#💬｜general-chat

1 messages · Page 7 of 1

pliant haven
#

if you can find that source be a life saver

#

I tried to reach out to them

#

as I had copyright query on it

#

but they never respond

#

they even followed my loab account at one point lol

#

@twin hullart on insta if yout interested

little plaza
#

i'll let you know if i find it! looking

pliant haven
balmy wasp
#

What I don't like about this particular statement is the decision to not even allow it in the workflow at all

#

which channel is appropriate to share a piece on?

peak orchid
#

That seems to be very biased speculation on their part

worthy marten
#

i have been using this script to generate midjourney prompts with chatgpt,
(write without wordwraps, headlines, connection words. Write back to back seperated with commas and spaces:

[1], [2], [3], [4], [5]

replace [1] with one or more random subjects
replace [2] with a list of detailed descriptions about [1] which can be totally unexpected
replace [3] with a list of detailed descriptions about the environment of a scene which can be totally different and not fitting [1]
replace [4] with a random list of detailed descriptions about the mood/feelings and atmosphere of the scene
replace [5] with a list of detailed descriptions about the way the prompt should be realized (eg. Photography with camera model and according settings, Painting with detailed descriptions about the materials and working material used, rendering with engine settings, and everything else that could be defined as a output type)

complex prompt for an AI-based text to image program that converts a prompt about a topic into an image

always start the prompt with "/imagine prompt: "
never write anything that is in [] brackets
don't use any line breaks

now generate the prompt)

any ideas how to improve or edit it to make it work better with stable diffusion ?

warped bear
#

Has anyone tried to make UI designs in stable diffusion?

#

I saw someone made some beautiful UIs in midjourney, but I can't seem to make them in stable diffusion

tame fern
#

when is the collab on 2.1?

pliant haven
#

did you get this working?

#

having the same issue

#

but mine is AMD

broken gull
pliant haven
pliant haven
pliant haven
#

this is why

fervent thunder
#

I was testing out Stable Diffusion 2.0, the Depth2Image method, and I was wondering if it'd be possible to somehow preserve or accentuate the shape of a thing in the original image so it preserves the shape of that too. Thanks in advance 😅

stoic linden
#

hey guys is there a list of potentially malicious models out there?

worthy marten
crimson magnet
#

I have a 3060ti and honestly while the speed is nice the lack of vram is a headache when it comes to training and deforum animations, 12gb 3060 seems to be the best bang for buck in terms of local use unless you can afford a 3090.

vapid tangle
hollow swallow
#

hey guys, we have made a super easy gui based installer / launcher, with option to switch branches, if you want to test our new autocanvas that extends the functions of our batch outpaint canvas.
https://github.com/XmYx/ainodes-installer
Please find it via the link, run start.bat, press install all, then press run aiNodes.
You can download models from within the app.

peak orchid
sour fog
fervent thunder
#

Hey guys

#

I’m new to stable diffusion

#

Why did you delete

timid gate
halcyon granite
#

any tips on turning one generated landscape into several more? I want them to look like they are the same region.

Just using the same seed doesn't help

rapid tulip
#

if I create 4 variations of the same image, what's the best AI software for blending all 4 of those into an animation? where the images transition from one to another

bleak matrix
#

Hello everyone

#

How are we all this evening?

peak fern
#

learning mood:)

bleak matrix
#

that's a great mood to be in!

#

I've certainly been in one myself

#

though I'm starting to get sleep lol

hollow swallow
jolly fog
#

guys what is the difference between textual inversion and hypernetwork ?

neat elm
#

hi guys, quick question on shivam dreambooth colab

#

how many regularization images should i use, how many training images of my selfie, and number of training step should i use?

old swift
#

does anyone have a good guide or explanation on how to inpaint hands?

neat rapids
#

I trained a model in SD Web ui 1.5 and now in SD 2.1 the same prompt is giving me blurry and distorted results. Should I exclusively train a new model for 768 res?

waxen swallow
true nymph
oblique pine
#

Does SD 2.2 have ETA? Or some roadmap for new releases? I saw some article for SD 3.0 training

lilac reef
#

they are on holiday so when they get back a lot of stuff is planned for Q1

oblique pine
lilac reef
signal dome
#

where can i find sd 2.1 depth model
i am trynna load this in blender plugin

sand wigeon
#

@signal dome

signal dome
#

ight ok

signal dome
#

and i saw somewhere 0.9 is out with texture projection

sand wigeon
signal dome
#

🙏

sand wigeon
#

have fun!

signal dome
#

yes

left knot
#

anyone here a Marvel comic fan? I need to consult someone's brain.

worthy marten
# timid gate Pin?

Wtf u talking about ?
Just write the script in chat gpt and it will give you a prompt for midjiurney/stable diffusion

tight dagger
#

Hey guys. If I use the discord but do I lose credit for it or is it free? It might be a dumb question 😅

spring surge
spark socket
#

How can I download the full quality full resolution image?
It downloads only in 512px

lilac reef
#

If I don't use xformers I get a totally black image?

#

@wise stratus Apparently 2.1 demands xformers or it doesn't work so TheLastBen telling me 2.1 is broken on Friday was spot on.

quick wraith
#

Hey. General curiosity, likely covered elsewhere can’t seem to find the answer … Will the opt-out list from haveibeentrained be implemented into next release? I seem to remember seeing that site and Stability are going to work together but having difficulty tracking it down. Anyone know anything on this?

I just want to make sure that when I inform my friends who are in the dataset, and want to opt out … I’m giving them accurate information

daring turtle
#

If i want to run a stable diffusion text to image on my website should i use the API or the stable diffusion docs , open source code ?

agile spoke
#

any idea what config file has the SAFETENSORS_FAST_GPU variable in the Automatic1111 release?

fervent thunder
#

If I have trained an embedding with dream artist, will it only work on one model or on all?

timid gate
lavish current
#

I feel like there is some irony to people being more critical of AI art errors than actual handdrawn art

#

Like that infamous Captain America image

lofty parcel
lilac reef
#

Anyone know python as I am getting some sort of file not found but I don't see what the file is?

lofty parcel
lilac reef
#

I have been trying to compile xformers for 11.5 hours now. I did it easily before (4 times) for 1.5 but now nothing but hell

lilac reef
#

Was a CUDA issue for the missing file

untold lava
timber raft
#

look pretty good for CG stylized projects in the future

#

the current learning procedual textures is kind of pain

cold siren
#

like, i literally had no idea who greg rutkowsky is, before all this

untold lava
untold lava
cold siren
#

yep 🤣 ban photoshops! use pencil

#

damn i learned about so many awesome IRL artists because of SD

untold lava
#

I also think it's a big "death of the author" pastiche.

#

If people saw an AI artwork, a good one, but thought it was made with another method, they would love it, but because the METHOD was "wrong" it doesn't deserve to exist, sigh...

cold siren
# untold lava Who?

perhaps SD users should "protest" by erasing all mentions of him from everywhere 🤔 that would be a funny kind of revenge))

timber raft
#

damn, you guys look like some of disrespect bitches _Worry_Think

untold lava
#

Hahaha, human stupidity just annoys me xD

#

But it's fine, 200 years of "this isn't not real art" shows that these kind of things always happen with a new art form 🙂

daring turtle
#

Anyone ,
"How to create stable diffusion website" link ?

timber raft
#

they have different opinions about AI, then let them have

#

if you found out Hayao Miyazaki also said about AI, are you gonna bring his family out and insult him? TKpepeHmmm

cold siren
#

all this situation, with machines being able to do something that only humans used to be able to do just highlights that maybe this stuff was (or now is?) a bit overpriced

timber raft
#

you are telling me, the expressions of human, the enjoy they go through drawing for thousand of years is fcking OVERPRICED?

cold siren
#

(...if we're speaking about poor artists losing their jobs ofc)

lofty parcel
#

Human labor is overpriced because there is nothing that a human mind can compute that a computer theoretically can't, and nothing that a human body can do that a mechanical body theoretically couldn't; welcome to the end of civilization

timber raft
#

maybe he was right "it's an insult to life itself", when we have people like you

untold lava
#

The way I see it, anything that let's more humans create beauty and share their vision with the world around them should be applauded ❤️

lofty parcel
#

I can't wait for text-capable image models so we can generate memes

cold siren
#

whats keeping you from using ms paint 🙄

lofty parcel
#

Having text seamlessly integrated into the image is nicer and makes for a more convincing setting

#

Besides slapping generic text captions over images is so 2010s smh my head

timber raft
#

just use photoshop sip

cold siren
#

one thing i'd agree with, is that "ai" generated stuff should be labeled as such & not mixed with the normal "art" 🤔 but we shouldn't stop there - let's also make everyone to clearly label all the paid promotions, lobbying, etc etc. and of course politicians should always clearly indicate that the speech they'll be reading is not written by themselves, but by special speechwriters

timber raft
#

find a job with only AI skill on Artstation

#

that sound kind of cringe

timber raft
cold siren
#

if only we (including all teh poor non-ai artists) did not have to pay all kinds of forced rent for literally everything...

timber raft
cold siren
#

well technically you could treat any corporate entity as an "artificial intelligence" running on human hardware.
so yes, there is 🤷 https://www.disney.com/

timber raft
#

the hell are you talking about? PogWeird

cold siren
broken gull
#

I am losing my mind

waxen swallow
pliant haven
celest flame
#

Is there a dreambooth guide

sleek osprey
#

How does sd plan to speed up the generation times?

