🤝｜tech-support

[balmy lily]

in the bat script should there be a space or no space here

set COMMANDLINE_ARGS=--xformers --no-half-vae

[ornate elk]

Doesn't matter but add --medvram-sdxl

[balmy lily]

betwixt the = and --

[ornate elk]

Then disable the VPN and delete the venv folder before Launching the webui-user.bat

And then everything should work hopefully

[balmy lily]

the pruned model with extension of .safetensor?

[crystal nacelle]

dude funny you mention this, whenever i use a NSFW lora alongside another pop character lora the render becomes nightmare fuel. i cnt get them to work together. gotta just use 1 lora and throw vulgar sexy prompts at it then hope for the best

[eternal lodge]

/game_stats A train on tracks, Surrounded by flowers, The expanse of verdant grass seems like a pleasant mat, Dreamy visual effects, A bright color, Bold shapes, Capture the moment

[balmy lily]

what do you think of the comfyui installation, do u know how to do it and stuff

[ornate elk]

Can't help much with comfyui. But the Installation would have the same problem as auto1111 as your VPN is blocking the download of torch

A guide for comfyui is linked in the Pinned Messages of this channel

[balmy lily]

disabled vpn, made a new ai folder in my c drive and redid the git download of it, and edited the bat script and put the model in the folder

[ornate elk]

Okay same error, is your connection stable over WiFi?

Can you try maybe over Ethernet?

[balmy lily]

its pretty stable, are u able to tell what the actual error is apart from memoryerror

[ornate elk]

It looks like it downloaded the 2.5gb torch file but then it errors out in a weird caching/memory error

[balmy lily]

when i reclick the bat script it trys to reinstall the torch thing, so perhap i could download the torch seperately from online, then put it into the stable diffusion folder?

[vocal burrow]

might be a paging file error

how much ram do you have ?

[ornate elk]

16gb

[vocal burrow]

ok so unless they have chrome opened with 149 tabs it s probably not that. But it s still going over fallback methods and uses "filewrapper" so maybe ....

doesn t hurt to check it, what s the size of your pagefile for c: ? @balmy lily cf https://mcci.com/support/guides/how-to-change-the-windows-pagefile-size/

[balmy lily]

what size should i set it to

[crystal nacelle]

how do i know whether my model is SDXL or 1.5, and how do i know which my LORA is compatible with?

[vocal burrow]

either set it to system managed and make sure you always have a few gbs free on c: or set it manually to something like 16gb

[ornate elk]

16000 min and 24000 max could be a good start

[balmy lily]

shouldnt i set the min to 0

also the pagefile of my d and e drives are 0, should i change that

[ornate elk]

No don't activate it for other drives

But yes you can also do min 0 and max 24000

[balmy lily]

y shouldnt it be activated for other drives

[ornate elk]

Because the pagefile should be on the OS drive and only one file should be used

Activating it on other drivers can slow down the system

Also it should be only activated on an SSD

[balmy lily]

but if i install the stable diffusion to my e drive then would changing the page file size still be able to help

[balmy lily]

u mean as opposed to flash drives

[ornate elk]

Is your e drive faster than your C drive?

[balmy lily]

ha u fuck

[ornate elk]

As opposed to hdds

[balmy lily]

that was it

what a smart lad

[ornate elk]

So it started?

[balmy lily]

this was when i ran the bat after changing pagefile size

[ornate elk]

Looks good

[balmy lily]

yea

[vocal burrow]

nice 👌

[balmy lily]

im not sure if ur interested in coding beyond stable diffusion stuff, but ive been wanting to make a customizable thing like this https://www.mersenne.org/ this is a application one can run on their computer thatll utilize their gpu to search for prime numbers, it partially relies on internet connection to keep track of the list of prime numbers and will update the logs so that no one who runs the app will be doing the same scanning that other peoples pcs have already done. but its specifically tailored for this act of searching for prime numbers, what i wanna try to make is a customizable rendition of this, so the act of keeping the logs somewhere online that constantly get updated while people run the application, so people can run the application to do a specific process that requires mass amount of gpu usage, and theyll all be collectively putting their gpus together on the same task but will never do the same work twice. but the actual process that it does will be fully customizable. so instead of searching for prime numbers it could instead search for numbers that appear in some specific graph, or search for pictures of tacos from youtube thumbnails. so what we'd make is an easy integratable way for someone to encorporate their exe or script into our thing, and theyll have something that multiple people can run on their own pcs, but will update to the same log so even when running on multiple pcs it wont do the same work more than once. im not that good at coding but im taking some programming courses now, but thats just some project i wanna do in the future and i was wondering if youd wanna do any free labor

[vocal burrow]

i was wondering if youd wanna do any free labor. That s a blunt way to put it :p. Thanks for your consideration but I already work a full time job and also have personal side projects. My schedule is definitely overbooked. Good luck with your project.

[balmy lily]

yea i thought its sounded funny to word it like that, but like im not really interested in hiring someone for the job, even if i had the money, cuz if its someone whos only motivation towards the thing is the money then they arnt gonna be able to do that good of a job and theyll charge alot and theyll work the minimum. id only be interested in letting in help from people who are equally interested in building the thing as i am. like theyd be doing it for themselves, even though technically theyre doing for me. but its not like i think theres a good way to make money off the thing so im doing free labor too, which doesnt mean anything cuz its my project, but if there was someone who wanted to build it just as much as me then it become their project aswell

[noble belfry]

The words youre looking for is passion project.

[balmy lily]

idk im not very passionate about anything

[vagrant charm]

Hey guys

your search and replace API has been down all day

Not showing up in your status tracker

I see this too

[fossil quiver]

Hey guys. Anyone here knowledgable about comfyui who could assist me with fixing my workflow and get it to work? I’ll pay you for it

[vocal burrow]

This discord is mostly for user to user / local installation. You ll have better chances getting support for the API using their support system / mailing them for that.

