🤝｜tech-support

[steady mountain]

There is documentation available online if you want to do an in depth investigation. In short, a lot, at least the last time I looked.

[gaunt vault]

If you host stability on a gpu server are you able to use the platform for free generating images and videos? Or do you still have to pay for credits?

[vocal burrow]

if you host it, it's your own, just like would be a local instance.

[gaunt vault]

Ok if I host it then I wouldn’t need to buy any credits correct?

[vocal burrow]

yup

[gaunt vault]

Also are you able to host stable diffusion 3 med and stable video diffusion?

[gaunt vault]

Thanks for clarifying

[vocal burrow]

sure

[gaunt vault]

Awesome thanks!

[vocal burrow]

stable video I haven't played with too much

[gaunt vault]

Looks good from what I see. Just wasn’t sure if this could be hosted on a gpu and used for free. Excited to hear that

[vocal burrow]

you'll need a big beefy gpu to run stable video tho

[glass sable]

I installed a custom node in comfy and now won'te start up. Windows 10 4090. The following is the console log:
Traceback (most recent call last):
File "C:\Program Files\comfy\ComfyUI\main.py", line 80, in <module>
import execution
File "C:\Program Files\comfy\ComfyUI\execution.py", line 11, in <module>
import nodes
File "C:\Program Files\comfy\ComfyUI\nodes.py", line 21, in <module>
import comfy.diffusers_load
File "C:\Program Files\comfy\ComfyUI\comfy\diffusers_load.py", line 3, in <module>
import comfy.sd
File "C:\Program Files\comfy\ComfyUI\comfy\sd.py", line 5, in <module>
from comfy import model_management
File "C:\Program Files\comfy\ComfyUI\comfy\model_management.py", line 119, in <module>
total_vram = get_total_memory(get_torch_device()) / (1024 * 1024)
File "C:\Program Files\comfy\ComfyUI\comfy\model_management.py", line 88, in get_torch_device
return torch.device(torch.cuda.current_device())
File "C:\Program Files\comfy\python_embeded\lib\site-packages\torch\cuda_init_.py", line 778, in current_device
lazy_init()
File "C:\Program Files\comfy\python_embeded\lib\site-packages\torch\cuda_init.py", line 284, in _lazy_init
raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled

C:\Program Files\comfy>pause

I have uninstalled and reinstalled Torch and CUDA toolkit.

When attempting to upgrade torch, I get "ERROR: Cannot uninstall 'TBB'. It is a distutils installed project and thus we cannot accurately determine which files belong to it which would lead to only a partial uninstall."

Upon uninstalling TBB, I am informed the installation can't proceed because TBB is absent.

[ivory bramble]

I've been stuck for a while. I reinstalled A1111 recently. It used to work fine before but I wanted a clean start. Now, when I'm generating images, my PC completely shuts down. I followed the guide for NVIDIA users. I have a 1080ti so I used --medvram-sdxl. I've tried messing around with my Windows power settings. There shouldn't be an issue, especially considering it worked fine before. I shouldn't have re-installed XD

[bright basin]

i have question in install a1111

RuntimeError: Couldn't checkout commit cf1d67a6fd5ea1aa600c4df58e5b47da45f6bdbf for Stable Diffusion.

what means for it?

And in pinned message "install guild"

i'm use in nvdia

is all should be download or choose 1 of it is be ok?

[ornate elk]

Only 1 is needed

[ornate elk]

Looks okay

It couldnt update because of the changes made to the webui-user.bat. but the second time it worked it seems

[ornate elk]

How much RAM do you have?

[gentle lava]

@ornate elk

im brazilian so the language is in portuguese

but the error is the same

[ornate elk]

you forgot the git clone before the https

[gentle lava]

i will try

@ornate elk like this?

[bright basin]

i try to run again and show up different error message

[vocal burrow]

Copy paste your webui-user.bat somewhere

git checkout --theirs webui-user.bat

git pull

Put back your webui-user.bat

[vocal burrow]

Yes

[gentle lava]

@vocal burrow man why

[ornate elk]

What's your GPU?

[gentle lava]

rx 580

amd

[ornate elk]

Then you installed the wrong webui version

Thats not the one for AMD

[gentle lava]

can you give me the link ?

[ornate elk]

Are you good with PC stuff?

Because I have two AMD guides.
One is easier and one requires multiple difficult steps

[gentle lava]

+/-

[ornate elk]

The difficult guide is better in performance

[gentle lava]

which one do you recommend

[ornate elk]

You can find the guides in the pinned messages of this channel. The first link.
There you can follow the
Automatic1111 ZLUDA guide
Thats the faster version.

The difficult one which require a manual step at the end

That I can give you

[gentle lava]

this?

[gentle lava]

give me

[ornate elk]

You need to follow my Zluda guide. When your done with the guide then do this:
Go into the stable-diffusion-webui-amdgpu folder.
Click in the explorer bar (not searchbar)
There Type cmd and hit enter.
Then type and run the three commands on by one:
venv\scripts\activate
pip uninstall torch torchvision torchaudio -y
pip install torch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1 --index-url https://download.pytorch.org/whl/cu118

[gentle lava]

first i have to see the guide, right?

[gentle lava]

after see the guide i do this, right?

[ornate elk]

Click the link in the pinned messages

[gentle lava]

i did

[final vector]

Is the API currently down?

[gentle lava]

@ornate elk i have to create a folder in user, right?

[ornate elk]

yes, on C drive, not directly in users

[gentle lava]

so i create here?

[ornate elk]

Yep

I'm AFK for a bit. Can help later

[torpid osprey]

Using SD next and does it just take its time to first generate or is something messed up on my end

I have it under my 2nd ssd A:\Ai\automatic

[ornate elk]

What's your GPU?

[gentle lava]

@ornate elk give me the easy mode

man

i cant

[ornate elk]

The easy mode is the Automatic1111 with Directml guide.
But its more limited in vram usage.

[torpid osprey]

using a 2070 max q

so its a laptop but im not running anything that requires any gpu at all

[gentle lava]

bro I can't download zluda

i did everything right

[ornate elk]

do you get an error?

[gentle lava]

yes

[ornate elk]

hmm okay, why not using Automatic1111 ? its easier to get working than sd.next (and easier to understand)

[ornate elk]

can you show it?

[torpid osprey]

yeah i went to fooocus

its working

[gentle lava]

''mat1 and mat2 must have the same dtype, but got Float and Half''

[torpid osprey]

i went to sd next from automatic1111 not working

[ornate elk]

what error did you get with auto1111?

[ornate elk]

can you show me the full cmd log?

[gentle lava]

okay

[torpid osprey]

couldn't install it at all

for some reason

my wifi is kinda shit

and it cuts off for like milliseconds which fucks with the uh

downloaders when i use cmd for anything

thats why i download the models on free download manager and install from there but idk what folder to put the installation to from automatic1111

fooocus is working though

[ornate elk]

you can install the AllInOne version that comes as a zip file from here:
https://github.com/AUTOMATIC1111/stable-diffusion-webui/releases/tag/v1.0.0-pre

sd.webui.zip

then run update.bat
then run the webui-user.bat (but before add --xformers --medvram --no-half-vae to the Commandline_ARGS= in it)

Models then go into models/stable-diffusion folder

[torpid osprey]

thank you 🙏

[torpid osprey]

no wait this is the error i get

[ornate elk]

whats your python version?

[torpid osprey]

uh

lemme check

i have

3.10 or smth

thats what it wanted

3.10.10

[ornate elk]

is that from my link or from your first attempt?

[torpid osprey]

both

it happened with the normal installer and this one

[ornate elk]

it makes no sense. have you run the update.bat before?

[torpid osprey]

ok

ill run again

ok its downloading torch

once again same issue

everything is up to date

i can try with a vpn

[ornate elk]

can you delete everything from auto1111 and then follow my install guide:?

[ornate elk]

are you from asia, russia, iran or phillipines?

[torpid osprey]

no 😭

[ornate elk]

then you shouldnt need a vpn

[torpid osprey]

i did the zip file

that error is from the zip code

[ornate elk]

okay show me the full cmd log

[torpid osprey]

tbf my wifi is shit and i can't download anything that needs command prompt

that is indeed everything

[ornate elk]

and you cant use ethernet?

