#💬｜general-chat

1 messages · Page 36 of 1

plush forge
#

@wide hare @lament steeple ?

#

btw my batch size is 1 and gradient accumulation steps 2

fervent thunder
#

use kaggle?

lament steeple
#

No. I had to buy a gaming pc just to do training on

plush forge
lament steeple
#

Not saying it can't be done, just I don't know how if it can

plush forge
plush forge
fervent thunder
plush forge
fervent thunder
#

Google collab is an option too

plush forge
#

do you have a tutorial on how to set it up?

fervent thunder
#

nope, you just register , find or make a notebook and you're ready to go

white niche
#

Do the stability Ai guys still pop past and chat? I can’t see emads account in the search anymore

plush forge
#

it seems unreal that it is so easy

fervent thunder
plush forge
#

where is auto1111 running?

plush forge
#

Oh ok, it is extremely complicated though by the looks of it

fervent thunder
#

Nothing complicated , just launch the code , wait for it to install and you're good

#

you can install anything into your a1111 too with command line

plush forge
#

ok how do you visualize the a1111?

fervent thunder
#

It has webui

plush forge
#

i assume it's like a virtualization right?

fervent thunder
#

I mean...a1111 = automatic1111 , it's your stable deffusion + webui

plush forge
#

but then how am i using someone else's gpu?

#

if it's on my computer?

fervent thunder
#

yea , that part is on their servers

#

You can use collab too, there are notebooks specifically for textual inversion too

#

But it's limited for 3-6hours a day on a free plan and can cut you off at any time, kinda risky for training

plush forge
#

Ok, if you have a tutorial it would be much appreciated!

plush forge
alpine bear
#

That is all

stable pawn
#

Is there any good open-source app for cropping and rescaling? I would like one where you scale and move a rectangle and the program then export and scale that to 512x512, with a sequential file name. I spent way to long time now cropping, friends, dogs, cats, teddy-bears and what god knows my friends will ask for here next.

alpine bear
#

It's free to use but not open source. If open-source is mandatory, there are Auto1111 extensions that can do it, though its one image at a time instead of many laid out.

stable pawn
#

Oh, am Automatic1111 thing sounds interesting, hope it can evolve, to use ai to crop and get images suited for training had been cool to have.

#

Ah, Brime was online, I am okay with that for photos of me, sending images of friends and family to an online service or how it work, I do not know. But thanks for the sugestion, will use that on less privacy important images.

warm junco
#

@stable pawn the Additional networks extension can crop images afaik

pastel fern
plush forge
pastel fern
#

I just bought a second hand gpu

#

Time to learn how to use this

static storm
kindred idol
#

Hello, is there a possibility to use Stable Foundation solutions to edit images ? Main goal which we are trying to achieve is to upload image of the room and using Stable API's change the interior style of the room

wide hare
flint fiber
#

is it just me or CivitAI kept getting error 404 every now and then?

#

are they getting DDoS or smt?

sudden ruin
#

how yall get previews in the extra networks tab? I dont get where it takes them from

sudden ruin
#

nvm figured it out dead

tepid cobalt
#

anybody know how to remove flags?

i got this error when try to installing extension:
AssertionError: extension access disabed because of commandline flags

note: I'm using stable diffusion web-ui

sudden ruin
#

what start up arguments are you using?

#

like --xformers etc

tepid cobalt
#

i'm actually double click on webui-user.bat

warm junco
vestal dew
#

Iam told xformers has a problem - makes images too different

warm junco
#

no not realy

vestal dew
#

OK

#

BTW, I never got that working yesterday, early this morning, but found the error I was having was an issue many others were having

warm junco
vestal dew
#

that it would corrupt SD, cause me to have to reinstall for it to work

#

after running unprompted to run that other thing...I forget the name now

warm junco
#

ah okay

#

now all works?

vestal dew
#

No

#

controlnet won't work

fervent thunder
vestal dew
#

ah

fervent thunder
#

well , or if you're testing prompts \ model \ sampler on same seed.
Xformers kinda ruins whole purpse of testing same seed with different values.

For regular generations - it doesn't matter

deft bronze
fervent thunder
#

About speed...I was thinking about 2 things...let me find it...

#

and maybe TPU generations on kaggle (since it gives it for free), there are few notebooks already implemented

#

both options should be faster

deft bronze
deft bronze
#

Is there any fun competition in Kaggle with sd? Like... vid2vid?

fervent thunder
#

Yea, I'm on same page, not feeling like fiddling around with it rn

fervent thunder
deft bronze
#

yeah fair

#

there is an "image to prompts"

#

which seems fair enough

#

(I would not want to try and compete on that one ugh)

#

but I'd be keen to see who wins

fervent thunder
#

I don't even know python, all my python and command line knowledge comes from google, that's about it, I'm a web dev

deft bronze
#

yeah I'm a python noob myself, just have done some basic tensorflow for some projects and such

#

but want to have an excuse to do more

#

(say, vid2vid)

fervent thunder
#

fair enough

runic cradle
#

Nowdays chatgpt is growing, and I am highly skilled in chatgpt api integration, model creation using fine tuning.

If you have new projects for chatgpt, I would gladly start with you.
Thank you.

vestal dew
#

LOL out loud at ChatGPT

dreamy igloo
deft bronze
#

tfw roko's basilisk will be unhappy now

vestal dew
#

I asked it to generate a worksheet to teach students comma use. It came up with many questions like this:

Bob is tall, and he is young.

A No Change
B tall, and
C tall; and
D tall. And

#

It claims the answer is B

deft bronze
#

tfw no one in the Internet actually uses commas correctly

vestal dew
#

I am trying to make work sheets for ACT/SAT

#

the answer can't be B

#

This is why ChatGPT is useless

deft bronze
#

You can try again and probably get some useful results

vestal dew
#

nope

#

multiple questions like that, and when I point out the mistake, ChatGPT just reworks the key to pick a different wrong answer as right.

split narwhal
#

spamming is against the rules

#

@livid bear

vestal dew
#

So if there's a way to get ChatGPT to correct that, if you even know what the error is, that would be quite helpful.

fervent thunder
vestal dew
#

I should make a worksheet that helps students with commas while they earn $72 K with Crypto in only 48 hours

split narwhal
#

@vestal dew good idea, lol

fervent thunder
#

Just go ahead and tell us, I want some lolz in da chat.

dreamy igloo
vestal dew
#

however, students do one very dumb thing

#

they fail to do the self study work I give them, then they cancel the tutor session because they haven't completed the work - doubling down on not doing work to prepare themselves

deft bronze
vestal dew
#

right, but one is understandable

#

why then double down?

deft bronze
#

tell them it's ok to attend tutor sessions

vestal dew
#

"I was out of town, so I didn't do the self study, so we can't meet this week" Which now means you missed the benefit of study and of the tutoring session

#

I do. They want to cancel because parents then feel it's a waste of time. However, it's like doubling down on doing nothing.

