#💬｜general-chat

1 messages · Page 25 of 1

fervent thunder
#

😛

cloud granite
#

Yeah

fervent thunder
# cloud granite Yeah

so would I just leave the prompt blank and just generate? how does it work exactly

#

im a bit lost hahaha

cloud granite
fervent thunder
#

but it should be a similar process I think

#

just with different models

cloud granite
#

You could do that

fervent thunder
cloud granite
#

I'm currently in 2 conversations right now, maybe later?

fervent thunder
#

sure!!

#

whenever you got time xD

cloud granite
carmine cedar
#

Im sure my stable diffusion is very outdated given I instaled it a few months ago

#

is there any easy way to update it?

#

(Ideally with better pixel art or videogame style art if possible)

forest apex
#

@carmine cedar it's usually not difficult to update, which UI do you use?

carmine cedar
#

not sure if thats what you mean

weak flower
#

What do you guys think about putting a watermark on AI-generated drawings that are uploaded to social networks? (announcing in the title that they are AI-generated) Do you think AI drawings should not have a watermark?

forest apex
#

there are a few different providers for UI's, automatic1111 being the most popular. if you last installed a few months ago it might be best to install a whole new version, but make sure of where you got it from

carmine cedar
#

this is where I got it from

#

how would I update it?

#

~10/10/2021 was when I installed it

forest apex
carmine cedar
#

sweet thanks 🙂

forest apex
#

you might get an error the first time you run it, but towards the end when pip runs. that's normal, run it again and you should get no errors.

#

one thing that script can't/doesn't do is upgrade pip - you may want to do that manually by running pip install --upgrade pip in the same directory as your installation before hand

carmine cedar
#

appreciate it 🙂

# 
rem place in webui directory
rem if pip produces errors in red text, re-run this script

call venv\scripts\activate.bat
git pull
pip install -U -r requirements.txt
#

I just place this in update.bat and then run it right?

#

(its throwing errors at the end each time I believe)

#

is there a way to pause and see when it ends/errors?

forest apex
#

@carmine cedar you can run it from the command prompt, navigate to your webui directory in explorer, hit Ctrl-L then type cmd.exe, hit enter, then run the batch file

carmine cedar
hard osprey
#

I had a 1050 2gb and my brother was kind enough to let me upgrade from his PC's 1060 3gb. Lol it's slow but noticably faster than before

carmine cedar
#

no clue why it worked that way and not from doubleclicking bat lol

hard osprey
#

We swapped GPUs

hard osprey
river maple
#

The Chinese Communist Party, the perfect model for all world legislation

lilac reef
signal nest
hard osprey
signal nest
#

unless they're worried about fake news along the lines of alien invasions or a rival power having colonized another solar system already

lilac reef
#

Pooh is just afraid of the incoming memes

hard osprey
#

There is a lot of fake news out there. I remember when there was a news story about Pakistan attacking with jets in Afghanistan with footage and it made waves, but later it came out that it was high quality gaming footage.

#

What kind of restrictions and consequences should be out in place for fake news?

lilac reef
#

govts put out fake news all the time (all of them) so in essence they just want ours to be labeled not theirs.

forest apex
lilac reef
#

CIA in the USA is famous (infamous) for Psyops which is fake news made to change how people think.

forest apex
hard osprey
lilac reef
hard osprey
#

America has a lot of soft power to control the narrative

river maple
forest apex
#

won't happen

river maple
#

No amount of watermarks will help that anyway

#

And the kind of people pushing fake news won't give a damn about any watermark legislation anyway

hard osprey
#

The fact of the matter is that if you actually call out fake news, it's acceptable by the masses but only within certain bounds. If you point it out beyond that, you'll be labeled something

lilac reef
#

The soft brains to no brains are destroying this world and personally I rather fight than to allow Nanny states to dictate such nonsense when they are the worst offenders.

hard osprey
#

For example, pointing out the conflicts of interests of politicians and the military industrial complex just after 9/11 would have gotten you branded a traitor to the USA

#

Anything that wasn't pro war was called bad

lilac reef
#

UnAmerican and Freedom Fries. You are with us or against us and that Bush laugh. ugh

hard osprey
#

It's easier to use propaganda by flooding stuff and leaving an impression than countering with debunks. The Narrative keeps churning content out

#

It's a bit like fast food. You can tell people not to but the masses do it anyway and don't care

#

The reason why nanny states control media is because the strongest media comes from the hegemon ie USA. It's content is too prolific and invasive to counter

#

All else being equal, the net media output will always be on the USA's side in general

verbal gyro
#

Any recommendations for free and easy over the web options for trying out prompts? Asking for a friend. I thought I saw something about an area on this discord that allowed prompts and generated images.

river maple
#

Copynot: The Case for Reducing Copyright Terms

forest apex
signal dome
#

Will there be a sd chatbot ? Thing

#

Like chatgpt

elder cradle
#

Hmm, nsfw images in training make prompt results better, myth, or fact?

sonic tundra
river maple
#

Stability AI is working on a ChatGPT competitor, yes

astral goblet
#

what do you mean, competitor? i think stability is working on the blue ocean strategy, where they intend to open source everything and not even compete with anyone. Like how Nintendo does business. Building their own market, not the open sourcing

#

ChatGPT is going to remain closed behind a openai's servers. That's not stability's game

sonic tundra
river maple
#

RaneteteoAn Chat?

forest apex
river maple
#

I believe it would result in better anatomy, though only based on unscientific tests right now

forest apex
#

feels like shadow wanderer is seeing matrix symbols flying around right now

sonic tundra
astral goblet
# forest apex very likely would increase anatomic rendering, but it's debatable right now

likely? not at all. the denoising network ins't building form and volume in it's process as it removes noise from an image. it only knows 2d planes.

The nsfw material helps in that it's high quality photographs of people in great lighting. Also, they tend to be hot people since nsfw training sets aren't exactly going to be ugohs. That'll help the refinement a lot, but you also end up with other nsfw elements being trained into the network which is... problematic in a lot of areas.

You can achieve the same photorealism quality just with any high quality photograph training set. If you use attractive photos, it'll train the model to denoise attractive photos.

#

if you put "laying down" into a nsfw trained set, you may not get a typical laying down pose. if you know what i mean.

#

even if it WERE building form and volume as it denoises, natural drawing models would be such a better representation of humanity than porn stars would be

forest apex
#

aight, I see

astral goblet
#

and while NSFW is broad enough to cover natural drawing imagery, the protests about it are regarding Stability's use of the filter on the training set. Which is intended to filter out smut and hardcore pornography, not something that is more vanilla nsfw

elder cradle
#

Oh, I see, so what are your general thoughts on people at Stable Diffusion Subreddit?

astral goblet
#

it's diverse and there are a lot of cross sections

forest apex
#

its all bots, reddit is all bots... discord too

#

I, however, am not a bot.

wide juniper
#

Is ‘by Greg Rutkowski’ good for improving all art or just art with dragons?

#

I’m doing a visual novel and I’m having him appear as a minor character

forest apex
#

and that'd work for anything generated, though artists who focus on a particular type of thing might produce cool results, I think it'd transfer pretty well

#

or just don't use an artist 🙂 depending on the prompting, it might appease some people.

wise flax
#

looking for realistic sounding text to speech ai,

unkempt hatch
#

sometimes I make an x/y grid with multiple checkpoints, and 2 unrelated checkpoints output exactly the same image. It must be some kind of glitch, right?

fervent thunder
#

I can't be the only one who thinks "this is so obviously not an 80s sitcom setup" whenever they see the "X as an 80s sitcom" videos.

glass lava
#

i am the only one who thinks

#

sry

unkempt hatch
#

automatic1111 constantly gives me errors, always something new.