[torpid osprey]

i am connected with ethernet LOL

[ornate elk]

why do you say your wifi is bad then xD

[torpid osprey]

because it doesn't work when i need it too and cuts off randomly

[ornate elk]

but now your on ethernet

so it doesnt use wifi

[torpid osprey]

i mean my internet*

[ornate elk]

then you may need to change your DNS server

that can help sometimes

[torpid osprey]

alright time to do that

i know how to do that

roughly

[ornate elk]

set primary to 1.1.1.1 and second to 8.8.8.8

[torpid osprey]

ok i forgot how ngl

i used to know how on windows

i know its control panel though

[ornate elk]

ethernet adapter settings

ipv4 settings

[torpid osprey]

ok

done

running

[ornate elk]

before running again do the following

[torpid osprey]

accidentally ran but instead of an error 1 its an error 2 this time

which means its progress

[ornate elk]

Delete the venv folder, then open up a cmd:
type and run:
pip cache purge
ipconfig /flushdns
sfc /scannow

first one deletes temp files of the broken installs
the second one clears old dns entrys
the third one checks and repairs corrupt system files

[torpid osprey]

venv where?

[ornate elk]

it should be inside the folder where the webui-user.bat is

inside the webui folder

[torpid osprey]

ok

doing scan now

[ornate elk]

hows it going?

[torpid osprey]

76%

atm

ok yeah it found shit that was corrupt

[torpid osprey]

opening the webui userbat

[gentle lava]

i press to generate and do it

[torpid osprey]

venv "A:\ai\sd.webui\webui\venv\Scripts\Python.exe"
Python 3.10.10 (tags/v3.10.10:aad5f6a, Feb 7 2023, 17:20:36) [MSC v.1929 64 bit (AMD64)]
Version: v1.9.4
Commit hash: feee37d75f1b168768014e4634dcb156ee649c05
Installing torch and torchvision
A:\ai\sd.webui\webui\venv\Scripts\python.exe: No module named pip
Traceback (most recent call last):
File "A:\ai\sd.webui\webui\launch.py", line 48, in <module>
main()
File "A:\ai\sd.webui\webui\launch.py", line 39, in main
prepare_environment()
File "A:\ai\sd.webui\webui\modules\launch_utils.py", line 380, in prepare_environment
run(f'"{python}" -m {torch_command}', "Installing torch and torchvision", "Couldn't install torch", live=True)
File "A:\ai\sd.webui\webui\modules\launch_utils.py", line 115, in run
raise RuntimeError("\n".join(error_bits))
RuntimeError: Couldn't install torch.
Command: "A:\ai\sd.webui\webui\venv\Scripts\python.exe" -m pip install torch==2.1.2 torchvision==0.16.2 --extra-index-url https://download.pytorch.org/whl/cu121
Error code: 1
Press any key to continue . . .

..

[ornate elk]

no module named pip

mostly a third party antivirus programm broke your python install

[torpid osprey]

yuh

i dont have an anti virus

i mean i have one installed but its never open / running

not in the background either

i leave it off unless needed

[ornate elk]

what the programm?

[torpid osprey]

malware bytes

like

[ornate elk]

mostly AVG, AVAST, MCAffe, Norton are the ones that make issue (and also are the ones nobody should ever use)

[torpid osprey]

im tech savvy in a lot of places so im not entirely dull on this

[ornate elk]

thats good

[torpid osprey]

but anything new to me like this specifically

lost on me

[ornate elk]

but you need to uninstall python 3.10.10
then install python 3.10.11 64bit fresh

[torpid osprey]

kk

[ornate elk]

then delete the venv folder again before launching it

[torpid osprey]

ill do that in a second

because sure sd isnt working BUT

foocus is

and im running that in the bg

[ornate elk]

whats the model and settings?

[torpid osprey]

and i'd rather use sd

[ornate elk]

yea fooocus doenst have that many featurs like auto1111

also its slow on lower end gpus (as it only uses sdxl)

[torpid osprey]

yea

thats why

ok

kk

10.10 is unstalling and i have 10.11 ready to install

ngl i hate when apps use python because for some reason it hates me and always fucks itself up

[ornate elk]

xD better than using conda

[torpid osprey]

like on my server pc it took me forever to get a discord bot running 😭

i had to uninstall reinstall and all that

then manually add the python packages by hand

it was a hassle 😭

yeah so chocolatey didnt want to install

it correctly

ok so its uninstalled now

other then my desktop killing itself

we're good so far

ok sick fixed it

k almost done

ok

opening webui-user

[ornate elk]

wait

[torpid osprey]

?

[ornate elk]

whats your gpu ?

[torpid osprey]

2070 super

8gb of vram

[ornate elk]

then you first need to edit the webui-user.bat

[torpid osprey]

oh yea

i forgot

ok

[ornate elk]

for the best performance args

[torpid osprey]

i did that last time

--use-cuda?

[ornate elk]

--xformers --medvram-sdxl --no-half-vae

[torpid osprey]

ok

added

[ornate elk]

ok then your ready

[torpid osprey]

still unable to find the requirement that satisifes torch

:sob

[ornate elk]

you installed the 32bit python :c

[torpid osprey]

did i

[gentle lava]

default

[ornate elk]

yep

[torpid osprey]

fuck

ok

im so smart fr fr

[ornate elk]

reinstall ypthon, then again delete the venv folder

[gentle lava]

@ornate elk

[ornate elk]

and the full cmd log? and i mean the complete one

[gentle lava]

its is big

i copy everything

[ornate elk]

copy and send it as text

[torpid osprey]

ok

trying again now

same issue 😭

as i said my internet hates me or its just not even me at this point 😔

[ornate elk]

try run pip cache purge in a cmd again and delete the venv folder again

[torpid osprey]

ok

[ornate elk]

you have to do the last step i mentioned to you

[torpid osprey]

trying again

[ornate elk]

@gentle lava Go into the stable-diffusion-webui-amdgpu folder.
Click in the explorer bar (not searchbar)
There Type cmd and hit enter.
Then type and run the three commands on by one:
venv\scripts\activate
pip uninstall torch torchvision torchaudio -y
pip install torch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1 --index-url https://download.pytorch.org/whl/cu118

[gentle lava]

oh

gonna try

[torpid osprey]

oh shit its installing 🔥

once in a lifetime event right here

ok taking its time on installing collected packages

[gentle lava]

@ornate elk can i put other comand or wait

[ornate elk]

you can there put the other two commands

[torpid osprey]

installing requirements

[gentle lava]

pip install and this link https is separated?

[ornate elk]

no

[gentle lava]

copy 2 and put right?

[ornate elk]

pip uninstall torch torchvision torchaudio -y

this

then this:
pip install torch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1 --index-url https://download.pytorch.org/whl/cu118

[gentle lava]

okay

[torpid osprey]

okay now its installing something else 🔥

[gentle lava]

@ornate elk everything okay?

[ornate elk]

looks good#

[gentle lava]

gonna try generate again

@ornate elk

[torpid osprey]

okay it works now

[gentle lava]

😭

[torpid osprey]

but i had to get rid of the low vram since it was kinda fucking up my models that i had and didnt let them work

[ornate elk]

have you installed hip sdk 5.7 or 6.1 ?

[torpid osprey]

changed port and added listen to it

🔥

thanks for all the help my dude

[ornate elk]

no problem 🙂

[ornate elk]

check it in the system control

[gentle lava]

where i see

[ornate elk]

systemcontrol

then on uninstall programms

then you get a list with everythin that is installed

[gentle lava]

this?

[ornate elk]

yes

[ornate elk]

okay

6.1 should work

did you replaced the files from library folder in rocm ?

in my guide thats step 4 and 4.5

[gentle lava]

no i think

[ornate elk]

you need the RocmLibs.zip

[ornate elk]

then thats the missing step

[gentle lava]

gonna try

@ornate elk whats the .zip file in step 4.5

[ornate elk]

RocmLibs.zip

[gentle lava]

man i dont understand

i have to rename the library to old library after that copy and wheres i put

[ornate elk]

You have copy the folder

So you have to library folders

Then rename one of them to old_lirbary

And in the other one you paste the content of the RocmLibs

[gentle lava]

but there is only one

i have to create?

