Ricky

[potent spire]

4090 is at above 2000 depending on manufacturer

Sometimes 2.300€

[peak portal]

yes, i know, i own one

[potent spire]

Would be nice to catch one for under 1,5k but lets see

[peak portal]

the W7900 is coming out at 64G VRAM and competes with the 4090 and is $4k

nvidia is only competing with themselves, they have no need to reduce the price

[potent spire]

They already reduced them after a while

Radically even

[peak portal]

eventually the cost will go to $0 but you'll be waiting a while

[potent spire]

1 year or 2

But depends

[peak portal]

i see no reason to reduce it any further and reduce their margins, they keep prices high Because They Can

[potent spire]

If not, well then i will have to spend 6-7.000€ for a PC build

[peak portal]

when Intel ARC is not crap maybe

[ebon brook]

I found PoseMaker2 to be a useful search term as well. Especially combined with Dreambooth. So: PoseMaker2+textualInversion+controlNet+dreambooth

[ebon brook]

I think this might be able to do it https://huggingface.co/spaces/ArtGAN/Stable-Diffusion-ControlNet-WebUI

[twin coyote]

can i use stable diffusion images for youtube videos

[hexed cliff]

Why is it that every chat I enter there from the bot, one or two are missing, every time I enter a chat, the other one is missing, is it some bug in my Discord? That's what I was confused about.🥹

[wild steppe]

Possibly your discord cause that's not normal haha

Try CTRL + R to restart discord and see if that fixes it

[honest zinc]

hmm

[crude willow]

what would you have to do to an image to make it not detectable as ai generated?

just because im a rebel. no particular reason

I imagine that the detectors are mostly looking at patterns in color values, samplers leave traces and have a sort of "brush stroke" look. where upon numerical inspection of exact color values there are repetitions and seams.

[fervent path]

I’m looking for some help trying to find the right terminology. What is it called if I want to take a picture of a person and then I want AI to do things like.

• replace the background with another photo.
• replace a background with an AI drawing that adds a lightning storm.
• makes the persons muscles bigger
• adds little gold wings coming out of ankles, or adds fire bolts coming out of fists

[potent spire]

draw and paint over it and then yeet the AI image out of existence 😛

[crude willow]

I was thinking add a light filter

maybe remove meta data

have a second ai model redo part of the image

rescale it

[potent spire]

who are you trying to hide it from, only detectors...or humans too?

for the detectors, they are still in early development from what i saw so in the future they may be far more advanced and with humans, especially those who are informed they can notice the artifacts and other anomalies in the AI image with or with a added filter

[crude willow]

if people think the detectors will catch all ai stuff, then getting around it puts you in a position of power/advantage in terms of convincing people: "No, donald trump really did get arrested. see look its not ai made!"

*not that thats what I would use it for

Im not even american

[potent spire]

as of now this isnt possible or not known to be possible but it also comes down to the future who will develop what, what courts have to say etc. etc.

we can truly speculate as of now

[crude willow]

what isn't possible?

[potent spire]

to make it impossible for someone to detect an AI image as what it is

you can make it harder ofc

[crude willow]

oh that quite doable

[potent spire]

but undetectable is another thing

the biggest chance would be if someone is very advanced in art and editing skills

[crude willow]

yeah, raw images out of an ai are 100% catchable. But after 20 mins of filters, compressing and upscaling, small manual adjustments etc it would not come up as AI

[potent spire]

depending on detector and how they are going to be updated and developed in the near future, for humans a different thing

[crude willow]

hey can we switch to #🏞｜general-with-images

i have something I want to show you

[potent spire]

you would need to edit a whole lot of things in the image

[slender mulch]

Hi all . i have a strange issue when i do img 2 img single images it does it quite fast. when i do a batch most of the time it barely starts or stops at 2nd frame any clue why .

[vague path]

How can I make a Lora file for a character?

[fervent thunder]

hi guys, i'm doing a digital painting and would like to get help from the ai to polish the image and render some parts of the image, but i do not know which tools to use with an already existing artwork as i dont want it to take too much liberties, i've stabble diffusion downloaded on my pc, anyone would be down to help me ?

i've already used img2img and control net but i tends to not respect my image who is already very detailed and dense

hey guys I did edit webui-user and wrote: set COMMANDLINE_ARGS= --no-half --no-half-vae --disable-nan-check does this affect image quality?

[vague path]

If I have a picture and I want the same character in that picture to look exactly the same, but in a different pose, how would I do that?

[broken smelt]

Could try a reference controlnet combined with an openpose one. May need to play with the weights a little to get it to work though idk.

[glacial mauve]

Can I sell stablestudio?

Just noticed MIT license so I'm good to go! Thanks

[bleak geode]

Hey I'm looking for a text to image with no censorship and I was wondering if Stable diffusion censors some prompts or not

[broken smelt]

Nahh have you seen civitai lol id wager most of those models are for 1.5. very nsfw.

[naive snow]

https://www.nasdaq.com/articles/openai-chief-goes-before-us-congress-to-propose-licenses-for-building-ai
https://technomancers.ai/eu-ai-act-to-target-us-open-source-software/#more-561

sigh

[opal oasis]

ola

[potent spire]

The second link is bullsh*t tbh

[eternal violet]

Dumb question, is there plans to release sdxl model once it's finished tuning or will it remain available only on dreamstudio?

[ebon brook]

iT's a morphic resonance lAnDiNg pAd for sOlidstaTe aliEns!

[teal hare]

yo

[upper anchor]

will stable diffusion forever be free?

many a.i art generators are no longer free, they advertise as free a.i generator, and later transform to paid server

[sick ermine]

u can run stable diffusion on ur local machine for free. its been distributed to millions of ppl already. if the github ever goes down, ppl can just reupload. art generators are not free because they have to use their own gpu resources to produce the image

[upper anchor]

I see, I am new to stable diffusion because I just came here. I used to use another A.I generator, but it got paid, I won't name it because I think its not allowed.
I plan on making money from A.I art that I generate and sell them online.
When I start earning, I would be more than happy to pay as well

[upper anchor]

I am currently using stable diffusion web version

I am not aware of the version we use on our local machines

Also, where can I show off the images I create?
It's much more fun, when there are people who you can share with, what you can create

Also, I want to ask, https://stablediffusionweb.com/#demo

this version,
does this use our GPU power too?

and what's the benefit of using a local version, vs this website version: https://stablediffusionweb.com/#demo

I understand, I am asking so many questions, but I promise, I won't ask much, when I know about this eco system

[sage coyote]

any one here to help i have some issues in instalation

[plucky shoal]

#🤝｜tech-support @sage coyote is better place

[sage coyote]

thnx @plucky shoal

[hallow plaza]

Does anyone know what the best clip model is? So far I've had some good luck with RN5x64.

[upper anchor]

is SDXL a better stable diffusion model?

[copper crystal]

its beta in preview rn. it's got a bigger foundational architecture. more params

[wild steppe]

see for yourself #🏅｜pantheon

[upper anchor]

definitely fascinating...much better than images generated in blue willow and other a.i generators..

This is definitely a great tool! Will it be free forever? Or will it get paid when it's in beta?

I am trying to understand this while eco system.
I recently came to know about stable diffusion.
So far, according to my knowledge, there is 1 web version of stable diffusion, which doesnt need us to host stable diffusion on our local pc.
And there is another stable diffusion such as 1.4,1.5,2.1, which needs to be installed on our local PCs, and are also free to use.
Then there is dreamstudio I believe and also stable studio, and finally the last one, and probably the best one : SDXL

Am I following right so far guys?

[wild steppe]

SDXL will be open sourced when it's ready just like 1.5, 2.1, etc ^^

[graceful spade]

Hi

Anyone can help me out?

I’ve seen some tutorials on youtube but unable to get it properly, can you please 🙏 provide me a tutorial link so i can create some superhero jump or anything similar cool animation videos?