#

Iguess I'll take these worksheet questions, edit them

#

at least chatgpt did some of the work

dreamy igloo
#

Does anyone follow stable diffusion development news? Any news on SD 3.0?

vestal dew
#

The Russians inwented SD.

young spruce
#

KGB

fervent thunder
#

wot

vestal dew
#

No. It was a Russian farmer who needed a way to keep his cows from wandering into neighboring fields. The cows would break through the fences.

fervent thunder
#

so he made AI to draw new fences , right, now I see.

vestal dew
#

LOL out loud, no

#

what kind of logic is that?

young spruce
#

is this conspiracy theory talk? ghost

vestal dew
#

think about the noise cows make

#

Can ChatGPT or SD help with these UFO's? U.S. government briefed governors, saying there are hundreds or even thousands of these objects flying over America - they are likely balloons released from used car lots

#

flying at 30,000 feet or higher, evading missiles

fervent thunder
vestal dew
fervent thunder
#

No idea where "thousands" comes from, but either way, that's a subject for #🌶｜off-topic talk.

#

and "evading missiles" part lol

vestal dew
#

I was asking about SD, ChatGPT, and other AI's in relation to this

vestal dew
young spruce
#

a balloon evaded a missil? catwhaaa

vestal dew
#

i read about people using AI's to make award winning books, images. I can't even get it to make one valid worksheet question for commas

fervent thunder
vestal dew
#

Do you even see the trouble with this? Bob went to the store, and he bought food.

A No Change
B store and
C store; and
D store, and

ChatGPT claims A is the answer.

Do you see the problem?

fervent thunder
#

i've been writing a program that uses AI to detect faces then sort images according to the amount of people in them and to the detected gender, hope to release it soon
it will help for people training large datasets and is multithreaded / works with CUDA if available

vestal dew
#

there will be complaints about such a program

fervent thunder
#

like what?

vestal dew
#

is the AI, the program, assuming the gender of people?

fervent thunder
#

lol

vestal dew
#

you laugh, but it's true

fervent thunder
#

it just looks at the faces and decides if its more likely a woman or a man it's not biased

vestal dew
#

people will claim it's misgendering

fervent thunder
#

it will still have a small error rate

worthy perch
#

Can anyone tell me how create image with prompt

fervent thunder
#

whatever , people can complain all they want

#

just don't use it if you find it offensive or something

vestal dew
#

in the recent Harry Potter video game, some people who play are modding to make a transgender character into a girl or boy. The company that makes the game is removing all such mods because they're "misgendering" a character

#

Ive done this before, but I can't match it now. Maybe someone has an idea. I saw a wrestling tournament that was done under UV light. All the wrestlers singlets and shoes glowed on logos and other areas. Looked cool. How would I get that look in SD? I tried the prompt "UV Light" and things like "glowing logo on singlet", but they do things like make the hair green or add light sticks as arms

fervent thunder
#

Hm, there are bots in image thread detecting erotic images?
Wondering why it doesn't let me post here.
Images aren't nsfw, everything covered

fervent thunder
wild steppe
vestal dew
#

does the prompt have something about NSFW...something like nude, sex,. erotic,. whatever?

#

I did find one of the images...it had "uv light" and "glowing logo"

fervent thunder
#

All my prompts has it, but I was posting just fine before

vestal dew
wild steppe
fervent thunder
#

oh yea, I see now...the reason was "nsfw" in the file name.

vestal dew
#

remove the exif info, the tags

fervent thunder
#

or was it different image...I'm confused now lol

vestal dew
#

otherwise your text prompt is saved in the image info

#

it doesn't have to be in the file name

#

it could be in the information saved in the file envelope

#

unless you remove it, for example, your prompt is saved in there

#

then bots will judge by that, not by the image you see

#

For example, if I want certain clothing, I mention it in the prompt and have "naked" in the negative, so it won't make naked people. Bot might remove that image because of the word "naked"

fervent thunder
#

After few tests, seems like it actually scans an image, yea, it's false positive

vestal dew
#

be careful, because if it removes too many images you post, it will ban you

fervent thunder
#

Timer?

vestal dew
#

no idea

#

it was in a different discord, but I was banned for 24 hours after the bot removed an image I posted

fervent thunder
#

Not a big deal

vestal dew
#

unless I posted another offending image...then it would be permaban

fervent thunder
proud nova
fervent thunder
#

It's fiine 😄

wild steppe
fervent thunder
#

It's fine then

wide hare
#

Hey guys I have to choose a project for a coding assignment
its the final assignment and I want to code something that incorporates SD
anyone has some ideas?
it should be a pretty big project for example a good project is a chess game with AI

trail ledge
#

how big is "big"? Are you expecting to implement the AI, or just training, or just UI/features around existing AI?

wide hare
#

I think implementing the AI

graceful lynx
#

Hello, Quick question does anyone know is there a Formula for Embedding Learning Rate and Batch Size for best results

wide hare
#

but its pretty big

slender kraken
#

Hi new to SD, and AI art in general. I have seen a few posts on reddit, where users say they upscale in SD, and it clears up the image, brings in more detail etc. I have played around with upscaling in img2img (.20 - .30 denoise) and with the upscaler in extra's

in both cases, yes, the image got bigger, but i cant seem to get it to add details etc. Am i missing something?

warm junco
#

Also rough edges get smooth etc

#

@slender kraken Ah maybe you just used img2img and not SD Upscale.
Look at the bottom and choose scripts, then SD Upscale, thats the real upscaler in img2img

fervent thunder
#

Going above 4k x 4k acts a bit weird too sometimes, it trying to draw new objects on tiles based on your prompt.
So...what happens when let's say there isn't a "girl" on a tile, but it is in your prompt?
It might attempt to draw one, while drawing additional small detail.

Higher denoise - more likely weird things will happen too.

slender kraken
#

ok thanks, so in theory, going from 512 512 img2im2 to lets say 700x700, the image is suppose to get crisper? or just bigger?

fervent thunder
#

that small difference won't really add new details

#

It should make higher resolution and make it look more "clean"

#

I guess crisper is a better word , yea

#

go x2, you'll see the difference

slender kraken
#

cool, thanks

slender kraken
warm junco
fervent thunder
plush forge
#

Does anybody have a good tutoria on how to use haggle (or alt) for training?

cerulean kraken
#

Anyone know if you can use a mask on instruct_pix2pix?

#

I want to use it to make an image a comic book style, but it doesn't do the faces well. Before when I would train my own model on the person I'm trying to generate, I'd always inpaint the face separately so it's a higher resolution. Can't seem to do that with pix2pix

kindred idol
#

Hello, is there a possibility to use Stable Foundation solutions to edit images ? Main goal which we are trying to achieve is to upload image of the room and using Stable API's change the interior style of the room

astral goblet
kindred idol
#

@astral goblet any chance you can give me the links (github/docs/etc) to read on ? All the stuff I google out is a little bit overwhelming and messy.

astral goblet
#

the system i use is nice but i'm not sure how good it is for server grade. can't attest to that. it's automatic1111's stable diffusion webui

#

a sort of all-in-one package. even has an api you can interface to

#

https://github.com/AUTOMATIC1111/stable-diffusion-webui i know github can be overwhelming, but this one is actually fairly well maintained and refined. relatively speaking. they have a big wiki on there and a discussion page and it's so widely used that many people know tricks to support it

#

your goal includes making a webapp, so i think this wouldn't be your endgame solution. but it's a great case study for you to understand all the needs of the technology

kindred idol
# astral goblet your goal includes making a webapp, so i think this wouldn't be your endgame sol...