[ornate elk]

No you have to copy it

Copy the Library folder

Then you have two

[gentle lava]

i did

[supple fable]

Uhm.. I just recently reinstalled SD after not using it for a year, why are all my images deep fried? I got the MSE 84000 vae or whatever, but everrything has this little red fringe to it and idk why

[gentle lava]

but continue 1

[gentle lava]

i copy but continue 1 library

i will record

[ornate elk]

I'm not at the PC RN to check

[gentle lava]

what i have to do

why its so hard bro 😭

[zealous portal]

Is it possible to negative prompt to stop text from being generated soley by doing it in the prompt? Right now im using a SDXL-lighting api, and for my project there cannot be text in the generated images.

[queen spindle]

pending generations on stable audio making me unable to generate

[ornate elk]

just paste it after copy it

lol

like copy and paste any file (right click copy, then right click paste)

[gentle lava]

oh

im so smart huh

[ornate elk]

and then the folder has to be renamed library

and the other one is the old_library

and you put the stuff from RocmLibs into the library folder

After that you need to restart the PC

and then try again to create an image

[gentle lava]

put these in libraly?

or 6.1

@ornate elk

[ornate elk]

not this

you only need the stuff from the RocmLibs.zip

put these files into the library folder

if you dont have the RocmLibs zip here you can download it:
https://github.com/brknsoul/ROCmLibs/raw/main/ROCmLibs.7z

[gentle lava]

thanks

i dont have it

@ornate elk error again

[ornate elk]

have you moved the files into library folder and restarted the PC ?

[gentle lava]

yes

[ornate elk]

okay then download this model. https://civitai.com/models/4384/dreamshaper
put it into the models/stable-diffusion folder
and delete the 1.5-ema pruned model.

maybe the 1.5 model is broken

[gentle lava]

When I did it, it said that I already had the files in the folder and if I should replace them and I did

[gentle lava]

okay

i have to put git pull in the webui-user? @ornate elk

[ornate elk]

no

but you can

for auto updates

[gentle lava]

@ornate elk

error again

'NoneType' object has no attribute 'lowvram'

[ornate elk]

whats in your webui-user.bat?

[gentle lava]

mine only have webui-user

not .bat

[ornate elk]

yea its the .bat

win11 just hides the extenions

[gentle lava]

@ornate elk

[ornate elk]

why is it empty lol
you need to add --use-zluda --medvram-sdxl --update-check --skip-ort
to the line Commandline_ARGS=

then save and relaunch it

[gentle lava]

in my mind it dint😭

@ornate elk same error

[ornate elk]

okay try edit the webui-user.bat again and there add --no-half --upcast-sampling

[gentle lava]

FINALLY

@ornate elk can a change medvram for xformers ?

its slow

[ornate elk]

xformers is for nvidia only. it wont work

[gentle lava]

is there another command then?

[ornate elk]

you can try to change --no-half to --no-half-vae

[gentle lava]

okay

[ornate elk]

Is the GPU a 4gb or 8gb rx580?

[gentle lava]

8gb rx 580

[gentle lava]

@ornate elk How many minutes does your stable difussion take to generate

[ornate elk]

How many does it takes for you?

It shouldn't take longer than 2 mins for a 512x512 image

[gentle lava]

10

[gentle lava]

it changes when i active hires fx

[ornate elk]

And without hires fix?

30 steps

512x512, Euler a

[gentle lava]

gonna try

[gentle lava]

i tried but show the same error

intel xeon e5 2680v4 and rx 580

why its dont work

my config is not bad

[ornate elk]

How much RAM?

[gentle lava]

16gb

[ornate elk]

Ok

[gentle lava]

3200

[ornate elk]

The rx580 isnt that great for stable diffusion as its an older gpu

The --no-half is slowing down but without you can't generate :/
Its strange as some other people with rx580 didnt needed it

Make sure your GPU driver is updated

[gentle lava]

i updated

[ornate elk]

@gentle lava when using hires fix always set the hires steps to 10

That is faster than 0

[gentle lava]

well

3 minutes

withouth hires fx

but

its ugly

@ornate elk where i can download upscalers

[ruby yoke]

you sure you're using the right VAE? what are your settings?

[vocal burrow]

https://openmodeldb.info/

[ruby yoke]

i might suggest trying a different sampler

[gentle lava]

its work just is slow

[gentle lava]

whem i try to put --no-half-vae gives error

[vocal burrow]

if you're already using --no-half, it includes --no-half-vae

But given how old your gpu is I wouldn't hold my breath

[gentle lava]

all rx 580 is slow in stable difussion?

[vocal burrow]

it's a 7 years old AMD gpu.

[gentle lava]

yeah but its good

im playing red dead redemption in 70 fps

red dead redemption 2 of course

[ornate elk]

Yes, they are not that fast for ai stuff

Maybe on Linux they would be a bit faster than on Windows

[vocal burrow]

AMD gpus do not support CUDA and everything that comes with it. Meaning it's not really built for non gaming kind of workflow. And the fact that's it's a old AMD gpu means it's not getting attention/support from the AMD community which prefer to focus on their latest cards

[gentle lava]

oohh

so

I'm Brazilian, things here are expensive, Rx 580 here is worth R$ 500 and practically 500 dollars

1 dolar is r$ 5

In other words, here in Brazil things are 5 times more expensive

[vocal burrow]

yeaaaaaah brazilian hardware market is ....

good luck

[gentle lava]

[ruby yoke]

can you check on newegg and see if you can find something there that can ship to you?

[gentle lava]

whats is newegg

[ruby yoke]

https://www.newegg.com/ a website that sells tech products

[gentle lava]

well dont have brazil in country options

[ruby yoke]

drat

[gentle lava]

Why did you want me to check?

i can search on youtube if have one way

[plain oak]

Does anyone know how can I get to the Foocus prompt screen again after shutting down my computer?

I accessed 127.0.0.1:7865, but it only tells me the site can't be reached.

[ruby yoke]

cause they are a very good site to get equpiment from, and i was hoping they'd be a better option than trying to go through a local shop that might charge you an arm and a leg

[ruby yoke]

start the server first?

[plain oak]

I don't know how.

[ruby yoke]

how did you run foocus before?

[plain oak]

Once I finished running run. bat, and installing the required files, it opened automatically.

[ruby yoke]

127.0.0.1:7865 <--- 127.0.0.1 is the ip of local host - i.e. your machine and 7865 is the port the server is running on. but you shut down your machine, and foocus isn't running on that port now

suggest doing a shut down, leave it off for 10 seconds, start it, and then start foocus however you normally would to get a nice, clean, reload

[plain oak]

Sorry, I neglected to clarify I'm a total beginner. This will be my second ever time using the program. I don't have a "normally would". After the run.bat installed all the files at first, Foocus started running on my machine, as though, from my perspective, some magic had happened.

I have no idea what to do to run the server again.

[ruby yoke]

okay. can you try the shutdown steps up to the point where I said "like you normaly would", and then try the run.bat file again? just to make sure that everything has loaded into memory correctly.

[plain oak]

"Start it" is one such step, but I don't know how to do that.

Is running run.bat again "starting it"?

I don't have to worry about a second intall process taking place?

[ruby yoke]

step one. close all the programs running on your machine. step two: shut down (not restart) your machine. step three: leave it powered off for at least 10 seconds so all the electronics discharge. step four: power the machine back on. step five: after it's powered back on and everything has finished loading, then find the run.bat file and try running it again.

[ruby yoke]

the only reason that would happen is if it didnt' finsh something

[ornate elk]

Why shutdown and not a clean restart (which restarts the drivers too) ?

[ruby yoke]

because windows cache. and we want to clear that

[normal birch]

is it normal that I'm getting 100% usage of a random CPU core constantly with forge?

GPU is being used to render, for sure, and I'm not noticing any bottleneck problems buuuut, seems kinda weird

[plain oak]

I think I mischaracterized my situation somehow. It was not necessary to restart my machine at all. The problem I was facing was just trivially fundamental: I had no idea how to start the program again. But running run.bat again worked.

Thank you for your patience.

[ruby yoke]

very good 🙂 glad you got it working

[gentle lava]

why the imagens have low quality

[ornate elk]

Because its 512x512 and it seems the Upscale didnt worked

What does the cmd shows?

[ornate elk]

Hmm looks okay

Try an other sampler

DPM++ 2m

[ornate elk]

Also for a speedup you need to disable/shutdown (not pausing) Wallpaper engine

[gentle lava]

okay

@ornate elk same

low quality

[torpid osprey]

ok im back 🔥

its half taking forever to do anything

oh yeah i forgot abt wallpaper engine ngl

forgot i had it enabled

nvm

i figured it out

but yeah ur cool

[ornate elk]

Looks normal.as its an 512x512 resolution

[gentle lava]

comcept in civitai is lora right? @ornate elk

[worn bane]

How i can generate images like this it's looks like anime drawing so what model or lora should i use?