[vocal burrow]

maybe ask (and explain your problem) in #🧣｜comfy-ui

[fossil quiver]

I had no shame and p. much asked everywhere, including there. Xd

[upper linden]

does anyone know why im getting a square error if i upscale a image? (Amd 7800xt)

[ornate elk]

Yes, your mostly using the Directml version of the webui and because you get Out of vram you activated the Tiled VAE feature, but you dont use the correct Encoder and Decoder Settings

You should follow my new AMD Install guide to get the webui running with Zluda instead of directml.
Its much much faster and with that you won't have any vram limitations at all.
The guide is in the Pinned Messages of this channel. Or direct linked here:
#🤝｜tech-support message

[upper linden]

its odd because before it was working fine

[ornate elk]

If you still want to use the slower webui you just need to set Decoder to 1024 and Encoder to 128

[upper linden]

i guess an update or something messed it up

[ornate elk]

It depends on the resolution

[upper linden]

i mean before i was doing the same thing same resolution too

i gues something got messed up

[ornate elk]

Is tiled vae enabled?

[upper linden]

yes

[ornate elk]

Yea then set the encoder to 1024 and Decoder 128

[ornate elk]

Also in the hires fix set the hires steps to 10

[upper linden]

i did alos for the zula instalation after following the intructions i get this

[crystal nacelle]

so im using zluda and everything is great. except my nsfw renders, they all look like nightmare fuel. the models and lora im using are all SD1.5, ive enabled fp8 weight for sdxl in setting

[ornate elk]

did you followed every step?

can you show me your system environment paths?

[upper linden]

i did

[upper linden]

^

[ornate elk]

thats not right

[upper linden]

oh

[ornate elk]

first you need to select Path at the system variables, not user variables

then click Edit

and there make 2 new entries

one for the zluda folder

and the other one should just be:
%HIP_PATH%bin

[upper linden]

like this?

[upper linden]

do i have to add the skip cuda line into the bat?

[ornate elk]

No that would not work, and instead break it

Instead try to delete the venv folder and rerun the webui-user.bat

[upper linden]

i did that gave me a error

[ornate elk]

Okay open up a cmd
And run
Pip cache purge
Then delete the venv folder again

Then relaunch the webui-user.bat

[upper linden]

in the zula folder?

[ornate elk]

Nope

It doesn't matter where you open the cmd for the pip cache purge command

And the venv folder is in the Stable-diffusion-webui folder

[upper linden]

alright it worked now waiting on first image

[wheat lichen]

Did stable diffusion remove being able to generate images on their web platform?

[rocky adder]

Hola! Many of my Loras don't show in Automatic1111, and there's this specific one:
https://civitai.com/models/245037/caravaggio

That I really want to use, can anybody help please?? I already googled but no good luck

[umbral plover]

How do I install diffusers model in comfyui, no matter in which dir i put the catalog with files it doesn't find them

[proven finch]

you make a folder in "models/diffusers" titled whatever its called, then in that folder you put the diffusers folders

e.g.
models/diffusers/sd1.5/text_encoder/blah.bin

[magic hollow]

So im trying to get automatic1111 to work with amd (rocm) on linux mint but i keep getting errors, im using stability matrix to organize all my models and just make things more easy, also i have no idea what this error message means or if my card is even noticed.

[umbral plover]

I'm using comfyui through pinokio, put it here \d1709756953901\diffusers\dreamshaper_8Diffusers and diffusers loader node doesn't see it...
EDIT// okay, in pinokio i have to put it directly in comfyui instead of virtual symlinked drives

[upbeat atlas]

Hi, I'm using an sdxl pony merged model on Zluda, it works somewhat fine, but for some reason I can't use loras with it. I downloaded specific loras for the model, but whenever I generate images using them it crashes the webui. It doesn't even give any error description, it just says press any key to continue and when you do it closes the webui. Anu idea how to fix it?

[ornate elk]

whats your gpu?

[upbeat atlas]

Radeon 6900XT on windows

[ornate elk]

hmm

and you have --medvram-sdxl in the webui-user.bat?

[upbeat atlas]

Yes I do

[ornate elk]

i will try it

didnt tested pony yet

[upbeat atlas]

Thank you 🙏

[balmy lily]

is there some kind of document of the details of everything, like what each thing is, like the vae and configs and everything thats not models, what are they what do they do how do i change them, and details about models and prompting too. so like just a comprehensive explanation of everything and instructions of how to modify and properly fine tune them in directions that point towards your desired result.

[noble belfry]

https://stable-diffusion-art.com/ a very good site

practical examples and an ocean deep.

still dont know who runs that or whatever, but its a very good mix of idiot notes and technobabble, well explained. Whoever writes it needs a backpat, its where most of reddit regurgitate stuff.

also bud, you've been asking questions here for days, most of us work SFW jobs, could you please change you name to something other than cumdetective on this server? I had to block you while on my phone because I kept getting notifications from cumdetective lol. thats fine for 4chan but if youre asking tech questions, sfw it up.

[balmy lily]

wdym, like ur work computer is risking getting flagged while ur on discord for work with stable diffusion

cuz u know theres like hentai an inch above us

i mean it may have bikinis or some kind of covering, but the lot of it is pretty scantilyclad

[chilly knot]

cs10 y0u there?

[balmy lily]

how do i teach it a character, like how can i provide it a bunch of instances of a character and give it a name so that itll learn it and what the character looks like. and without makeing a checkpoint model of the character, how can i tell it in the positive prompt "orange toned, painted, (character name) dabbing" while using a model i didnt make thats not based around that character

[chilly knot]

hm\

embending maybe

lora

[balmy lily]

also what should i put in the lora folder

[chilly knot]

lora file

[ornate elk]

whats up?

[chilly knot]

yor cany model works

thank you

🤗

🎁 its for you present

its not give me errors like before

[ornate elk]

nice thx ^^

[ornate elk]

hey i tested it and it doesnt crash for me

with pony and 2 loras

[fair oxide]

okay so this might be a stretch but like, say i had two separate storage devices hooked to my computer. how would i get a1111 to pull from like both of those and the storage on my laptop

just so i can use ALL my space

[upbeat atlas]

That's odd, any guess on what could be causing this? It happens right at the last step before hires.fix, I tried without hires.fix and it's still the same

[ornate elk]

when did you installed the zluda webui?

maybe you need the new zluda files

or maybe the loras are broken?

does it work without the loras?