Thank you very much! I am eager to try this out. Just to clarify, creating a web app or UI is just a small part of my goal. I have a good understanding of the tech stack, so I am more interested in finding backend solutions, regardless of the technology, such as Python, C, or JavaScript. If possible, I would like to use a ready-to-use API, regardless of the cost, but if that's not an option, I am interested in self-hosting, training, or generating models ourselves. I am new to AI and I lack knowledge of the possible ways to achieve my goals. However, I am more focused on completing the end product with the easiest solution without going into too many specifics or details

astral goblet
#

for this stuff the backend is generally all python, and that canbe wrapped a number of ways. there may even be node js modules for this already

chilly sparrow
#

This server has grown

astral goblet
#

you been here since before the surge i see

chilly sparrow
astral goblet
fervent garden
#

Hey guys, can we buy this token now?

astral goblet
#

lol

lost flicker
#

hiya everyone!

#

man, anyone playing with ControlNet?

pastel fern
#

Is it normal my webui shows cuda:0

#

Using a a 3060

vestal lynx
#

how do I download Stable Diffusion? Do I do it directly from the github?

trail ledge
vestal lynx
wild steppe
trail ledge
#

keep in mind stuff is moving fast, so even a month old video may start to get out of date

trail ledge
lost flicker
trail ledge
#

nice, I haven't seen it yet, but yeah, he beat me to animation

#

I was going to try that tonight

lost flicker
#

It's super exciting.. So question then. How do you run a sequence in SD? what UI are you using?

#

i've done Deforum, but havent worked sequencing in SD, whats the secret?

trail ledge
#

I don't think there are any publicly available tools yet for sequencing controlnet renders, but things are moving so damn fast there could be 20 extensions out now and I'm not aware of it 😆

lost flicker
#

ok.. i'll keep looking. 🙂 I got the ability to drive OpenPose with motion capture or video worked out, and now I am eager to use that in ControlNet as toyxyz has done very successfully. I didn't realize he rolled his own sequencer tho...

trail ledge
#

well, you're further than I am, so I'm not going to be of any help to you 😄

lost flicker
#

I just learned this was even possible!

tepid cobalt
solid holly
#

Anyone know if SD can blend two or more images like in Midjourney?

#

I copied a MJ prompt which it had two image URLs at the front, into SD text2img and it seemed to use at least one of the images and a guide

#

I haven't tested it further, but I thought I'd ask inf anyone is in the know

warm junco
#

Maybe with an extension, there should be one that can blend

dire lichen
#

How to install ControNet for mac?

solid holly
#

I didn't see one in the "Available" list. Got any suggestions?

dire lichen
#

I've watched 6 tutorials now, ALL usijg windows

forest apex
dire lichen
solid holly
dire lichen
#

Yes

#

The tutorials begin with people downloading something from huggingface

solid holly
#

Are you using web-ui?

dire lichen
#

They then proceed to type in a command prompt

#

But they are all using Windows. I have no idea what the equivalent command would be for mac os

#

All 7 tutorials I've seen so far, the demonstrator is using Windows, so it leaves me in a bind.

solid holly
#

Let me see if I can find the guide I used - BRB

warm junco
solid holly
dire lichen
#

I just tried to install it within the webui via the extensions tab

dire lichen
solid holly
#

Instead of the fill 5.5GB files, these are pruned down to about 750mb

#

*full

dire lichen
#

Oh wow

solid holly
dire lichen
#

I just installed it via the extentions tab in like 2 minutes

#

O.O

dire lichen
solid holly
dire lichen
#

Now I have no "models" in the model dropdown

#

Is this normal?

solid holly
#

Did you download the models?

dire lichen
#

Here's a screenshot

#

I'm supposed to download the models on my local machine to use in the webui?

solid holly
#

For ControlNet? Absolutely

#

All your models will be local

dire lichen
#

I'm not running stable diffusion locally though

solid holly
#

D/L and put in extensions/sd-webui-controlnet/models

dire lichen
#

Very confusing

solid holly
#

Ah, I see. Colabs?

dire lichen
#

Yessir

#

Is there a forum for that so I don't bother you guys any more? Sorry, never ran it locally

solid holly
#

Sorry, noob here, so IDK where you'd go for support

dire lichen
#

Ok thanks anyways. Too bad

solid holly
#

But You'd D/L the models the way you d/l all others

#

And then move them into the right directory

dire lichen
#

The other models I uploaded to google drive

solid holly
#

Eitehr command line, or use the files panel at the left of Colabs

dire lichen
#

Then I manually reupload them via the google drive link directly to Jupyter Lab

dire lichen
trail ledge
#

well, if you're using colab, then mac command line won't help you

solid holly
#

If you're using colabs, command line is universal

trail ledge
#

probably all of the instructions out there are for local installs

solid holly
#

linux, I think

dire lichen
solid holly
#

Just look for Linux instructions

dire lichen
#

I have a command line within the web ui

dire lichen
solid holly
#

That will prolly be linux as well

dire lichen
#

Ok

solid holly
#

Mac Commands are for your machine, only local machine

dire lichen
solid holly
#

Colabs are all linux or windows boxes, I think

dire lichen
trail ledge
#

what's the OS on the VM?

dire lichen
#

That virtual machine has a vast array of folders and programs, including a command line. I install my models via the same command line everytimr

dire lichen
#

Jupyterlab 3.5

#

I have folders titles "embeddings, configs, extensions, html, models, scripts" etc

#

The minimum requirements on the github say "Windows"

#

Rip

#

The github page says if the model list is empty, refresh it, if still empty, reload ui

#

I have done these things and yet my model list remains empty and therefore unusable

#

What's the most active stable diffusion forum?

warm junco
#

Did you copied the files into the extensions/Controlnet/models folder?

dire lichen
#

Which files?

#

Sorry, the github page is hard to navigate

warm junco
#

The model files of controlnet

dire lichen
#

The 5.7 GB files?

warm junco
#

Yes

dire lichen
#

Is there a way to upload them directly to the web ui as opposed to downlaoding to my local machine and the reuploading?

#

I'm paying $3 per minute for this host and uploads directly from my machine over 1 gigabyte takes several hours

warm junco
#

I dont know 😮 but there are pruned models with less filesize

dire lichen
#

Whereas uploading a model directly from google drive takes 30 seconds

dire lichen
trail ledge
dire lichen
#

Thank you happyfunball

trail ledge
#

I haven't tried them though, so ymmv

warm junco
#

He was faster xD

trail ledge
#

I had the tab open 😆

dire lichen
#

Is there a difference in results?

warm junco
warm junco
dire lichen
#

Whyyyy tf are they forcing people to download 5GB files?

trail wing
#

control net is mind bending me

#

I just cannot believe what I am doing with it

#

damn

#

it even "tames" models that I almost couldn't make work with some TI stuff

dire lichen
#

I'm uploading 1 trimmed model

#

Estimated time of upload is 1 hour 45 minutes

dire lichen
#

Is there a way to upload the model to the webui directly from huggingface?