[gentle lava]

@vocal burrow How do I put upscaler in stable diffusion?

[worn bane]

LoL i found the The way

[bright basin]

so i need to clear user.bat and clone this two(git checkout --theirs webui-user.bat

git pull) in user.bat and run again?

[glass sable]

Anybody here have experience with IF_AI Nodes?

[quiet geyser]

Лисе́нок играет в снегу

[fluid grove]

Hey guys! I just installed SD with pinoccio and I'm getting messed up faces

any idea whats going on?

[dusty apex]

Is it possible to use and train SD models effectively using a 7800xt?
I consider buying them in the near future to have a local gaming machine and to also use and train AI on. If anyone knows.

[deft violet]

what is the best anime model for sd 1.5

[hybrid bison]

Hello ! May someome help with this ? Have a nice day

[vocal burrow]

ESRGAN upscalers go into stable-diffusion-webui\models\ESRGAN

[fair oxide]

anyone knows to to make text not blurry ?

[jagged glade]

what the latest version of sd its 2.1 or sdxl or anything else?

[ornate elk]

Currently its SD3

[jagged glade]

It's out for automatic 1111?

[ornate elk]

With the next update

[old parrot]

Is SD Comfy better for video to video generations?

Or what UI is suggested on a 4060 Ti?

[gusty salmon]

I cant get ip-adapter to show up as control type in controlnet v1.1.313... I downloaded ip-adapter-faceid-plusv2_sdxl.bin and put it in both stable-diffusion-webui\extensions\sd-webui-controlnet\models and stable-diffusion-webui/models/ControlNet and restarted the UI.

Am i missing something? Do i have to install something additional first?

This is driving me nuts for 2 hours already 😵‍💫

[jagged glade]

really bad question after 1 year of SD but i fotgot how to get same seed as last image?

[gusty salmon]

i think you press one of thes two buttons,

should be the green one

(after generating the image)

[jagged glade]

ohhh right

[jagged glade]

how to make same background stable diffusion? any videos about it?

[gentle lava]

thanks

[river ivy]

Do you guys know why i get unmatched colors? I set denoise to 0.00 on purpose because i wanted to get the same output as input, but somehow I get faded colors

[solemn moon]

Inpaint

Not masked

Speaking of which

In paint tools are so 2023

😅

[jagged glade]

its too lengthy process for few hundred images

[river ivy]

I have found that the custom node is not the problem, but vae decode. Just why?

[gentle lava]

why its ugly

[dry ivy]

bad vae or model

[gentle lava]

its xxmix9

[dry ivy]

You need to provide more information than just an image

still, more is needed

[gentle lava]

what u wannt

[karmic crown]

copy/paste what you got below the image

or this part

[jovial mango]

i downloaded this lora and it wont show up under my lora tab in the app, but all my other loras do, does anyone know why? https://civitai.com/models/43516/megami-no-cafe-terrace-pdxl-and-sd15

[karmic crown]

btw. since that is not a technical issue or bug (it is a prompting issue) better ask in #📝｜prompting-help

[karmic crown]

it will be shown only with 1.5 models, with xl models it will not be shown

1.5 models show only 1.5 loras
xl models show only xl loras

[jovial mango]

so if i change my checkpoint it'll show up? im new to this lol i dont understand a lot

[river ivy]

u're right. Something is wrong with model. Not sure what.
I tested it on other models (non VAE and VAE baked) and everything work perfect

[karmic crown]

yes, if you switch to 1.5 model it should be there

[jovial mango]

https://civitai.com/models/257749/pony-diffusion-v6-xl do you know if this is a 1.5 model?

[karmic crown]

no

[river ivy]

if you see XL it means its stable diffusion XL model

[jovial mango]

oh ok thanks

[karmic crown]

xl, pony, turbo, lightning... none of those

[jovial mango]

do you know any good 1.5 models to use cause i think all of mine arent

[karmic crown]

realistic? anime? both?

https://civitai.com/models/4384/dreamshaper?modelVersionId=128713

[jovial mango]

anime

[dry ivy]

Anydiffusion or abyssal maybe?

[jovial mango]

could you send me a link

[gentle lava]

it was the model

idk why its ugly in xxmix9

[river ivy]

anything v3 or above

[karmic crown]

https://civitai.com/models/6755/cetus-mix?modelVersionId=105924

https://civitai.com/models/66/anything-v3

[jovial mango]

thank you 🙏

[karmic crown]

btw. what is new link for https://civitai.com/models/118561/anything-kl-f8-anime2-vae-ft-mse-840000-ema-pruned-blessed-clearvae-fp16cleaned

@jovial mango try this vae https://civitai.com/models/23906?modelVersionId=28569

to get better colors

[jovial mango]

👍 appreciate it

[karmic crown]

please doublecheck if it .safetensor version

if it is not, download kl-f8-anime2.safetensors from here https://huggingface.co/hakurei/waifu-diffusion-v1-4/tree/92e0470e5a9601cdd85a3db349a591df9ed7e2ad/vae

[ivory bramble]

16GB of RAM

[jovial mango]

it is

[gentle lava]

@vocal burrow i have to put these in lora right?

[vocal burrow]

No, those are embeddings

they go in stable-diffusion-webui\embeddings

[jovial mango]

i used the cetus mix and anything v3 models but the lora still didnt show up, do you know a fix?

[gentle lava]

who i put embeddings in negative prompt

forget it, I already got it, but thanks anyway

[ornate elk]

Then you should check your Windows Pagefile. Make sure its only enabled for the C drive and disabled for other drives.
Set it to 16000min and 24000max
Then restart the PC

[karmic crown]

doublecheck if you downloaded file correctly and completely and check if you put it in right folder

it should go in Lora folder inside models

https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/16223#issuecomment-2241501357 <- this is how you can check if file is downloaded correctly

be sure that you are comparing the right hash

i explained how to change them

[restive sparrow]

I unfortunately ran into the bastard NansException bug yesterday out of no where. Didnt do a damn thing to my setup. So as suggested I started using that modfied Vae and disabled nohalfvae per its instructions. Hope these things dont mess with checkpoints outputs or performance.

[gentle lava]

@ornate elk bro why my imagens looks ugly

(masterpiece, best quality),1girl with long white hair sitting in a field of green plants and flowers, her hand under her chin, warm lighting, white dress, blurry foreground
Negative prompt: bad-hands-5 negative_hand-neg poorly Rendered face
poorly drawn face
poor facial details
poorly drawn hands
poorly rendered hands
low resolution
Images cut out at the top, left, right, bottom.
bad composition
mutated body parts
blurry image
disfigured
oversaturated
bad anatomy
deformed body features
Steps: 25, Sampler: DPM++ 2M, Schedule type: Karras, CFG scale: 10, Seed: 801135180, Size: 258x390, Model hash: cbfba64e66, Model: counterfeitV30_v30, Denoising strength: 0.5, Hires upscale: 2, Hires upscaler: R-ESRGAN 4x+ Anime6B, TI hashes: "bad-hands-5: aa7651be154c, negative_hand-neg: 73b524a2da12", Version: v1.9.3-amd-32-g67fdead6

Time taken: 10 min. 24.4 sec.

[gentle lava]

@vocal burrow

[prime mulch]

ive noticed theres no pre processor review

[gentle lava]

what is this

[hollow ginkgo]

Hi, maybe an unrelated question, but is it possible to automatically recharge credits for sd 3 api? right now we have to top up our api credits, which is a bit cumbersome and doesnt scale well.

or is there a pay per generation option ?

[thorn sage]

What UI would I need to use for videos like this? And would 16gb VRAM be enough?

[fair oxide]

trying to run stable diffusion on automatic1111 with zluda and getting this rust error

my gpu is a rx 6600 and I'm on windows 11, anybody went through this?

I followed this guide

[gentle lava]

tu e br?

[fair oxide]

sim 🤔

[spring pike]

my whole pc freezes when stable diffusion gets too 99 percent and im not sure whats wrong

[gentle lava]

finalmente achei um br

alias

[gentle lava]

tu sabe me dizer o pq a imagen sai assim?