[upbeat atlas]

Yeah it works without the loras, though I do get the same error every now and then. It never happens with 1.5 models, with sdxl it happens quite often, but I still get to generate a handful of images before I get that error. But when I add lora it immediately just breaks. I installed Zluda I THINK about 6 weeks ago

[ornate elk]

@upbeat atlas okay then you can try to use the newer zluda files

[upbeat atlas]

Can I update straight from the webui?

[ornate elk]

nope

you would have to download the zluda files (ZLUDA-windows-amd64.zip) and replace your zluda folder with the new files (but name it the same as before) and then delete the venv folder

zluda 3.6 from here is stable:
https://github.com/lshqqytiger/ZLUDA/releases/tag/rel.cd1e0a3d50c135f13a0a919b07133ad18c7c9447

[fiery mural]

Soooooooo I am using TripoSR and I use a converter to get files like this when I export from blender I get this

not surw what i am mussing

Full ticket https://github.com/VAST-AI-Research/TripoSR/issues/77

what am I missing

[sly jetty]

hey tech support, can anyone look at failed diffusion channel

[balmy lily]

i went to https://stable-diffusion-art.com and search for Tertiary model and i didnt find anything, it kinda seems like this isnt the comprehensive documentation i was looking for with explanations of every feature on stable diffusion, do u know of any other. also do you know why when i inpaint something it just colors in the part that i inpainted gray

[ornate elk]

You can checkout this documentation of the auto1111 features:
https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features

[fervent sorrel]

hi please how can i run it via colab gradio --share publink link

[chilly knot]

g00gl c0lab?

[balmy lily]

every image i make is solid gray

[uncut heron]

can someone maybe explain what i did wrong here?
GitCommandError: Cmd('git') failed due to: exit code(128) cmdline: git clone -v --filter=blob:none -- https://github.com/Mikubill/sd-webui-controlnet.git C:\programs mine\stable difusion everything\Automatic 1111\stable-diffusion-webui\tmp\sd-webui-controlnet stderr: 'Cloning into 'C:\programs mine\stable difusion everything\Automatic 1111\stable-diffusion-webui\tmp\sd-webui-controlnet'... fatal: unable to access 'https://github.com/Mikubill/sd-webui-controlnet.git/': Could not resolve host: github.com '

[balmy lily]

i looked at the metadata of this image and i cant find what checkpoint was used, how can i find what checkpoint was used

masterpiece, best quality, 1girl, solo, long_hair, looking_at_viewer, smile, bangs, skirt, shirt, long_sleeves, hat, dress, bow, holding, closed_mouth, flower, frills, hair_flower, petals, bouquet, holding_flower, center_frills, bonnet, holding_bouquet, flower field, flower field, lineart, monochrome, <lora:animeoutlineV4_16:1>
Negative prompt: EasyNegative, badhandv4
Steps: 20, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: 1541450048, Size: 512x768, Model hash: 6e430eb514, Denoising strength: 0.5, Clip skip: 2, ENSD: 31337, Hires upscale: 2.5, Hires steps: 10, Hires upscaler: SwinIR_4x, Regional Prompter: "mode:Horizontal,divide ratio : 1,1, Use base : False, Base ratio : 0.2, Use common : False, Use N-common : False"```

[ornate elk]

Copy the model hash into the civitai search field

Thats your only chance to find it

[balmy lily]

why am i getting solid grey images generated

[noble belfry]

or even a search engine

[ornate elk]

Probably your using a Lora as model

That would generate Gray images

[balmy lily]

always.

[ornate elk]

Yea thats not how it works

Models are always larger than 1.98gb (~2gb)

[balmy lily]

https://tenor.com/view/keith-stone-keystone-always-smooth-keystone-light-gif-13717453

[balmy lily]

but so is it like an error or what, also i thought it might be that but i changed the checkpoint to a checkpoint and the lora the a lora and it still made a grey image

[balmy lily]

keith stone says 'always' btw thats why i posted that gif

[ornate elk]

You need to make your all your checkpoints are 2gb or larger

If not they are not models

Loras go into models/lora

Checkpoints (models)
Go into stable-diffusion/models

[balmy lily]

i noticed on civitai it says on the thumbnail of things whether or not theyre loras or checkpoints and whatnot

[balmy lily]

but is that the only difference is the size

[ornate elk]

Yes the size is the only indicator to tell whether its a model or a lora

They are different and used for different things

[balmy lily]

also how am i supposed to call upon a lora, is it just through the positive prompt, also how do i find out what to refer to the lora as

cuz lora:animeoutlineV4_16:1 isnt the file name

[vocal burrow]

it's the alias

[balmy lily]

so the only indicator is size cuz like theyre both .safetensor files, but like inside what actually differentiates them and makes them uncompatible to work as one or the other

[vocal burrow]

you can either set a1111 to invoke loras by alias or filename (somewhere in the settings)

[chilly knot]

can somebody tell what effect canny in i choise with preprocessor and model and only model

in txt image

[balmy lily]

nvm its working now.

so if you were to try to fill in this girls face to make it complete and look normal and well fitting, how exactly would you do it. like what model and settings and prompts and stuff would you use

[abstract gate]

Hi all, please could someone help with the REST API inpainting, i have asked a question about it here before (as have others) - but all Rest API questions seem to be ignored. I have sent Stability AI emails and tech support questions, but again ignored. Has anyone been able. to. get the REST API inpainting to work? if so what did you do? please help! this is my original message: #🤝｜tech-support message
if this is the wrong channel to be posting where should i. post? Thanks in advance!

[tiny abyss]

Hey
Getting this error

User friendly error message:
Error: 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.. Please, check your schedules/ init values.

[tiny abyss]

and when i use --no-half
i get this

Error: 'A tensor with all NaNs was produced in Unet. Use --disable-nan-check commandline argument to disable this check.'. Before reporting, please check your schedules/ init values. Full error message is in your terminal/ cli.

is this a space issue??

[chilly knot]

if you have videocard like my 1050 maybe only changing the model can help

[tiny abyss]

i have nvidea RTX 3050

[chilly knot]

4 gb?

[tiny abyss]

yeah

[chilly knot]

model?

xl?

i have same if i use xl models

[shut bear]

so my automatic1111 AI kinda ignores the models I set, and the prompts arent always listen on the civitai pages, it just ignores the model and generates the basic no model picture

I use checkpoints. With Lora prompts is doesn't really do anything either

[teal pivot]

Noob here using Stability Matrix and am playing around with the Gradio sharing. It's working great and I have username/password set.