#

Is there a command line for that?

#

Otherwise I'm paying 90$ just to upload this thing

#

Due to server time

astral goblet
trail ledge
dire lichen
#

All good

#

Thanks anyways everyone. I'll have to wait for an easier method. 5 hours of my life

trail wing
lost flicker
#

anyone work out how to incorporate controlNet into an img2img batch?

trail ledge
#

not yet. What happens when you try to run it in batch mode?

lost flicker
#

after frame one it errors: AttributeError: 'numpy.ndarray' object has no attribute 'save'

trail ledge
#

hrm

#

that reminds me, I should fork the code 😆

dire lichen
#

Ok

#

I spend an hour uploading the model to google drive

#

The filename ends with ".safetensors"

dire lichen
#

Then, once it was uploaded to google drive, I uploaded it directly to my webui

#

And then I dragged it into the models folder under the stable diffusion folder

#

And now it's showing up in the models section at the top of Automatic1111 webui, where I normally choose a model

#

But not under the controlnet box

#

Please help. Where am I supposed to put the ".safetensors" file?

trail ledge
#

yes, it's fine, you don't want to put it in the regular models folder, you want to put it in extensions\sd-webui-controlnet\models

dire lichen
#

Omfggggg

#

It showed up in the list

#

Getting the error

#

"Runtime error: Mat1 and Mat2 shapes cannot be multiplied (154x1024 and 768x320)"

#

My uploaded image is 500x500

#

Image width and height parameters in the upper model is 500x500

#

And the same for canvas height and width (500x500)

#

Any help?

dire lichen
#

🇵 🇱 🇪 🇦 🇸 e

dire lichen
#

This whole time I thought this was the stable diffusion discord

#

Welp

fervent thunder
#

Hi

astral goblet
pastel fern
#

Man automatic1111

#

Is not automatic at all

#

I cant seem to generate any images with it

vestal dew
#

true

dire lichen
trail ledge
#

did you install it right? There should be one.

trail salmon
#

Heya gang! So it the SD discord dream bot been completely closed now? Is there any other place to continue that? Thank you! ❤️

dire lichen
dire lichen
trail ledge
#

ok, I'm confused. We're talking about the controlnet extension?

azure nebula
#

which one was pose and sketch model?

vestal dew
#

I cant get controlnet to work

dire lichen
dire lichen
trail ledge
vestal dew
#

I started trying yesterday.,..mo luck. Had to keep reinstalling. Then read that other people were having the same issue

dire lichen
#

I got everything working in 2.1

#

But when I tried to use it

#

I got an error

trail ledge
dire lichen
#

People advised me that it's not compatible with 2.1

#

So I tried using 1.5

trail ledge
#

I've managed to put it into a state where it has Assertion Error

dire lichen
#

But 1.5 doesnt contain that same folder as 2.1 does

trail ledge
#

but it worked for a fair bit before that

dire lichen
#

So there's nowhere to put the model

#

My error was weird I forget the exact lingo

trail ledge
#

1.5 and 2.1 are models. The extension is for Automatic1111

#

not model related

#

do you have two instances of Automatic1111 or something on colab?

dire lichen
#

I'm hosting these remotely using a virtual machine

trail ledge
#

right

vestal dew
#

for some people, after running controlnet, whatever model was loaded would be corrupted

dire lichen
#

I just fired up 2 machines, paying for both

#

1.5 and 2.1

trail ledge
#

do you have two different virtual machines, one with 1.5 models, and one with 2.1 models?

#

there you go

#

install it on the 1.5 machine too then

dire lichen
#

1.5 machine doesnt have the folder described

vestal dew
#

either had delete it, reinstall it, or reinstall a1111 completely

dire lichen
#

2.1 does

trail ledge
dire lichen
trail ledge
#

I've seen it choke on inpainting models, but once I switch to the non-inpainting ones, it started working

#

that is, until it got to the state it's in now

trail ledge
dire lichen
#

It looks like this

vestal dew
trail ledge
#

if you're paying for 2 machines, one running 1.5 and one running 2.1, that sounds like you're running two instances of automatic1111

dire lichen
#

/extensions/stablediffusionwebuiimagesbrowser/

trail ledge
#

so you need to install the extension on both

dire lichen
#

That's where "model" should be

#

But the "model" subfolder is nonexistent

#

But it's there on 2.1

dire lichen
#

When I'm running "stable diffusion v2.1" the correct folders are there

#

But not one 1.5

high spire
#

i cannot using stable diffision bot

dire lichen
#

I've tried multiple machines and multiple OS

high spire
#

can someone help me

trail ledge
#

so I don't know what else to tell you

dire lichen
trail ledge
#

I'm using 1.5 and it works fine for me

dire lichen
#

Well thanks for trying.

trail ledge
dire lichen
#

The only thing in the folder you're describing is "javascript" and "scripts"

#

But no "models" folder

trail ledge
dire lichen
#

Ok

brave matrix
#

new episode featuring Paul Rudd and Chris Rock

fervent thunder
#

How made pix here

opaque hazel
#

so if I want to create something what do I do?

brisk bloom
lament steeple
#

bangs head against wall

lament steeple
#

It is officially make stupid mistakes day

plush forge
#

Hi there! Does anybody know how to use kaggle for model training (or any other of that style of services)? A tutorial of any type would very much be appreciated

lament steeple
#

points at previous comment

#

wrong day to ask me. Hopefully someone else can chime in

plush forge
#

lol yeah I know you don't know hehe but maybe someone knows

lament steeple
#

But I have gotten TI training working locally

wide juniper
#

Any ETA for the next version of SD?

astral pivot
#

can someone help, how do i use google colab for sd ?

astral goblet
pastel fern
#

I like how hentai is now like advertising itself as non ai generated lol

severe idol
#

@quick bluff

pastel fern
#

Can someone tldr me what's the difference between the model types

unborn echo
plush forge
#

@fervent thunder the problem here is images cant be uploaded

fervent thunder
#

@unborn echo I didn't have such error , I don't think, try resetting and launching code again

fervent thunder
unborn echo
#

tutorial

plush forge
fervent thunder
fervent thunder
unborn echo
fervent thunder
#

if everything is alright - it will drop little gui where you can chose model and extensions and launch webui

unborn echo
#

i sent the error here

fervent thunder
#

At which step is it happening? Did you chose model before launching webui?

unborn echo
#

this is just the 2nd cell

#

i deleted it and made a new copy, this time when i launch the 2nd cell i get no errors but the gui doesnt pop up

fervent thunder
unborn echo
#

sad

dreamy igloo
#

i am trying the axelfar one now, it says it allows you to download models from civitai

#

22minutes still running the 1 set of codes, downloading, cloning....