[upper anchor]

Absolutely amazing! Midjourney is going to have a tough competitor. That was the only thing holding stable diffusion back, it's prompting capabilities, but with the release of SDXL, I see that will no longer be the case.
That's really something

And stable diffusion much better customization abilities

[graceful spade]

Can you help

[upper anchor]

i am new to stable diffusion also

[graceful spade]

Okay

[fervent thunder]

hello, can i use stable diffusion in my phone

[upper anchor]

you can use the web version

[upper anchor]

https://stablediffusionweb.com/#demo

[twilit brook]

Hi all, pls what command can i use to create a custom size like height=300 and width=900?

[void falcon]

how do you make the ai picture

[spare plume]

Hi, what do you think about this : https://www.reddit.com/r/ArtistHate/comments/13l8s2p/the_first_court_rulling_against_using_ml_being/ ?

To me it sounds like straight up misinformation, this lawsuit was about stealing art and running it through a simple Instagram filter, and they're trying to make people believe it makes AI non fair-use.

[fresh iron]

how to add file to generate ai ? i can't

[outer ginkgo]

Hello! Do you know the website with the list of artist styles?

[tulip yoke]

what was the thing called people make loras with

[tulip yoke]

add file?????

[ember steppe]

hi guys i have a question

can i remove an object from an image and add another image in that place for example: a girl wearing a watch, remove that watch and add another one from an image not a prompt

[outer crescent]

https://www.youtube.com/watch?v=zmobGnOjnAE skim through this video and you might get sme ideas 🙂

you can also use controlnet and masking.

[ember steppe]

thanks

[tender gull]

Hey everyone 👋

I was wondering if you can generate NSFW images with any model? or is it prohibited?

[spare plume]

You absolutely can with most of them. You'll have to run a local version of SD though, as most online platforms have filters. Don't forget to read the licenses of each model to make sure you're not going against it, but most of them are fine with that as long as it's not meant for illegal, harmful or discriminatory content.

[tender gull]

If i use the API for stability model it should be possible?

Thats what im worried about because in terms of service for stable diffusion it says to not generate any images of pornographic nature ( bare breasts, buttocks etc. ).

[tender gull]

Maybe these terms are only meant for public bots? I will verify

[spare plume]

iirc, since SD2, there isn't any nsfw content, even in the training data, so this would not work.

If you want to check the license in details : https://github.com/CompVis/stable-diffusion/blob/main/LICENSE

Here are the restrictions :

Use Restrictions

You agree not to use the Model or Derivatives of the Model:
- In any way that violates any applicable national, federal, state, local or international law or regulation;
- For the purpose of exploiting, harming or attempting to exploit or harm minors in any way;
- To generate or disseminate verifiably false information and/or content with the purpose of harming others;
- To generate or disseminate personal identifiable information that can be used to harm an individual;
- To defame, disparage or otherwise harass others;
- For fully automated decision making that adversely impacts an individual’s legal rights or otherwise creates or modifies a binding, enforceable obligation;
- For any use intended to or which has the effect of discriminating against or harming individuals or groups based on online or offline social behavior or known or predicted personal or personality characteristics;
- To exploit any of the vulnerabilities of a specific group of persons based on their age, social, physical or mental characteristics, in order to materially distort the behavior of a person pertaining to that group in a manner that causes or is likely to cause that person or another person physical or psychological harm;
- For any use intended to or which has the effect of discriminating against individuals or groups based on legally protected characteristics or categories;
- To provide medical advice and medical results interpretation;
- To generate or disseminate information for the purpose to be used for administration of justice, law enforcement, immigration or asylum processes, such as predicting an individual will commit fraud/crime commitment (e.g. by text profiling, drawing causal relationships between assertions made in documents, indiscriminate and arbitrarily-targeted use).

[tender gull]

So which would be the best suited model to use for NSFW content, do you have a suggestion?

[spare plume]

It depends, photorealism or anime?

[ember steppe]

i didn't find the solution is there anything else please

[outer crescent]

you can use inpainting and put a mask over the watch, and then describe the watch in the prompt very detailed

[tender gull]

Photorealism preferably

[spare plume]

Then on CivitAI you have many models for this. ChillOutMix is great, Lofi is also good, UUPM is made for nsfw especially, but I've never tried it. You should try adding LoRAs, also. This way you'll get more coherent images and closer to the intended output.

[ember steppe]

but i wanna add a picture of my watch with the logo of the brand instead of writing a prompt is that possible ? it's like blending two images

[tender gull]

Wow thank you for the detailed answer i'll try around with this this Weekend:)

[spare plume]

np, have fun :)

[ember steppe]

is there any help guys ?

[plucky shoal]

#🤝｜tech-support

[quick crow]

Hi All, I am very new to this and came across this AI which generates random faces could anyone please explain how this works? Where can I do it? What else can it do?

[outer crescent]

Its very difficult to explain but it can create pretty much anything you can describe to it in enough detail.

try playing with this https://beta.dreamstudio.ai/generate

[fervent thunder]

whats the best way to create brutalist style poster art for the game I'm making?

[nocturne grail]

hey how do you make stable diffusion animations not zoom in, i set the zoom to 0 on a 2d animation but it still just slowly zooms in

[slender cave]

Morin

[outer crescent]

keep emphasizing the words, like brutalism, concrete, bleak etc. Read up on brutalist architecture and steal some cool words

of you want it to look slightly bleak, i find "decrepit" and "poverty" really good power words, they add a bit of lived in wear and tear to most images, portrait, scene or design.

[nocturne grail]

how bout my animation issue, how do i stop the zoom, so that i can have a normal gif type animation

[outer crescent]

it might be a bit woo, but i find obscure but not uncommon words work well, not sure how CLIP was trained but it seems to really nail terms if they sound like they came from the atlantic or some upper kinda source, possibly because they're less poppy and less cluttered. i dunno. i expect sdxl to be less like that

[nocturne grail]

: (

[outer crescent]

@nocturne grail i dont know anything about animation, maybe try sd-chat

[nocturne grail]

i suppose i shall, but if anyone else does know please inform me, it seems to just zoom regardless of the camera settings

[severe sonnet]

Hello
Recently, when I want to work with Dreambooth, in automatic 1111, when I click on train, it gives this message:
Exception training model: 'No executable batch size found, reached zero
I didn't have this problem before and it was running fine. Recently, this problem happened to me. Does anyone have a solution to fix this error?

[tender gull]

u know if i can run the webUI with an intel GPU?

[novel timber]

Its funny how unstable diffusion has 100k more members

[spare plume]

I know some people had issues with AMD GPUs, never heard of anyone using an intel GPU actually. Doesn't mean you can't, but it certainly means you'll run in more trouble along the way

[spare plume]

I don't even get the point of unstable diffusion

[tidal thorn]

Anyone know what "Cuda error : an illegal memory access was encountered" could mean when doing higher resolutions on a 4090

[forest trout]

If it's cuda related you probably are running out of VRAM.

[tidal thorn]

thats weird because I could do 2048x2048 on my 3080ti with 12gb sometimes

I just got the 4090 with 24gb and doing the same stuff

[forest trout]

I'd try rebooting your PC and trying again.

[tidal thorn]

oke thanks, will try that

[forest trout]

It might not fix the issue but maybe it will help free up some ram.

[tidal thorn]

edit: nvm

[nocturne grail]

hey i tried to make an animation 1080x1080 and it says it resized itself to 701x701 any particular reason or is that like the max

[split rain]

can i ask questions here ?

[fervent thunder]

you just did, but it depends on the question. Tech stuff is probably better in a tech sub channel for example :)

[split rain]

u got me.. i wonder if i can find info on models here , something i can use as a personal assistant , im not skilled or uptodate with ai but im hoping to find a decent model to run local

[fervent thunder]

what do you mean by "personal assistant?"