What I'm wondering now: is there a way to have a separate, paired down "guest" instance that doesn't include the dozens of items I have in Users/user/AppData/Roaming.

Basically I want a super simple set of tools that I can show off to friends on their phones or whatever without them seeing all my waifu shit.

[noble belfry]

could you elaborate, screenshots and stuff. I had that happen once, and what it was was i had two instances open.

check the cmd, especially for any async related messages

[heady pine]

Can someone explain to me what the white marker does in ControlNet please

[noble belfry]

im not sure what you mean, do you mean the mask that you paint on things? if so that is the selection of the part of the image you want to work on.

[heady pine]

Alright, thank you @noble belfry

[noble belfry]

are you on forge or a1111?

https://www.youtube.com/watch?v=mHvnHA04sbc my boy seb kampf has a quick instruction for masking through controlnet through forge

@2 mins in

[rapid knoll]

关羽

guan yu

[noble belfry]

english if possible, sorry 🙂

[red cradle]

any1 knows how to make controlent model fit in image size in comfyUI

[heady pine]

I watched that video a while ago, I tried downloading forge, but my files wouldn't show up - loras and Textual Inversion files, Sebastian Kamph is great with his videos. Thank you for the recommendation though @noble belfry

[noble belfry]

well anyway, the white area from what i understand you mean is the masking area. if your controlnet allows it, it will only use that area to execute on

very handy if you say, use a photo of a smiling face but the person is doing something else with their body, you can just select the face part

[safe inlet]

looking at the tags and the resolution looks like SD 1.5 which can certainly be smaller than 2GB, dont delete em!

[ornate elk]

Nope

There is no 1.5 model smaller than 1.98gb rounded 2gb

Loras are never larger than 1gb is also a good indicator

[safe inlet]

you are very helpfull here, but that is not coherent

[ornate elk]

Oh okay

He got grey images and thats mostly because he tried to use a lora as model.
And a simple trick is to check the sizes of them.
But yea there are some self trained models on stars etc that can be 1.5gb or smaller but thats a rare minority.

Thanks for remininding me on that

[shut bear]

Ai takes like 2 minutes fir a single Pic, I'm at 16gb VRAM and a 6800 any clue ?

and takes extremely long to start generating too

for the progress bar to even appear

[heady pine]

Alright, thank you for your help, I will be sure to remember what you told me for the future 🙂

[empty vector]

hey guys, I've installed the SuperMerger extension to my Stable Diffusion web ui but it doesn't show in the web-ui. What should i do? I use runpod secure cloud Stable Diffusion template. I hope anyone could help me. It seems like no extension that I install is showing in the tabs for me, what could I do wrong? I use the "install from URL" option

[noble belfry]

not all of them have big fancy tabs or anything, check the documentation and see if it has a UI

from reading about it it should just sit in the merging tab as a script

reminder: theres people who make UIs to look cool and awesome like tiktok and theres people making cool awesome things. they rarely work together, especially open source

[upbeat atlas]

@ornate elk hi, so I did what you told me about ZLUDA and venv, it took me a while to get it working. Now I can generate one or two images with lora before it crashes, but it still does crash. Any idea how to permanently fix this? It's annoying having to rerun the webui after a couple of image generations.
Also, I just noticed this while running the webui, could this be causing the problem?

[lyric valley]

#🏞｜general-with-images

[ornate elk]

Are you using zluda or not?

[shut bear]

i thjnm so

[ornate elk]

What's in your webui-user.bat

[shut bear]

I have medvram, the one for amd for it to work, I have autolaunch and that's about it, cant use xformers because it doesn't generate and gives me an error

I tried with zluda but it wouldn't launch so I reinstalled everything, not sure if the files still have zluda or not

[balmy lily]

why are there no previews of the models on the ui

[ornate elk]

Then you dont have zluda

And you should try to get it to work, then you won't have vram problems

My guide in the pinned messages covers the steps needed

[gloomy wave]

I am gonna try install forge ui again but i need to uninstall python and previous git program so i can use the "one click" option?

Right now when i try run update from the extracted folder its just a black CMD window.

And thats after i command it "run anyway" due windows says its unsafe

[ornate elk]

You dont need to uninstall git

And for python make sure you have 3.10.11 64bit

[gloomy wave]

At the moment a1111 works just fine. I am pretty sure i have the older version is instructed in for the A1111 installation, otherwise how do i check?

when runned as admin.

[ornate elk]

At the bottom of the webui, you see the versions

[ornate elk]

Thats okay too

Should also work with forge

[gloomy wave]

Nothing happens whn i click update file, its just a black window.

[ornate elk]

Which update file?

[gloomy wave]

The downloaded Forge ui folder has update and install before using, the update showed a black CMD window but i believe i fixed the issue. Its always the VPN.

[ornate elk]

True xD

VPN can cause issues with git links and other needed urls

[gloomy wave]

Thank you anyways! 😉

[red warren]

hi, I've been trying to use stable diffusion via the autonomic1111 ui, but when I attempt to use it I get the error "connection errored out". anyone else ran into this issue before?

[ornate elk]

Hey, make sure you didn't closed the cmd after Launching the webui-user.bat

[red warren]

I just ran the web-ui.bat

[ornate elk]

Then you probably should checkout my latest guide on how to install automatic1111 the right way.
Here: #🤝｜tech-support message

Its also in the pinned messages of this channel

[soft fiber]

pretty sure we want to run the webui-user.bat right

[red warren]

when I do that it tells me python is not a recognised command

I have version 3.10 but usually have to use py instead of python

in my command line

[soft fiber]

oh yeah in that case cs1o has a good suggestion to double check the installation with his guide

[ornate elk]

Thats because python isnt added in your system path.

[red warren]

i'm pretty sure I added it but let me checkj

[ornate elk]

You have to rerun the python installer

Click modify

Next

And check "add python to system variables"

[red warren]

I added it to path in my environment variables

[ornate elk]

Thats not how it works

[red warren]

i thought it was 😅

[soft fiber]

also disregard this, im uninformed and never questioned just simply using webui-user.bat lmao. i figured the default webui.bat had a specific usage case.