I can't get invoke running, but is it worth the initial effort to make it work? Will it really be stable and not with constant errors?

astral goblet
#

its just the core tech in general. it's all very mvp

unkempt hatch
#

so I'll still get errors; better to keep on automatic1111 and try to solve whatever problems I'm coming across, thanks @astral goblet

astral goblet
#

yeah keep on the track, but also, if you see a better track, feel free to experiment.

i tried installing invoke too but it's convoluted. i'll try it again later

fierce charm
#

@zealous wedge can you send me a DM with more information about that. Thank you

unkempt hatch
#

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

So I'm getting this now... and when I try cmdr-ui it only lets me use CPU.... so something's hogging my GPU I guess? Everything's working one minute, then this happens

#

I guess I'll ask in tech help

thorn hare
#

so i updated auto1111 but i cant find hypernetwork selection ?

cobalt cliff
#

whats deep floyd? the latest psyop? new model coming out?

elder cradle
sonic tundra
brazen fox
sonic tundra
severe field
odd stag
#

Why is midjourney so good with text and stable diffusion is so bad with it??

#

Is there a specific model that can create correct words and letters?

ebon pewter
#

i tried copying a a prompt and its settings, matching PNG info 100%, but my outcome still different than the original..whats the cause?

valid river
#

I want to train stable diffusion on a specifik Chair product so that I can make design iterations on that chair. Which method should I use?

odd stag
odd stag
ebon pewter
odd stag
#

Is it the same model/ckpt ??

ebon pewter
#

even downloaded upscaler used by the original owner

#

if u mean by model hash..YEP

odd stag
#

I have no idea then

forest apex
proud nova
#

Is there any way to recreate almost the same image? I have same prompts, same seed but the image is always different from the original

forest apex
#

@proud nova maybe try img2img and mess around with a lower denoising value?

cold siren
#

check your model

split dock
#

Good morning for all! I searched a group in the Telegram about Stable Diffusion and don't find, so I created a channel for we to participate and share our arts: https://t.me/stablediffusioncommunity

cold siren
#

or hypernetwork & stuff that isnt shown on the main ui

proud nova
#

Alright. Thx

cold siren
#

IIRC one time i changed the model but somehow it saved the previous model's name & hash with teh newly generated image

hallow rain
#

who TF is chevydude36?

#

he just invited me to a bluewillow discord?

elder cradle
#

Bluewillow seems to be MJ competitor

hallow rain
#

that's odd

dusk bane
#

Sky

hallow rain
#

their art is just 1.4 quality

dense pine
#

yo

dense pine
#

is midjourney using the same stuff as sd?

severe field
#

why do people join voice chats and they always have a default discord pfp

#

some of them just made their account

feral cairn
#

Just blocked some scammer with the username clissia

latent flower
#

that started after the bluewillow raid

#

for that reason I no longer join the VC

#

like

#

what if they are recording idk

feral cairn
#

They could be those bluewillow scammers like the one I just had blocked

latent flower
#

yeah I dont wanna @ mods but I think there should be like an "age limit" for an account to join... That would prevent scam dms/raids

forest apex
#

I think you should @ mods if you're spammed

feral cairn
latent flower
forest apex
#

wonders if there are discord policies regarding spamming a server?

spark kite
severe field
#

who cares if they are recording

latent flower
severe field
#

block em

#

oh

latent flower
#

like what happended in the bluewillow server

severe field
#

Idk anything about it

#

id never expect this place to be raided

latent flower
#

ok, if you guys agree with my suggestion maybe react for the mods to see ppl are interested in that

severe field
#

also VPNs

#

not sure how they do it

latent flower
severe field
latent flower
#

oh, I see

severe field
#

A bot called "Double Counter" prevents people from joining on multiple accounts

#

it would also be nice for people to have an account age limit

#

but I havent seen a bot that does it

latent flower
forest apex
#

they are discussing it on the bluewillow discord, according to one of their mods they are having a contest and encouraged people to invite friends, but people began spamming dms. allegedly

normal dock
#

Hi everyone, I would really appreciate the help. I would like to install and use stable diffusion to make my images and creative stuff, but after watching dozen of vidoes I am even more confused. There seems to be so many versions, can anyone explain in simple English which one should I install on my machine? If you have a link to good tutorial I would be grateful. Tnx In advance

dry trellis
proud nova
#

Agreed

bleak matrix
#

If you see any spam, just let us know.

forest apex
astral goblet
#

makes sense. spammers are trying all the old tricks again as the ai community grows so rapidly. there's no internal community immunity against these kind of attacks yet

forest apex
#

stealth founding, don't want attention while they are starting up, no information on where they are founded out of, hmmm

lilac reef
forest apex
#

maybe operating it over a vpn? many mods seem to speak a specific type of broken english

lilac reef
astral goblet
#

no not crypto mining. that used to require actual work. crypto scams are just acting like they're part of it, and then since there's absolutely zero regulation in that space, they don't have to do any work and just get away with it

lilac reef
astral goblet
#

automation of the absolute bare minimum to pass as legitimate is the goal

forest apex
#

well. questions raised. we'll see where this goes...

lilac reef
#

I just know if they were attached with crypto in any form I'm not touching them.

astral goblet
#

proof of stake took over. where you pay enough to become a key node, and if you disagree with any of the network and publish wrong results, you lose all your money

lilac reef
astral goblet
#

not pay enough, stake enough. you can take it back anytime

#

its good incentive for nodes to behave according to the protocol

#

on a technical level i think it was the way to go. it sure fucked their economy up though lol

lilac reef
#

Only good thing with proof of stake is that the days of video cards costing the price of cars is over, only Nvidia can't let it go.

astral goblet
#

yeah. it's the same goal. an incentive to behave. work is just, well, energy intensive by definition

forest apex
#

now its a recession, wonder if they ever will

astral goblet
#

the idea in the past was work costs money so people are incentived to behave

lilac reef
astral goblet
#

proof of work came out of spam preventing. hashcash is something emaiil servers use to make sure spammers weren't firing 10k emails a second. it was a forced delay because to send the emial, they had to do some work that took time

lilac reef
#

proof of work I do like as you have to proove you did something

#

pressing one button and off goes 10k spam emails is not work

static badge
#

where can i find the latest models to use with automatic?

astral goblet
vocal maple
#

anyone els got some guy trying to invite you to use bluewillow Ai?

forest apex
#

I guess the last weird thing is they've got 90k users, and were founded a sometime in dec 2022? hmmmm

#

goes to look for the privacy policy

grim aspen
#

has anyone here gotten stable diffusion 2 working with automatic1111 in linux?

astral pivot
#

does anyone have a prob where their file explorer keeps getting stuck on "working on it" after downloading sd?

wispy jasper
forest apex
#

apparently hal could do it, sweet

fervent thunder
#

anyone have suggestions for prompts or otherwise for generating terrain heightmaps

#

im using sd-1.5 currently

peak orchid
#

They are also calling for people to start engaging in frivolous lawsuits

fervent thunder
#

could wired get sued for encouraging people to waste court time

#

or penalized

peak orchid
#

Probably not. News normally gets a pass on things (sometimes for good reasons) that could get others into hot water

#

Reporters also have a ton of influence on what legislation gets passed, company policies, what public opinion is, and other stuff with realistically zero accountability

fervent thunder
#

ok so for heightmap to get a decent one made for terrain i made a color terrain map from above with txt2img (and img2img of that a few times to get desired results)

#

then used depthmap on that with res101

#

bonus is i can use the color terrain photo as a colour map

#

im gonna try using textual inversion to train a style embedding on heightmap samples

bitter shard
#

where is hypernetwork option ?

harsh mist
past vortex
#

It's the same Blue Willow thing others above are posting about

bleak matrix
#

And thank you!

past vortex
#

uh... if I already ignored the message request without screenshotting, is there a way to pull it back up?

bleak matrix
#

I don't actually know. Was it the same wording about the contest?

past vortex
#

I can't seem to pull it back up

#

yes

bleak matrix
#

okay

austere tundra
#

Can someone be so kind to point me in the right direction of where people are making those animated AI clips. Seems to be like the same subject but the exterior kinda morphs and there’s a little motion?

serene vessel
#

hello

peak orchid
#

The BlueWillow people are also active on the Unstable Diffusion and Midjourney Discords

wispy jasper
#

Bluewillow literally just seems like SD 1.4

river maple
#

what's blue willow?

fervent thunder
#

idfk but they are spamming people im pretty sure

#

prob another porn model

#

hopefully we get a new model next week

#

#stalediffusion

forest apex
#

How'd they get 90k users in a lil over a month if they aren't behind the spamming is my question

river maple
#

Also, it's not spamming

#

It's a multilevel marketing reverse funnel system

#

Rewards for inviting people

#

That's how it circumvents ToS

forest apex
#

A mediocre product with no new features or internet buzz did that organically?