[split rain]

i should be able to have conversations and search on web bit like chatgpt but local on my pc

[fervent thunder]

this discord, from my knowledge, is mostly about making art with ai, and not anything similar to chatgtp

[tidal thorn]

Why is img2img trying to allocate 16gb for a 1500x800 generate on my 4090? on my 3080ti it worked fine with less vram

[split rain]

oo im sorry stupid me

[turbid whale]

chan has a local models general where you could probably find links to info you're looking for. I don't know if I can post links here, though.

[split rain]

slide them in DM ?

[tender gull]

alright ill just use my pc with Nvidia then 😛 seems like intel causing trouble with pytorch

[turbid whale]

Sure, give me a min.

[turbid whale]

Your DM's are closed.

[karmic vigil]

Super interesting stuff.
https://www.reddit.com/r/StableDiffusion/comments/y0t4pd/a_bizarre_experiment_with_negative_prompts/

[lethal nest]

hi guys i have a question about controlling camera movement, angles and rotation ... i have generated a realistic figure ... can anyone give me some suggestions ?

[turbid whale]

You can try detailing that into your prompt. For example: "front-facing view" or "head-to-waist view"

Wide-Angle Shot
Ultra-Wide Angle
Satellite View
Eye-Level Shot
Far-Shot Angle
Medium-Shot Angle
Ground-Shot Angle
Low-Angle Shot
Full-Shot Angle
Full-Body Shot
Glamour Shot
Cinematic

[naive snow]

So, is XL going to be as badly censored as 2.0?

[tawny galleon]

Hello everyone!, just started with this whole thing yesterday and man its overwhelming but its so fun to see stuff come to life as I change the prompts.
I have a question: If I wish to use a picture I have taken with a camera and want to use it as foundation to turn it in to something more stylized how do I go about doing that?

[turbid whale]

Drop it into img2img and develop your prompt with the details you would like to see.

That being the most basic way. Of course, there may be different models, embeddings and Loras that you wanna try, but that's up to your taste.

[viral tulip]

hello everyone, does anyone know the name of the neural network that is able to search for images by prompts? It is not to create, but to search from existing ones on the Internet.

[fervent thunder]

you described basic google image search

[balmy jungle]

idk if it's a network but there's a site that lets you

(i seem to be very popular in this channel, judging by mentions)

[broken smelt]

Mabey he's talking about booru sites?

[lethal nest]

I want to make an animation like I have an image of a realistic figure which I have generated with ai ... And slowly make the camera zoom in and change the camera angle like rotation around the character

[lethal nest]

Do you think it is wishful thinking or is it possible?

[viral tulip]

What site?

[lethal nest]

🥲

[viral tulip]

Nah, I heard about neural network, that can do that. So and decide to ask here

[balmy jungle]

LoftySky the name of the site refuses to remain in my brain, again.

[viral tulip]

Ahhh, okay... thanks for trying to help

[nocturne grail]

hey ive been generating different stuff and at some point it just started giving me errors, and now whenever i make a fresh copy of the deforum and hit connect it says cannot connect to backend, but sometimes does anyways, as well as the first issue that maybe caused it all, at some point i tried generating without changing the batch name and it then said that i was missing a batch output name, idk what to do

and even if it does say its connected if i try running the nvidia gpu part it says this ``---------------------------------------------------------------------------
FileNotFoundError Traceback (most recent call last)
<ipython-input-1-fc229b26f876> in <cell line: 3>()
1 #@markdown NVIDIA GPU
2 import subprocess
----> 3 sub_p_res = subprocess.run(['nvidia-smi', '--query-gpu=name,memory.total,memory.free', '--format=csv,noheader'], stdout=subprocess.PIPE).stdout.decode('utf-8')
4 print(sub_p_res)

2 frames
/usr/lib/python3.10/subprocess.py in _execute_child(self, args, executable, preexec_fn, close_fds, pass_fds, cwd, env, startupinfo, creationflags, shell, p2cread, p2cwrite, c2pread, c2pwrite, errread, errwrite, restore_signals, gid, gids, uid, umask, start_new_session)
1861 if errno_num != 0:
1862 err_msg = os.strerror(errno_num)
-> 1863 raise child_exception_type(errno_num, err_msg, err_filename)
1864 raise child_exception_type(err_msg)
1865

FileNotFoundError: [Errno 2] No such file or directory: 'nvidia-smi'``

nothing changed on my pc so i dont get why its saying no nvidia

[covert magnet]

hi there! 🙂

[nocturne grail]

hello

[covert magnet]

just a newbie loorking around, and tryind to get into the massive revolution of AI-powered apps

[nocturne grail]

ah, then dont do what i did, cause i broke mine, even tho i dont think i did anything

[merry gull]

How does everyone have good artworks? Everytime I Generate the art is bad quality and looks like paint drawn. Anyway I can Fix this?? Please help

[turbid whale]

Gotta provide us with more details than that. For example, what samplers, models, embeddings, extensions are you using? What's your workflow?

How do your prompts look?

[turbid whale]

Oh, well I haven't delved into animations/videos yet so I can't suggest anything it that regard, my apologies. Best bet browse #🎥｜animation for tips, check out the ebsynth and deforum post on reddit, and what tuts on youtube.

[sick ermine]

how long does it take to generate a simple 30 second deforum video? havent had time to try it out yet

[copper crystal]

thats just why its popular. stability as a company does all things machine learning

[fervent thunder]

do stability have a chatai? :O

[copper crystal]

#1098025024541167646 . it's good. not great.

[fervent thunder]

thanks for telling me. Hope someone corrected me from earlier so that my mistake didn't end up being the only answer :(

[copper crystal]

eh he'll find out what he needs if he's determined. theres a massive resource explosion for people . technological boom going on

[blazing moss]

anyone know why I cant paste a code in cmd when trying to copy my hugging face token?

[fervent thunder]

Do you have shortcuts enabled in cmd properties?

[blazing moss]

son of

a

ty

@fervent thunder

[copper crystal]

shift+insert is the traditional terminal paste command as opposed to ctrl+v

[blazing moss]

gotcha ty much

[copper crystal]

np! also, i think it's a new windows 10/11 thing, but right clicking the window pastes too

[true stream]

How do you actually get onto the website

I forgot the website

[fervent thunder]

What website

Dreamstudio?

[inner flint]

Hey all, so I stepped away from SD for like a week, updated to ControlNet v1.1 the other day and there's like, 20+ new models. Are there any resources on this Discord that fives a little run down of the new ones, I did a little searching on the github but mostly found old info. Just wanted to ask before I download and go through them all myself

[topaz dirge]

hi there!
If you don't mind, I'd like to hear what you guys think. Help us out.
We are developing a system that allows you to rent or lease Stable Diffusion on your personal PC to others.
Like renting a room on Airbnb, here you share GPU power and AI models with others.

https://forms.gle/7mQotdDXi7NYFcZi9
If you would like to help, please fill out the development survey.
It takes about 1-2 minutes to complete the survey.

[proud nova]

Maximum resolution I am able to generate with RTX 3060Ti 8GB?

[cursive veldt]

Will this version be available on dreamstudio.ai?

[gleaming topaz]

GPTE is a really good AI resource. You can search over 2,000 tools https://gpte.ai

[bleak matrix]

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

[tulip yoke]

does fp16 and fp32 make any difference in finger accuracy

[empty rivet]

Hello everyone I need to know a thing, I have create a model with dreambooth, the model working but now I want to add a style in stable diffusion checkpoint on this model. But how can I doing that ?

[copper crystal]

there's a new github page for the 1.1 models. explains them all

https://github.com/lllyasviel/ControlNet-v1-1-nightly

[grim sedge]

https://youtu.be/FtbHRuIP9_k Exploring generative models in video Art with artificial intelligence, If you follow me you followed me.😜🥘🌞

[empty rivet]

I think I've found, its with "Show/hide extras network" isn't it ?