[fair oxide]

tô aprendendo a usar o stable diffusion agora

tô tendo só problemas com placa de video amd

se arrependimento matasse tinha comprado NVIDIA

[gentle lava]

kkkkkkkkkk

dboa

a minha e amd e ta dboa

[fair oxide]

pô, posso te dar pm?

[fair oxide]

anyone?

[gentle lava]

nn entendi

[fair oxide]

se poderia te mandar mensagem privada

[gentle lava]

e q eu nao sei explicar pow

um cara q me esinou a baixar

ele ta off

@ornate elk

[fair oxide]

qual é a sua placa de video?

[gentle lava]

rx 580

[fair oxide]

tá no windows?

[gentle lava]

10

[fair oxide]

[limpid horizon]

Hello, what would be the best model/pipeline for inpainting new hairstyles on men and women of all hairstyles? Thanks.

[fair oxide]

why do my images always come out like this trying to use an SDXL model? I'm using Automatic1111 Master branch is there something I gotta do besides DL a model?

[main roost]

I just wanted to give my condolences on Biden, hes withdrawing from the race. I guess we wont see in November after all...There is a God after all though.

[onyx solar]

You're arguing a point I wasn't making. That comment wasn't about whether or not Biden would remain the candidate, it was about who the winner will be in November, hence why I said what I did.

But again, not trying to continue the conversation in #🤝｜tech-support. So I'll drop another one of these for ya: ✌️

[jovial mango]

yeah i have it in the lora folder

its with all my other loras that work fine but this one wont show up with any of the models

[fair oxide]

Nevermind I'm an idiot and forgot to remove the vae

[solemn skiff]

Any Kohya experts around, I'm having trouble getting it back up and running after coming back for a while

Its setup all the same, except now I'm getting permission errors

[slender fossil]

I'm trying to use the Kohya GUI to train SDXL LoRAs and I'm just getting this error in the terminal any time I try. I've tried uninstalling/reinstalling accelerate, updating it, and installing it through activate as someone told me to do, but nothing's been working. This is also the only error message that shows up in the terminal, everything else is just Kohya booting up normally. Anyone know what to do?

[empty cipher]

Why do my LoRA always come out like crap?

[slender fossil]

Could be a whole lotta reasons, from bad datasets, settings being wrongly set, or even just a problem with whatever program you're using tbh

[empty cipher]

Let me put it another way

I've read many many guides and even done some research into the underlying mechanics of how SD and model training works

I've assembled an excellent dataset. 150 high definition images, hand-captioned. And I've tried training the LoRA at least 20 times with different parameters. High learning rate, low learning rate, higher rank, lower rank, 512x512, 1024x1024

I've done a handful of fine-grained comparisons, creating the same LoRA but only adjusting a single parameters at a time. No matter what I do I can't get the LoRA to even enter the uncanny valley let alone come out the other side

XD I've reached the point, I think, where I need some help

The only thing I can think of at this point is that I'm writing bad captions, but they should be appropriate. Maybe a bit too detailed?

[slender fossil]

Damn, yeah I'm not too sure myself what the problem could be. Might be a weird contradiction with whatever model you're using, or maybe the prompts are just messing it up? I dunno, someone more knowledgeable would need to help with that

[slender fossil]

That could also def be the issue, from what I've seen/heard the captions should be somewhat simple but that might not always be true don't take my word on it

[empty cipher]

low-angle-shot of adult ram wearing sports-jersey and grinning, white-background, simple-background, low-angle, three-quarter-turn, knees-up, solo-focus, male-focus, adult, male, ram, mountain-goat, grey-fur, brown-fur, horns, curly-horns, red-eyes, furrowed-brow, 4-fingers, red-and-black-shirt, tank-top, red-and-black-shorts, smug, smiling, smirking

Yes, I'm a furry XD

I guess I could strip all the camera angle and other "meta" stuff

[late tree]

The ADetailer extension is enabled and is showing in the UI, but it doesn't seem to be working and there is an error about it in the log:

[trim isle]

how can I clear pending generations on the stableaudio.com website? I have stuck generations that have been pending for days and even with a paid account now it will not let me make more generations

[worn bane]

What is the name of this model?

[ornate elk]

Can you post the full cmd log?

[ornate elk]

What's your GPU and what's inside your webui-user.bat?

[ornate elk]

Use cfg 7 and maybe an other VAE. Counterfeit 3.0 has a bad vae

[gentle lava]

@ornate elk why 512 x 768 go to 1024x1436

its normal?

when i generate its like 3x more long to generate

[spring pike]

My gpu is a rtx 3060 and what do you mean by inside my Web user

[ornate elk]

When you edit the webui-user.bat. whats inside it?

[restive sparrow]

Would anyone know why during processing my inpaints look fine but once done its broken garbage. thank you
EDIT: Adetailer helps but i still wonder whwy

[spring pike]

sampling steps 25 or more and some hires steps to 10? if thats it

[empty cipher]

I've always wondered this

My images often look best mid-processing XD

They have this gorgeous "rounded" look that gets lost when the crispness of the image is finalized

[slate sinew]

upscale video

[ornate elk]

No the file webui-user.bat

There you need to add performance Arguments like xformers etc

[ornate elk]

Do you have an example?

[stuck flume]

Any suggestions on this? I'm getting this intermittently even on very reasonable resolutions like 832x832.

"OutOfMemoryError: CUDA out of memory. Tried to allocate 6.97 GiB (GPU 0; 23.99 GiB total capacity; 10.43 GiB already allocated; 3.48 GiB free; 17.47 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF"

[solemn moon]

Check your drivers are updated

[ornate elk]

What's your GPU and what's inside your webui-user.bat?

[stuck flume]

Yes, latest drivers.

[stuck flume]

It is too long to post, what should I look for?
GPU is a watercooled RTX 4090 24 Gb.

[dark barn]

Hello, i'm trying to add the extension mov2mov but after downloading it, i don't see the tab and i don't know what to do 😦
I got some error and it's probably related but i still can't resolve them

[vocal burrow]

1/ you didn t install a1111 correctly (with git)
run those in your sdwebui folder

git init
git remote add origin https://github.com/AUTOMATIC1111/stable-diffusion-webui.git
git fetch --all
git reset --hard origin/master
git branch --set-upstream-to=origin/master master
git pull

2/ your roop and openpose-editor are also broken, make sure they re updated
3/ mov2mov hasn t been updated in, 8 months, it s deprecated

[vocal burrow]

webui-user.bat not .sh

just drag and drop the file in here

[solemn moon]

Webui-user.bat
Open that file with notepad and copy and paste here

[vocal burrow]

drag and droping the file will work too

(sometimes asking people to open a .bat is already too much :p)

[ornate elk]

In the webui-user.bat you need to add --xformers to the line commandline_args=

[stuck flume]

I was more concerned with cluttering the chat with code. 😉

[ornate elk]

Can't see that on mobile

[stuck flume]

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --xformers

call webui.bat

[ornate elk]

Looks good

Now it should generate faster and use less vram

[stuck flume]

I have changed nothing. I added --xformers two GPUs ago... 😉

[ornate elk]

Ohh okay

[ornate elk]

Then when do you get the error?

[stuck flume]

Sometimes I can generate at 2048x2048 for hours. Sometimes I can't even do 768x768 without getting the out of memory error.

[unkempt dawn]

Hi, Does anyone know how to make it so when i click on the preview img it doesent switch to the last image generated?

[vocal burrow]

looks good to me btw

might want to switch to another model or prompt differently

[fair oxide]

I got it working with SD.Next, thank you anyways

I am using stable diffusion 1.5, trying to generate a design ui for my mobile application, but it keeps generating phone images, what am I doing wrong

I'm still new to this, sorry

I asked chatgpt to generate this prompt for me

[gusty salmon]

correct me if i'm wrong, but i think SD is not meant to generate this kind of stuff

[fair oxide]

isn't it what people been using to generate stuff?

[gusty salmon]

from what i know, it cant generate proper text at all. thats why you see these weird cryptic stuff

also its only able to create generate stuff that doesn't require perfect symmetry/shapes, which is the case for UI/UX design

it can generate creative art, or human bodies that can be thicker or thinner etc

not sure if that makes sense

basically ui/ux design is too detailed and structured for SD, at least for now

[fair oxide]

my base model is set as stable diffusion 1.5

am I missing something for this

[gusty salmon]

i dont think there is any model for this

you just cant generate this tye of stuff with SD

you know what, sorry, seems like im talking shit

https://civitai.com/models/98715/ui-ux-lora-sd-15-and-sdxl

[fair oxide]

is it some sd 1.5 with more stuff?