#

Seen a handful of posts and videos, about it

river maple
#

There are services that will force-join 1,000s of users to your Discord server

#

They're on black hat forums

#

Uses real (but hacked) accounts to do it

#

Have you ever looked up and found yourself in some random server you don't remember joining?

forest apex
#

Nah.

#

But I get it. Lol I'll try and drop it. Initially wanted to check their features out but there's nothing of substance at all there.

iron otter
#

mattvidpro has 30000 views on his video about it and it's free + upscaler

forest apex
#

So Matt's to blame, I see. 😄

river maple
#

30,000 views ain't nothing

old marten
#

Thats what gives it its distinctive "acid trip" look

old marten
#

They are helping poor folks generate images for free.....for now

iron otter
old marten
#

Bluewillow is also sponsoring folks on YT to promote their discord

river maple
#

even at $0.3/hr, you can't afford to provide for free indefinitely

old marten
#

Theyre probably trying to build up their base for now

#

But look at it this way, competition is always good. If MJ gets decent competition they might lower their absurd monthly membership prices

#

$60 per month for 150 fast images suck my dick.

#

I suspect Bluewillow is based in China, theyre a Chinese company

#

Pure speculation though.

forest apex
#

Felt

river maple
#

wait MJ is $60/month now

#

unlimited was like $30/month

#

what the hell

fervent thunder
#

people must be paying it

#

but yeah the 30 was bonkers to me

#

60 is bonkers squared i guess

river maple
#

holy shit I ain't payin that

#

inb4 someone NovelAI-izes Midjourney

fervent thunder
#

has dalle ever updated?

#

i mean beyond the outpainting thing

#

i always thought midjourney was an afterthought last summer, now it seems to have overtaken stable and dalle

old marten
#

ive never heard of a single person use dall-e

#

I always thought of dall-e as this thing in beta but never an actual product

fervent thunder
#

it was pretty popular when dalle2 dropped, everyone was on the waitlist

old marten
#

well im glad it flopped. anything that isnt local can suck my testys

river maple
#

DALL-E 2 is out, but it's not as big as before

fervent thunder
#

i dont know if it flopped even

river maple
#

MJ is more flexible and productized

old marten
#

MJ flexible lmfao.

fervent thunder
#

so i guess dalle never updated then, they lost the steam they had going

river maple
fervent thunder
#

MJ lets you link jpgs i love that

#

the results on MJ are great imho

#

but im not paying for it

old marten
#

Well if MJ is flexible compared to dalle then I honestly dont even want to try dalle

fervent thunder
#

dalle gives you like 20 free gens, may as well check those out

old marten
#

Protogen and Vision 1.2 > Any MJ model

#

I can put my money where my mouth is

river maple
#

DALL-E harkens back to the basic text-to-image generator: you type in a prompt, and you get an image with a few variations

zenith thunder
#

Mj costing money is weird

#

Is mj created using images they own the rights to?

river maple
#

But things get weird very fast when it comes to the law

#

Are machine learning model weights copyrightable? Who knows!

old marten
#

imagine if they leak out the .ckpt MJ uses, I assume they only use one, or at least the v4

#

People would start merging the shit out of that thing with SD models

#

And MJ would collapse lmfao

river maple
#

I'd imagine all of their code would have to leak too for anything useful to be done

#

It's probably easier to train a model competitive with MJ on your own, if you have the funds to do so

#

or still easier to raise the money

old marten
#

Or just merge models. 90% of those on civitai are merges

#

and theyre already better than MJ (personal opinion)

river maple
#

MJ's magic is that any person can spam a bunch of characters and still get something half-decent

wispy jasper
#

PPPPPPPPPPPPPPPPLLLEAAASE give us tools to train 2.1 better! I loooooove this model but my God it's impossible to train well with Dreambooth lol

cobalt cliff
#

3.0 when?

#

will we get more updates? or is the lawsuit freezing all further public releases?

fervent thunder
#

does the 8400 or whatever it is vae or any vae go in the vae folder or the model folder?

#

or both?

cloud granite
charred pike
#

extremely new to stable diffusion. how do i check what version i have? lol

zenith thunder
#

Not sure how mj would defend using copyrighted works to train their models. At least with stable diffusion you can train it on things you own

river maple
#

Which UI are you using?

misty quest
#

Hi guys, I just updated my stable diffusion
and can't find the option to select hypernetworks on the Settings tab
did it change its location?
how can I use hypernetwork now?

zenith thunder
#

I've been trying to reach the ai how to draw otters

#

It's really bad at it for some reason

#

What's a hyper network?

misty quest
#

I'm not really sure how to explain it technically, but it is kind of an "extension" to use on your stable diffusion

#

it's different from a model

#

you can use it combined with a model, to generate better results

#

or results with a specific design or style

#

it's pretty nice and the files are small, compared to models

zenith thunder
#

Oh interesting 🤔

forest aspen
#

hey anyone can tell me a good script for A1111 to draw inside the webui? like in photoshop with colors and such (like invokeai) or something like that

bleak adder
#

Is the model used here in this discord server any good or should I forget about using SD until I have a computer powerful enough to use custom models locally?

forest apex
#

this is fun

regal comet
proud nova
river maple
#

It should be at the bottom of the main UI page

#

It's a short hexadecimal number (numbers 0-9 and letters a-f) after the word "commit:"

#

if you don't have that at the bottom of the page, you're running a really old version

fervent thunder
#

Hey guys, something crazy just happened

peak orchid
#

The anti-AI artists seem to be cheering over ArtStation's new ToS, but after reading it seems that scraping the content for training is still permitted. They are just reaffirming that you can't sell the scraped dataset of downloaded images (links to ArtStation are okay), and they also state that they aren't planning to make their own AI image generator like DeviantArt

vapid tangle
bleak adder
#

Learned a lot, thx

#

So is the Windows 97 look of most images a matter of resolution?

regal comet
#

Np! I actually have more guides but they aren't updated for 2.0 yet

bleak adder
#

In Midjourney, you upscale 1 of the 4 images it gives you and it usually comes out incredible, then you "upscale" it with another software to just make it bigger. Seems same name for two different concepts are being used interchangeably.

regal comet
#

Resolution, steps, keywords, iteration

bleak adder
#

I noticed the envelope emoji trick doesn't work. boo.

regal comet
#

Usually, just iterating gets better results

bleak adder
#

I don't see how regenerating will eventually get me the smooth, intricately detailed, 4k-looking, properly centered look of a Midjourney generation. But I know it's possible.

#

Via this server's version, not so confident

#

Locally, of course it can.