[fallen hinge]

hey guys I have a problem I just installed stable diffusion and have no background in coding. I have an acer nitro AN515-57 with rtx 3050 laptop gpt and i7-11800H which has 4 gb ram. I am getting the cuda out of memory error. How do I fix this?

[spare plume]

I still don't get how Stable Diffusion has less members than unstable Diffusion. What's even the point of unstable now? iirc it was about producing NSFW, but SD models already do it

[pallid flame]

whats reference_adain mean in contolnet

[marsh river]

Are nsfw art prompts still allowed?

[neon sedge]

Hello, how do I create an animation like that TikTok trend around here? Is it possible?

?

[karmic brook]

check out #🎥｜animation lots of stuff in here

[knotty turtle]

Do you guys happen to know what happens in A1111 when you set the VAE to "automatic"?

I tried setting the VAE to "automatic" and all my images are completely washed out. Wondering if "automatic" is the same as "none"?

[rough canopy]

Bang bang! 🤍

[frosty mist]

hey dose anyone one know how to train my own custom model for stable diffusion

i am trying to create a stabletnsor file not a embedding

[median birch]

Guys, the bots stopped working?

[karmic brook]

yes, #1047610792226340935

[frosty mist]

?

[median birch]

Why ?

[grizzled spire]

Hi. What is this channel about? Found It when I was browsing channels.

Is this 3d Modelling tool, Game development or what?

Found a website about this program. Looks really interesting. Got to check this one out what I can do.

[inner flint]

Exactly what I was looking for, thanks!

[tidal thorn]

Hey guys, for inpainting can i import my own image mask black and white? I cannot find where to put it

[spice shadow]

Fun showcase of creating a video using stable diffusion. I'm sure the models used for video creation are probably old by now but stuck with my workflow until it was finished 😉 . https://www.youtube.com/watch?v=4a8-6tcSU1I&ab_channel=VisualConcepts

[copper crystal]

in automatic1111 webui, it's the "inpaint upload" tab

[tidal thorn]

Thank you so much

[broken vault]

help guys i can't open the webui-user.bat file

[steep star]

Hey, I am getting some errors while setting up minigpt, can you help me?

[cunning saddle]

Hey everybody. I just have a question for y'all...
Is it possible to "complete" a LoRa model After its creation ?
I created a LoRa, but I want to add more pictures to it now...is there a way? If you Say yes, I will search for a method....ThanK YoU in advance 🙏🙏🙏

[warm bane]

In which channel can I talk about models ?

[onyx sundial]

is there a tool that can give me a word cloud of a safetensor model?
I want to know what words have the most samples associated with them
I assume there would be more or less samples depending on the word used
I.e "office" samples vs "backroom" samples
Kinda new to this, sorry if im not using the correct terms

[civic forge]

Anyone good with portraits interested in doing a paid commission? I've got 8 portraits I'd like made up as proof of concept. Real people faces on superhero bodies of my own design.
Please hit me up!

[fervent thunder]

4080 or 7900 xtx?

[alpine bear]

4080

[hollow beacon]

wondering if there's an image classifier out there than can detect aesthetic or deformed faces?

[oak depot]

Did they delete the entire "notebooks" channel?

[rugged cobalt]

Bot still down?

[rare fjord]

im looking for hints how to prompt images for ancient civilizations cities visualization during ther prime times, something like time travel. Im getting ruins all the time no metter what model or prompt Im trying.

Ancient Greece a city from the time of splendor and power

ruins

maybe i should use some img2img frames/images from games or films?

[solemn marsh]

anyone else using kandinsky 2.1?

[hearty grail]

same, did u fix it?

[broken vault]

can't

[plucky shoal]

edit or run it?

[wide acorn]

When trying to vote shows me an error message.
Does anybody knows, why this happens?

[fervent thunder]

Shhhhhhh

[analog mango]

NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.
Time taken: 0.34sTorch active/reserved: 2393/2620 MiB, Sys VRAM: 4657/12288 MiB (37.9%)

how can i fix that

can someone help me pls

[muted rain]

Hey Guys

When can we expect SD3 ? Is it Delayed ?

[finite cloak]

it's called SDXL and you can see work on it in the SDXL Beta channels, or on dreamstudio, or on clipdrop

[stray nymph]

Some models have BakedVAE's so they then recommend setting VAE to auto

[muted rain]

How to use SDXL ?

[finite cloak]

"automatic" and "none" are the same for VAEs usually, the difference is if there's a .vae.pt file, which was the legacy method of overriding vaes

[finite cloak]

can use DreamStudio or ClipDrop. We have a bot in this discord but it's currently down

[upper anchor]

once it is opensourced...free forever?

[finite cloak]

yeh

[tidal lichen]

Sorry for asking this 😛 I dunno where else to go... is there an open source LLM that runs on GPU? I've found articles with tons of LLMs listed, but they don't specify whether they run on GPU or CPU.

[solemn marsh]

what is the stable diffusion bot here sourced from?

[alpine bear]

Yessir. Lookup oobabooga.

Its a webui that hosts llms that you download

Gpt4-alpaca is an unrestricted llm that was finetuned using gpt4 responses, originally based on the leaked llama model

There's also Vicuna, probably the highest-end open source model, but restricted.

[muted rain]

So clipdrop is free and dreamstudio chargable ?

Also can we download something to our Mac ? or is it not possible

[rugged mirage]

didn't this discord have a gpu channel?

[tidal lichen]

Thanks a lot! I'm less interested in the UI (I have a repo that I'm working with) but need a model I can download that uses GPU. But if it helps me find a good one, I'll try it 😅

[finite cloak]

GPTQ label is usually attached to the preferrable GPU-runnable models

(GGML can mix cpu/gpu, -HF is GPU but runs in 16/8bit mode which is VRAM hungry)

[tough fog]

hi, theres some gallery ?

[torpid zephyr]

Is the Stable Diffusion v1.5 going away once they're fully on the new DreamStudio website? I didn't see it as an option from the model listing on the new GUI.

While I see a quality improvement on the newer models, 1.5 is still the most flexible model they have, IMHO. Sometimes you don't want the prompt fulfilled too perfectly. It's that imperfection that makes for the best images sometimes.

[sturdy patrol]

In every channel I get: You do not have permission,etc, etc. I only joined via an invite a few days ago.

[neon kelp]

Fun fact: The RTX 3060 Ti is 63600000 times faster than the IBM 7094, which was one of the most powerful mainframes of it's time, and was the first computer to sing a song

ControlNet seems to work on any SD model

Any SD 1.4 or 1.5 model I should say

[glacial mauve]

dang it bro, missing the bots so badly

https://tenor.com/view/lazy-stressed-sigh-gif-25786527

[glacial mauve]

"...It's that imperfection that makes for the best images sometimes...", I mean, I understand you wanna be philosophical at a point, I do see that sometimes you'd want that little creative side from the AI to take over, but IMHO.... The only reason why StabilityAI keeps coming up with newer models is to actually improve on them, and to actually make them better than the predecessors, essentially moving towards making a model near perfect or better than the competitors because... ofc, the AI Master Race, amazing features, achieving photographic realism, etc., etc., while also contributing to the Open Source ecosystem. Just my 2₵

[glacial mauve]

welp... bots still down... back to DeeplFloydIF lolz

[sinful elk]

Does SD work with AMD cards, thinking about getting the 6800xt card.
But if it wont work with Sd im out 😅

[glacial mauve]

#1108055793674227782 message

[north nimbus]

I scanned my extensions folder with malwarebytes and it found no threats. Am I safe?

[upper anchor]

where can I try out SDXL, since the bot is not working here?