#

I hope this is a 1 time deal

unborn echo
dreamy igloo
#

I have persistance off...btw i am in gradio finally

#

24 minutes

#

since we were chatting with Tito Ortis, I'm gonna make an image out of that UFC fighter

fervent thunder
# unborn echo am i supposed to have persistance on

It worked without it for me when I first started , it shouldn't matter.
What persistence does - it keeps all your files between sessions.

I'd suggest keeping it off while you're trying to figure out why it doesn't work, so you don't keep anything broken between sessions

unborn echo
#

alright

fervent thunder
#

keeping it off* - edited, that might help

unborn echo
#

btw where is the gui supposed to show

dreamy igloo
#

manny try the axelfar one later, just 1 play button click and it loaded gradio in 24 minutes for me.

unborn echo
#

24 minutes damn

dreamy igloo
#

the GUI is just like the typical Auto1111 one I used on google colab

#

but I dont see image browser for some reason on this GUI...

fervent thunder
#

What I was doing - is making .rar archieve and downloading it from file explorer on right side, but to launch another block of code - you have to stop currently working one

#

That's most annoying thing on kaggle , unless someone will figure out better way to save

dreamy igloo
fervent thunder
#

I think it's possible to use collab to mount your google drive and save your images there

#

But I didn't want to waste collab time just to save files lol

dreamy igloo
#

maybe i use kaggle for training in the future

fervent thunder
#

Yea it's better if you want to do something for long time, since you know exactly how much time you have and don't risk to get disconnected at random moment as on collab

#

for short operations collab is easier option

wide juniper
astral goblet
#

not quite a niche. i'd say it's so game changing that it'll be a standard part of the toolset going forward.

#

just like 2.1 came with depth awareness, this is towards that

#

all future models will have systems like these in place, since this upgrades the hell outta 1.5 beyond all current capabilities

#

depth guidence was just the start, this brings full resolution depth maps to 1.5, as well as pose maps, edge detection, straight up pencil sketches, and a number of other controls to any existing 1.5 model

quartz night
astral goblet
#

like a hypernetwork, or a lora, or a embedding. like how they're another network that can loop in. control net does that too. but its a novel kind of extra network where it's not pushing clip vectors

quartz night
#

and not to mention while StabilityAI took 2000GPU hours on large scale A100 clusters for the depth2img model on 12M images, this depth-based ControlNet Model was trained on a single 3090TI for less than a week on 200k images.

astral goblet
#

bro..

#

thats not much

#

one could in theory train it on previous frame sort of interpolation, and use it to drive motion

quartz night
astral goblet
#

thhat's very efficient and the card is very cheap comparitively

quartz night
#

I would like to see a new section for ControlNet models on Civitai

astral goblet
#

someone will train it for uv maps

#

point data

#

low poly iinput

#

i'm sure theres tons more of useful inputs

fervent thunder
#

I cant update auto1111!=??!??!

final gorge
#

My colors look very dull on multiple models, anyone have any idea whats going on ? sometimes its sort of grainy or smudged

warm junco
final gorge
#

Thought i read somewhere it came built in, ok ill dot hat , thanks!

still glacier
final gorge
#

Im such a noob but having loads of fun

storm marten
#

hi! I have a weird problem I need to solve.

I have an existing image and I want to effectively decohere this image into noise, while capturing that process as a video.

I've been experimenting with the steps extension in Automatic (creating videos then reversing those in premiere) but it's shonky as heck.

It's a longshot, but anyone have any thoughts on how to achieve this?

chilly sparrow
exotic acorn
#

Any extensions suggest for stable diffusion ?

#

Or must have once ?

pastel fern
#

I just go to civitai and download the most popular ones

#

Is there a blog for like small tips and tutorials

exotic acorn
#

Not an extensions

pastel fern
#

oh extensions

#

maybe u can look through here idk

neat condor
#

guys is there a channel to drop tutorial video?

analog crag
#

is it possible for the web version (not dream) to only supply THREE images?

chilly valve
#

morning Coffee

shadow birch
#

Hello everyone, does anyone know the difference between the public link and the local link that is generated at the end of the stable diffusion connection by Google Colab?

lament steeple
#

In google colab, you won't be able to access the local link. you have to be on that computer which is remote.

shadow birch
lament steeple
#

not if you set a password

#

There really isn't a built in way to share images directly anyways. needs an extension which are difficult to install in colab

thin heath
lament steeple
#

but if you don't set a password, someone else can access your interface and share it with you whether you want them to or not

shadow birch
#

👍

neat condor
fervent thunder
#

Unstable Infusion

stable pawn
#

I use Automatic1111 and the "Train" tab to embeddings for styles, I try to make it on Images made in Second-Life, and on images of colorful fur, and also just images of the street where I live, I use 30 to 50 images and it often take very long time with the original settings and the result is never as I want it to be. Do any have suggestion or a good link to a easy tutorial.

#

Due to a stroke a year ago I can sometime miss logical steps, if I missed something or there is something I need to pay attention to, point it out too even it may sound too simple for you 🙂

lament steeple
stable pawn
#

That sound a bit long for the Training tab, but I am willing to test it all 😄

lament steeple
#

for styles, I use a vector size of 4 and the styles_filewords.txt template. That is about the only difference

#

lol like training is easy 😛

stable pawn
#

I know, just learn new after my stroke period is very hard, so I can not understand why I excepted there should be a easy way to learn this. Thank you.

lament steeple
#

I had to learn calculus to understand nn training

jovial maple
#

Hey there

red gorge
#

Hi will this project have a cryptocurrency running?

ebon pewter
#

hi..how to do this "-m pip install --upgrade pip" ? automatic1111

dark jackal
#

Hello friends. My Deforum extension tab is missing. Who knows how to restore it?

autumn portal
#

Glad I got to this community! I’m trying to learn SD and it’s very intimidating!! 😄

ocean prawn
#

Anyone knows where to get the access to 10million+ images generated on SD by users? Where is the database for that? How can I access it?

empty barn
alpine lava
#

what is a negative promt?

#

prompt*

ocean prawn
empty barn
ocean prawn
empty barn
foggy elm
#

What is a good place to get models for SD ?

empty barn
autumn portal
#

This feels a bit like pirating video games - downloading cracks and inserting files in other folders with fingers crossed hoping it actually works 😅

#

But with a whole glossary of new terms and functions

ocean prawn
foggy elm
rare quail
#

hi

#

anyone from saopaulo

#

enjoyed ?

maiden hornet
#

Hello, I am a designer with over 20 years of experience creating brands. I have come up with a method to train artificial intelligence in design and strategy to create high-quality brands.

Currently, available platforms have very low quality and leave everything in the hands of the client's personal preferences (without providing expert and strategic guidance).

I am willing to train the AI for thousands of hours to transfer my knowledge and experience.

Within this server, can someone help me find the right people who are interested in developing a project like this?