#

I saw some tweet that said its gonna get upto 30 images per second

grizzled grove
#

Hi can someone help me ? I installed InvokeAI it never produces results as good as the ones i see here .. what am i doing wrong ?

manic igloo
#

Hey guys I am trying to install analog diffusion, I put the ckpt file into a folder called Analog in stable diffusion ( \stable-diffusion-webui\models ) but that didn't do the trick, anyone can help me?

versed urchin
#

Anyone know a good repo to finetune 2.x locally?

versed urchin
manic igloo
winter pike
#

below 768 will produce very distorted images

manic igloo
safe sentinel
fathom olive
#

On Auto1111's wiki, under features at the top there's a short guide on setting it up for 2.0/2.1

safe sentinel
#

ah, thank you!

wicked jungle
#

How are you people getting crisp images? im getting fuzzier renderings

north stirrup
#

I usually find a sweet spot in the 60–80 steps range for the stuff I like, but that's often overkill and sometimes "overbakes" the image

wispy palm
#

Does Deforum have a Discord sever or does this server deal with deforum too

brittle moat
#

How do I DM the bot to generate images?

earnest fractal
#

with the prompt syntax, what does '' mean? using that deepbooru thing, (medium) was generated with a portrait of person standing straight, taking up most of the picture. i get it must mean 'medium distance' or medium range, but does the \ bit add any strength? I'\m using 1.5 local install btw.

astral goblet
#

hi friends. I just rebuilt my PC with a new 4080 and windows 11. Big changes today. How do i best configure things so that webui uses this to the best ability? i was using a vega 64 through linux before this, not because i'm loyal to linux, but because amd windows drivers began to suck a lot

velvet jackal
wicked jungle
golden patrol
#

Heya i tried to generate some animations with the depth2 image model earlier today with zdepth passes from a 3d software. But i'm having a really hard time to get a consistent animation from frame to frame. If anyone has some infos or a link to a tutorial it would be highly appreciated. Thanks in advance

north stirrup
scarlet tapir
#

Any reccs for generating decent channukah images?

wicked jungle
#

Any tips on how to get a single person to be rendered? whenever i try to do a random generation i get 2-5 people in this image

halcyon granite
#

anyone have experience running SD remotely? I want to upgrade my desktop 1080 but also need a laptop since I plan to work on the go. I don't know if it's worth getting a beefy laptop / GPU or just upgrade the desktop and connect remote to it

unreal cargo
#

(or 768 with 2.1's 768 model)

wicked jungle
unreal cargo
# wicked jungle i am, didnt think it would do that

Yeah, basically since it was trained on 512 x 512 (or 768 for the 768 model) anything greater will often do that. You can get around this by either seed searching (basically some seeds won't have this issue) or something like the "Highres. fix"

#

You can also get around it by using img2img

#

basically you can do giant blobs and then run it under very high denoising

#

kinda similar to finding a good seed, without having to find the good seed

wicked jungle
#

sadly this is all gibberish to me lol. im trying my damnest to learn how to use this

unreal cargo
#

oh I can't share screenshots here

#

forgot

crimson magnet
ornate talon
#

is stable diffusion connected to the internet?

crimson magnet
#

It's really just downloading some files, putting them in a folder, and installing python and git

crimson magnet
unreal cargo
#

@wicked jungle

wicked jungle
unreal cargo
#

one sec

wicked jungle
unreal cargo
vapid tangle
#

Anyone able to successfully train a model with 6gb vram? I get cuda errors using automatic1111 and the model doesnt do any change

old marten
#

i'm at 12 GB and I already have to reduce to 420x420 and even then it doesn't work with 1.5 only 1.4

vapid tangle
#

@old marten I tried playing around with those settings. I used "performance wizard" to optimize settings but basically got in this back and forth loop from the program's demands

old marten
#

I have never had any success with the wizard

#

as a rule of thumb 8bit Adam should always always be on

vapid tangle
#

Yes

old marten
#

There is no good reason to have it turned off, mix position should also be FP16 since bF-16 takes up ridiculous ram

vapid tangle
old marten
#

Now that's weird

vapid tangle
#

yeah

#

alternates between what mixed precision setting idk

old marten
#

What repository are you using to train? Are you not using the one from stable diffusion dream booth?

vapid tangle
mellow knot
#

i forgot how often should SD 2.0 be updating

#

was it per week or pe rmonth

old marten
#

Omg

#

U doing CPU training lol

#

Why did you toggle on that option

mellow knot
#

o

#

u must be talking to the other guy

golden nebula
#

I found out about stable diffusion about 2 weeks ago and I am addicted. I think this is the first time I've released my true creative outlet and I just can't stop creating art I see fit in my mind

#

Seeing the pieces I've edited come together brings me so much joy

rancid wren
#

is it possible to do image to image without having to use a prompt?

vapid tangle
#

Anyone know of a collab that is recently working to train SD images?

fervent thunder
#

Hye everyone

#

Im struggling my ass off to get xformers to work on Windows.

#

Anyone have any suggestions?

Collecting xformers
  Using cached xformers-0.0.13.tar.gz (292 kB)
  Preparing metadata (setup.py) ... error
  error: subprocess-exited-with-error

  × python setup.py egg_info did not run successfully.
  │ exit code: 1
  ╰─> [8 lines of output]
      Traceback (most recent call last):
        File "<string>", line 2, in <module>
        File "<pip-setuptools-caller>", line 34, in <module>
        File "C:\Users\Dev\AppData\Local\Temp\pip-install-8yny034i\xformers_48ac54ddc03048448be8cb5876f51fb6\setup.py", line 239, in <module>
          ext_modules=get_extensions(),
        File "C:\Users\Dev\AppData\Local\Temp\pip-install-8yny034i\xformers_48ac54ddc03048448be8cb5876f51fb6\setup.py", line 157, in get_extensions
          raise RuntimeError(
      RuntimeError: CUTLASS submodule not found. Did you forget to run `git submodule update --init --recursive` ?
      [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
error: metadata-generation-failed

× Encountered error while generating package metadata.
╰─> See above for output.

note: This is an issue with the package mentioned above, not pip.
hint: See above for details.
PS C:\stable-diffusion-webui>```
timber radish
left knot
#

someone ftlog help

#

where the does Remacri upscaler go. Its pretrained model is "none", not esrgan, or the other.

#

send halp

safe sapphire
solemn hill
#

What's checkpoint?

chrome spindle
#

Are there any instagram pods where people promote each others AI stuff?

mortal narwhal
#

There are very good midjourney artwork. Instead of taking artist work it is possible to take those images and train on it. What is issue with it

latent widget
#

Hello everyone, I am new to this community and server.

viscid lava
#

Hey! Can anyone explain how to reference a model I made on huggingface into my images on stable diffusion? Feel free to pm

bleak matrix
#

Good morning, everyone!

#

How are we all this beautiful day?

bleak matrix
bleak matrix
latent widget
bleak matrix
#

Yw!