[gusty salmon]

im also fairly new, so i cant tell you

but check the description, they have a discord linked, maybe they can help you out

[dark barn]

i've manage to install an old version 1.7 of stable diffusion to make mov2mov work, do you know how can i get the compatible version of reactor?

[vocal burrow]

found an old mov2mov commit from when a1111 1.7 was the latest version and git clone this specific commit.... But I would not recommend going this route. It s gonna be a pain to install any extension cause they ll all be written for the latest a1111 version and not 1,7, you ll be missing new features, etc....

There are other ways to generate video with stable diffusion. If mov2mov has not been maintained it s most likely because there was another extension doing the same thing in a similar or better way.

deforum, animatedif, sdvideo, etc

[delicate citrus]

what means this error message?
OutOfMemoryError: CUDA out of memory. Tried to allocate 1.57 GiB. GPU

[long oak]

hi folks! I feel silly asking this but how do I access the regional prompter feature on SD WebUI Forge?

is this an extension, am I blind

[delicate citrus]

forge is a version of automatic 1111

I think rss

[delicate citrus]

https://github.com/Panchovix/stable-diffusion-webui-reForge

forge was descontinuated this is the sucessor

[long oak]

yeah I know I'm just not switching over yet

[delicate citrus]

sorry i dont read correctly your question

there are two

they are extensions

https://github.com/opparco/stable-diffusion-webui-two-shot

[long oak]

perfect, thank you! I couldn't find them in the main extensions index.

[cosmic furnace]

can i use lycoris as lora?

when i use any lora it seems speed drops sygnificanrly , can installed lycoris be a reason? hm

indeed probab installed rycolis fault hm 😦

yet again cuda out of memory & fast error

[shy charm]

i know it's kind of an outdated concept, but is there any way to do something like this on a1111?
https://github.com/cysmith/neural-style-tf

i want to turn pictures of anime girls into spaghetti:

i feel like there's likely some way to do it with img2img and controlnet but i wouldn't begin to know how

[cinder warren]

this is a pretty stupid question, but I know I have stable diffusion installed on my pc, but I've forgotten where I've installed it. Any ideas on how to find it? Thanks

[ornate elk]

you need the sdxl fp16 vae
https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/tree/main

[cosmic furnace]

indeed wanted to try it out, but forgot somehow)

[torpid osprey]

Currently trying to use SD.Next and its not working at all, loading models or not. Trying to use a simple prompt i can show what the logs are in a second but im restarting it to check

DirectML initialization failed: No module named 'torch_directml'

trying to use directml

without directml

it get stucks on loading

it gets stuck on "Progress ?it/s 0% 0/20 00:00 ? Base"
(please ping me if someone replies to this)

[quick forge]

so I run SD on a Linux box on my home network, the same network that the computer I'm using now is on.
Is there a way to use SD from this computer (presumably IP) and a way to be able to drop files from here to the linux machine.
This machine is windows 10

[onyx solar]

I'd rather see you turn pictures of spaghetti into anime girls.

[torpid osprey]

assuming --listen and the specific port

[wise flame]

foocus runing in amd?

[ripe sable]

where should you put SDXL loras? just right in the regular lora folder? i havent used SD in a year or so and trying to get back into it

[slender fossil]

Trying to train LoRAs with the Kohya GUI and it keeps shutting down the training almost immediately. Anyone know what to do to fix it? This is all that shows up in the terminal

[slender fossil]

Stared at it for long enough and was somehow able to figure it out, nvm

[restive sparrow]

the option tiling randomly enable after togglong a completely unrelated setting and hand me scrating my head for 10min as to why my images were broken

[long radish]

This took like 2 hours of shitty photoshop edits and img to img generation to do. There has to be a better way

[ruby yoke]

there is

[long radish]

what do

[ruby yoke]

take a photo and then style transfer

[long radish]

how do I do style transfer?

[fair oxide]

let's try this then: RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 1 but got size 2 for tensor number 1 in the list.

I can't load any SDXL inpainting model getting this error

and it's only the inpainting models

[halcyon raft]

is there a way you can use style.csv from a1111 in the comfyui?

i just switch to comfy ui

[elder sable]

anybody know a good discord bot i can host so geneate images with SD? every one i try gets a stupid discord api error 😡

[restive sparrow]

Broken inpainting? pardon the many images
1 during
2 result
3 post AD result
456 parameters

[vocal burrow]

try tweaking the inpainting settings so the modifications made by adetailers get lessen. (lower the denoising strength for instance)
You can also activate controlnet (pose or depth) to restrict its impact.

[safe rover]

How do you do that?

[solemn moon]

Hi friends, is SD3 out yet for Auto?

I had a "set ...." line of text added to my .bat file and SD3 worked.
Then it didnt, so i removed that added line, and still doesnt work so just wondering

[ornate elk]

Hey, nope only when the 1.10.1 gets a stable release

[solemn moon]

okay cool thanks

[solemn moon]

Ip adapter

[safe rover]

ty

[crimson rover]

hey, this is kinda off-topic but i was doing something in diffusion, then pressed something on keyboard and everytime i open "Volume" on my PC or in Youtube video it starts to decrease on it's own, never had this issue before 😄 any idea how to fix it

https://gyazo.com/08d18a84bc13c159532c0bdfbf940498

[crimson rover]

Fixed : It was caused by wireless mouse, after unplugging everything works as usual.

[regal lava]

Hi, after running: python -m scripts.demo.gradio_app, I get "Running on local URL: http://127.0.0.1:7860", how can I change the host url here ?

I try to use the video models, I'm usually an invoke user who doesn't mess with the sd repo directly, but since there is no video support by now, I have to 🙂

[thorn sage]

How can I get a VAE dropdown to show here like all the vids I see instead it's in my settings?

version: v1.9.4  •  python: 3.10.6  •  torch: 2.1.2+cu121  •  xformers: N/A

[ruby yoke]

Put more VAEs in the van folder

[thorn sage]

Thanks! I have 2, how many more do I need?

[ruby yoke]

the question is more why do you want others? you don't just pick them at random, they have a specific purpose

[thorn sage]

Nvm I fixed it, this was the answer not more VAE

[thorn sage]

I don't want "more" I wanted an easier way to toggle the 2 on or off

[ruby yoke]

are you using comfy or auto1111?

[thorn sage]

A1

[ruby yoke]

okay, this is a question that @ornate elk is going to need to answer. i use comfy, and i would just set the node to bypass

[thorn sage]

I already got it fixed with the pic above, thanks tho!

[thorn sage]

Is comfy better in your opinion? I have been considering switching, seems better for video

[ruby yoke]

i prefer comfy because i like the graphical node interface - but they're both just interfaces, running the same models and functions.

[tough seal]

Does anyone know of a way to export a full portable comfyui from a workflow? some sort of prepackager that exports only what is used by a workflow into a new portable env separate from the original one?

[ornate elk]

You should enable xformers for a massive performance improvement

What's your GPU?

[thorn sage]

By using --xformers?
I am using 4060 TI

Is this all I need? set COMMANDLINE_ARGS= --xformers

[ornate elk]

Yess

[timber plank]

I want to be able to effectively run this model but I'm having difficulties loading the fp16 safetensors into comfyui
Comfyui doesn't seem to be able to detect the model type
ERROR: Could not detect model type of: C:\Users\dbs_5\ComfyUI\models\checkpoints\auraflow\transformer\diffusion_pytorch_model-00001-of-00002.fp16.safetensors

[humble blade]

hi, quick question, i set up vladmandic on my system, tried out the normal stable diffusion model which works fine
however, whenever i try to generate images with an sdxl model, namely chimera in this case, i just get broken images like this
anything i'm missing in the setup, or something i need to configure?
generating in 1024x1024

[karmic crown]

show your parameteres

[humble blade]

none right now
i wanted to try xformers but.. that just seems to have messed up my installation, welp

[karmic crown]

not that

[humble blade]

what do you mean then?

ah
one second

[karmic crown]

can you try euler a, as a sampler?

all example i see on civitai use euler a and clip skip 2

[humble blade]

huh... interesting
that does actually produce an image

weird, thats not an issue i had before

[karmic crown]

some models does not work with some samplers

it is usually written in model description,

but i do not see anything about it here

i would recommend you to always check what they used in examples

and to try the same thing

i would also add a comment and ask the author
i have an issue with this sampler while this one works
what do you recommend

[humble blade]

yeah, thats fair
its been a bit since the last time i played around with it, so i thought it'd just have to do with it being an sdxl model

[karmic crown]

when i have an issue i always do the same thnings

first i check if model is downloaded corectly

https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/16223#issuecomment-2241501357 <- this is how you can check if file is downloaded correctly

than i check if there is some known limitation written in model description

then i check examples and copy the prompt, sampler, steps, vae

i ignore loras and other things

also, do not use upscale and similar things while you are trying to figure out what is wrong

[fair oxide]

Will using my all ram(16gb) and vram(8gb) can damage my system or hardware? the temperatures are fine but I am still afraid if it is okay or not

I am running sdxl on automatic 1111

[vocal burrow]

no problem as long as thermal are ok. You've paid for the whole computer you can use it to its full potential.