#

It keeps marking my content as NSFW when I'm pretty sure it's not

#

I mean damn every time you get an image of a hole or a slit is it gonna blur me out?

#

Next thing you know it'll think eyes are boobs and every curve's a booty

#

Anyway, there's this old computer glitchy half-rendered chopped-up pixelated look to all these images. Does upscaling fix that?

regal comet
#

Learning how to prompt is a process. But you can use things like Clip Interrogator, #1045349359044280360

#

Etc, to help you get a fundamental understanding of the relationship between words

#

There are tons of really great examples to build off of

river maple
regal comet
#

SD is also continually improving, and there's also plenty of people who are willing to help, share examples, etc. Also

#

You can run SD via Colab Notebooks, too

#

You can still get great results with the Dreambot, though

#

I would just say find something you like, maybe learn, and build off of it

peak orchid
#

In more UK news (Where Stability AI is based), the authoritarian "Online Safety" is apparently well on its way to becoming law in the UK, and its apparently set to get even worse before it becomes law.

#

Its going to impact this Discord and future models if Stability AI wants to keep it accessible to UK users

knotty hinge
#

will stable diffusion implement a tool to blend two images like midjourney has done?

river maple
#

As an American, I do not comply with foreign "Online Safety" legislation and my government will not extradite me over the laws of foreign countries in which I do not do business.

regal comet
peak orchid
river maple
bleak adder
#

Only the rich and profitably perverted will come out on top, as usual

#

so...what's an embedding

river maple
#

The code to train image generation models is also open source

#

and non-UK companies are never gonna demand KYC for their domestic and foreign non-UK users

peak orchid
river maple
knotty hinge
#

image1 + image2 = and get a mix

mellow canopy
#

when sd distilled

knotty hinge
#

a example:

mellow canopy
crisp oasis
#

I cannot access my creation History, is this an issue for anyone else? I've asked tech abut it and they gave me explanation of new Beta version problem, that was 2 weeks ago and still no history to see - I never clear my cache.

river maple
#

If this is the Automatic1111 web UI, try to update and keep in mind that things randomly break due to the lack of any unit tests or integration tests.

whole mortar
fervent thunder
#

does [chris evans/ben affleck] try and create a 50/50 version?

#

trying to remember the command and not at my generating machine atm

waxen pasture
#

I just finished my first somewhat successful embedding! I am super proud of how it came out. You can take a look here #1066559052328472596 message

elder cradle
#

But still, if the "Online Safety" thingy got implemented, will sd models become illegal and get removed?

river maple
old marten
frank rune
#

Can you use merged models in dreambooth?

astral goblet
#

try. it' shard to say. merged models get really weird. some might work. some might not. some might give weird unpredictable results. go for it

plush haven
#

Would anyone have any advice with approach for training a style in the vein of “lego city adventures” tv series. At the moment I am attempting via a hyper network with about 200 images from the show. About 10k iterations (still early I guess) getting concerned perhaps it’s not learning very well. I’ve manually edited all the classifications so that they say “lego man” or “lego woman” if there is a character in the shot. And prefixed the text classifications with “a 3D render of…”

pure yoke
#

Can you code tradetron option strategy

astral goblet
#

remove artists.csv (but what does it mean????????????????????)

In the recentcommit. More of automatic's cheeky attitude towards BLM culture that is honestly hard to stomach. It was removed because people kept pointing at "n" category that he made in the csv

#

he'll never officially say, but i suspect it's exactly what people suspected

#

another personal note of his. some say child. doesn't seem to align. i can't say i don't believe they'd be capable of that

#

intelligent people often are giant dickheads

#

human condition and all

fervent thunder
#

so auto is a racist

#

we need another option for us peons to use

plush haven
#

Is this the right channel to ask for advice on trainin a model, embedding, or hypernetwork?

fervent thunder
#

go for it. or try tech support i guess

plush haven
#

Copied from above: Would anyone have any advice with approach for training a style in the vein of “lego city adventures” tv series. At the moment I am attempting via a hyper network with about 200 images from the show. About 10k iterations (still early I guess) getting concerned perhaps it’s not learning very well. I’ve manually edited all the classifications so that they say “lego man” or “lego woman” if there is a character in the shot. And prefixed the text classifications with “a 3D render of…”

fervent thunder
fervent thunder
narrow sluice
fervent thunder
narrow sluice
fervent thunder
narrow sluice
#

Heck, nobody even knew about the topic until it was leaked

fervent thunder
#

But what really separates proprietary code from open source code?

narrow sluice
fervent thunder
narrow sluice
#

It's if you copy very specific programs and large snippets that does specific stuff that nobody else has already done that could get you into legal trouble

#

Especially when you don't have a license for the code at all.

astral goblet
narrow sluice
#

And if you copy paste entire sections of the code and you only change round brackets to curly ones, as well as using a bunch of experimental stuff that does absolutely nothing...that's the equivalent of changing Sonic the Hedgehog and turning him yellow and calling him Sanic

astral goblet
#

its a good idea for anyone that's looking into being a professional creative, to get a primer course on ip law. a local lawyer might consult you or offer a course

narrow sluice
#

I was here during the entire debacle at worst, so I am used to explaining wth happened

#

Also, I knew the NovelAI lead dev, who also explained what the differences in his vs A1111's code as well.

#

But yeah, it was pretty sad, when that happened.

#

I described it as "If a bunch of teenage DreamSMP stans plus keyboard warriors with 0 programming knowledge tried to prove how A1111 wasn't wrong."

#

One even told me in a debate in this chat how I should pretend that I knew what I didn't know, instead of admitting I wasn't the most knowledgeable about programming.

astral goblet
#

just fake it bro, for the internet points

severe field
#

is it big?

astral goblet
#

michaelscott.gif

boreal cosmos
#

good evening

#

I'm looking for a bare minimum PC that can run SD

signal nest
neat quest
next spade
#

what is the basic image size for portrait?

brazen fox
thorn snow
#

Hey everyone!
I've been playing with SD for a little while now. And I was wondering if it was possible to keep the structure of an image and only change its look. I'm planning to do that on a basic viewport render from a CAD model
Is there any way to do that inside stable diffusion, should I use Dreambooth or any other extension ?

zealous wedge
thorn snow
fervent thunder
zealous wedge
thorn snow
radiant linden
# narrow sluice One even told me in a debate in this chat how I should pretend that I knew what ...

I made a reddit post with my view on it: it was 11 lines copied, and they were just dead code, not used in impentation. I did implement their supporty for hypernets, I disd not copy code. Current implementation for hypermnets is many hundreds of lines. The concept existed long, long before: et the time we didn't know this but it's Lora: https://arxiv.org/abs/2106.09685 - the formula is exactly the same as in those 11 lines; so code is not stole and idea is not theirs

#

so please don't go making statements about me stealing code

#

cheers

lavish current
#

I really want to continue to see the growth of this new technology but real life stuff is making me feel like it's my time to go

#

This is very sad for me. I love AI and seeing the images and text it can produce, but I can't deal with the pain of being alive for much longer

stuck depot
forest aspen
#

hey guys

#

anyone using a1111?

#

i need some help

#

im using img2img with a1111 and i get this

#

wired thing

#

it starts to inpaint the area i want, but it blends back in with the original image, and i have no idea what's causing it

brisk bloom
forest aspen
#

noup

#

wansnt that @brisk bloom

#

finally i found it

#

its the clip skip

#

value in Stable diffusion settings

#

needs to be 5 or above (to 12)

#

other wise blends back the original image no matter how much noise you add,now it works like before

odd stag
#

Is there a way to train an embedding online in a collab as I don't have the bandwidth to train a ckpt and download it back and forth

radiant linden
#

are you on latest version spybg?