#

What are you up to today?

fickle kindle
#

Good morning sir

fickle kindle
bleak matrix
#

Good morning! I'm Sunny! And I'm in the Eastern US

#

It's 7AM here

fickle kindle
#

Excuse me Where is Eastern US?

bleak matrix
#

It's the right side of the United States, so think New York

fickle kindle
#

yeah I see
Can we private chat?

bleak matrix
#

I don't mind DM's, but I'm actually on the road today, so general chat is best. 🙂

#

You're free to talk about anything here, though!

cursive veldt
#

How do I use a seed from one of 4 images?

bleak matrix
gloomy haven
#

It iterates

cursive veldt
gloomy haven
#

So the 4th pic is probably your first pic seed +3

bleak matrix
#

Just clicking on the button uses that seed.

cursive veldt
bleak matrix
#

Think of it like a starting point, or a branch. You can always go back to it.

cursive veldt
#

I want to do outpainting on just one

bleak matrix
#

So go back to the seed you like with that thread, and do outpainting.

cursive veldt
#

So just continue on all 4?

bleak matrix
#

Yeah

#

You can always type in the number, too, manually

sand wigeon
fervent thunder
#

if it's not your page, then it's probably some sort of scammer trying to set something up :/

vivid plinth
#

Hmmm having problem with 2.1 and 2 models when using automatic webui, results are too shiny or saturated and plastic like.

winter pike
vivid plinth
winter pike
vivid plinth
winter pike
vivid plinth
winter pike
winter pike
#

i never got it working tho, so im using 768x768

sand wigeon
vivid plinth
fervent thunder
winter pike
fervent thunder
winter pike
sand wigeon
fervent thunder
fervent thunder
sand wigeon
#

? What email address?

fervent thunder
# sand wigeon ? What email address?

<a class="inline-flex justify-center rounded-2xl bg-blue-600 p-4 text-base font-semibold text-white hover:bg-blue-500 focus:outline-none focus-visible:outline-2 focus-visible:outline-offset-2 focus-visible:outline-blue-500 active:text-white/70" href="mailto:hi@healthydiffusion.com">Reach Out!</a>

sand wigeon
#

Ah yeah true, ty. I kinda doubt it's real but sure

#

creates protonmail account

fervent thunder
#

something is always better than nothing ;3

fervent thunder
lofty parcel
#

What in Greg's name

fervent thunder
# sand wigeon Ah yeah true, ty. I kinda doubt it's real but sure

<a class="Link Header-businessLink Link--primary" href="https://healthydiffusion.com/" aria-label="Back to OGull.com" title="OGull.com" target="_self"><div style="position: relative;"><div class="flex-container align-items-center">

That one was very weird :P

#

as there should be info from them if you check, but I don't know how open they just give out that stuff

crimson violet
#

Are there any models for writing? Since every time my outputs generate text on a sign or shirt it's always gibberish I'm wondering if there is somewhere I can inpaint some actual writing.

shy juniper
#

Any updates on when the bot is going to get an open source release?

broken gull
rare vector
#

hello guys , am new here and have some doubts, can anyone help

magic widget
#

What are your doubts, @rare vector?

fickle kindle
#

hello guys, I am looking for person who located in UK or USA.
please dm me

magic widget
#

For what, @fickle kindle?

worn patio
#

Does anyone know what settings astria uses ?

fickle kindle
spiral rapids
fickle kindle
#

hello guys, I am looking for person who located in UK or USA.
please dm me

acoustic tulip
#

Hey guys, I'm having trouble standing up 2.1, I keep getting images with a beige background and a slight outline

#

No actual images

#

any suggestions?

#

Also I couldn't get the xformers library downloaded, could this be the issue?

#

Oh, jk, I'll head to tech-support, I think my question is more relevant there

winter pike
acoustic tulip
#

is it the v2-inference-v.yaml that I need to specify?

winter pike
acoustic tulip
rare vector
#

what makes it diffrent

steady geyser
#

Folks can now also follow me on Mastodon if desired since Musk is blocking other social medias... https://cr8r.gg/@Sarath

fickle kindle
#

@steady geyser hi

magic widget
# rare vector is it same as midjourney?

Midjourney seems targeted more at art, where SD seems more general purpose. The main distinguishing feature is that SD is open and reusable pretty much as you please, where the others are all locked behind private walls owned by corporations.

fickle kindle
#

@magic widget bro where are you from?

rare vector
magic widget
#

@fickle kindle Why are you asking people where they're from with no other reason for them to give you that information, then ignoring them when they ask you why?

fickle kindle
#

I am looking for partner to work together

magic widget
#

If you say that up front, I promise you'll get way more bites

magic widget
steady raptor
#

Hello guys, I'm new here, I was wondering if there is a library or a bank of images to reference, I know midjourney does, but I was wondering if SD had something like it

static badge
#

where is 2.1?

#

i that?

winter pike
fickle kindle
fervent thunder
fierce musk
sand wigeon
#

Hey all! FYI:

The Harmonai team in the Stability family is working on audio diffusion and have released a public 24/7 stream of AI generated music. Some aphex twin madness, some chill beats, and a lot of Dubstep/DnB bangers!
https://www.youtube.com/watch?v=uGRLOMf2hSc

YouTube
Harmonai
🎵 Diffusion Radio - 24/7 AI music from Harmonai 🎵

🎵 Introducing Diffusion Radio - A 24/7 livestream of AI-generated music! 🎵

The Harmonai team has been working hard on diffusion models for music generation. We're happy to bring you this stream to share the progress of our experiments!

We definitely appreciate any feedback you have on the music. These models aren't fully trained and are being...

▶ Play video
#

Official announcement coming soon!

limpid ferry
#

if im generating 4 images at once, should i chnage batch count or batch size? (webui)

forest trout
#

Who here is listening to the Diffusion Radio station?

fervent thunder
#

quick q - whats the diff between dream channels on discord and dreamstudio?

#

seems like the same thing to me

winter pike
lunar wasp
#

what would be the best ai model for anime? Im wanting to merge a dreambooth model and anime model together

waxen pasture
untold chasm
#

.

winter pike
#

Anything v3 is the most broad model for anime

lunar wasp
#

thanks

#

I hope my dreambooth model works okay with it

sweet frost
#

okay guys just for clarification- if I´m poor I have to stick to stable diffusion, as in, there is no way to get around the expenses of dalle or midjourney, right?

minor dune
#

Are there any good music generating AI bots, similar to what something like Stable Diffusion is to art?

bleak matrix
#

You can also use the bot here in chat for free

sweet frost
#

I use SD all the time and it´s great, I would just love to use the other programs as well

winter pike
bleak matrix
#

Oh, I misread, sorry

winter pike
#

dalle currently gives you 15 free credits monthly, its very little but you can use them to test them

bleak matrix
#

Too

#

For other AI communities

sweet frost
bleak matrix
#

Np

scarlet tapir
#

How are people using Dreambooth?

bleak matrix
#

Dreambooth is a pretty complicated topic.

sonic carbon
#

can someone run a photo through gigapixel for me

winter pike
#

wait

#

oh oh its too big already

#

my gpu cant handle it

sonic carbon
#

forgot that runs locally

#

im tryna get it to be huuuger

#

i got this far but my next 2x upscale is estimated 4 hours

#

im thinking a hogh batch size

stoic hinge
#

Got bad news, ArtStation is trying to protest against AI Art

#

It is so sad... Only because they do not know anything about it, they start protesting...

thin widget
#

Anyone know if you can import custom models into poseit?

stoic hinge
#

They really need o stop doing that

#

We did nothing wrong...

hasty bramble
#

guys, I need help. why when I write "/dream" and enter my request, I am always asked to write "image_init"? why can't I generate a unique picture, instead they force me to change the one in the gallery?

thin widget
#

AI protests are just getting started 🤷‍♂️

stoic hinge
#

I hope it will not got too far

#

AIs like Stable Diffusion finally allow me to live my dream and they try to destroy it...