[glacial mauve]

Either in, DreamStudio, NightCafe, or ClipDrop

[upper anchor]

Tried in dreamstudio, ran out of credits

[tidal thorn]

Hi, how could I have img2img work within a mask? When I upload a image with a black background, it also fills that in, I would want it to use only whats within a mask. I do upload a image to controlnet.

[left solstice]

question is the stable diffusion bot usable like midjourneys bot?

[torpid zephyr]

I'm definitely not opposed to improvement, but I'm more than just being philosophical. I think sometimes the best art created from DreamStudio is the result of where the AI failed to understand to an extent. The failure to understand often leads to creative interpretations of the prompt, which lead to fascinating results.

It's similar to the whole auto-tune debate in music. Auto-tune gets you closer to being fully in key, but the appeal that some vocalists have is when they may drift a little bit away from the 12 tone equal temperament. You don't want to be completely out of tune, but some drift can be good.

I guess my take on it is come up with better models, but leave the old models as options at least.

[glacial mauve]

I second this also... 👌

[slate pebble]

When I tried to use my 1050ti graphics card to train the model, it crashed.

I shouldn't have treated my old companion that way.

[left solstice]

?

[glacial mauve]

I second this also... 👌

[spice plank]

Quick question, anyone with a 1660ti, can you guys use waifu diffusion? I'm getting out of vram errors

[oblique mauve]

Guys? I need someone to do one for me. Is this okay or is it against the rules?

[gloomy hedge]

i have no idea what im doing]:

[fervent thunder]

those are my daily thoughts over the last 20 years

[north nimbus]

What are good settings for running EveryDream with 12GB of RAM?

[tulip yoke]

Twisted my foot while walking

[young dust]

w

[hallow plaza]

Does anyone ever share their data sets? I'm looking to boost my regularization images with high quality faces.

[fervent thunder]

we don't have bot.

I mean, technically we do have it, but it's mostly up for testing and not for a long time.
If you want to generate something - you use services, like Dreamstudio or install a1111 or invokeAI or comfyUI or any other SD webUI on your PC (or on someone else's PC)

[left solstice]

tried installin it like a month ago but i got intel.

[fervent thunder]

There are always other options

[left solstice]

like what webversion. if so then that version is ass

[harsh palm]

Hey all! I’d like to play around with Stable Diffusion a bit and I’m in the market for a new laptop (lucky coincidence). I’m exploring options, and one option is a second-hand MacBook Pro 16”, M1 Pro, 10 CPU cores, 16 GPU cores, 16GB RAM and 512GB disk. I have no ideas what the “comfortable threshold” is for hardware, so hoping to get some insights here. Would this suffice? Sort of? Not at all? Thanks.

[unkempt leaf]

Hey 🙂 ... Is there a method to use a premade Image mask, så SD/CN only affect areas not masked (Visa versa ?)

[steel bay]

Hello guys. Do you have mobile app for iOS?

[warm junco]

SD works on Apple Devices with M1 chip. But it would be better on a Laptop with a dedicated nvidia gpu with 8gb vram or more.

[warm junco]

Hey there is an Community made App for IOS.
https://apps.apple.com/us/app/draw-things-ai-generation/id6444050820

It generated images localy on your device.

[steel bay]

Thanks 😊

[next haven]

Launching Web UI with arguments: --theme dark --xformers --medvram --opt-split-attention-v1 --autolaunch
Traceback (most recent call last):
File "D:\SDWebUI\launch.py", line 356, in <module>
start()
File "D:\SDWebUI\launch.py", line 347, in start
import webui
File "D:\SDWebUI\webui.py", line 31, in <module>
from modules import extra_networks, ui_extra_networks_checkpoints
File "D:\SDWebUI\modules\ui_extra_networks_checkpoints.py", line 5, in <module>
from modules import shared, ui_extra_networks, sd_models
File "D:\SDWebUI\modules\shared.py", line 580, in <module>
opts.load(config_filename)
File "D:\SDWebUI\modules\shared.py", line 517, in load
self.data = json.load(file)
File "C:\Users\Bujm\AppData\Local\Programs\Python\Python310\lib\json_init_.py", line 293, in load
return loads(fp.read(),
File "C:\Users\Bujm\AppData\Local\Programs\Python\Python310\lib\json_init_.py", line 346, in loads
return _default_decoder.decode(s)
File "C:\Users\Bujm\AppData\Local\Programs\Python\Python310\lib\json\decoder.py", line 340, in decode
raise JSONDecodeError("Extra data", s, end)
json.decoder.JSONDecodeError: Extra data: line 235 column 2 (char 9700)

Tell me why the program does not start?

[sacred onyx]

Anyone knows good models for generating portals?

Like rick and morty styled ones

[warm junco]

Pls post that in #🤝｜tech-support with your gpu and Python Version

[sacred onyx]

I want to make intro for my videos with SD as well as outro

[graceful spade]

hi

[wicked tinsel]

hi

[fervent thunder]

is highres fix usely much slower? its taking over 10X longer per iteration then normal generation.

[timid bloom]

check if upscale is set at 2 or more

[fervent thunder]

it is set to 2

would that explain it?

[timid bloom]

yep.

set it to 1.

[fervent thunder]

also it ran out of memeory and crashed on the last step 😭

[timid bloom]

you will need more than 8gb for 2 and more

[fervent thunder]

i only got 6

[timid bloom]

well depends on initial res but usually more

[warm junco]

Hey you need to make sure you have --xformers --medvram in your webui-user.bat

Then you can upscale by 2 if you set the hires steps to 10

[fervent thunder]

good to know, i do have those

i was trying to see if i coul;d speed it up by removing --no-half, but it actualy made it 3x slower to remove it

welp it worked

[fervent thunder]

Hey guys! Sorry to bother you but I'm conducting a survey about AI for one of my projects and I was wondering if you guys would mind answering it for me (considering this is a server for AI art, I figured you guys would have a unique and valuable perspective on the topic). It's really short and shouldn't take more that 3-4 mins. All information is completely anonymous and won't be shared to anyone else.
Thank you
https://forms.gle/hNii83T9upQkPA2N8

[scenic crest]

Done

[bleak matrix]

Good morning, everyone

How are we all today?

[ebon gazelle]

Evening to all. Just recently joind the club

Learning the maths there about the models...

Seems a bit a lot to take in at first

Lots of math stuff, And I don't like math :))

[bold sparrow]

Hi everyone

[ebon gazelle]

hey.

[potent cloak]

guys i used stablediffusionweb.com and it says getty images on my image, is the website fake

[alpine bear]

100%

[potent cloak]

damn

everything looked so cool though

is there somewhere i can post it because it still looks ai-ish

i just want someones opinion on how fake it is

[alpine bear]

#🏞｜general-with-images

[broken frigate]

Hey, planning to build a new PC, need to know what kind of component I need to focus on if i want to run SD better

Is the most important thing graphic card? Like should I prioritize getting a 40 series, or 3070 is good enough??

[fervent thunder]

depends on how badly you want each image created in time vs cost
is +1 second too much? what about +1.5 seconds? etc

And do you want to pay top dollars for "just" being able to make faster ai images?

[broken frigate]

well currently 1 image takes more than 2 minute

planning to have thing more managable, like 3 image less than 1 minute is good enough, or maybe 1 image but almost instanteus

also I assume having a good PC for ai generation mean that my PC is good enough for moderb 1440p gaming

[fervent thunder]

yes, but a 3070 or a 40xx won't have those number differences. Or I think anyway :P

(between them)

[broken frigate]

oh I see, so u mean like, after a certain point GPU don't really have an impact

[umbral ether]

anybody here want to teach me how to make good promts

[sly pecan]

Opa

[fervent thunder]

more like, if you isntead had a 3070 and create an image. Would you regret not having the 40xx instead? meaning that if it takes 2 seconds to make a image using that 3070, will you regret not having it 0.5 seconds faster if you instead got the 40xx for (I assume) more money?

either way, it will be a lot faster than what you currently have? :D

[broken frigate]

yes this is what I'm currently at, so changing a new graphic card whether its 30 or 40 do changed a lot. also, I don't play in 4k, so assume most games would work fine in 60fps

going from 2 minute + to less than 30 sec is already fast!

btw, its 3060 not 3070 , dont know if it downgrade much

[fervent thunder]

I play using my 3090 with v-synch at 60 frames because a lot of games starts to "screen-tear" because of a billion different reason over different games.