#

black can be the result of generation failing with Nan

#

and there is detection for it in latest version

short tapir
#

doe sanyone have a list of negative promts i can use? i had a hude one for copy and paste but uit updated and now my stufs all gone

radiant linden
#

so it will still fail but you will get a specific messasge

#

anyway if that's it, running with --no-half can be a solution for SD2.1

#

increasing clip skip should not be done

main sparrow
#

lol "lowres liquid breasts" is any original image ever captioned that way?

radiant linden
#

it produces much worse results

main sparrow
#

some of these negatives can't possibly help. maybe the lowres part. I dunno. I just guess at stuff most of the time.

forest apex
radiant linden
#

they don't call it schizo negative for no reason

#

and you are meant to paste it in entirety into the neg prompt, not pick and choose

#

all 1000+ tokens

forest apex
#

yea I wondered, if adding a ton of negative prompts affects generation speed

radiant linden
#

2834 tokens

#

it generation time goes from 3.2s to 5.5s for me

forest apex
#

yup, ok

bleak matrix
#

Morning, folks! Hope you're all well today!

rancid needle
bleak matrix
#

👋

pliant reef
#

AI image gen will improve so much

mossy crater
#

Hey everyone, someone here that has experience with running Dreambooth in Automatic11111 UI on Runpod? Cuz I can't see the usual cmd window on runpod so it's harder to debug. But basically when I click "Train" to start training of my dreambooth model I get this error/warning:

"Exception training model: 'CUDA error: invalid argument CUDA kernel errors might be asynchronously reported at some other API call"

Does anyone have any idea how I can fix this? I've been browsing the forums and tried some suggestions but nothing working so far
Images of my settings/setup:

leaden tiger
#

i can´t figure out how to use a hypernetwork .pt file, every ressource tells me i should add it on the settings page and apply, but i can´t find that option anywhere

#

(using automatic1111)

forest apex
radiant linden
#

it's the card button under Generate

#

@leaden tiger

solid dragon
#

Does anyone know of a checkpoint model trained on art history paintings? Like from the Renaissance through to the 1890s? Just as an example.

bleak matrix
forest apex
#

it was directed towards @pliant reef . and good morning 🙂

bleak matrix
#

Ah, lmao

#

I'm awake, woo!

#

Good morning!

#

How are you doing today?

leaden tiger
forest apex
#

starting out with the difficult questions, ey

#

how often is it recommended to do a pip module upgrade for the webui? every git pull?

bleak matrix
#

Periods, techniques, etc

solid dragon
latent flower
#

Good morning ppl!

latent flower
bleak matrix
#

Use prompting to guide what you want

#

If it's a painting, and the style is romanticism, you might go "oil painting, in the style of romanticism"

#

You can get very detailed

#

And use key words like realism, detailed, and pick palettes, etc

solid dragon
bleak matrix
#

If it's a sculpture, you can use words like photography, etc

#

You can use either model

#

There are really good examples in both

#

I suggest checking out other people's prompts, the various guides, etc, to get a better feel on how to get what you want

#

As well to help

solid dragon
bleak matrix
#

It's a WIP guide on how to start

radiant linden
fervent thunder
#

where I can generate images?

zealous crypt
peak portal
#

Hey question, i'm rendering a batch of a PNG sequence from a video. Is there a way to make each frame a little more similar so the video is a little more cohesive vs. now where every frame is img2img from scratch? Would that be the looping script?

bleak matrix
mental axle
#

How can i make it so the AI makes anime style art

last hinge
#

Hi guys, im new to this server, Im looking for space landscapes or kind of sci-fi images, what's the best way to look for this?

forest aspen
#

question, how can you turn OFF clip skip in a1111's web ui?

barren pumice
#

Awwww... How on the holy earth i m getting this msg ...

Woah there! The owner of Stable Diffusion has requested that Discord block any messages our mostly-accurate robots deem to be explicit, so your message has not been sent.

forest apex
#

holy cow, dynamic prompts has jinja2 template support

runic magnet
#

Do I need both files for "anything" new model? I see a file .ckpt and file .safetensors. Version of stable dif I had so far didn't used the safetensors so I'm nos sure if I add both of them to models or just one of them

peak portal
#

can I pause and resume a batch img2img?

#

If I want to play some games, I imagine i can't run both at once

bleak matrix
tough heron
#

There is a guy going around DMing people inviting them to an AI generation server for contest? Is it a scam?

#

He said he found me in this server.

forest apex
#

@tough heron ping a mod

bleak matrix
tough heron
#

Yeah where is the ticket dispenser?

bleak matrix
forest vale
#

new preview of dreamstudio pro in the latest corridor crew

#

looks nice

#

gimme access :P

forest apex
#

@bleak matrix is it possible the spammers are using the same message each time?

bleak matrix
#

Sending a screenshot helps us verify everything. 👍 So it's just best to send a ticket with a screenshot.

#

Regardless, if someone sends you spam for other servers, regardless if it's Bluewillow or not, you can always report it to us.

forest apex
#

I got the most insincere excuse from a mod on their server, but not after a response regarding the contest

bleak matrix
#

Humbug!

#

🙂

forest apex
#

gotcha

#

this jinja script is creating some trippy animations. I was using latent blending before, but this seems a little more coherent, just slower

bleak matrix
#

I love seeing any animations with SD

forest apex
#

alright, sometime tommorow after 3am this should be done generating lol

bleak matrix
#

There have been a lot of sweet music videos have made and some part of my brain is like, "I'll get back to that...someday."

#

Sometime...forever in the future.

forest apex
#

enigmatic_e has some of the best workflows I think

#

but this stuff is like electricsheep

bleak matrix
#

There's a lot of really great workflows now. And it just keeps getting better and better.

#

I love every moment of this.

forest apex
#

I too am surrounded by a fire and smiling

bleak matrix
#

Everything is always expanding and it's great to see all the community made stuff.

untold chasm
#

has there been any word on 2.2?

robust coral
fervent thunder
#

grats

queen kite
#

@alpine slate thanks for the invite! I'm already there:)

queen kite
cold sage
#

Even if the intentions are good and the project is legit (which I have no idea about), it's still spam.

forest apex
#

lol same message

#

@bleak matrix

amber cairn
forest apex
#

@tight quartz3 spammers here now

river maple
#

I'm somehow lucky enough to have avoided spam

narrow sluice
narrow sluice
amber cairn
#

issue is as well, @tender blaze joined today
so I have a feeling either a bot, or hacked account

#

no idea if there are more victims

forest apex
#

@thorny depot

thorny depot
#

Sup?

forest apex
#

3 spammers on the server rn

thorny depot
#

Unghck.... I've been getting spammed by these clowns nonstop for like three days.

forest apex
#

discord can't do anything? got a weird feelin lol

thorny depot
#

There's been some talk but unfortunately, options are limited.

forest apex
#

irc server wars all over again ...

thorny depot
#

I imagine this will peter off eventually.

narrow sluice
#

I’ve seen rules for not DMing folks that doesn’t know you that is ban worthy from the server

#

Ofc, we don’t need that level of restraint, but it is an option

forest apex
#

odds that it's one of their users trying to win the contest? or that there is no contest?

narrow sluice
#

Yes, even compromising accounts doesn’t seem out of the realm of possibility.

drowsy burrow
#

Please tell me where can I download ready-made Styles? 🙏

river maple
astral goblet
#

i doubt anyone will take home the money and it's all just a giant user farming scam, to get an audience for the rest of their scams

river maple
#

So it's a multilevel marketing scheme

forest apex
#

aren't they taking credit cards for image generation credits?

fervent thunder
#

I hope we get some new model this week 🙏

#

i wanna be wowed you know

astral goblet
#

the new rpg model refined on 1.5 came out nice

fervent thunder
#

on civ?