#

Only because they think we take their jobs away which is not true

thin widget
#

It'll probably last as long as govts refuse to implement a good universal income system

stoic hinge
#

I see

#

Then we only need to hope for the best

#

Oh, did not see the clock, its very late, see ya all, have a nice dayAE_HeartHug

golden nebula
#

Came up with an Act name the (Digital Ethical Art Rights)

#

DEAR

#

It would be hilarious if many artists raided the SD server

minor hatch
#

Hello

wild steppe
#

Hello

winter pike
#

"/dream prompt:"

hasty bramble
# winter pike i think youre missing "prompt:"

no, I wrote everything correctly, "prompt" did not forget to write, but at the same time the bot makes me write "image_init" every time. if I don't write "image_init" and select a photo from the gallery, then I just won't be able to generate anything

summer patrol
#

Sirs after a couple days of experimenting

#

I just can't get the results I want

#

I'm trying to finetune 1.5 on Pepe the frog to make a meme generator

#

But it just won't work

#

Dreambooth completely does something else

dawn yarrow
#

its slash dream prompt colon

peak orchid
#

lol, I was using BLIP to label my dataset and it kept giving me outputs like this: "sitting on a table with a gun in its hands and a gun in its mouth". WTF was the training dataset used for BLIP?

solar swan
#

Is there a go-to paid Dreambooth/Fine Tune service where one would still have access to low level settings, and still gets the chpt model file at the end? A111+DB extension is great, but running out of memory for 768 SD2.0 fine tuning so wondering what else is out there

vapid tangle
#

Is there a way to train a model besides just adding images? Like can you tell it, this image is bad, this is good, etc

#

im using dreambooth btw

winter pike
verbal violet
fervent thunder
#

what did the pruned version mean?

peak orchid
scarlet tapir
#

Want to make images of celebrities lighting channukah menorah - any suggestions for best practice?

signal dome
#

The survey form Have bug ? I can't submit feedback

patent pollen
#

does any one know how i can make seamless patterns?

astral goblet
#

I just caught a 4080 and having a lot of fun with it

hushed edge
#

Anyone who sells AI art as stock, do you ever profit? What generator do you use?

crimson magnet
neat bronze
clever sand
solar wyvern
#

can someone link me to a tool that gives a caption when submitting an image? I remember using this but cant remember what its called

astral goblet
# neat bronze Nice! How many it/s?

how do i get a number that makes sense for that? different jobs give different amounts, and it feels like i'm not reving it full for some reason

patent pollen
astral goblet
#

it's not always perfect but running the result through img2img with tiling on as well helps

patent pollen
astral goblet
#

when was the last time you updated it ? i don't remember a time when a1111's webui didn't have it, but i've only been using it a couple months now

patent pollen
#

underwidth and height is cfg scale

astral goblet
#

no batch count, no batch size?

patent pollen
#

updated? its a program and not a website?

astral goblet
#

oh. mb. you're not using a1111's software. no but that's how to do it. You run your own version of SD and configure it to tile

patent pollen
#

where can i find this please? thank you

astral goblet
patent pollen
#

thank you

astral goblet
old marten
#

I think it used to be that way, but now that is obsolete. You can train with the lighter checkpoints too just the same

#

Unless someone can prove me wrong

compact anvil
#

Hello.
Can someone explain me what ENSD means?

I see it sometimes when people share settings and prompts.

I've tried googling it, but no luck.

crimson magnet
old marten
#

Probably luck of the draw, although maybe there is a difference. I have memory issues with using 1.5 so I stick to 1.4, although I did try 1.5 in runpod with more VRAM and saw no difference in the output.

#

Also, I cannot run the 7 GB checkpoints with my own 12 GB VRAM card so even if there was a difference, I couldn't even use it without paying up 😂

astral goblet
sonic sparrow
#

Can anyone give me an idea of what prompts I would use to make a photograph of me as gandalf? I've already trained my face.

sonic carbon
torn plaza
#

will locally run SD ever be as good as midjourney is now ?

fervent thunder
#

i fucking love yoyu

lofty parcel
#

Google developed a 20B parameter model that was able to consistently generate coherent text
In practice I believe that requires about 40 GB of VRAM

#

Far beyond current consumer GPUs

golden nebula
#

does storage drive speed matter?

wide juniper
#

What does {{}} do? I thought (()) is emphasis

fervent thunder
#

ill eat it

sand wigeon
# wide juniper What does {{}} do? I thought (()) is emphasis

It does nothing (or not something reliable) on our official StaleDiffusion stuff, e.g. beta.dreamstudio.ai - if you want emphasis you can do multiprompting ^^ Reliable and transparent. Check the guides there.

I'm actually trying to figure out where these practices came from. Any help would be appreciated!
Was it a myth that took a life of its own, an artifact from how text was encoded/tokenised, or whether there's some repos/webuis who do depend on this formatting?

Thanks in advance!

wide juniper
#

As in ^emphasized phease^?

sand wigeon
#

yeah that kind of stuff. I know what it's supposed to do, but I'm not sure how the practice came about

#

And people think it works for DS too, but it doesn't reliably. There's hypotheses on it effectively using a bug/artifact from tokenisation, but i havent gotten confirmation

robust coral
# wide juniper What does {{}} do? I thought (()) is emphasis

a very quick rundown:
{keyword} is a syntax that the Novel AI platform uses to emphasize tokens / keywords in a prompt. but it's also used for example by "dynamic prompts", an extension for the automatic1111 webui that you can run on your own computer.

also automatic1111 uses this syntax (keyword) to emphasize a keyword and [keyword] to de-emphasize a token. there are more, but these are the most common

sand wigeon
#

MPV, thanks Mass!

#

Do you know context of !!emphasis!! as well?

robust coral
sand wigeon
#

also send me dat email ❤️

#

"GUITard" wtf lol

robust coral
sand wigeon
#

haha ok understandable, jsut to make sure it's not unnecessary hesitation - good luck!

robust coral
# sand wigeon Do you know context of !!emphasis!! as well?

so afaik the !!!!keyword!!!! syntax was used because it was also supposedly emphasizing a word as well. it was a practice also used on midjourney.

I can't say exactly if it still works. I used it in the Stable Diffusion Beta (before the official release) on this Discord and I think it did something - but whether it was making something stronger or just feeding random tokens to the prompt parser I can't say. you find it mostly in older prompts. most have since then shifted to brackets and repeating tokens.

strange flume
#

Sup

sand wigeon
strange flume
#

What do you say if i buy a new laptop for prompting

sand wigeon
# strange flume Sup

Just doing some work while vibing on Harmonai radio https://www.youtube.com/watch?v=uGRLOMf2hSc

YouTube
Harmonai
🎵 Diffusion Radio - 24/7 AI music from Harmonai 🎵

🎵 Introducing Diffusion Radio - A 24/7 livestream of AI-generated music! 🎵

The Harmonai team has been working hard on diffusion models for music generation. We're happy to bring you this stream to share the progress of our experiments!

We definitely appreciate any feedback you have on the music. These models aren't fully trained and are being...

▶ Play video
strange flume
#

What's that

#

Again

sand wigeon
#

Stability has a 'family' of research orgs. Harmonai focuses on Audio diffusion. This is all fully AI generated audio from pure noise

#

Mix of their accumulated library of songs and newly generated ones

robust coral
sand wigeon
#

They'd super appreciate it! Especially if you have audio aligned text/annotation, e.g production stems project files (i dont make music, dunno if thats the right terminology)

robust coral
left knot
brazen fox
sand wigeon
brazen fox
bright cradle
#

why interaction failed

sand wigeon
# brazen fox Will there be something like that for SD as well? I have a ton of photos I'd be ...

https://images.cv/ seems a good place to donate images to research datasets

images.cv | Image datasets for computer vision and machine learning

images.cv | Image datasets for computer vision | Object classification dataset | Machine earning and deep learning | Big selection of image datasets to build your custom image dataset | Use our built-in tools for dataset pre-processing: image color, data split, image size, and data augmentation

brazen fox
tacit trout
#

Does anyone know of any extensions or scripts that “remaster” an image during upscale like in Midjourney, where it adds detains in the upscale process rather then upscaling only what’s there.

robust coral
tacit trout
robust coral
tacit trout
barren bridge
#

What does 'batch size' do? whats the difference between generating 10 images with batch count vs 10 with 5 count and 2 size? is it changing the seed for the second lot of 5 or something similar?