[ornate elk]

You shouldnt add it manually xD the installer will do it for you

So delete the entry

Before running the installer again

[red warren]

sure

should I uninstall python as well?

[ornate elk]

Yea

[red warren]

cool

[ornate elk]

And then checkout my guide

It covers every step

[red warren]

kk

should probably get better at reading

the new python install worked

but i'm still getting the connection errored out message

heres the cmd log

when i looked on google other ppl were having this issue but i didn't see any solutions

[ornate elk]

You should update Your gpu driver and restart the pc

Then delete the venv folder and rerun the webui-user.bat

[red warren]

sure - doing that now

btw is it worth deleting this and installing forge cus apparently its better..

[ornate elk]

If you want you can.
I still prefer auto1111 as some stuff doesn't work in forge

But you can also have boath installed

[red warren]

i'll try to get auto to work then

but if I can't I can use forge as a last resort

[red warren]

ok I've done all that and I'm not getting errors... but also not getting images

how long do they normally take? I have an i9 9900k and a 2070

[proven finch]

turing....

[proven finch]

what resolution and what do you see in the cmd

[red warren]

oh never mind I got the error again

it was weird the image appeared and then it said connection errored out

[ornate elk]

Whats your resolution?

[red warren]

and my monitor flashed grey

[ornate elk]

Strange error

[red warren]

all my settings are default

could it be browser? im using chromium

[ornate elk]

What does the cmd shows?

Any browser should work

[red warren]

nothing

the bottom line is a little glitched out

[balmy lily]

I keep getting an "opera has run out of memory" error and it shuts down

[red warren]

doesn't opera have memory limiters?

[ornate elk]

Looks normal

Can you try out an other browser?

[red warren]

ima have to install one

[ornate elk]

Firefox worked good for me

[balmy lily]

im not sure, well i suppose it does, but im not sure if its any more than any other browser, but thats why i was asking if anyone else gets this or knows how to avoid it and whats the best browser

[red warren]

and I just set it as defaut right>

[ornate elk]

That or copy the url

[red warren]

yh

i'll try it out

[balmy lily]

how do i make stable diffusion open on say firefox instead of my default browser, when i click on the webui bat

[red warren]

copy the link across

still got the error on edge

so I dont think its browser

just "connection errored out"

[ornate elk]

Do you have any other stuff running in the background, like games? Or wallpaper engine?

[red warren]

i do have wallpaper engine

[ornate elk]

Close that when using SD

[red warren]

ok

[ornate elk]

That can cause issues

[red warren]

just closed a bunch of helper apps (steam, razer, etc.)

Thx guys for ur help but I'm still getting errors

I have to go to bed tonight, might have another crack at it tm

[ornate elk]

Sure gn

We can troubleshoot further tomorrow

[ionic mantle]

ERROR: Could not find a version that satisfies the requirement torch==2.1.2 (from versions: 2.2.0, 2.2.0+cu121, 2.2.1, 2.2.1+cu121, 2.2.2, 2.2.2+cu121)
ERROR: No matching distribution found for torch==2.1.2

anyone have any ideas on how to fix this error?

Im using an AMD radeon gpu if that has any relevence and ive tried putting --reinstall-torch as a command line argument

[red warren]

I had it, install it manually in cmd

[proven finch]

windows issue

enable --medvram or something alike

[red warren]

I have

[vestal sage]

my stable diffusion is got strucked with CMD "commit hash"
how can i fix this issue

[late tree]

Is it possible to run an already generated image through Adetailer without changing anything else about it? I know I can put it through img2img, but that would do the whole image unless I mask it, which would basically defeat the purpose of using Adetailer

[limpid monolith]

It depends! If you want to use an Adetailer for faces, you can set the denoising strength to 0.01 literally and just activate the Adetailer in img2img 🙂

[late tree]

But does setting it to that also affect Adetailer? Or does Adetailer have it's own denoising that it uses?

[limpid monolith]

Adetailer has its own denoising strength in the settings 🙂 The default is usually 0.4, which you can change.

[late tree]

I'll try that, thanks.

[lethal comet]

requires a personal CivitAI API to be downloaded, you can set your own API key in the CivitAI Browser+ settings in the SD-WebUI settings tab

Anyone knows how to fix that ?

[empty vector]

does anyone know why Supermerger is not showing in the tabs?

https://github.com/hako-mikan/sd-webui-supermerger/issues/350

[ornate elk]

Adetailer only changes the things it recognises based on the adetailer model selectet

You can also load the image into the PNG-INFO tab and then press Send to txt2img and you get every settings back. Then activate Adetailer

[ornate elk]

Hey, what's your gpu and what guide did you followed?

[bitter sorrel]

hello, i need someone to help me install stable difusion to my computer 🙂
can anyone please help me

the problem is my internet speed is slow, when i trying to install stable difusion

with cmd
it broke and not install
Microsoft Windows [Version 10.0.22631.3374]
(c) Microsoft Corporation. All rights reserved.

C:\A1111\A1>git clone https://github.com/vladmandic/automatic
Cloning into 'automatic'...
remote: Enumerating objects: 43156, done.
remote: Counting objects: 100% (260/260), done.
remote: Compressing objects: 100% (135/135), done.
error: RPC failed; curl 92 HTTP/2 stream 5 was not closed cleanly: CANCEL (err 8)
error: 6766 bytes of body are still expected
fetch-pack: unexpected disconnect while reading sideband packet
fatal: early EOF
fatal: fetch-pack: invalid index-pack output

C:\A1111\A1>
Please help me .

[fair oxide]

wtf

[fair oxide]

bro i literally did the same thing 2 minutes ago

i just started downloading this slop replacement

and had the same issue

[bitter sorrel]

yes , someone say i can find help from here.

[ornate elk]

Hey, for a good install guides checkout the Pinned Messages of this channel.
But you should try to connect via Ethernet for the download

[bitter sorrel]

told*

[fair oxide]

https://tenor.com/view/matrix-the-agents-know-dont-gif-18536773

[bitter sorrel]

yes i have ethernet.