So I just gave up and enabled v-synch :P

[fervent thunder]

nah, it only degrades my memory, or rather, the none memory of my memory :P

[broken frigate]

processor doesn't really matter right? wheter its 15, or ryzen 7

[fervent thunder]

the only issue you should think about is overheating the card.

everything matters, but the vram matters so much more than anything else in my opinion. Then reality hits me with semantics :P

[broken frigate]

yes that is also an issue. but with my current rig, playing games is hotter than generating art

I assume, like generating a few image would be less taxing than playing games yes?

like, if u play say 9 hours a day nonstop, its less taxing than 1 -3 hour of generating image

[fervent thunder]

no, generating images is so, soooo much more demanding for the card. Games can have more or less happening at times while generating art always maxes out

[broken frigate]

i see i see, how do people manage that then usually?

like only doing it a few times an hour at atimes?

say 1 project need like 500 run, do u split that in 1 day

[fervent thunder]

I use MSI afterburner and HWMonitor to check the temps.
I don't trust the ai art generating to let it just "do its thing for over 2 minutes without supervision" but I'm probably more worried than the average user. Hopefully :P

hot spots shouldn't really be reaching over 90C on most cards, and it's always different because each computer is set up different. Then you need to think about the air in the room, location in general, and all other boring stuff

But that comes AFTER you have it all set, so it's not really something you should think/worry about yet :)

just get the hardware first

[broken frigate]

I see I see ok ok

[quiet bramble]

#1010934719455707218 guys any one knows how to extract trigger words from a model?

[restive prawn]

hey guys

I'm reading an article and it says you can use more efficient frameworks for SD than PyTorch

is this so?

what do you guys use?

[hollow pebble]

Google colab isn't running SD anymore?

[ember folio]

nope, at least not on UI

Anyone have any killer prompts that I can use to showcase my model on civitai?

[main salmon]

does sd have an api ? if not how can i integrate it in my website

does anyone know?

[fervent thunder]

it is someone pretending to be sd.
They have ai here, looks like 1.4 or 1.5 base or poorly trained model

[tidal thorn]

Howcome controlnet doesn't get applied when dragging image into prompt?

[undone salmon]

where can i ask questions about installing SD?

i'm having errors

[hollow topaz]

Hi what are the strongest prompt-based image editing papers today?
So there is of course stable diffusion, text2live, instructpix2pix, imagic. what else?

[noble beacon]

Guys where I can generate pics?

Which channel is

[sacred onyx]

For me it works

On the Colab

If you refer to the cuda error

[noble beacon]

Trash server

[sacred onyx]

No you

[alpine bear]

I blame Jae for deleting miku role

[west echo]

I've done multiple "testing" sessions on my 1080TI and it can run for hours maxing out at about 36c in an ambient of 20c with a max hotspot of around 44/45c.

Then again it is water cooled 🙂

[young ivy]

I'm more interested in this type of cooling https://www.icegiantcooling.com/products/prosiphon-elite, I want to see it for GPU

[fervent thunder]

sounds like black magic to me as my temps STARTS at almost 45C before I even started the AI :P

[west echo]

better not show you my idle temps then ;o) (oh I cant in here lol, 24c @ idle)

[main sluice]

hey guys

i have a problem with stable diffusion ui

i keep getting a gray or white screen after render on deform

i have tried all things i found a didnt got a solution 😦

[glacial mauve]

Crap... I think the era for the SDXL bot is long gone now.. Few days have passed already and no single announcement on the incident, we just get that's an issue with discord itself... That StabilityAi can't do anything... 🤷‍♂️

I bought credits on DreamStudio to keep generating there but, I noticed the bot version was a better version or something, Cuz I try the same prompts and they look extremely different (uglier)

[fervent thunder]

I've defaulted to "legal issues" whenever a company doesn't say anything, or tries to ignore it like it doesn't exist. Most lawyers recommend people not to talk about those things at all because it might be used against them. :P

[west echo]

a quick "silent movie" showing idle & load GPU & CPU temps
https://www.youtube.com/watch?v=k_lE4yM_DFU
(sorry for the delay, waiting on YT to process something better than 360p!!)

[fervent thunder]

I'm using a RTX 3090 and I run it for around 2 minutes before letting it cool down for maybe 10 seconds, and it reaches around 87C on hot spots when first using it, then it slowly rise toward 89C with each new "generate" click

[copper crystal]

i live in the pacific north west. must be something about the air, cause i've only ever had a gpu over heat ona summer day while gaming for an hour.

[west echo]

which part of the world are you in? (and those temps are nuts lol)

[fervent thunder]

south sweden, but yeah, I know the issue. It's both a design of the card thing as well as my computer case

[west echo]

have you considered water cooling ?

Found a French company that does custom waterblocks fro any GPU. If the 4090Ti 16GB lands at a good price I may get one and then use these guys to add a waterblock (even if it doesnt "need" it)

https://hybridcooling-modding.fr/gpu/58-waterblock-custom.html

[fervent thunder]

I heard from people that water cooling doesn't really do that much in the end.

[west echo]

my 1080TI would beg to differ lol although to parapharse an answer I use a lot these days.

YMMV, this is the way

[fervent thunder]

yeah, there's probably people on both sides on this. There's so much into this and I tried to find the best help for it but I learned that my temperatures for the card and other components are fine.

Can't really compare one card to another, but that's probably the obvious part :P

[west echo]

TBH apart form the visual aesthetic my primary use case for making my rig water cooled was keeping it quiet.

My normal day to day settings I really cant hear it at all, the nosiest part is probably the water pump itself and thats mainly because its sat around 65-70 cm from my right ear lol

[fervent thunder]

yeah, that's my issue with all 11 fans in my case. It sounds like a helicopter lifting off. And sometimes it also looks like one as well ;P

[potent cloak]

does anyone know which upscaler should i use for hires. fix if i want to make desktop backgrounds? im doing 512x512 upscaled to 1920x1080

im using the stable diffusion web ui thing on windows 10

I must have a bad gpu because i think it takes almost 2 minutes to upscale to full hd from 512x512

[warm junco]

Whats your gpu ?

[potent cloak]

3060

[warm junco]

Thats good

[potent cloak]

512x512 doesn't take long but 1920x1080 takes over a minute

[warm junco]

Wait the 6gb or the 12gb Version ?

[potent cloak]

12gb

[warm junco]

Depends on the hires steps, you should use 10-15

[potent cloak]

it's 0

[warm junco]

Then it uses the same amount as normal steps

Thats mostly bad xD

You would wait to long

[potent cloak]

idk why it takes so long then

When I try to upscale a 512x512 it seems to zoom in which is annoying

[warm junco]

What are your normal steps ?

[potent cloak]

sampling?

its 20

[warm junco]

Okay yea try hires steps 10-15

[potent cloak]

this is my first time using this btw

[warm junco]

And make sure you use --xformers

[potent cloak]

but uh how do i fix the upscale zooming in?

[potent cloak]

no idea what this is

[warm junco]

You Shouldnt upscale 512x512 to Full HD. You Should use
540x960 and then select Upscale by 2
Then it wouldnt zoom in

[potent cloak]

Is Euler a the best sampling method?