#

sorry for the super noob question

bleak matrix
#

Good morning, everyone! I hope everyone is having a fantastic Tuesday

forest apex
#

yup, good day, scored a new rpi within a minute of it being in stock, lol

broken gull
bleak matrix
gentle ruin
#

Can anyone help me with some of my basketball photos

north stirrup
grizzled harness
#

I noticed the running python process fired off a DLL under run32 at some point ... any ideas what this is?

ocean mango
broken gull
#

Now

gentle ruin
#

Idk how to do it myself

north stirrup
#

What kind of “edits” are you looking for?

hushed tangle
#

where is 2.0 depth yaml?

bleak matrix
#

Are you trying to do inpainting/outpainting?

hot dragon
#

is automatic1111 compatible with 2.1?

strong moat
#

Hey guys, i am trying to make the same image that i made with Dreambot in Automatic1111, but it doesnt come out the same. Any idea why that could be? I am using the same seed, prompt, sampler etc.

wicked hollow
#

Automatic will be different by default. There are a number of reasons why this is the case, but it's likely you'll never be able to get a 1:1 from things created elsewhere.

north stirrup
strong moat
wicked hollow
strong moat
#

Hm, I see. So if I use vanilla StableDiffusion instead I should get the same image,right?

wicked hollow
strong moat
wicked hollow
manic igloo
#

Hey guys how do I remove blur and depth of field, I tried blur: -1 but that just changed the image that is generated completely and gave me something completely unrelated to what I wrote. What am Idoing wrong?

north stirrup
#

remember that diffusers don't know about composition, they only know about keywords

dull hare
#

i have a question is there anyway to make a vector art ?

north stirrup
timid gate
feral kelp
#

I am blown away by how much adding "ugly" to negative prompts has improves some of my prompts. I really recommend giving it a try

old marten
#

I stopped using ugly in my negatives just because it's very subjective

#

Anything that is open ended and subjective is a no-no for me for negatives

dull hare
rough ridge
#

I must have something wrong with me, I'm not really an artist, but I kinda love seeing artists seeth over AI art

north stirrup
rough ridge
#

I didn't say wanting other to suffer

north stirrup
#

well, aren't they suffering?

rough ridge
#

That's subjective

#

How many of these people were making any kinda of actual cash flow, and how much was that affected by AI art

north stirrup
#

"enough"

rough ridge
#

Most ppl can't seem to agree on it tho, I see people whining and complaining that it's going to take jobs, then the other half of the time they say it looks awful and you can tell right away it's AI

#

So which is it lol

feral kelp
north stirrup
rough ridge
#

Why should ppl care about art

north stirrup
#

I mean, we want to do art using SD

rough ridge
#

But you can still do that tho

north stirrup
#

Sure, but it might be harder to make a living from it, and I can't blame people for being scared about it being harder to make a living doing art

rough ridge
#

Im not blaming ppl for being scared

#

I just hate all the "stop using AI art" gatekeepint

north stirrup
rough ridge
#

It's like veganism, the meat is still on the shelf if you choose not to buy it

north stirrup
#

just be clear you're against gatekeeping and not antipathic to the artists themselves

#

that's certainly where I'd draw a line

copper pivot
#

Hey guys, suggestions on which models to use for scenery/scenic shots?

rough ridge
#

Well I think I kinda am tho, like I understand why it's like stressful ig, I just don't really care

copper pivot
#

Been mostly creating characters

rough ridge
#

Are more people not enjoying art now then they were years ago?

dull hare
#

guys is SD 2.1 has a vae file , should i download that 355MB file in the vae folder ?

#

first time using SD

north stirrup
#

I think we take art more for granted, and SD doesn't exactly help there

rough ridge
#

Also AI arts been around for a few years now

#

Why are ppl just vocalizing how upset they are now

north stirrup
#

Eh, it's only gotten particularly good now, so artists feel threatened

dull hare
north stirrup
#

because the people who would previously commission art might try to get away just using cheap generated crap

rough ridge
#

It's gonna push the boundaries of art imo, like ppl should just b excited

north stirrup
magic widget
#

Every tool is an encoding of the toolmaker's skill. Every creation is an extension of what came before. Nobody has the right to own an idea or a pattern because they didn't create it in a void, they created it from what came before.

north stirrup
#

I do agree with the "artists' side" that there's a lot of danger and a lot of people doing shitty stuff

#

I just don't think that's inherent to diffusers

magic widget
#

Training a model on public data and keeping it private is bullshit though. Intellectual property is a scam.

dull hare
#

@fervent thunder is there anyway to add a vector section to the art collection that would be awesome 🙂

north stirrup
#

and there should be more exceptions carved out for the public interest

magic widget
#

Copyright law doesn't protect small creators, @north stirrup. It protects Disney.

#

Large companies steal from small creators 24/7.

#

The people who would argue that copyright protects us are the same ones hollering about their copyrights being infringed.

north stirrup
magic widget
#

It's just not true. You can't even take action without a lawyer. How many small creators can afford lawyers?

#

How often do you see news organizations stealing videos from people on social media?

#

Copyright law doesn't do shit for average people, it's only for corporations.

#

But even if it did, it's unethical by design

#

Nobody owns an idea. Intellectual property is the most obscene form of an already ugly idea.

north stirrup
#

I think we need much stronger law for those "asymmetric" cases and weaker law for things like fair use and fanart

magic widget
#

Companies have governments imprison people for violating copyright.

#

Copyright is not our friend. That's something artists really need to understand.

#

It does nothing to protect us and everything to prohibit us.

old marten
#

Yeah, I'm already prepared for uncle Sam to drop a massive Dookie with ridiculous laws prohibiting AI

#

At the end of the day, it's all about the money baby and AI isn't making money for Sam

carmine topaz
#

hello

#

how come this is free to use?

old marten
#

And because AI and art should be free.

carmine topaz
old marten
magic widget
#

People aren't crybabies, they just don't understand.

old marten
magic widget
#

What I'm saying is, it does nobody any good to be a dismissive ass hole to them.

old marten
#

Well crybabies is kind of the wrong word to use. I mean SD can get sued so they're trying to do what they can to avoid it.

magic widget
#

There's definitely a level of hypocrisy there, because as I said before everything comes from something. But we're all programmed to think in silly ways that benefit rich people.

old marten
#

But even mid journey, lets you use celebrities I mean come on

north stirrup
old marten
#

And AI isn't even copy pasting lol

north stirrup
magic widget
#

I think it's crucial that we distinguish between law and ethics, which is an opportunity SD uniquely gives us because of its open model.

north stirrup
#

It might or might not be illegal depending on jurisdiction

#

if you don't distribute it … it's probably unethical but no one will care too much

old marten
#

Photoshop fakes and deep fakes were already there before AI

#

Not to mention, it falls under the first amendment, which is freedom of speech. Censoring someone from making fun of someone is a violation of that amendment.

#

Yeah, it's unethical If you try to pass it is real but come on. Anyone would be a fool not to question a deep fake these days.

untold chasm
jovial portal
#

Yeah but defamation isn't legal either and that's up to interpretation of the individual being represented usually.
Not that I really know anything law-wise about anything really

old marten
#

So because certain parts of the world don't know the deep fakes exist We have to ban all AI from making deep fakes.

#

No way.

untold chasm
#

Sorry did I say that?

#

You instantly lost my interest when you put words into my mouth lmao.

astral goblet
#

sounds like a bad idea

#

maybe the law should embrace ai

old marten
#

Yep that's exactly my point. I guess we'll see what happens.

astral goblet
#

jyeah its gonna get neat

north stirrup
astral goblet
#

real humans will develop our own secret language vernacular that machines can't emulate , or maybe they can and our language patterns will just have to dissolve into chaos. its gonna be so cool what comes next

north stirrup
#

the point is getting fewer people to do [insert thing here]

magic widget
#

Sometimes you can mitigate something with a ban, but you'll also be creating a black market with many unforeseen consequences.

astral goblet
#

same with the war on drugs. people wanted to do drugs and legal businesses couldn't assist them. so criminal empires thrived on that demand

magic widget
astral goblet
#

if ai gets banned, criminals WILL use it

astral goblet
magic widget
#

I think they're very closely correlated though, because power flows between corporations and government.

north stirrup
astral goblet
#

they'll use it either way anyways. a lot of high frequency stock trading with insider knowledge is done through machine learned algorithms and proprietary models

fervent thunder
#

Does anybody have any idea about how much electricity stable diffusion consumes to generate a single 512x512 img ?
Dummy me generated more 2000 images + about 400+ videos with deforum and now i am worrying about my electricity billcatwhaaa

north stirrup
#

(FWIW I'm dual-citizen Canada and US)

magic widget
#

We have cultural problems that banning things won't fix.

astral goblet
#

making sure ai is accessible by all ASAP is going to sting a bit, but the future of democratic society depends on it imo

magic widget
#

Letting millions of people go hungry or without shelter while Elon Musk blows $40bn to prove he's a worthless dipshit requires an extremely fucked up society

astral goblet
#

allowing any ruling class to control this tech seems really bad

magic widget
#

It should infuriate people that SD is the only open large model.