[fair oxide]

Lets give this server a shot

so i have an issue that occurs when i swap to any XL model that my ram floods and crashes the webui

i've tried many webui's

[ornate elk]

What's your gpu and what's your webui ?

[fair oxide]

3060

Webui Is A1111

i tried invokeAI

same issue

i tried a1111 forge and it fucking worked for 1 day

[bitter sorrel]

but i have intel arc a750 gpu.

[fair oxide]

i switched to ponydiffusionxl and it worked

[ornate elk]

3060 with 12 or 6gb ?

[fair oxide]

12

with 32g ram

[bitter sorrel]

8 gb, and ram 32 gb

[fair oxide]

You know your cooked when you couldn't find anyone having the issue in reddit or github

[ornate elk]

Then you need to edit the webui-user.bat and at the line COMMANDLINE_ARGS=
You need to add --xformers --no-half-vae

Then save and relaunch the webui-user.bat

[fair oxide]

alright

ill try that

[ornate elk]

For that card checkout this guide:
https://github.com/vladmandic/automatic/wiki/Intel-ARC

[fair oxide]

how big is xformers?

[ornate elk]

A few hundred MB I think

Maybe less

[fair oxide]

alright ill see if your solution works

[bitter sorrel]

yes i try this one, but the problem is when file downloading then it broke and stop.

i am searching for the main file without using cmd.

is there a direct file that i can download on my computer? without using command promt

[fair oxide]

it still crashes

this is the last thing it says before it crashes

Loading weights [ac006fdd7e] from J:\New folder\Terminate\2005463692\Content\Items\yup\Aib\webui\models\Stable-diffusion\autismmixSDXL_autismmixConfetti.safetensors
model_type EPS
UNet ADM Dimension 2816
Using xformers attention in VAE
Working with z of shape (1, 4, 32, 32) = 4096 dimensions.
Using xformers attention in VAE

[ornate elk]

Does 1.5 models work?

[fair oxide]

yeah

[ornate elk]

Have you tried an other sdxl model?

[turbid raptor]

I'm trying to install SD 1.5 but am gettin this error. Any ideas how to fix this?

[soft fiber]

Installing into OneDrive causes issues doesn't it

[vocal burrow]

yup, don't install it in any kind of sync/cloud folder

also you're not installing """SD 1.5""" you're installing stable-diffusion-webui 1.8.0

which allow you to use SD1.5, SD2.0, SD2.1, SDXL, Stable diffusion cascade, etc models

[balmy lily]

why is it that when i add the lowvram command in the bat file, i get more cudo memory overload errors than i do when i have it set higher or get rid of the command entirely

[trim raven]

hello sir, an 6950 xt don't need --no-half ?

[balmy lily]

yea what is no half vae

[ornate elk]

If you use directml then only if you want to use Inpaint or Adetailer.
But you should install the zluda version for better performance and then your not vram limited

The guide is in the Pinned Message of this channel

[trim raven]

ok i don't need it👍

[balmy lily]

is ur gpu better than a 4060ti, and how good are u able to generate images, especially at a high resolution, i just wanna gauge how much my gpu is my problem and how much anything else might be the problem

[signal sigil]

Can anyone tell me why the seeds with SD have changed so much since two years ago? I have noticed this when using A1111 for ages now. I have always messed with settings that i knew would changed seeds, even with the same models loaded. I can't seem to get the seeds to look right. I have matched the settings in the image data on old images as closely as possible. I've never got to the bottom of it. SD1.5 feels less coherent to prompts because of this change too.

[balmy lily]

how can i avoid this ``` torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 4.00 GiB. GPU 0 has a total capacty of 8.00 GiB of which 0 bytes is free. Of the allocated memory 11.74 GiB is allocated by PyTorch, and 1.26 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

[ornate elk]

What settings did you used?

[balmy lily]

like prompt?

and model

[ornate elk]

Yes and resolution, steps etc

[balmy lily]

well it happens every time that i run it, eventually. and with a variety of models and prompts. i mean if i set the resolution to low then the image generates faster, and if i set it to high then i goes slower, logically, but it just kinda feels janky, and its constantly either 'erroring out' with no info in the command prompt about it, or getting a cuda memory error and displaying the text i pasted. so i was wondering if there was any good way to avoid overloading type errors, cuz alot of times all the applications on my computer closes, and the web browser says memory error, not as much now that i do it on firefox instead of opera. but i just wanna make it not error out just becuz its overloading, and instead just take processing power from everywhere else, but never closes itself or errors itself out from over need of memory or power and what not, and just go slower if it needs to if theres not enough processing power

[vocal burrow]

https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Seed-breaking-changes

[ornate elk]

Make sure you have --xformers --medvram-sdxl --no-half-vae in your webui-user.bat

And to always launch the webui-user.bat

And not the webui.bat

[balmy lily]

yea, also what do each of those do btw

[vocal burrow]

cf https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Command-Line-Arguments-and-Settings#all-command-line-arguments

[balmy lily]

cuz i when i got rid of --medvram-sdxl my stable diffusion stopped erroring out, but i started getting more cuda memory errors

[balmy lily]

would using the python launch and manual installation without a virtual environment and running stable diffusion from a command prompt window instead of using the browser ui help lower the amount of erroring out that happens

[vocal burrow]

probably not

While SD runs on your gpu, the webui is only using your cpu and ram (and some tiny little part of your gpu if you've enabled gpu acceleration for your browser)

[broken bramble]

Hey guys can someone tell me why im getting this error! and hot to clear it?
"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."

[vocal burrow]

what's your gpu ?

[broken bramble]

1660 Super

[vocal burrow]

use --xformers --medvram --no-half command line arguments in webui-user.bat

[broken bramble]

is 1660 super good enough?

[vocal burrow]

good enough is the correct term

with those arguments it should run

[broken bramble]

Thank you

[balmy lily]

whats up with this

--use-cpu {all, sd, interrogate, gfpgan, bsrgan, esrgan, scunet, codeformer} None Use CPU as torch device for specified modules.

what do each of these do

all, sd, interrogate, gfpgan, bsrgan, esrgan, scunet, codeformer

also should i maybe use the Alternative installation on Windows using Conda instead

[upbeat atlas]

Hi, can someone help me fix this error?

stderr: ERROR: Could not install packages due to an OSError: [WinError 5] Access is denied: 'D:\SD-Zluda\stable-diffusion-webui-directml\venv\Lib\site-packages\~6mpy\.libs\libopenblas.FB5AE2TYXYH2IJRDKGDGQ3XBKLKTF43H.gfortran-win_amd64.dll'
Check the permissions.