[stable dust]

Anyone have a tutorial for regional prompter with loras that actually works? looked on youtue and reddit and nothing I do helps 😦 ughh so frustrating

tried prompts posted on here that someone posted when they tried to help hut hat didnt work eother

[thorn sage]

Which versions are compatible with SD 1.9.4? Seems not many work with animated diff as well

[wise flame]

https://github.com/CS1o/Stable-Diffusion-Info/wiki/Installation-Guides#amd-fooocus-with-zluda i follow all step but receive this error after open run.bat RuntimeError: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx using amd

anyone can help me?

[ornate elk]

Why it didnt worked? I showed you the exact settings

[ornate elk]

did you replaced the zluda files in torch lib folder?

[thorn sage]

modules.devices.NansException: A tensor with all NaNs was produced in VAE. This could be because there's not enough precision to represent the picture. Try adding --no-half-vae commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.

Is this from incompatible models with animated diff? I'm not using a VAE

[ornate elk]

the model you use does have a vae integrated

[thorn sage]

How do I stop the error? This happens with the suggested model from YT tutorial as well

[ornate elk]

you can add --no-half-vae to the webui-user.bat

[thorn sage]

I tried that D:

Anything else here look off? I use ToonyouBeta5 model

[ornate elk]

make sure you also and only run the webui-user.bat and not thr webui.bat

[thorn sage]

I've been using the "run.bat" included in the download

[ornate elk]

okay, that should run the webui-user.bat but to verify show the full cmd log

also you can try to generate an image without animatediff enabled

[thorn sage]

It is very long, which part do you need?

And yes it works without animatediff

[ornate elk]

all

there it says it didt launched with --no-half-vae

[thorn sage]

Right I took it off when the same error showed

[ornate elk]

next error

it says you used pony with animatediff?

[thorn sage]

Interesting, I was using the same things the guide suggested

but I did test all models

[ornate elk]

please add it back

[thorn sage]

Oki

[ornate elk]

the motion module only works with 1.5 based models like the toonyou beta 5

so if everything is set up correctly you shouldnt get an error now

[thorn sage]

Thank you! I will try. How can I see which motion modules are compatible with each model?

[ornate elk]

There is only on motion module for sdxl. Its called something like sdxlBeta

Motion module v2 and v3 are for 1.5 models

[thorn sage]

Thank youuuu!!

It works, time to learn more

[ornate elk]

No problem 🙂

[solemn garnet]

Hey folks, I'm having trouble installing ControlNet (I installed ADetailer earlier) but realized I never installed 'ControlNet' in SD1.5.
So I've updated my python and told SD to install it from the url https://github.com/Mikubill/sd-webui-controlnet.git
I get the error pictured.
Now the first thing I did here was search for this and found a few others with the issue but.. so far I'm not finding any solution that was ever posted (and I'm seeing A LOT of the "don't run sd or the cmd prompt as admin" for good reason).
Anyone else that's run into this installation issue and was able to get through it- what did you end up doing? What's the proper solution?

Edit: I've also reviewed the "issues#1024" on that github page that says to 'run the cmd as an admin' (which is contrary to what most everyone here said not to do in past postings).
So follow-up question: If I'm having Automatic1111 install this via the URL, how do I make it do so using admin? Since the cmd is already running in the background that Automatic1111 is loaded in to, I mean. :/

[stable dust]

what settings?

oof been trying for 8 hours

to get it to work

[surreal sentinel]

what's the extension that gives all the buttons and tabs with a bunch of prompts? I can't seem to remember the name.

[sour gulch]

Why isn't VAE working

also why is this happening (separate issue)

[ornate elk]

Here I linked it:
#💬｜general-chat message

[granite sonnet]

Hey how do i fix this? I went to the settings and checked the upcast box but this still keeps popping up

NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.

[ornate elk]

What's your GPU?

[fair oxide]

hi, i seem to have an issue where when installing extensions for a1111 that effect the ui [e.g. stuff like taggers], they just dont appear anywhere on the ui? they'll show up on the extension section as installed but then don't seem to change anything

has anyone else experienced this or

[ornate elk]

You mean the extension boorutag autocomplete?

Make sure you restarted the whole webui (cmd+browser) after Installing extension

[fair oxide]

i dont think i tried that one i tried like 3 different ones that ended up not working even after restarting

but ill try that thanks

[fair oxide]

this one appears in the uncategorised section, is there a way to move it to the top ui selection or

nvm im a lil silly that serves a diff purpose

meant stuff like wd 1.4 just doesnt function for me at all

[tight zinc]

is it normal for ponyxl to take 100% gpu and nearly crash my computer. laptop 4070, so like, 8gb vram, and 16 gigs ram

using cooomfy rn but it's happened on other things too. also using stabilitymatrix

not cooomfy wtf. sorry I'm tired

uhh

fooocus

that's the one with a lotta o's lol

[rain helm]

Create a vibrant and dynamic text logo for a YouTube channel named "Miles’ Chill Tunes." The text should be in a modern, urban font with vibrant colors such as electric blue, red, and purple. Incorporate elements inspired by the comic book style of "Spider-Man: Into the Spider-Verse," including halftone dots, action lines, and a graffiti-like texture. The overall look should be fresh, visually appealing, and energetic, representing music for young people and relaxation. The text should be clear and readable, with a stylish, youthful vibe. The background should be black. -

[granite sonnet]

Nvidia GeForce GTX 1660 Ti

[tight zinc]

we're good update just needed t oclose out some tabs n close steam n stuff

[ornate elk]

You mean the tagger system from an input image?

[ornate elk]

Then you need to edit the webui-user.bat and add --xformers --medvram --no-half
To the line commandline_args=
Then save and relaunch

[orchid stag]

i initially signed up with my gmail account, and now I can't link my discord account to access the bots. How do I resolve this? website didn't say anything about not being able to link discord if you use gmail initially.

[karmic crown]

this is a help channel, you cant generate images here

[granite sonnet]

I have no idea how to do that

[rigid siren]

which samping method is the best? for anime

[ornate elk]

Right click the webui-user.bat and click edit

[ornate elk]

Euler a and DPM++ 2m

[granite sonnet]

Bet I'll try it when I get off work

[rigid siren]

and shedule type?

[ornate elk]

automatic

it will use karras

[stable dust]

oof Webuiforge keeps crashing on me, so annoying

tied adjusung my pagefile size, still happens a ton

st it to the recommended amount someone said, sill crashed. Tried ivreasing it to 20gb min 30gb max, still happens

set it to automaic, same thing

[stable dust]

What are your regional prompter settings for that prompt you said earlier with 3 characters

where you used loras

[wraith ice]

Hi everyone!
I'm using Stability.AI API, with Stable Image Core and SDXL 1.0
I have a question about styles, when I want an image in a certail style, is adding "in style of {style}" to prompt enough and analogous to using style_preset parameter when my style matches the list the parameter can take?

also If anyone could give me a rundown of how styles work and how to use them Id be really grateful

[ornate elk]

they are shown in the screenshots i linked to you twice

[thorn sage]

Hi I just upgraded to 32gb of RAM what pagefile amount do you suggest? I raised it for my 16gb when you suggested it to someone else and it worked grrrreat

[ornate elk]

with 32 you can set it back to syste managed

[thorn sage]

Thank youuuuuuuuuuuuuuuuuuuuu

[upper crystal]

Hello, I am wanting to try out kohya_ss new masked loss, I think it should work simularly to how OneTrainer does it but I cannot find how I pass the mask directory in the webgui
I tried adding conditioning_images --conditioning_data_dir "<PATH_TO_MASKS>" in the Additional parameters field.