#

These corporations are using public data with algorithms developed at public institutions.

astral goblet
#

openai just put out that cool new 3d one today

magic widget
#

Oh?

astral goblet
#

first glance it looks like the model and code are available

magic widget
#

Ya, looking at it

astral goblet
#

its basic but cool

#

i personally am of the mind that the models we're using today are all going to be obsolete in a year, so what's out today isn't worth stressing over. the overall vision though, everyone should be moving towards an open model, but more importantly awareness of the tech is huge. the models being developed and deployed in secret are the worst offenders

#

eventually, training base models is going to be easy enough on a cellphone sized device, so as long as we carry momentum and awareness growth, the peopleshould prevail

opal root
#

H

tiny gazelle
#

Is it possible to take the image of my avatar and add an AI-generated santa hat with Stable Diffusion?

dawn yarrow
#

you want image to image

#

so yes

tiny gazelle
#

Oh, can't share images in this channel. Well it was just 'me' with some difference mustaches.

#

Max Cfg Scale just made it... somewhat gorey.

compact anvil
bleak matrix
astral goblet
golden sentinel
#

Recommended method to fix a wonky eye? I'm finally generating some good pictures, but definitely need to re-do my dreambooth model with better photos. It keeps giving me one wonky eyeball due to light in one of the pictures.

compact anvil
astral goblet
compact anvil
#

Ok then.
Thanks for the help :)

coral wind
#

If you drive to school, do you ever get pulled over

steep path
#

how are embeddings actually used after the .pt file is placed to embeddings folder? using webui

fervent thunder
#

use the name of the embedding as a word in your prompt

placid wing
#

if been downloading some new models but can't find the yaml files for them anywhere could someone help? they are the Van-Gogh-Style-Ivngvncnt-v2.ckpt and knollingcase-v1.ckpt. thanks in advance

astral goblet
#

those are embeddings if i'm not mistaken

astral goblet
placid wing
#

embeddings are .pt files to my knowledge. ckpt are model right?

astral goblet
#

my bad i thought knolling was an embedding but the first version wasn't

#

the one i use is

#

most of the 1.0 models don't need yamls, at least on webui, in my experience. I only have ever had to do that for 2.x models

copper pivot
#

Hey guys I'm trying to generate scenery for my brother's book, which is the best model for scenic art and which is the best sampling method?

viscid rune
#

Hi guys, I had a question. Let's say I have around 10 different sci fi chairs and I want stable diffusion to make infinite variations of chairs using those 10 different sci fi chairs, like mutations. Is it possible?

mossy fulcrum
#

is it just me or has the quality of SD 1.4 on HF significantly dipped

hearty vault
#

Hey guys, can't find an answer to this question.. with deforum, can you please tell me if it's possible to connect the ending of an animation up to the beginning of another animation? like, by using the first frame of the second animation as a final seed or something?

Surely its possible to make an intro to a video, and have it seamlessly connect up to the rest?

fierce grail
#

Does anyone have any guesses as to why my trained model seems to have “weak” weights?
Specifically, I used Dreambooth to train images of my cat as Fnnck404. But now if I ask for an image of “Fnnck404 as a lion, trending on artstation, etc”, I only get lions that bear NO resemblance to my cat. I even tried weighting the prompt like [Fnnck404:9, lion:1] as I’d seen in a tutorial but that had no noticeable affect.
I know the model works, though, because Fnnck404 on it’s own usually usually produces a halfway decent image.

cobalt cape
#

yoo

#

habe you seen the OpenAI text to 3d model?

fervent thunder
slender badge
slender badge
sonic notch
#

Hello! I'm using the pharmapsychotic Colab to test some Analog Images, do you know how to add negative prompts on it? It only have one prompt field to put the text.

patent pollen
#

Anyone have or know of any tutorial to get realistic renders? Everything I try it looks like a damn alien.

I even copy prompts off lexica and get jibberish

unreal eagle
#

im using sd 2.1 with 768 model but all the images i make arent that clear they all look unfocused and bad, idk why

feral kelp
patent pollen
feral kelp
unreal eagle
#

i come from midjourney and the image quality outputted there is much higher compared to the 2.1 sd model....any way i can improve it? cant tolerate midjourney anymore

patent pollen
patent pollen
feral kelp
# patent pollen Can you give me some examples of negative prompts Thank you

it really just depends on what you're going for. like if your outputs aren't realistic enough you can try adding terms like "illustration, cartoon, drawing" to negative prompts. I haven't done much in the way of patterns either, so sadly can't help much there either. Hopefully someone can chime in who has more experience with these kinds of styles!

#

definitely do lots of tinkering though, and try to find others who have posted workflows for styles similar to what you're going for, to get a running start

patent pollen
#

I'm not even sure if I have everything installed correctly lol

Anyway to confirm that?

long marten
#

I sure do know lots of communist artists

#

The Karla Ortiz go fund me has raised almost 200k to lobby the government against SD

pliant locust
#

hi, I talked to someone earlier on this server about a thing called distilled stable diffusion, and was shown this tweet about it: https://twitter.com/EMostaque/status/1598131202044866560
I just wanted to know if there was any new information about it because I cant find any aside from that tweet, and this paper about it: https://arxiv.org/pdf/2202.00512.pdf
Is there any update on its release, or its going to be released publicly at all? and if so, are there any special steps you need to take to run it? I have a gaming laptop that currently takes ~5min for one 512x512 img on the Basu Jindal version, so would I be able to use this for a ~15s img generate time?

unreal eagle
#

hey im trying to use deforum diffusion

#

and when i get to the last step it says the file not found? on the colab? why is this?

#

did i break something

#

i tried to join the deforum server but i cant even get past the main page ....broken verificator

unreal eagle
#

nvm figured it out myself

signal marten
#

Anyone got any good resources they would recommend for getting better results? I see a lot of people's stuff just looks way better, so I am wondering how I can get similar results.

unreal eagle
signal marten
unreal eagle
signal marten
#

My results typically all look like melted wax XD

unreal eagle
#

which one are u using??

#

& what settings/prompt

signal marten
#

Just using 1.5 SD model. Idk I've tried a bunch of different prompts and settings.

unreal eagle
rough ridge
#

i feel quite dumb but like, how do i use img to img lol
ive got stable diffusion download, 1.5 and 2.1
but like i drop an image in there, hit generate, and it looks nothing like my image

#

feel like im missing somethin

unreal eagle
#

mess around with the denoising

#

more denoising strength = less like your image i think

rough ridge
#

i downloaded the 2.1 and, launched webui, and it took a while downloading, and thats working

#

but then i go and download both the 1.5 and 1.5 pruned versions, put them in the same folder, and when i run webui there wasnt a long download, and when i click on either 1.5 version in webui it just doesnt load

#

but it still shows up

#

in the dropdown list

#

LOL

#

nvm

unreal eagle
rough ridge
#

im impatient

#

problem solved, it was actually loading lol

unreal eagle
#

im trying to use this

#

and im getting an error

#

on step 12 ``` 0%| | 0/6 [00:00<?, ?it/s]

TypeError Traceback (most recent call last)

<ipython-input-13-f50195ec4328> in <module>
16
17 with torch.autocast("cuda"):
---> 18 merged = inpaint(
19 prompt=prompt,
20 init_image=new,

/usr/local/lib/python3.8/dist-packages/torch/autograd/grad_mode.py in decorate_context(*args, **kwargs)
25 def decorate_context(*args, **kwargs):
26 with self.clone():
---> 27 return func(*args, **kwargs)
28 return cast(F, decorate_context)
29

TypeError: call() got an unexpected keyword argument 'init_image'```

atomic zephyr
#

Are there any strong reasons for using a CLIP model as the text encoder for SD 2.0? Are there plans to use a pure text based model (one of the GPT models) as the text encoder, like the Imagen paper did?

lunar shadow
#

Hey can Automatic1111 run on CPU only, i kinda have a 1060 6GB but it always cuda errors on me even set COMMANDLINE_ARGS=--opt-split-attention --always-batch-cond-uncond --precision full --no-half --no-half-vae --medvram

i have tried using TorchCPU version and set COMMANDLINE_ARGS=--skip-torch-cuda-test --precision full --no-half --use-cpu all in another directoy and still cuda errors but a very long one like it expected cuda, so i think the cpu part isnt taking?