[fossil dew]

What does this mean:

OutOfMemoryError: CUDA out of memory. Tried to allocate 3.40 GiB (GPU 0; 8.00 GiB total capacity; 10.44 GiB already allocated; 0 bytes free; 13.62 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

[fossil dew]

my batch size is 1, maybe its that im resizing 2x 🤔

[balmy lily]

@ornate elk @vocal burrow whats you guys' gpu, and how well does you stable diffusion run

[vocal burrow]

it's giving you option to use cpu instead of gpu for all of the listed sd functionalities. No point doing so if you have a functional gpu

[balmy lily]

how can i run stable diffusion in both google collab and my own gpu, with the regular stable diffusion ui, so that google collabs gpu can assist

[vocal burrow]

there's no benefits using conda

[balmy lily]

yea but it keeps crashing and working jankily i feel like

[vocal burrow]

might be because you're aiming for too high of a resolution

You're running out of vram/hitting the limit of your GPU. How to fix it (Nvidia edition) :
1/ If you're using a 2000/3000/4000 RTX series cards then you can try to use the --opt-sdp-attention command line argument or --xformers argument. To do so, Edit your webui-user.bat and append it to the set COMMANDLINE_ARGS= line. Performances between --opt-sdp-attention and --xformers can varry depending of your GPU and usecase. Please note, if you're using vladmandic's fork then you won't be able to use--xformers.
2/ Also throw --medvram in there if you've got a GPU with less than 8gb of vram, --lowvram if less than 4gb. (each argument must be seperated by a space). If you intend to use SDXL with a gpu that has less than 12gb add --medvram-sdxl or --lowvram-sdxl if you have less than 8gb. (Alternatively you can try to turn on fp8 support in settings)
3/ If you're using --precision full, --no-half and don't know what they're doing. Then you probably don't need them and therefore they're increasing your vram usage for nothing. Remove them. Notoriously --no-half is necessary for GTX1600 cards.
4/ Make Batch size lower, use Batch count instead if you want to make multiple generations.
5/ Output at a lower resolution (and eventually upscale later on with hires.fix or img2img feature)
6/ Write less complex prompt
7/ Use lighter model (use fp16 pruned variant of your model if available)
8/ If you're using a VAE and your generation crash at the very end. You should try to install multidiffusion-upscaler-for-automatic1111 extension (available in the default extension list) and enable the TiledVAE features (cf screenshot)
https://media.discordapp.net/attachments/1002602742667280404/1111097170150563881/image.png?width=810&height=228

PS : keep in mind that SD1 models are trained with 512x512 images, SD2 models with 768x768 and SDXL 1024x1024. So stick close to those resolutions before doing any upscaling,

[vocal burrow]

2070S 8gb run just fine for what I do with it.

[balmy lily]

whats ur it/s

[balmy lily]

also what r some images u ve made

[vocal burrow]

dependsm it varries from 2 to 14 it/s

[fossil dew]

Thank you for the input!!

[fossil dew]

Could this error in my box thing when I boot up stable diffusion be part of the issue?

i dont understand most of this 🥲

ok this is a different error something about a file having multiple terms in it

which is weird cuz thats completely new

[balmy lily]

this is the error i keep getting, most of the time it happens right when i open SD and try to generate an image.

[soft fiber]

What does the CMD show?

[balmy lily]

i have a nvidia 4060 ti, which is pretty close to what urs is in terms of power, so i wanna try to test what my problem is, could you try to make an image for me. for a test i wanted to make a realistic mockup of the jeff the killer image, not to hoax a fake original version of it, its just something i was interested in trying to make a realistic face version of it using inpainting and stuff, thats not why i have stable diffusion but its what ive been trying to start with, so could u quickly try to generate one, and tell me ur exact prompt and model and settigns that u used and stuff, and ill try to do that exactly, and if it doesnt work for me then ill know its some other fundamental thing wrong in how mine is set up rather than my processing power

[balmy lily]

nothing, not for this error

it just says, "press any key"

[balmy lily]

the reason why im asking for this specific task is becuz it involves inpainting and modifying an already existing picture to achieve a specific goal of a realistic well fitting face. which is what i havnt been able to do very well so far. but i have been able to generate a few regular high quality ai generated looking images of amiguous girls,

but u dont need to do it with this picture specifically, u could just find something of similar nature that you wanna do

[signal sigil]

Oh thank you! I thought it was just me going insane. Haha! 😅

[ornate elk]

Im running a AMD RX 7900XTX 24gb
And it runs very good with zluda on windows
Can be as fast as 16 it/s

[fair oxide]

@ornate elk have you looked into ip adapter v2 that was released a few days ago?

[ornate elk]

Oh no, do you have a link?

[fair oxide]

https://github.com/cubiq/ComfyUI_IPAdapter_plus

im testing it on a1111

[ornate elk]

Where does it say new that there are New IP adapter models?

On huggingface I didn't found any new ones too

[fair oxide]

they are calling it an upgrade on top

[ornate elk]

For me it looks like this comfyui node just got updated to work with v2 while in auto1111 I was able to use v2 the whole time

[fair oxide]

you could be right about that, tho i downoaded some of those files which i didnt have before

[ornate elk]

Tencent also didn't released any new IP adapter models since January

[fair oxide]

ahh ok

[ornate elk]

The models can be found here:
https://huggingface.co/h94/IP-Adapter/tree/main

Let me know if you find anything that could have changed.
Maybe they updated their scripts in the background but then we would need to wait for a controlnet update I guess

[fair oxide]

btw do you notice any weird results with plus models?

i kinda like the outputs with this combo

rest of the others have some weird results

[earnest jackal]

Hello, I installed some embeddings but they are not showing up in my textual inversion

[onyx solar]

Did you refresh/restart after? Or perhaps are you trying with an XL model by accident?

[earnest jackal]

I used it before it was fine

oh yea nvm my model says xl

[onyx solar]

See. 😉

[earnest jackal]

so If I change it to any other one it should show up right?