[stable dust]

ugh idk why ys not working.It just does 2 judyhopps and conpletely ignores Zhen, idk why

got it set to common rrompt. have use loha or other as im using a lora. Pronpt is "2girls, on bed, lying, on back, viewed from above, looking at viewer, hoodie, shorts, score 9 up, score 8 up, score 7 up score 6 up, source_furry (this is the first line)
BREAK judyhopps lora:judyhopps:1
BREAK Zhen, Lora:zhen:1

i adjusted judyhopps lora strength and still, judy hopps's features is on both characters. wtf

heres screenshots

btw disabling adetailer doesnt change anything either, so ts not that

Ive also tried adjusting cfg scale with no luck

just feels entirely hopeless oof

even tried using the prompts t gives you fir the regions, that also didnt help

also triedturnig offthe loha setting, little to no diference

[vernal wolf]

Yeah sometimes that happens in sd, try inpainting half of the image using the same prompt, then when the result is decent use that image for inpainting and then change the prompt (and of course inpaint the other half of the image)

[stable dust]

how do I do that. Also whats the point of regional prompter if its not gonna work oof

[tight zinc]

how do you change clip skip with swarmUI?

nvm, figured it out :)

[stable dust]

what are other srvers tovet help? struggling to find any oof

[delicate citrus]

anyone know that extension for a11 HUB that allows to install controlnet, checkpoint and anything by it?

[delicate citrus]

I'm tryng to load a sd3 checkpoint but noone works, give this message, anyone knows what must I do?

[cinder warren]

this is a pretty stupid question, but I know I have stable diffusion installed on my pc, but I've forgotten where I've installed it. Any ideas on how to find it? Thanks

[night crown]

What to do here

[knotty merlin]

Recently moved my installation folder to a thumb drive and now every time I try to run webui-user.bat it gets stuck at "Commit hash: <none>"

[night crown]

I have the same thing but it loads after, I'm assuming it takes more time to load from a thumb drive.

[knotty merlin]

Alright, it continued but didn't actually load further

NVM just had to add exceptions for the directories that are in /repositories/

[north locust]

I keep getting a 400 error with this

curl -f -sS -X POST "https://api.stability.ai/v2beta/image-to-video" \
  -H "Authorization: Bearer sk-key" \
  -F image=@"./neworleans-architectural-v1.jpeg" \
  -F seed=0 \
  -F cfg_scale=1.8 \
  -F motion_bucket_id=127 \
  -i

[fierce hollow]

Good day to all, I read the news that SD has released a new video generation, can I find out more about this? We are talking about 4D video Stable Video 4D, Is it possible to install this neural network on your computer and work locally?

[unkempt dawn]

is it possible to run xl models on a 3060ti?

[onyx solar]

Yup

[unkempt dawn]

and by run i mean not wait over 5 minutes

[ornate elk]

Yes

[unkempt dawn]

is there anything i need to do? or am i simply just using bad promts?

[ornate elk]

What webui do you use?
For auto1111 you need to add some launch args into the webui-user.bat

[unkempt dawn]

i dont remember what the ui is called but i think its one of the standard ones (stable diff ui on github), as for my args i just use --xformers

[ornate elk]

Okay, yes thats Auto1111 webui.
Then you need to add: --medvram-sdxl --no-half-vae

[unkempt dawn]

I have a 3060ti so should i use medvram or high?

[glass sable]

Is anyone else having issues upscaling in A11 since the last update? I'm pulling my hair out. Upscaling X2 on R-Esrgan-4XG from 1344x768 is taking almost an hour. At only 40 steps on 3dpmpp++ Karras. On a 4090. This is garbage.

I'm on Winblows 10

[frank comet]

Hello guys, I have a question about Stability.ai, Is it possible to send a request with C#?

[solemn moon]

Check your drivers are updated?

I had this, started a new instance and it was fine..

[glass sable]

Just ran an update.

[storm spruce]

Is there a plugin or something like that for auto1111 webui that puts the generation parameters in the meta data of the image? I think something like that would be cool

[ornate elk]

Only --xformers --medvram-sdxl --no-half-vae

[ornate elk]

Thats not normal

What's in your webui-user.bat?

[ornate elk]

Auto1111 does that nativly. You can check the images with exif tools or by loading it into the PNG-Info tab

[glass sable]

Just --xformers. It got wiped during the update. Don't even remember what needs to be in there.

[ornate elk]

You need the 3 I told you then you have the best performance

[glass sable]

Oh, same setup as a 3060 for a 4090?

[ornate elk]

No a 4090 only needs --xformers

[storm spruce]

Wait, really? Even the prompt and model used?

[ornate elk]

Oh sry I mixed something up xD

[glass sable]

Oh. Yeah. Dassame. What else could I check?

[ornate elk]

Yes everything

[ornate elk]

Make sure your hires steps are set to 10-15

[storm spruce]

Okay, I guess I need proper software to read it then, the windows ui gives nothing

[ornate elk]

Yea windows image viewer can't do it.
You can load the image in Auto1111 PNG-Info tab.
Or download and setup ImageGlass with its ExifTool

On mobile I use Aves Gallery to see the metadata

[glass sable]

YOU FIXED IT!

Three years now I've been using str8 Esrgan with no Hires steps.

My eyes are open to the truth, now.

[storm spruce]

That's pretty neat, makes it a lot easier to generate similar images

[ornate elk]

No problem ^^
If its set to 0 it will use the same amount as normal steps, which is to much. 10-15 are enough

[karmic crown]

@vocal burrow on windows 11 arm in virtual machine on Mac 😂

what should i try next? nokia 3310?

now seriously, if anyone came here with a problem with pyenv, i have a simple solution https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/16261#issuecomment-2250806110

[past tapir]

when training checkpoints, do you prefer captioning dataset with BLIP, WD14 or manually? I just tried manually and got really detailed but have a hard time triggering the LoRA:S

[past tapir]

and manual caption took like 2 days so yeah, not doing that again 😄

[steep cape]

Guys, how should I go about upgrading Pyhon from 3.10.6 to 3.10.10 without breaking my entire Automatic1111 installation? I used 3.10.6 because the A1111 guide said so, but now I can't install Kohya because it needs 3.10.9 at minmum... 😫

[ornate elk]

you need to install 3.10.11 64bit without uninstalling 3.10.6, you can just overwrite it. then you need to delete the venv folder of Auto1111 and khoya

[steep cape]

So I install 3.10.11 and overwrite, delete venv folders, and then when I open the UI bat file, they will be remade?

[ornate elk]

yep

[steep cape]

Thanks, I'll give it a go. In terms of CuDNN files, should these be kept fully up to date?

[ornate elk]

i dont think thats important

[dim fractal]

*************** EP Error ***************
EP Error D:\a_work\1\s\onnxruntime\python\onnxruntime_pybind_state.cc:891 onnxruntime::python::CreateExecutionProviderInstance CUDA_PATH is set but CUDA wasnt able to be loaded. Please install the correct version of CUDA andcuDNN as mentioned in the GPU requirements page (https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements), make sure they're in

---

2024-07-25 17:09:30.1755186 [E:onnxruntime:Default, provider_bridge_ort.cc:1745 onnxruntime::TryGetProviderInfo_CUDA] D:\a_work\1\s\onnxruntime\core\session\provider_bridge_ort.cc:1426 onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : LoadLibrary failed with error 126 "" when trying to load "C:\Users\Eternus\Documents\stable diffusion\webui\venv\lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll"

onnxruntime::python::CreateExecutionProviderInstance CUDA_PATH is set but CUDA wasnt able to be loaded. Please install the correct version of CUDA andcuDNN as mentioned in the GPU requirements page (https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements), make sure they're in the PATH, and that your GPU is supported.

i cannot use CUDA for anything.. and its required for face swaps with reactor.. ive spent 3 days now trying and trying to fix this, im at my wits end, can anyone please, please help me

it works JUST FINE for 6 months, then 3 days ago i had to wipe my PC, and now it doesnt work

(trying to use REACTOR for faceswapping, which is what i always, always use!)

[ornate elk]

hey, whats your gpu?

[dim fractal]

updated - i have an RTX 4060 - worked fine for 6 months, now it doesnt, 3 days ago wiped PC and reinstalled.. and have had trouble since

[ornate elk]

and you installed the latest nvidia driver?