astral goblet
#

always new models coming out

fervent thunder
#

yeah theres some good ones but i still want to see what the big dogs have cooking

astral goblet
#

she? i'm not sure

fervent thunder
#

noice

bleak matrix
#

Lol

fervent thunder
#

apparently artstation has updated their tos to disallow AI generation to use it to train things

#

:(

river maple
fervent thunder
river maple
#

ToS is a voluntary agreement only entered if you create an account

leaden tiger
river maple
#

Courts have ruled that simple "browsewrap" ToS are generally not enforceable

fervent thunder
bleak matrix
#

All taken care of 👍

bleak matrix
#

Lol, it seems it was right in time haha ❤️

forest apex
#

I actually just got one, I reported and blocked now it's disappeared.. @regal basin something, now there are 3 of them showing up

bleak matrix
#

Maybe Discord being goofy

forest apex
last iris
#

Add @fleet birch to the list of scumbags sending out unwanted spam invites

fervent thunder
#

I feel so unloved, I havent gotten a single spam message

forest apex
#

loves @fervent thunder

astral goblet
river maple
#

Courts are starting to crack down on unfair consumer contracts nowadays (especially in the EU)

boreal apex
#

Do u guys think it's worth going from 1.5 to 2?

forest apex
#

it's not either or, each has strengths

fervent thunder
#

sure, i mean its just dropping another model in your folder, why not play with it

#

not like you have to erase all evidence of 1.5 if you use 2.0/1

astral goblet
#

2.1 is very much a base model like 1.1 was, it isn't refined. if you want to refine and explore the new latent space it's worth it. your goals depend

boreal apex
#

Gotcha alright thanks

#

I'll play with it

astral goblet
#

embeds, TI, bringa lot of capability to 2.x

fervent thunder
#

yeah the PhotoHelper and CinemaHelper are pretty great imho

#

there was a midjourney embed for 2.0 that was pretty tight too

last iris
#

Haven't even got around to setting it up locally yet, just been downloading the various needed bits today/yesterday, want to make sure I have everything I need in case the courts decide to get spicy.

bleak matrix
#

Just for you

fervent thunder
#

has anyone ever made like a Batman embed? I know theres lots of hero models, but i wonder how an embed for one character would turn out. and i know i could make one, but im lazy as hell bro, i need this shit spoonfed

low rain
#

is the person who made instruct-pix2pix on this server?

#

@mint hare is it u mayb? lol

mint hare
#

No

fervent thunder
#

hello

#

im currently finding a research topic regarding AI

#

is discussing the ethics of AI art something worth researching about in a qualitative format?

#

only you can answer that incabs

#

or are you asking a rhetorical

fervent thunder
#

it doesnt have to be "ai art" but id like to have some suggestions

#

Well there's been a lot of discussion on the topic, I dont know how many academic sources there are discussing it

#

my general answer would be, yes, it would be

fervent thunder
#

idk really, Im just a dummy with this stuff

#

i think the current heated controversy of AI art and artists is too controversial and can easily be biased

#

but theres lots of people around that probably have some good ideas

#

@orchid wadi or maybe @astral goblet

#

sorry to @ you guys lol

orchid wadi
#

Uhh.... Whats the topic here?

#

AI ethics? Batman TI?

orchid wadi
fervent thunder
orchid wadi
#

Technical or theoretical?

fervent thunder
#

lol feel free to pipe in about a Batman TI too if you want (jk)

#

that isnt broiled in controversy like the discussion of the ethics of AI art is

orchid wadi
#

I mean like... You want a topic on the technology or social subjects?

fervent thunder
orchid wadi
#

OK... Biases is always a juicy one.

astral goblet
#

Conjuring the power of thhe eldritch through machines. Deus ex Machina

orchid wadi
#

Like the fact that the biases come from datasets.

#

They aren't even intentional.

#

They often reflect attitudes but not reality.

#

You can go as deep as wittgenstein if you want

#

like if you are desperate.

#

Like lets say you prompt Somethingh simple. And end up getting like... Only Asian people.

#

Like some... Asian sport.

fervent thunder
#

no, im interested in tackling on the more general affects of AI

orchid wadi
#

Ahhh...

#

Sure...

#

How specific you want to be?

fervent thunder
#

not that specific, but still neutral

orchid wadi
#

Because I got like.... Nautical metric fuck ton of ideas you could do.

fervent thunder
#

more opinionated

#

im not that technically literate, and this is for our research subject lol

orchid wadi
#

No I mean like... In job settings? Industry? Media?

fervent thunder
#

hmm, pick ur poison lmao

orchid wadi
#

Well... If i'd start I'd start from the perspective of Automation and DDD-jobs. Automation has been increasing gloablly for DDD jobs. And it hasn't really "replaced" jobs because those were jobs that people didn't want to do to begin with. I wrote about this myself in automation.

DDD is Dirty Dangerous Dull.

#

However in industrial setting we really have automated as far as we can. Because now automation requires workflows specifically designed for it.

#

This leads to a middle ground of "job that could be automated but is not worth it because the poject is small".

#

HOWEVER!

#

When I wrote about this (6 pages) I brought up how automation is used in industry to get competitive edge. I'm in 10 man machine show, even we got a robot and welding mechanisation.

#

However consider something like... A bartender who just pours or mixes basic drinks in a busy venue.

#

You don't need the "human" in that because they don't have time to chat to begin with.

#

There are some robots but they are show pieces.

#

Like attractions.

fervent thunder
#

hmm, i see ur point\

orchid wadi
#

You could easily start investigating something like... How shitty jobs that really humans shouldn't be doing should be replaced by AI.

#

Like... Do we really need data-entry people level stuff.

#

You can even take the framework of bullshit jobs

#

David Graeber

#

There are lots of jobs which... Well they don't produce or make anything.

#

They just like... Manage for the sake of management.

#

Do we truly need a person to fill an time slot excel sheet? Do we need person with biases for that?

fervent thunder
#

interesting thought experiment as to what areas can AI fill, considering both the limitations of said AI and limitations of the company

orchid wadi
#

When we could have AN ai that can do complex calculations to figure out more optimal configuration. Where in people could be given more free time or made so taht they are more efficient in those tasks or just burn out less.

#

Well "ai" is bitty silly in industrial settings. We don't use AI we use advanced algorithms.

#

Like machine vision. WHich is what even SD really is about.

fervent thunder
#

but i think if we go too "vague" with AI it makes it become this "miracle technology"

orchid wadi
#

I think you should approach it from practical sense.

#

It doesn't need to be a miracle technlogy.

fervent thunder
#

i guess we can moreso focus on AI with visible limitations

#

like ai image generations

#

and the impact of that in the creative industry

orchid wadi
#

It is just a framework that can process more information and variables than human.

#

Well... If you like wittgenstein then you could go to about how you can't make an AI to make a picture you can't describe with words.

#

Because our language and thinking is fundamentally limited by our linguistic skills.

#

Don't know if you have read wittgenstein...

#

It is uh...

#

Heavy stuff.

fervent thunder
#

i guess thats a bit too deep for me

orchid wadi
#

Like very often when you prompt something, you get stuff that is linguistically correct, however in context and intention incorrect.

fervent thunder
#

thats more on the theoretics

orchid wadi
#

I don't think it is.

fervent thunder
#

i guess a simpler topic is asking experts in different industries about so-and-so with the AI

orchid wadi
#

Like... Going back to the last rant we had on 2.1 channel.

fervent thunder
#

?

orchid wadi
#

If you only use google. You might think that "Pampers Diaper" is just a word describing a cardboard box.

#

And the AI agrees

#

SD AI model that

#

is

#

Because it is trained on LAION google scrape.

#

So if you wanted to prompt like a disposable baby nappy. You can't use those terms.