Thank you!

blissful mist
#

Hi, someone knows how to change the ubication or link my lora's folder outside sd folder in Automatic1111?

fervent thunder
winter pike
stuck shoal
#

lol yeah I found mine too

ocean prawn
ocean prawn
ocean prawn
toxic brook
#

is dreambooth still a better option than lora for training specific person's face ?

winter pike
winter pike
#

SD 1.5 testing lasted a couple of days

crystal patrol
#

Hi guys, is anyone using the rt 7000 series to use with sd+dreambooth?

rugged sail
#

Hello, can stable diffusion generate text too, or images only?

trail ledge
#

it struggles with text

exotic acorn
#

Anyone know where i can download style for stable diffusion ?

left dew
#

is using the vae ft mse 840000 ema pruned still important? should i use it for every model? some come with a baked vae now

exotic acorn
#

This place spamed with with to@much of the same picture

astral goblet
#

As a toy now an then. their model is great at really nailing a concept down, but at the same time it is really over trained on one style

exotic acorn
#

Write digital art .. all the same

astral goblet
#

their model is really great but it sure does have a "lexica" look on every image. also, i cna't download it to use on my own machine and their generation parameters are really limiting

astral goblet
#

middle of february in the middle of the day in north america. not primetime hours at all

fervent thunder
#

I noticed it too, I think its the rearrangement of the channels, now all the art display channels are not as easy to reach. Can't speak for everyone but it's definitely made me browse the server less.

#

Are there plans to release a video open source model?

fervent thunder
astral goblet
fervent thunder
opal jacinth
#

Is there a Discord for Deforum and Automatic1111?

tardy steeple
#

why this **** installs every time i refresh my web ui? "Installing openmim
Installing mmcv-full
Installing mmdet"

astral goblet
opal jacinth
#

Is this the right forum to ask for some advice running Deforum animations?

solid mortar
#

Is there any repository for specific styles prompts? Maybe with some examples. Like anime of specific studio and bunch of prompts for sd to follow them? Basically I’m in search for library of some sorts for this kinda things. To feed it to styles and then just apply in one click with different creations.

broken gull
wild steppe
broken gull
wild steppe
#

Any chat channel

broken gull
#

And also hey long time no see

#

Do I need to @ someone in the command?

wild steppe
#

Yep you need someone to show love to, could be yourself too!

fervent thunder
#

hey guise.

is there a way to use the A1111 setup but to send prompts and parameters via commandline?

#

I'd like to automate the generations

#

especially img2img

#

I wanna input a sequence of frames and output img2img images from stable diffusion

#

I don't want to recreate the whole A1111 setup from scratch as it's already setup for me

#

I just want to automate it

#

all the packages it needs are already setup

#

all the models downloaded

#

I hope automatic 1111 is not called automatic for no reason

#

or maybe there's a way to interact with it through the javascript console?

#

that would also be neat

celest phoenix
#

I know with Invoke you can run it in terminal without the GUI

fervent thunder
#

InvokeAI?

celest phoenix
#

yeah

fervent thunder
#

does it have to load the model everytime you prompt or it keeps it in memory?

celest phoenix
#

No idea i havnt really explored it. But at the beginning it asks you how you want to run it

#

Doubt it will though

fervent thunder
#

I didn't know Invoke was a commandline tool as well

#

interesting

#

I can't find any videos on it

celest phoenix
#

In A1111 for text2img you can import prompts from file or text box and it sequentially generates the various prompts

fervent thunder
#

how is it supposed to be structured?

#

the file

celest phoenix
#

Just a note pad new line new prompt

fervent thunder
#

what about parameters

#

and negative prompts

#

etc

celest phoenix
#

No idea, have a play with it. if you look at the bottom its under scripts labled Prompts from file or textbox

#

never used it just know its there

fervent thunder
#

I can't find any videos on it

celest phoenix
fervent thunder
#

ah there's no way to manually specify seed per line

#

it's one of these 2

#

ah I can't post images here

# 
Use same random seed for all lines
celest phoenix
#

dont you want to refine one before you start the other?

fervent thunder
#

I want complete control over the prompt

#

plus img2img batch

#

also with complete control

#

I'm planning to generate these lines with a script

#

procedural AI animation

celest phoenix
#

what more control do you need?

#

you want a diff seed each img?

#

and to tell it what seed and model for each img in a batch gen?

fervent thunder
#
  1. prompt
  2. negative prompt
  3. seed
  4. cfg scale
  5. denoising strength
celest phoenix
#

so every image has unique parameters

fervent thunder
#

I want these parameters per generation

#

yes

celest phoenix
#

yeah ok

fervent thunder
#

they will be changing

celest phoenix
#

you want them to all render at once or sequentially?

fervent thunder
#

sequentially

celest phoenix
#

makes sense

#

seems very possible

fervent thunder
#

but that's why I ask, I don't see these kind of tutorials on the internet

#

it's like people are not interested in doing it

celest phoenix
#

because it doesnt exist yet from what i know

fervent thunder
#

but A1111 is capable of it

#

why not expose these parameters

celest phoenix
#

think an extension will need to be made where you could create new tabs and the run all tabs

fervent thunder
#

to a commandline interface

celest phoenix
#

because the point is that is a GUI for SD

fervent thunder
#

which language are extensions in?

#

JS?

celest phoenix
#

yeah

fervent thunder
#

great, I'm a JS dev

#

now

#

how tf

#

do I make an extension

celest phoenix
#

most extensions pull from git

astral goblet
#

i'm pretty sure there are extensions for that kind of file batching

#

let me look into a couple

fervent thunder
#

is there a website hosting those extensions with their info

#

so I can look through

celest phoenix
#

well not that link hang on aha

fervent thunder
#

ye it's JSON perfectly readable for me

# 
 {
            "name": "Deforum",
            "url": "https://github.com/deforum-art/deforum-for-automatic1111-webui.git",
            "description": "The official port of Deforum, an extensive script for 2D and 3D animations, supporting keyframable sequences, dynamic math parameters (even inside the prompts), dynamic masking, depth estimation and warping.",
            "added": "2022-11-01",
            "tags": ["script", "tab"]
        },
#

this sounds interesting

celest phoenix
#

check if they're not depreciated though

fervent thunder
#

ayy there's actually multiple of these

# 
{
            "name": "Animator",
            "url": "https://github.com/Animator-Anon/animator_extension.git",
            "description": "A basic img2img script that will dump frames and build a video file. Suitable for creating interesting zoom-in warping movies. This is intended to be a versatile toolset to help you automate some img2img tasks.",
            "added": "2023-01-11",
            "tags": ["script", "tab"]
        },
celest phoenix
#

for animation yeah

fervent thunder
# 
{
            "name": "Video Loopback",
            "url": "https://github.com/fishslot/video_loopback_for_webui",
            "description": "A video2video script that tries to improve on the temporal consistency and flexibility of normal vid2vid.",
            "added": "2023-02-13",
            "tags": ["script"]
        },

wait what's "normal vid2vid"?

astral goblet
#

I'mnot seeing any batch job extensions that allow that. thought maybe i'd look but didn't dig any up

fervent thunder
#

there's vid2vid in a1111?

celest phoenix
#

not natively

#

was added 3 days ago

fervent thunder
#

Deforum looks like exactly what I'm looking for I think. tho not sure it can do img2img yet

#

however I can try video2video later

astral goblet
#

i think people use deforum with img 2 img.

fervent thunder
#

ayy that's good news

#

thanks everyone

astral goblet
fervent thunder
#

that's deforum?

astral goblet
#

yup. op confirms in comments

fervent thunder
#

Basic workflow: edit video --> run through Deforum video input --> create transitions in Deforum using 3D animation --> add a bunch of effects on top.