[potent cloak]

oh okay

[warm junco]

Yea then you will get a FullHD too but not streched or cut off

[potent cloak]

is high res steps being 0 making the quality bad?

[warm junco]

There is no best method. Euler a, dpm++ 2m Karras and SDE are the most used

[potent cloak]

my 512x512 looks so much better than upscaled...

[warm junco]

Like i said it would use the same amount like the normal steps so just set hires steps to 10 or 15

The normal steps to 20-30

[warm junco]

Set the denois to 0.5 and choose ESRGAN 4x+ as upscaler

[potent cloak]

is 0.5 better than 0.7?

[warm junco]

Yes for esrgan based upscaler dont go over 0.5
The higher you go the more will be changed

[potent cloak]

also, im still not sure what --xformers does, I just put it at the end of the prompt?

[warm junco]

No --xformers goes into the webui-user.bat behind
Commandline_ARGS=

after a restart it will be installed

Its for better vram usage and a little more speed

[potent cloak]

idk what you mean by "behind", i am new to this

[warm junco]

After the =

[potent cloak]

set COMMANDLINE_ARGS=--xformers

[potent cloak]

like this?

[warm junco]

You can also add --autolaunch there so the webui opens in Browser after launching the bat

[thick otter]

Uh guys, do you need both the yaml and the pth files for controlnet directly. I seem to have only copied over the pth files

To the requisite directory

They seem to work but would I have needed to copy the yaml files over ?

[feral stone]

Hello, does anyone know when the server will be back up and running?

[deep stone]

Hey everyone, is there a forum or place where I can ask questions pertaining to one topic of my interest with Stable Diffusion use?

I mean I can ask here, but its a free for all, while I'm looking for to make a topic for specific discussion to a subject I want to learn about

[shy arch]

hey I'm new , i tried to download the automatic1111 but it doesn't work for me, I think because I don't have a gpu. Is that what you guys use or is there other ways to use stable diffusion?

[hallow plaza]

Is there an alternative to this project? It seems to be broken / abandoned. https://github.com/Tps-F/sd-webui-blip2

[west echo]

oh dear boohoo lol

a whole 2 minutes , my that's practically an eternity :p

Seriously though , why are so many people hung up on instant gratification? Are you the same people who look at the instructions on a Pot Noodle and say "nah 3 minutes in the microwave is too long, it should be quicker?"

Is patience a dying commodity? (asking as a community member likely at the upper end of the bell curve for the demographic here)

/Me wanders back to his < £200 1080TI 11Gb running 24c at idle and 36c under load to check on his 25 step 768x432 hi res fixed at 2.5 using the 4x ultrasharp upscaler to 1920x1080

Ping she's done

*(SFW),(RAW Photo),

A product photoshoot of a Pot Noodle container

(soothing tones, insane details, intricate details, hyper detailed,photorealistic, 8k, ultra realistic, volumetric lighting,(film grain:1.4)),
(Taken using a 35mm Hasselblad 500C/M camera using Lomography colour 400 film at f/1.8)
Negative prompt: Yes there was one but I took it out to keep under message limit

Time taken: 2m 37.94sTorch active/reserved: 6984/10228 MiB, Sys VRAM: 11264/11264 MiB (100.0%)*

[potent cloak]

i ain't reading allat

[west echo]

Case Proven m'lud.

People have no patience these days lol

[potent cloak]

I've got limited free time these days

[west echo]

#multitask

[slate pebble]

It said "couldn't find lidpython39.dll" when i click webui-user.bat . I have already reinstalled python about 10 times. But. I can launch Stable Diffusion with "Launch.py"

[kind kraken]

Is there an Model/ AI tool where if i give it a content, then it can make some sort of diagram / Fun sketches / Workflow from it? Doesn't have to be accurate..

[scenic moth]

Huh

[bleak matrix]

You can find more information about prompting in #📝｜prompting-help You can use img2img as well for this, etc.

[tulip yoke]

Rip bozo

[tulip yoke]

Lora

[clever patio]

Are these bots for video

[tulip yoke]

No

[warm junco]

You need to delete the venv folder. And make sure you have a Python 3.10 Version installed like 3.10.6-3.10.11

[tulip yoke]

Why does the python version matter tho

Like

Why not 3.11

[warm junco]

Because some features rely on the 3.10 Version.
Until auto1111 mentiones a newer supported Version we Should stay to what works reliable

[tidal lichen]

Thanks again for this! Finally got it set up today. Downloading models like a mofo.

I also like how I can tell it how much VRAM to use ... but how does it even work, if a model is 40 GB? 🤔

[fervent thunder]

https://stable-diffusion-art.com/samplers/

Choose one depending on your needs

I'm mostly ++ SDE Karras guy, DDIM or Euler a for quick tests

[west echo]

depends on the prompt and the model and the look you're after.

Consider looking at using XYZ scripting so you can generate tables of images from the same prompt with different settings (such as sampler, steps etc) and pick the combination you like.

eg this is a relatively simple one comparing 2 Samplers against multiple models

#🏞｜general-with-images message

[tulip yoke]

Anime is mid on unipc

[alpine bear]

If a model is 40gb you need 40gb of vram. AKA an a100. I recommend using the quantized models that are just a few GB.

Unless of course you're a chad with an a100 or several. In which case go ham

[tidal lichen]

The UI has this lovely slider that lets me specify how much VRAM to use. I have no idea how it works but somehow I can run prompts...

[ebon gazelle]

I am wondering what do you think about this software:
"Drag your Gain"

I see it as a direct continuation in the logic of working with the generator

[real zodiac]

It's "Drag Your GAN" and there's no code released so it's hard to think anything other than "oh cool screenshots".

[ebon gazelle]

true there

Then again one apon a time thede was no Stable diffusion....

[real zodiac]

Actually I found this "unofficial implementation" https://github.com/Zeqiang-Lai/DragGAN but I can't get it to run locally or in colab

judging by the issues it's not working very well though

[wide flicker]

Hey @hidden dagger may I know how to report security vulnerability to stability?

[ebon gazelle]

we will wait for it

[hidden dagger]

i've checked in with our CISO to ask where you should report, thanks for your patience

[balmy slate]

How long is the bot generally down for

[waxen pasture]

They took it away a loong time ago. They aren't bringing it back anytime soon probably.

[balmy slate]

Well that’s disappointing

[surreal perch]

@balmy slate

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

Currently, there is a public bot on the server that generates images available as a research beta for SDXL, you can find the current status of the bot in #1047610792226340935. There are plenty of ways to use Stable Diffusion such as the official https://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!

[waxen pasture]

oh, guess the sdxl bot is out? I thought that was only for scpecific people

[balmy slate]

So I can’t use it through Discord, but I can on the website,?

[waxen pasture]

Nevermind that is down anyway too

[waxen pasture]

I guess so

[balmy slate]

Thanks for your help man

[hidden dagger]

please send it to [e-mail], thanks again

[fervent thunder]

Most people install it locally

[stable pawn]

I been out of the loop for a month or so, I started my Stable Diffusion and Automatic1111 WebUI and get this message, "WARNING[XFORMERS]: xFormers can't load C++/CUDA extensions. xFormers was built for:
PyTorch 1.13.1+cu117 with CUDA 1107 (you have 2.0.1+cu118)
Python 3.10.10 (you have 3.10.11)".
How important is it to have just the right version, would I gain anything on stepping bay a version of Python?