#

Even tho if I say to you right not "Man wearing pampers" you know what I mean.

#

The AI has NO FUCKING idea.

#

You could hunt examples like this and compare them

#

Like ask people "What do you think these terms describe?" and then basically see what the SD thinks it is

#

Sample like 100 generations.

#

There are all sorts of things like this.

#

Granted I could prompt "jump suit" or "boilersuite" but those aren't terms we use in common discourse.

#

If I wrote about soemthing I'd write about that

silk sparrow
#

yo could someone tell me how to set share=true in ds

#

sd

cosmic crescent
#

Is it possible to seed a textual inversion embedding with a text prompt?

Eg start with "40 year old man" before training it on pictures of me? I'm finding my results seem fail at understanding I'm a man. For example, "man cooking" will give good results but "jarail cooking" will not. It might have me and cooking in view but I'm not actually cooking.

But something like "40 year old man cooking that looks like jarail" does kinda work. So trying to get it to understand that jarail is a man first, then train on that to hopefully get appearance without forgetting everything it knows about people.

#

To be fair, I don't fully grasp how textual inversion layers on top of models. So if I'm going about this the wrong way, that's possible too.

forest apex
#

@cosmic crescent it depends on how you captioned your images? you can add the '40 year old man' to your captions

cosmic crescent
#

I'm using the subject_filewords.txt template and used BLIP for generating image captions

forest apex
#

how many steps?

cosmic crescent
#

I think I've got up to around 3000. My first attempt corrupted and all the images it produces are garbage now. So I'm starting again. But I get similar lack of understanding at 10 steps and 1000. So I feel like I'm doing something wrong.

forest apex
#

try twice as many steps

cosmic crescent
#

When you're asking about steps, I'm doing batches. So 3 image batches and 16 gradient accumulation. So not sure if that counts as ~50 steps compared to 1x1

forest apex
#

that I don't know I've never done batches

cosmic crescent
#

It is generating images that very strongly resemble me.

#

I'll keep experimenting but feel like I lack some fundamental knowledge here.

#

Maybe I'll go through the generated BLIP captions and try to improve those.

#

And the template. Maybe some of the wording there could use improvement. It doesn't seem entirely templated out for a person.

forest apex
#

yes, and make sure to describe only what you want the embedding to produce

#

I included shot type, race, age, hair style

cosmic crescent
#

Oh, so if there's something like "a man standing in the grass", I should remove the bit about the background?

forest apex
#

yes

cosmic crescent
#

Okay. Then I'll have to drastically edit the BLIP captions because they generate a lot about backgrounds

#

thanks for the help

#

Gives me another days worth of stuff to direct my next round of experiments

jaunty hazel
#

Is there an "erase" function on A1111 inpainting, or just "draw" and "go back"?

shadow birch
#

hey guys, i have 100 credits a month, or after it's over?

maiden magnet
#

someone dm'd advertised a scam to me I think

halcyon granite
#

whats the ip address used to open automatic1111, i can't seem to find it anywhere

waxen pasture
halcyon granite
timber dock
#

are there any good AI comics?

#

I'm of the opinion that AI is bad at that from what I've seen but I haven't really set out to find any good examples.

forest apex
#

@timber dock you mean like completely AI generated? I haven't seen any but I think soon

timber dock
#

Hmm I mean I can’t really judge it cuz it’s in German but it seems pretty similar to others I’ve seen

#

You probably need post generation work or a tightly trained model for comics

forest apex
timber dock
#

That really shows AIs ability to make things that at first glance look correct

forest apex
#

hehe. well full generation of something like that probably won't get very good.
the text was probably manually written. but there is also https://www.reddit.com/r/AIcomics/

timber dock
#

the webtoons format fits AI

timber dock
#

??????????????

floral zealot
#

hello every one i have a question,this is for somone having a 3060 +,can you generate a img on 1920*1080

#

or its not powerfull to make it

fervent thunder
#

Im sure you can, but it might take a few minutes

halcyon granite
#

is it possible to increase auto1111 batch count above 100? If I set it to say 400 it automatically changes back to 100

brisk bloom
fervent thunder
#

seems like that would be a lot of constant stress on a video card but maybe I'm just overly wary

#

anyone here use stable diffusion on run pod?

green sandal
#

is there a way to pay to get SD installed on my pc? I've been trying for weeks. I can not follow the guides. I don't understand it. It's above my computer literacy level. All I want to do is upload some reference images to train a model, and start generated some new art in the style of those reference images. I don't understand why it has to be so complicated.

fervent thunder
#

welcome to hell

radiant linden
#

the installation should not be difficult if so

green sandal
#

AMD

#

AMD Radeon R7

fervent thunder
#

yeah the install is pretty straight forward, just read and go step by step

#

seriously, if I can pull it off, anyone can

green sandal
#

It's not. I have tried at least 10 times. I'm not a computer person.

fervent thunder
#

I'm not really either

#

what part are you struggling with?

green sandal
#

I would have to start over. last time i tried was last week. but let me see real quick

fervent thunder
#

I mean the first step is to get python, follow that link, start the install and make sure when you do to check the box for 'add python to path'

#

surely youve installed programs before

#

do you know if you are running 32 or 64 bit os? then get your git

green sandal
#

so for example i tried following this one

fervent thunder
#

might be cleaner to follow

#

scroll down to Automatic Installation on Windows

#

i mean its essentially those first 3 steps

green sandal
#

does it work on amd?

old gull
#

can you generate in larger than 512x512, without hires, on SD 1.5?

fervent thunder
#

oh amd might be your problem

green sandal
old gull
green sandal
#

and that answers that

#

no wonder.

fervent thunder
#

ye thats likely been your issue

old gull
#

you need cuda cores to run SD, but amd doesnt have those, and windows doesnt have an emulator, rocm:pytorch for linux can emulate cuda cores, and let you run the webui (LIKELY INNACURATE DESCRIPTION MEANT FOR BASIC UNDERSATNDING IM NO EXPERT ON THAT STUFF)

green sandal
#

yea you may as well be speaking chinese.

old gull
#

luckily too it usually auto pulls the correct pytorch for your card when you launch it

fervent thunder
#

well if you need to upgrade your video card, go nvidia next time i guess

#

theres also the official DreamStudio site if you havent used it

#

thikn they lowered the price of gens and you get like 200 free

green sandal
#

yea in 5 years when i get my next computer

#

thanks anyway guys

old gull
#

hey i could kinda help

#

ya get it runnin on linux

#

unless youre just not about that

green sandal
#

yea thats beyond what im willing to do. im not gonna install a new oos

#

os*

fervent thunder
#

I feel that. Id be the same way

old gull
#

well, as an alternative, have you heard of lexica?