#

there's a video input!

#

perfect

astral goblet
#

I use the Deforum colab notebook where you just need to reference the video file and it will separate all the frames and set it up to process each one.

#

i dont knwo deforum too well yet but it looks like thats what he's doing

fervent thunder
#

ah, colab notebook.

#

I hope the extension exposes that

#

I also hope it doesn't require more than 8GB VRAM

#

colab notebooks are fine but horribly slow

#

can't be compared to a 3070

celest phoenix
#

hate the UI on collab notebooks as well

fervent thunder
#

indeed

astral goblet
#

i mean, the extension is a better version of the collab notebook aint it? like, it's what collab developers chop down and make work

#

really it should be doable either way. all needs to be done is to split a video into images using any script that does that

#

batch process on that folder and then video it all back together

#

video is nothing morethan a sequence of images

fervent thunder
#

oh I don't need a script I can split video with ffmpeg

#

been a while since I used that fucker

#

but I can google the commands

#

and then put the video back

#

I'll just write a .bat file

#

I absolutely love deforum so far (watching the video I linked above)

#

it's exactly what I was looking for

#

math formulas with time variable

#

damn for a procedural artist this is a goldmine

astral goblet
#

a script would just use that too probably

astral goblet
fervent thunder
#

😂

#

a lot of things in ffmpeg can be done in 1 call

astral goblet
#

bat would idealy index the files through a loop and then have knowledge of what to do with your filenames that come out of deforum

fervent thunder
#

yeah I can do that

astral goblet
#

also, weird science a babe in your closet. well, at least thats what i wished it could do when i first learned batch files

fervent thunder
#

I had to learn bash and then I switched back to windows and now I'm doing this shit everywhere what a mess

astral goblet
#

mmm bash. so glad that powershell got that going on now. and its got tabs now!

fervent thunder
#

fuck powershell

#

it's a snail

astral goblet
#

you know how long i wanted windows to have a tabbed terminal?!

#

no its super duper now

fervent thunder
#

I do regular cmd scripts

astral goblet
#

really come up

fervent thunder
#

it's slow as fuck

astral goblet
#

for running .bat files, yeh. for .ps1 files, neigh

fervent thunder
#

just the regular dir command

#

takes like 3 seconds

#

with cmd it's instant

astral goblet
#

let me powershell into my ext4 partition an see what happens

fervent thunder
#

well ok it's not as slow on my top end machine but back when I used a shittier hardware it was terrible

#

I know it's doing a lot of stuff under the hood it's not supposed to

astral goblet
#

i just dir'd my stable diffusion generation folder

#

pretty instant

fervent thunder
#

so I just avoid using it

astral goblet
#

i just came back to win11 after two years on arch linux. i'm pretty iimpressed

fervent thunder
#

I prefer something that executes direct directives

astral goblet
#

did i mention tabbed terminals yet??? HOLY SHIT

#

lol also i keep using unix commands like ls and it knows those

fervent thunder
#

look at mine it even got frosted glass look

astral goblet
#

i do it like "shit! .. oh its working.. okay"

fervent thunder
#

I have WSL

astral goblet
#

yeah i gots to customize mine again. its sick what windows 11 can do

fervent thunder
#

I can do linux commands pretty sure

#

I just don't know how

#

I can install Ubuntu on my windows

astral goblet
#

no no i mean like, when i'm using powershell on windows environment, it still knows unix commands in there

fervent thunder
#

yeah I know

astral goblet
#

yeah i've got the default ubuntu environment setup so that i can use my ext4 drive

fervent thunder
#

but PS is a beast I'm not touching

#

bash is instant, unlike powershell. try running it on older hardware

#

maybe they fixed it in windows 11 but I doubt it

astral goblet
#

it mounts in ubuntu and then i get it in explorer under "linux" but it behaves exactly like any folder

#

can mount it as a network share to make pathing more accessible

#

well i mean, i use new hardware so why would i?

fervent thunder
#

have you tried WSL?

astral goblet
#

yes i have ubuntu mounting my ext4 drive

fervent thunder
#

I am paranoid about stuff eating more performance than it should

#

especially since i'm on a laptop

astral goblet
#

why have resources if you don't want processes stretching their legs?

#

i got a i7 alderlake and 4080

#

all old storage devices though. none of them were from this system build. scavenged from 2 dead pcs

fervent thunder
#

sorry this probably got too offtopic for this channel

astral goblet
#

"general" its all good i feel. if someone else got a better topic we can always allow the interjection

#

i think terminals are a big part of usage too. why not rant about them

fervent thunder
#

yeah I actually thought A1111 has an interface to access it via commandline

#

or some portal to stable diffusion

#

wrapper process

astral goblet
fervent thunder
#

but things aren't designed the proper way nowadays

astral goblet
#

mmmmmmmmm

astral goblet
celest phoenix
#

and split tabs

#

right click it

fervent thunder
#

making it a plug-and-use solution

cerulean kraken
#

man I wish instruct pix2pix worked better with higher resolutions

fervent thunder
#

if A1111 was a fork of an existing wrapper that did just the assembly part and had CMD access that would be a good project structure

#

but A1111 devs are probably not very smart

celest phoenix
#

smarter than me

#

but doesnt take much

bold hemlock
#

Hey

#

How to use stable difussion bot in discord?

#

to generate images

celest phoenix
#

you dont, they moved away from discord generations to their own UI called dream studio. or you can install it on your local machine via python in command prompt

astral goblet
#

at least it's not entirely programmer ui like deepfake tools were when they first landed

#

i've seen some cool PRs for webui that attempt to rebuild the ux on it. neat work going on. not sure if tey've got "it" figured out yet, whatever "it" is

celest phoenix
#

its the balance between making it user friendly but allowing the freedom to do pretty much what you want. Issue with that is user friendly tend to mean dumbing down.

astral goblet
fervent thunder
astral goblet
#

lot of ongoing work

fervent thunder
#

commandline access is the most flexibility it can grant

astral goblet
#

might be dead in the water though

fervent thunder
#

it would let anything possible to the guys that know the batch

celest phoenix
#

but 20% of users will use it

fervent thunder
#

without learning any extension API

astral goblet
#

i think too, another team is probably redoing a big part of what a1111 has got together. perhaps a way to support extensions for that ui too, just to transition

#

it is janky, but such is the way with this edge stuff

#

6months old now?