Awww, I did try to generate an image and did get this long message; "NotImplementedError: No operator found for memory_efficient_attention_forward with inputs: query : shape=(2, 4096, 8, 40) (torch.float16) key : shape=(2, 4096, 8, 40) (torch.float16) value : shape=(2, 4096, 8, 40) (torch.float16) attn_bias : <class 'NoneType'> p : 0.0 cutlassF is not supported because: xFormers wasn't build with CUDA support flshattF is not supported because: xFormers wasn't build with CUDA support tritonflashattF is not supported because: xFormers wasn't build with CUDA support triton is not available requires A100 GPU smallkF is not supported because: xFormers wasn't build with CUDA support dtype=torch.float16 (supported: {torch.float32}) max(query.shape[-1] != value.shape[-1]) > 32 unsupported embed per head: 40
Time taken: 1.30sTorch active/reserved: 2125/2148 MiB, Sys VRAM: 4490/24576 MiB (18.27%)".

So it looks like I have to re install a bit more than just a older Python to get it to work again. But any tips is welcome if other have same trouble.

[sinful elk]

Guys will auto1111 work on windows using an AMD card, im thinking of getting the 6800xt...?

Have an rtx2060 super now but want to upgrade

[stable pawn]

Well, I manage to really destroy it now, I get this error "Couldn't launch python

exit code: 9009

stderr:
'python' is not recognized as an internal or external command,
operable program or batch file.

Launch unsuccessful. Exiting."

Do some have a good link to a Installation guide to newest Automatic1111 and Stable Diffusion, it look like I have to unstall tons of things and start over.

[alpine bear]

@wild steppe pin it^

[karmic brook]

man cmon it's Tuesday lol

[alpine bear]

[karmic brook]

[fervent thunder]

https://www.reddit.com/r/pcmasterrace/comments/13p8rv5/budget_build/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button

@peak portal this is your PC

[peak portal]

-_-

Generating class images: 1%| | 394/35085

😭

[sinful elk]

Can anyone answer this please, really need to know b4 i buy a new card, if auto1111 wont work with AMD its not an option for me 😆

[peak portal]

it works with AMD

[sinful elk]

Sweet, anything special i need to do on install or just switch cards and use my current install?

[peak portal]

for windows there's specific install structions for each GPU type

for Linux it auto detects AMD

[sinful elk]

Ahight, thanks.
Ill have a look on the github page 👍

[fervent thunder]

https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Install-and-Run-on-AMD-GPUs

[sinful elk]

THANKS

[broken smelt]

Python mIght not be the issue, the latest a1111 git pull uses a new pytorch version. Try --reinstall-pytorch (and xformers if you use that) first before reinstalling your python

[narrow hamlet]

When stable animation comes out 😊

[peak portal]

when tis done baking, maybe

[regal plover]

https://colab.research.google.com/drive/1dMhxE9J-o7XJZc5foXgfgKopnMX3hbQX?usp=sharing

here ^

you need colab pro

to run it

[valid atlas]

The model feels ok, but it's missing all the extensions that make it great

Especially the hand, artifact and pose corrections

[narrow hamlet]

It is the original version, do you have any tutorial, thanks

[regal plover]

it's not the original version

it is like beta

[regal plover]

what about this https://github.com/dmarx/notebooks/blob/main/Stable_Diffusion_KLMC2_Animation.ipynb

[broken gull]

so random question what's the different between a domain like .mov .zip and .ckpt?

is .ckpt also a domain?

[regal plover]

google it, or snr

[broken gull]

I tried

getting confusing answers

[regal plover]

.ckpt is a file type, it is a checkpoint basically

[broken gull]

No, file extensions and domains are not the same. A file extension is a suffix that indicates the type and format of a file, such as .doc for a Word document or .jpg for an image file. A domain is a name that identifies a website on the internet, such as godaddy.com or kaspersky.com. A domain can have different extensions or top-level domains (TLDs) that indicate its category or purpose, such as .com for commercial, .org for organization, or .edu for education1. However, some domain extensions, such as .zip and .mov, can be confusing because they are also file extensions for archive and video files2. This can create problems for users and applications that may not be able to distinguish between a file and a website2. Therefore, it is important to be careful when using these domain extensions and avoid clicking on suspicious links that may lead to phishing or malware sites2.

this correct?

[broken gull]

.mov is a file type as well no? some people are calling it a domain

same for .zip

[regal plover]

not everything is a domain

the old openai whisper outputs .ass files

[broken gull]

🤭

[young ivy]

I still have ass subtitles

[gleaming topaz]

i dont know if this is the best place to ask this but im looking to hire a prompt engineer for a project. Does anyone know the best place to look for someone?

[regal plover]

https://t.co/RW88i1aAkW

[dull hare]

how can i make something like pokemon or digimon with stable diffusion cartonish style ? or if there any model that can do that ?

[hasty bane]

Does anyone know a tool that writes the title and tags for an image?

[hollow robin]

Anybody know some sort of app or apk that I can run on android for AI images? I'd love to experiment but it isn't always convenient to have my PC

[north nimbus]

Depends if you want to run it "locally" or on the internet

[hollow robin]

Preferably something "local" but I'd be fine using something online

Also preferably free, but it's not a deal breaker

[north nimbus]

With the app, do you mean at home or outside?

Also jesus christ my models folder is almost 100 GB

[hollow robin]

What do you mean?

[fervent thunder]

My models folder is almost 2TB :P

[hollow robin]

My laptop filled up so fast XD

Gonna transfer everything to my 2TB drive tonight

[north nimbus]

Do you want to use SD in your phone at your home network, or in another place?

[hollow robin]

Another place

If I'm home I can use my pc

I don't expect it to be the full SD, but maybe something like Lexica but free

[north nimbus]

iirc you can add "--public=true" to auto1111 webui and make it a public gradio server

I used the public thing to run auto1111 on collab when my GPU was from the ice age

You'd probably need to find a way to use https and passwords so nobody uses it other than you

[hollow robin]

Sounds like a bit of a doozy, ill see what I can figure out

Thank you

[maiden steeple]

Question, if I wanted to have better controlled compositions and camera placements for stable diffusion, how would I do that? control nets obv seem to be the way to go (tbh I haven't used them that much so I'm a bit unsure). because for example, I want to make storyboards for a film, and I need an image that (semi) accurately shows what the script would say. Im assuming better prompts are a given, but not sure where to start. I tried searching for custom Loras or models but didn't find anything. Could anyone help?

[peak portal]

captionr

[north nimbus]

Is anyone here experienced with EveryDream? I saw that it's a powerful fine-tuner for thousands of images. Does anyone have more info (other than the github repo) about it and on running it on Google Collab (even if it runs on my GPU, I'm not fond of burning it rn)

[void ridge]

hey gamers, any good images to use with controlnet? need some wholesome stuff featuring two fellas

[fervent thunder]

hello is this a good place to talk troublshooting errors with automatic1111 webui?

[urban mason]

I'm making a video on the future impact of AI, anyone up for a chat on a voice channel?

[balmy slate]

https://tenor.com/view/blues-clues-blue-spiral-notebook-drawing-kid-show-gif-15621890

[void ridge]

I get what this is implying, however I am uncreative

[balmy slate]

Me too that’s why I like AI

Have you tried DALL-E

[void ridge]

I was in the beta actually!

Ah I get what you're suggesting

Still though, someone must have a folder full of good images for use

[violet prairie]

#992359629419991142 message

[wary topaz]

What channel do I dream in?

[peak portal]

#1110564013710000258 message

[torpid zephyr]

Ack. No more legacy version of DreamStudio?

[lone fox]

hi gays,good day~ as a long time sd user, i do really have a worried，sd is really good at drawing portrait and art style, because of that could be train into lora. but the most basic topic of Ai drawing i think must be can Draw common items correctly。so which mode can do this?

[warm junco]

Sure in #🤝｜tech-support

[wanton vigil]

Does token merging really work? I’m using it with Token merging ratio at 0.5 and Merging ratio(high-res pass) at 0.5.

I’m using 496x608 hi res 2.75. It doesn’t really seem to change much. I’m having a 4.3 s/it with to 4.5 s/it without. Not sure if it’s really affecting

[hushed quarry]

#1100170312106127410 - #1101178553900478464