Feel free to @ ME or DM if you have it working with same or worse specs lol 🙂 or got cpu working too.

unreal cargo
#

You might have better luck without the "--precision full" alicatHm2 and you could always try --lowvram instead

#

precision full is important for 1660 cards

#

could also try:

set OPTIMIZED_TURBO=true```
carmine topaz
#

Struggling to understand why overall quality is so bad even with trying different adjustements compared to different apps that use forks

reef epoch
#

since we wont get anything new until after the end of the year

#

I'd like to announce that i am releasing and open sourcing Stable Diffusion 2.1

#

here ya go

#

go have fun

lunar shadow
#

Ali, both medvram and lowvram failed cuda ran out memory. i really wish the cpu option worked on Automatic1111 and didnt spit cuda errors

carmine topaz
#

anyone can help share some light onto why the resolution is so god damn awful

#

for most of what i m generating in dreambot pages

#

i tried different RFG and steps

silent parcel
#

hey hi everyone! am new to this server

rough ridge
#

is midjounrney just the best for making characters?

fervent thunder
#

How do i actually use the damn thing :/

rough ridge
#

u gonna need to be more specific

fervent thunder
#

How do i create images

#

With text

#

What link do i go to

carmine fulcrum
#

then read the section "Automatic Installation on Windows"

median matrix
#

hey, are gens using img2img prompts that have multiple celebrities names with different weights, to blend their features allowed to be shared?

lilac reef
#

Anyone do embeddings using captions?

knotty raptor
#

hi frens

wide juniper
#

How can i get pictures of modern train stations and train interiors? Train gives me the old, traditional trains, and subway gives me the fast food chains

knotty raptor
carmine topaz
#

hey guys

#

how do you change AR

lilac reef
#

We do not use that term with SD

wide juniper
#

is there a list of artists that work with prompts? I'm not sure if 'hayao' has been working

robust coral
vestal raven
#

I'm testing Dream studio's SD 2.1 and also set it up locally. I'm observing difference in quality in both models while every parameter is same. What could be the reason for better results provided by Dream studio. Does anyone can guide me?

radiant bronze
true halo
#

Hello, is anyone her good in imitating a Newsreporter Voice? If so, please contact me private

vestal raven
fervent thunder
#

i've started using the automatic1111 colab because of how much faster it is than local
i have 1 google account with the 100 GB drive upgrade and like 4 more accounts that the first one has shared the stable diffusion folder to
everything gets stored on the 100 GB drive

#

when colab kicks me off the gpus on one account, i can switch to another

dapper ferry
#

omg my C: is out of disk space. Again!!!!!

#

its one of those days where i have to hunt down the culprit

signal marten
steel mist
#

how do i train my own stable diffusion mod

#

with my own selection of images

silent parcel
#

hello

steel mist
#

hello

fervent thunder
upbeat tangle
#

so when is the next sd model going to come out ?

warm pasture
#

Is there any easy way to get in touch with a dev to ask a quick question about V2?

wide juniper
#

What’s a negative prompt that y’all swear by?

fervent thunder
#

lowres, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username

wide juniper
#

So bad hands actually work? Doesn’t seem to make a difference

manic igloo
#

What is the model that creates the most realistic humans to the point it's non distinguishable? Analog diffusion is great but I can't seem the remove depth the field and smooth skin which makes it a little bit filtery

remote osprey
#

Hello, i've a little problem with the SD installation, someone is free to helping me pls ? 🙂 thanks

lofty jungle
#

Hey everyone, I'm planning to develop an app which uses Stable Diffusion. I've been searching about what should I use in order to run Stable Diffusion. Some of the similar projects use AWS and some of them use DreamStudio. According my researches, DreamStudio is ready to start but it is more expensive. In the other hand AWS is cheaper but I should spend decent amount of time in order to learn AWS and implement it in my app. I'm confused right now. So I want to hear from you. Which one should I use? Or if you have any alternative I'm also open to alternatives.

marble swallow
#

Hey guys! I was wondering, can this AI make textures that can be repeated?

#

As in the texture picture made in such a way that can be copied and placed against it on x/y axes. Can this do a picture like that?

vital osprey
#

are there any actual artist here?

#

or is there

#

fuck english

brave basalt
#

Good day, anybody have used spectrograms from another softwares like Audacity or Adobe Audition and used in img2img for generate similar music? I have tried but this doesn't work for me.

gusty oasis
#

Did we get any update on distilled diffusion

manic igloo
north stirrup
brave basalt
#

Is there a plugin for batch images that resize every single image according to the input size image?

fickle kindle
#

@manic igloo hi

frank moon
#

If I use IMG to IMG with a photo is there a way to use/get the look of rendered image on another photo?

summer nexus
#

hey fellas! how can I use batch img2img in automatic1111 in a way that it uses the prompt and seed of each original image?

summer nexus
plucky basin
#

where do i go to ask questions?

north stirrup
plucky basin
#

for the text to image generator

north stirrup
frank moon
#

any good pre trained aesthetic-embeddings for automatic1111?

cerulean rivet
#

Does anyone know where I can download regularization images for objects? I am specifically looking for a guitar.

opal root
#

hello

signal dome
#

do we have more depth 2.0 models

#

other then sd official

#

i am trynna load this up more models in blender

static wadi
#

has SD cought up with midjourney yet 🤔

north stirrup
agile tinsel
#

Has anyone finetuned an embedding/TI on top of the depth-to-image model? is there any modifications to the colab that i need to do .-.

rough ridge
#

what kinda prompts should i use to make the ai like, recognize two different peoples in a piece

#

like for example im trying to do A group of dwarves wearing black uniforms and wielding muskets, Fighting a group of Agitated chimpanzee's , in a large forest, Mårten Eskil Winge, epic painting, clear faces, correct proportions

#

Blurred, Disfigured, overlapping, mixed faces for negative prompts

#

but i get a lot of dwarves with chimp heads and vice versa

peak orchid
#

Well Karla Ortiz just got Kickstarter to ban Unstable Diffusion for "unresolved IP issues". So I guess no one is allowed to crowd fund diffusion models on the site anymore

fickle kindle
#

Hello Everyone I am looking for long term partner to work together.
stable income monthly.
If anyone intersted , please dm me.❤️ ❤️❤️❤️

peak orchid
#

How can we as a community hope to work together creating new large models, if the crowd funding and payment sites ban us?

sudden isle
peak orchid
#

Everyone should message Kickstarter and calmy explain that they made the wrong decision ban AI models from the services: support@kickstarter.com Use OpenAI's ChatGPT to help you organize your words if you suck at writing messages like this. Also make sure to tag Kickstarter on social media!

valid river
#

Lund University - Research AI in Design Process

I have a survey with some questions regarding AI art generators used for design. I don't know if I'm allowed to post links here so if you want to contribute to the research please DM and I'll send a link to the survey!

fervent thunder
#

So, in training models with Dreambooth, does it make sense that training with more high quality images would make for a better model?

signal nest
#

Could a diffusion model be trained on video (instead of labelled images) i.e. on a task like: "Predict a feasible futures N seconds into the future from this frame" - with some state vector associated with that then being translated to a language model

serene hare
#

Kickstarter hasn't banned AI models, they just require at this time that data sets are sourced in a way where artists know and consent to being in that dataset.

peak orchid
serene hare
peak orchid
serene hare
#

I think that it's perfectly reasonable that a platform that very much relies on independent IP from small creators to sustain itself to take a stance that favors consensual datasets. Especially when there is so much negative press whirling around about this.

sudden isle
#

Kickstarter can't reverse this position without losing face so I think it's just best to move on

peak orchid
neat bronze
grizzled harness
#

google doesn't notify or source any of the data it uses

#

it crawls the internet for everything

#

it'll literally serve you the exact images from its own site

#

it takes all that info to build its own models for search, ads, building profiles about you

#

uses your browsing data, or data about you online, to feed its algorithms. Did you consent to that?