#

6 more days and it's 6 months old

bold hemlock
#

How to use stable diffusion bot in discord to create images?

wild steppe
bold hemlock
#

oh

#

thanks

astral goblet
#

man oh man. over 20k iimages. so much of it are just random experiments and junk lol. every batch i ever made had a grid saved too. i gotta get this stuff under control

trail ledge
#

you never know when you'll want to whip out an old fail to impress family and friends 😆

fervent thunder
#

just purge it all

#

everytime I make something worth I just post it on my server

#

the rest is pretty much trash

brisk bloom
#

I delete the majority of output as its not great and have grids turned off

astral goblet
#

like not that many images, that was the speed it operated at

fervent thunder
#

ah that's slow

#

how did you get to 20k images

astral goblet
#

bought a 4080

fervent thunder
#

I wanna get a 4090 someday

#

but not because my GPU is slow or something

#

it's simply to be able to do training

#

dreambooth

astral goblet
#

somedaymeans the 5090 will be out by then 😮

#

get that

fervent thunder
#

yeah

astral goblet
#

maybe... even.. a hundred 90

fervent thunder
#

in that case I'll get something cheaper

astral goblet
#

train skynet

#

things are going to get stupid different in 5 years we'll see

fervent thunder
#

the only reason I don't wanna simply pay for colab dreambooth training is because I have a shitty internet connection and it's a pain to download and upload big files like 5GB

fervent thunder
astral goblet
#

renting to train sounds like a great option for quick entry, but if you want to do it regularly and experiment and waste time with it... well...

brisk bloom
astral goblet
fervent thunder
#

I wanna watch James Cameron's Avatar 2 in the style of first

astral goblet
#

i want to see how it all interweaves with the sarah conner chronicles

brisk bloom
#

amazing series

astral goblet
#

john was a little emo but otherwise great

#

hedley brought it

#

oh and 90210 was in it!!

brisk bloom
#

they need to put life back into that show

astral goblet
#

90210 was ahuge hit

fervent thunder
#

I want "The Crow" extended

astral goblet
#

hell even more bruce lee movies

astral goblet
brisk bloom
#

shes a little old for that now

#

past her prime

astral goblet
#

i thhink she still could rock it. most of her stunts are wirework, not actual kickin butt. she's just a really good dancer that it looks like she's got that mass so well

#

shit she sure dont look 40

brisk bloom
#

there would be plenty of young upstart actors who could fill that role

astral goblet
#

yeah like dark legacy right?meh

brisk bloom
#

lol

astral goblet
#

plus we wont need stars. "Computer. Make for me, terminator 2.5"

#

"Judgement Day Returns"

brisk bloom
#

its crazy they killed firefly back in the day, but glad Summer got more roles after that

astral goblet
#

"This time it's personal"

#

"in 3d"

brisk bloom
#

yeah they could make anything or any one

#

its mostly all cg these days

pastel fern
astral goblet
#

don't really have a good bearing there to give a good recommend. i bought hardware instead. might rent if i have a job i can qualify for that. i'd look into raw compute power, possibly amazon cloud or something like azure

heady thunder
#

Have you guys seen that Raiden webm floating around that talks about anime waifus?

plush forge
#

does anyone do textual inversions?

river maple
#

When will Bing Chat be made a moderator of this Discord?

plush forge
#

where are styles saved in a1111?

hard iron
#

hi

lament steeple
cerulean kraken
#

anyone know if I can inpaint with control net but just the masked area?
not the whole image
I've tried.. but I get the error image doesn't match

fervent thunder
#

aftereffects adobe

spring brook
#

hi

grizzled rose
#

yo, how do i get an ig: page to work in the prompt

verbal osprey
#

why did the name change to "Stable Foundation" ?

wide juniper
#

is there any way to run 2 instances of SD? I was told to edit the settings to cache models in the ram, but that's not enough

empty barn
wide juniper
#

Thanks. How may I do that

empty barn
#

Of course you also need enough VRAM to fit both instances.

wide juniper
#

Yep I have rtx 3090. Thanks!

pastel fern
#

My friend is a big raiden fanboy

#

He needs to see it

wide juniper
lament steeple
#

That's how I use it

#

you can also use localdns where you use <pcname>.local

wide juniper
lament steeple
#

add --listen after "set COMMANDLINE_ARGS=" in your webui-user.bat

#

then to connect from your laptop, go to"http://<computername>.local:7860/"

#

replace <computername> with whatever you named your pc

#

checkpoint frustration: when you open up the cygwin terminal and type wget -c

#

I said download the damn file!

vestal dew
#

I'm adding high quality full body images to mangled hands and feet as far as SD limitations. Eyes do get mangled, but those can be produced at high quality quite often.

lament steeple
#

I'm working on collecting data for a TI for full body shots

vestal dew
#

Generate images in 3D software like Daz

#

you could generate thousands, millions of images

#

instead of rendering, just pose and screenshot...very fast

lament steeple
#

Gives me an excuse to start using iClone again

vestal dew
#

iclone same same

#

Although, when I try to use Daz images as feeder image for img2img, for some reason it always generates the character hunched over in some way

lament steeple
#

except I can do some things with I iclone I can't do with daz

vestal dew
#

I haven't tried t or a pose, just walking characters

lament steeple
#

Try lowering your denoising strength

#

That's one of the places where iclone shines. I can make a prompt edit character head shot, and use headshot in character creator to bring that character into iclone.

#

same character in my source image

vestal dew
#

wait what?

#

you can make an AI character with SD, then bring it into iClone?

lament steeple
#

Yes

vestal dew
#

Ill have to try that

lament steeple
#

lets take this to general with images. I'll show you

plush forge
#

Anyone here trained lora's with colab or kaggle?

vestal dew
#

I saw a video today showing making a talking avatar with an AI generated character

#

Ill try that

#

the question is, what can you do with the character once in iClone? It's expensive for animation for head and body...even just talkimg

lament steeple
#

Yeah, but I already have it

#

Great for motion editing

astral goblet
#

thin plate motion splines? i think its called

#

been meaning to look inot that

vestal dew
#

You have something like Faceware and some motion capture?

lament steeple
#

I have a lot of mo-cap files

vestal dew
#

ah

lament steeple
#

also have a mo cap rig

astral goblet
vestal dew
#

I do too, but never the motions I need

#

Mo Cap rig is expensive

mellow bison
#

/dreaming shark robot swims underwater

lament steeple
#

no generator in here ;0

#

🙂

vestal dew
mellow bison
#

//dream shark robot swims underwater

lament steeple
vestal dew
#

the image generators are gone

astral goblet
plush forge
#

Anyone here trained lora's with kohya-LoRA-finetuner?

vestal dew
#

I cant get controlnet to work

plush forge
#

@olive crystal is it your's?

covert axle
#

Hi everyone! Quick question, clip guidance when running local, do we have that yet?

wide juniper
#

Some LoRAs fro Civit.ai don't come with trigger words on the page. Does that mean those LoRAs don't need trigger words?

hollow scaffold
#

Does anyone know of an Ai tool (SD or otherwise) that can take a wide variety of different quality product or portrait pictures and make them more consistent with each other?

So for example, let’s say I have 100 pictures of products of roughly the same size on roughly the same location but they all have different quality and lighting, could an Ai analyse them and make them into a consistent series for a webshop?

low rain
wide juniper