#

click the menu slider icon on the right side of the search bar, and switch from lexica aperture, to SD 1.5

green sandal
#

i havent but my goal was to use dreambooth or lora to train a model with my own paintings.

fervent thunder
#

cool idea

old gull
#

man that would be cool.. what if i said i got linux running on my pc in under 10 minutes.. and its not as clunky or difficult as it was a decade ago when you heard about it in a tech class

green sandal
#

ill think about it. thanks anyway, im gonna head out.

old gull
#

take it easy!

fervent thunder
#

laters

elder cradle
#

Ehh, what are your thoughts on Bluewillow AI?

wide juniper
#

When I don’t get some thing that I want, when should I try different samplers instead of changing my model or modifying my prompt? When would I ever want to try a different sampler?

fervent thunder
#

Good question on the samplers, and Ive never heard a good reason for why you'd want to use x,y, z, seems basically like a crap shoot beyond some of them are more efficient at lower steps like sde++ or whatever

radiant linden
#

The larger the resolution, the more likely you are to get weirdness, on 1.5

#

something like 512x768 should be just fine

#

larger - very likely to see repeating/similar patterns in the image

fervent thunder
#

which folder do i download textural inversions into?

forest apex
fervent thunder
fervent thunder
forest apex
#

you refer to them in the prompt with the token they were trained with, often it is the filename of the embedding

fervent thunder
#

ok

slim relic
#

Hello guys I'm new to Stable Diffusion. I only understand the website version yet. I heard that Stable Diffusion can be used with other models. How can I achieve that?

forest apex
#

@slim relic you can either access them through various https://huggingface.co spaces or oher sites, or download a webui, or interact with SD through python, etc

fervent thunder
#

do i put VAE files in the same folder as the model?

slim relic
#

which version of stable diffusion has the best performance?

forest apex
#

@slim relic performance is mostly related to your video processor, but some webui's have more features, others have more streamlined workflows - Automatic 1111 and InvokeAI are the big ones right now

warm junco
fervent thunder
#

ok

fervent thunder
#

i'll be a chad when ai makes a normal human hand

grave gale
#

anyoneknow where i can get an updated version of the " midjourney " model ? i think mine is outdated

fervent thunder
#

Hi guys! I want to change my girlfriend's photo according to her zodiac sign. Can you help me with prompt and version?

forest apex
grave gale
#

ah

#

Thanks for link i will try this

#

i imght have outdated SD also

forest apex
#

which UI do you use?

grave gale
#

where do I see this

forest apex
#

^_^ where did you get your interface from?

grave gale
#

when I run it it says 1 August 2022

#

i havent used SD since last year i august

ruby jacinth
grave gale
#

and also I tried adding those " .pt " into my models folder but it cant select time in my UI

forest apex
#

@grave gale you just need to refer to them in the prompt with the keyword given on the site

grave gale
#

hm im not 100% sure what you mean by that 😅

forest apex
grave gale
#

ok sorry I dont want ot ask dumb questions

forest apex
#

after that, check the site you downloaded your .pt from and see what keyword they mention for using the embedding

grave gale
#

i have this

#

ah i cant post images

forest apex
grave gale
#

i get that sorry my ADHD makes it hard for me to search

forest apex
#

what is a curse to some is a blessing to others

grave gale
#

true for me in my artwork its a blessing 😄

forest apex
#

yea that UI is essentially ancient at the rate things have been moving

grave gale
#

but i need vidoes to understand who to learn

grave gale
#

so there is a very different setup now ?

forest apex
#

yes

grave gale
#

gocha

#

well like I said i got it august last year

forest apex
#

you'll probably be able to just git update and pip update modules

grave gale
#

hmm ok git is a very strange place for me i get Lost fast there

forest apex
#

@grave gale you can use this batch file to update - place it in your webui directory

grave gale
#

since its so much test

#

txt

#

ok so i press " get file"

#

i honeslty dont understand how to get files from gethub

forest apex
#

first refresh the page, I just updated it to pause after it's done.
You can copy/paste and create a file or click on "Raw" (next to "Blame") and save it to your webui directory

grave gale
#

ok refrshed page

shy thicket
#

So like a lot of others I'm still on 1.5, tried 2.1 but it's just not as good yet. Everyone still seems to be mainly making stuff for 1.5 rather than 2.1 which is understandable.
So my question is will there be a SD 2.2 or will it just leap to SD 3.0 ? And any news on when this might be? 🤷🏼‍♂️

hardy gust
#

Heyo guys, do you have any experience with CKPTs created with TheLastBen's fast-Dreambooth colab pynb?

#

a few months ago I trained a ckpt based on my face, worked out pretty great, now I decided to make a new one, automatic1111 webui tells me of an error, I check the console, I read it's potentially malicious - how I can further pinpoint if it's really malicious or if I can totally disable the warning?

#

(I mean TLB's training seems to be one that's well-used, so I wouldn't necessary associate anything malicious to it)

vast minnow
vast minnow
fervent thunder
#

why does it fuck up the words

#

Anyone able to modify my current logo image? With different variations! Willing to pay just DM

forest apex
elder cradle
#

What's the point of saying that?

silver pebble
#

yo fellas

normal dock
#

Hey. I got a simple question as a beginner SD. I see tutorial of people and many have this black interface of SD. I got white interface. Am I using different version or is it something I can change?

brazen fox
normal dock
#

I think so, but I'm not sure

#

How do I know?

brazen fox
normal dock
#

I do have model selection

#

Menu

brazen fox
#

anyway, if you go to the folder where Stable Diffusion is installed, there's a file called webui-user.bat . Right click on it, choose edit

#

And after the words set COMMANDLINE_ARGS= add a space and then --theme dark

winter pike
normal dock
#

Awesome. Thanks

brazen fox
#

save it and you're done- every time you launch stable diffusion with webui-user.bat you'll have the theme dark 🙂

normal dock
brazen fox
warm junco
normal dock
left knot
#

bruh

#

on the new update of auto's webui, where tf did hypernetwork go

#

its legit gone from settings

#

cant find the dropdown to swap it

#

ahh moved.

fervent thunder
#

kind of interesting I guess

bold perch
#

Hi, i heard a lot of times from Emad Mostaque interviews that they working on educational AI applications maybe someone can tell me what is the name of this project or where i can find more information about it?

tawdry summit
#

I got the Blue Willow ping. Is it anything legit or is it just a scam?

wide juniper
#

On 1111, on the img2img tab, how do I save the image on the left side? i need this when i decide it's not worth generating a better picture and just settle with what I have

forest apex
fringe creek
#

hello everyone , i have problem to install stable Diffusion . can u help me please

bleak matrix
mossy crater
#

Hello everyone, I have question about training Textual Embeddings in the Automatic1111 WebUI
When creating an embedding you can select "Number of Vectors per token"

  • I have found information saying that the this numbe depends on the amount of Tokens your initialization text takes up. So if you write “photorealistic”, then you need to set VectorsPerToken to 2, because Photorealistic consists of 2 tokens “Photo” and “realistic”.
  • on the other hand I have found information saying that the number depends on the amount of training images you have. The more training images, the higher the number should be:
    • 2-3 Vectors Per Token for <10 Training images
    • 5-6 Vectors Per Token for 10-30 Training images
    • 8-10 Vectors Per Token for 40-60 Training images
    • 10-12 Vectors Per Token for 60-100 Training images
    • 12-16 Vectors Per Token for 100+ Training images

If anyone can help me with clarifying this, that would be really helpful

I turned my question into a reddit post as well: https://www.reddit.com/r/StableDiffusion/comments/10jfn5f/how_do_number_of_vectors_per_token_work_when/

bleak matrix
#

That would be great!

bleak matrix
#

Area! Someone might be able to assist you there

floral umbra
#

Odd how i always see people in SD discord voice, but all scattered to one chat alone each Thunk

empty barn
#

Oh hey Duckers. Got the Steam Deck to work yet?

stable pawn
#

I am a newbie and total noob, I followed a video to install Automatic1111 WebUi, but now it say.

"You are running torch 1.12.1+cu113.
The program is tested to work with torch 1.13.1.
To reinstall the desired version, run with commandline flag --reinstall-torch".

But where should those flags be used? a smaller stroke a year ago affected my logical thinking, so sorry if I ask in wrong place and can not understand simple things.

mossy crater
warm junco
stable pawn
#

thank you all.

chilly valve
#

gm gm, is there any channel we can post our work to share with community?

forest apex
#

good morning, there is another general chat with images enabled, a channel for animations, and any of the dreamer communities are appropriate

steel mulch
#

hey fam how to update upscaler? i only have few options

hardy gust
#

good morning, has anyone ran into false positives when creating checkpoints with thelastben's dreambooth google colab script (via automatic1111 webui)?