#

Does anyone care? No.

copper pivot
#

The AI really doesn't understand cigars

jade garnet
quick narwhal
#

RIP unstable diffusion campaign

#

That’s why I wasn’t a fan of the opt out thing @wise stratus was flexing, can’t give these people an inch on fair use tbh, they’ll take a mile and still have shit to say

agile tinsel
#

I feel like the kickstarter statement will just discourage even more the sharing of training data / process in the creation of embeddings/dreambooths/hypernetworks

#

I think the best solution isn't to ban artist styles / images since thats p much impossible, but to create a legitimate centralized authority that can properly distribute stylistic rights

wise stratus
#

I am surprised they didn’t just go straight to crypto funding tbh seems even DAO sensible for this one

sudden isle
#

Idk if they can raise 25k there, they can raise 25k anywhere so this won't stop Unstable Diffusion in any case

wise stratus
#

NSFW finds it hard to stick with conventional systems

#

Due to liability concerns

jade garnet
#

it's not just about nsfw, though....
The artstation protest surely wasn't

wise stratus
#

Yeah but excuses people give

sudden isle
#

I'm also pretty sure image generation as a whole will be solved next year no matter what

wise stratus
#

Doesn’t take too many complainers

serene hare
#

I mean, kickstarter would look pretty bad if UD ended up blowing up and got attached to unethical NSFW generation.

peak orchid
jade garnet
#

I think people should be doing their best to counter all of the artist voices that are continually attacking AI art

serene hare
#

I'm frankly surprised that kickstarter didn't just ban it based on their no porn policy.

wise stratus
#

I think just like listen to valid concerns and ignore the rest

jade garnet
quick narwhal
wise stratus
#

Most folk are just scared but tech is inevitable. That’s why you get silly things like the lobbyist go fund me

#

If that worked we would just get more lobbyists

#

But the legislation they are pushing for centralised creative control

jade garnet
wise stratus
#

To big media

jade garnet
sudden isle
#

Yeah I just think AI generation is inevitable so any meltdowns now ultimately do not effect the outcome

jade garnet
#

It would harm their own interests as individual artists, yes. But it would also cripple AI art.

north stirrup
wise stratus
#

We have spent $$s on legal prep checking stuff just to make sure

#

Where $$s is dos commas

north stirrup
#

but inpainted/composited works can still qualify, probably?

quick narwhal
#

I try to be sympathetic to the artists about to lose their jobs. Though sometimes a few go out of their way to come at me and it kinda brings me joy they’re so mad.

wise stratus
#

Nah it’s sad people are mad

jade garnet
wise stratus
#

Life is too short to engage with those that don’t want to engage

#

Just be positive and make cool things

sudden isle
#

Think about this, if anyone can generate an infinite amount of compelling IP in all mediums like games and movies in the privacy of their local PCs, what use will copyright be?

quick narwhal
north stirrup
#

There's a responsibility on our end to avoid exploitative behaviour, and we can be sympathetic to artists

#

But we don't have to take bullshit

wise stratus
#

When you can create anything it’s something that has value as we are inherently pro social narrative driven as a species

#

Folk don’t realise this is the dawn of seamless visual communication for humanity which is an amazing step forward

#

They’d rather keep their monopoly on visual communication

quick narwhal
jade garnet
serene hare
#

I really think it's likely that copyright is going to be revised in the advent of this new tech, not all the techniques that are popular right now are on an equal playing field ethically. Hopefully we'll be able to thread the needle between making room for innovation and respecting peoples ability to produce creative works.

jade garnet
#

The concept art association is still going around talking about how stability is data laundering and only doing all this to make cash off a service and etc

wise stratus
#

You know it’s like lower classes shouldn’t be allowed to learn to read as writing is the privelege of the elite and they could use it for bad stuff. They would also learn to copy the style of the elite which should be banned

peak orchid
quick narwhal
jade garnet
wise stratus
#

It’s very anti American and free speech

#

Tbh

quick narwhal
sudden isle
#

If you project the pace of AI forward 5 years you'll realize how over it is

jade garnet
sudden isle
#

Those Ws are rare and fleeting

solar sleet
#

I feel like some pr focusing on how disabled people can use AI to create art that they could not before (sometimes even after they lost the ability to do so) would go a long way to show amazing applications of the tech that would take a lot of wind out of the sails of opponents

quick narwhal
jade garnet
#

the state at August was a free reign for AI art on any and every art aggregation website

#

The state we have now...

little plaza
bleak matrix
#

Good evening, everyone. How are we all this evening?

peak orchid
north stirrup
#

The big PR win would be policing our own and disowning the “lol I’m going to train on your art, get mad” assholes

jade garnet
jade garnet
little plaza
sudden isle
#

Back in April when DALL-E 2 released there was hella meltdowns lol

little plaza
#

for normal people

quick narwhal
jade garnet
little plaza
jade garnet
little plaza
#

ya lemme find

jade garnet
#

I didn't see it sry

north stirrup
little plaza
jade garnet
#

Don't sleep on it

north stirrup
#

I’m pretty sure part of the MJ “secret sauce” is training heavily on art

north stirrup
#

Expensive/difficult to make

little plaza
jade garnet
#

Thx

solar sleet
#

I don't think there is a chance that they get hugging face to stop hosting ai art models, that would be a horrible business decision for hugging face

jade garnet
#

???

#

What money is HF making off of hosting rando models

quick narwhal
sudden isle
#

Doesn't hf charge to set up ml infrastructure?

jade garnet
#

innovation is very reversible and the people that think progress is a linear unstoppable force have not reflected on history adequately

quick narwhal
sudden isle
#

I hear it but at the same time I think AI is different

solar sleet
#

A machine learning site banning ai art models would be the worst PR for their userbase, it would prob kill the site

jade garnet
#

I thought it was free or something

boreal chasm
#

How does a new person get started with this

jade garnet
sudden isle
#

Entreprises

jade garnet
#

Companies don't need anime-booba-v3.ckpt lol

sudden isle
#

The anime booba companies are on the horizon though

jade garnet
#

frankly I only expect it to not happen if stability is economically partnered with HF

solar sleet
jade garnet
jade garnet
#

People here were defending kickstarter's decision

#

What negative PR?

#

The neg PR will come if they don't

#

Not if they do

sudden isle
solar sleet
#

The negative PR comes from a side they don't care about

jade garnet
solar sleet
#

Why tf would hugging face care about artists and their normie sympathizers?

jade garnet
#

aka machine learning engineers that are defecting

solar sleet
peak orchid
jade garnet
#

I am very aware of what HF is in a general sense (I use their other tech for things)

peak orchid
jade garnet
#

That is the difference

jade garnet
solar sleet
#

Why would that make the difference? People are raging against all Ai art, not just Stable. Most artists against AI barely even know what AIs exist.

jade garnet
peak orchid
jade garnet
#

the drama was much much smaller in the days of just MJ and DALLE before SD came out

peak orchid
#

They've been stealing the scene from Github lately

lost pond
#

Hey is there a way to combine two different images into one?

jade garnet
#

everything other than base SD could be axed and they'd be fine

solar sleet
# jade garnet they are raging much less

The drama might have been smaller, but only because it was not open to the public. 90% of the rage I've seen, and I've seen a lot, was directed against Ai art in general.

jade garnet
#

I dunno about that one

solar sleet
#

Either they all sink, or none of them. Do you really think that will happen though?

jade garnet
#

i foresee a number of horrible outcomes

#

to give one example

#

forced construction of some kind of "AI art ethics" division in all of the big SaaS companies

sudden isle
#

The saviour they're appealing to, the Walt Disney Corporation, is probably working on an AI Manhattan project of their own

jade garnet
#

that's one dystopia

#

another dystopia is that LAION becomes illegal, and only private corporate datasets are allowed

solar sleet
#

Could be, but that could still be toothless pr effort, like a lot of diversity efforts these days.

peak orchid
jade garnet
#

another dystopia is just stability feeling the fight isn't worth it and backing out of open source AI models

#

stability works on much more than just AI art and i wouldn't fault them

#

a lot of their other projects are quite important too

jade garnet
#

One piece of good news

sudden isle
#

Where's the dystopia where I get put in the anime Matrix

jade garnet
peak orchid
wise stratus
jade garnet
wise stratus
#

We are all prepped

solar sleet
#

There's so much money in AI though, and even though AI art by itself could be sacrificed, the danger that some legal precedent hits many more models is probably far to high for big tech companies to take the chance, they will put a lot of money on the line to stop Anti-ai art efforts

wise stratus
#

Plus we have official models with top content providers too coming

#

My basic thing is to get to a public company asap with dozens of important institutions backing and loads and of public invested

dreamy valve
#

any legal precedents affecting AI art would probably end up hurting innocent artists too no?

jade garnet