[onyx solar]

If it's not XL.

[earnest jackal]

got it

is there a way to use it with xl?

[onyx solar]

No

[earnest jackal]

its fine thank you for the info

[onyx solar]

👍

[loud aspen]

hey yall, new to comfy and trying to get this running:
https://civitai.com/models/335070/simple-lcm-img2vid-workflow-or-comfyui

i'm on a fresh install and I believe I have all the constituent parts installed and loading correctly, but this workflow is just spitting out stuff like this for me:

that's an actual example

after several hours

lol

no clue what i'm missing, if this kinda noise generation has a commonly known cause i'd be interested

[crystal mortar]

are you using macos?

[loud aspen]

windows

[last lotus]

Hi guys, my generation's previews look like they are coming along nicely, then right as it's finished generating the images turns out like this, any idea why?

[onyx solar]

Use the correct VAE.

[last lotus]

how could i figure out which is the correct one?

[onyx solar]

What model is it?

[last lotus]

SDXL 1.0

[onyx solar]

You have to use the SDXL VAE.

[last lotus]

gotcha, thank you

[boreal cradle]

Must I have a NVIDA GPU to use Stable Diffusion? I have an Intel UHD Graphic 620 GPU. Is there a way to make Stable Diffusion work with this hardware?

[faint widget]

What ui are people using with stable diffusion now? is auto1111 still the go-to, or would vlad still have better support for newer things

[pseudo nebula]

whys it making a blakc image

[sinful halo]

Can someone pls help me, i dont understand why it keeps generating these extra limbs and fingers if i have a extensive negative prompt with embedding like badhands and easynegative, im using embeddings wrong ?

last render, the eyes are so strange too

[onyx solar]

Because even with negatives, embeddings, etc., it's still randomized mathematical image generation. Sometimes you will just get poor coherence on images depending upon the combinations.

[pseudo nebula]

im not sure but i think the number of steps is the number of times it tries to make an image, and if it hits a negative promt, it will skip that step. So if it just keeps making extra limbs on every attempt, ur gonna end up with extra limbs

so u could try changing the denoising strength, increase number of steps, ect

[crystal mortar]

thats not how diffusion works

[sinful halo]

so its a matter of trying more ?

[sinful halo]

okay thank you, ill try

[pseudo nebula]

lolololol i love how to make it depressed it gave her blood shot meme eyes

[sinful halo]

LOL true

[onyx solar]

Patterns have been trained to be associated with words using machine leaning algorithms across vast sets of data. That is turned into mathematical models. You input words that are tokenized and used as vectors across those models in conjunction with the seed value which is sampled over the number of steps provided on a schedule algorithm you've selected. Hands are complex as there are exponentially far more positions they can be in than, for instance, a nose can. But misshapen hands are also much easier to spot than something like a single strand of hair is to human eyes. So without using additional tricks/methods, they tend to present a problem. You can solve that in a few ways and simply rerolling a new seed is one of them.

[sinful halo]

oh i see, thank you so much for the detailed explanation, ill find a good one, and use that seed to keep generating more until i get a really good one

[onyx solar]

There's also methods like inpainting, which you can focus specifically on an area like the hand, auto detailers that can detect and attempt to fix hands and faces, controlnets for hand posing, and more. It all depends on what you want to do and what kind of investment you want to make to learn.

[sinful halo]

i really want to put some effort into this, i love Yin from darker than black but theres so much good wallpapers of her on the internet

you helped me a lot, thanks, ill search on these things you citaded

[onyx solar]

Look some of the things up that I mentioned and you can deep dive into any of them.

[sinful halo]

surely i will, thanks

[pseudo nebula]

do u need insight face for anything?

its not building it, and i think its missing vs 2019

[ornate elk]

For Faceswap extension and IP-Adapter

[pseudo nebula]

is ip intellectual property?

[ornate elk]

Only if you need the Faceswap aspect of it

The normal IP adapter models will still work

[pseudo nebula]

i think im just trying to copy the pose

[ornate elk]

Therefore you should use openpose

[lyric valley]

#🎥｜animation

#🌠｜show-and-tell

[vestal plank]

Im try img2img SDXL with SD model. And i get this red-yellow noise garbage. Why?

[ornate elk]

Because you have Photon model loaded too and thats 1.5 so the sdxl VAE gets in conflict

[upbeat atlas]

Hi, how do I fix this error?

stderr: ERROR: Could not install packages due to an OSError: [WinError 5] Access is denied: 'D:\SD-Zluda\stable-diffusion-webui-directml\venv\Lib\site-packages\~6mpy\.libs\libopenblas.FB5AE2TYXYH2IJRDKGDGQ3XBKLKTF43H.gfortran-win_amd64.dll'
Check the permissions.

[ornate elk]

When do you get this?

[upbeat atlas]

When I start the webui

When I run the webui user bat without internet connection it doesn't open, and when I do connect, before the interface shows up I get that error in the cmd

[vestal plank]

Im try embedded VAE from photon - and its same

[ornate elk]

Why do you have one sdxl and one 1.5 model loaded in the same workflow?

[ornate elk]

You can try to delete the venv folder and relaunch.
That will redownload the torch stuff

[vestal plank]

SDXL make so blurred images, so i go to use img2img with SD model. I hope it will help.

[ornate elk]

Do you have an example of a blurry image?

[vestal plank]

Second in this message

[ornate elk]

Maybe its just a wrong setting

[vestal plank]

It not so blurry, but i want sharp, like real photo on good camera

[ornate elk]

Maybe try changing the prompt to test

Like
Photo of an bear in the forest,
should get you a sharp image

[loud aspen]

bump 🫣

[solemn atlas]

https://www.facebook.com/share/r/d65Z4oDVedhdCrFS/ how is this done? the quality is really good

Is this done via SD?

[languid kelp]

how to fix this?

I searched for this error fix but couldn t understand

[ornate elk]

What's your python version?

[hazy wigeon]

running comfyui for the first time, got this error

using the extra_models_paths.yaml to point comfyui at a models folder in D drive ( I already have auto1111)

[balmy lily]

@ornate elk would u wanna do something like this #🤝｜tech-support message

also do you have any tutorial for the comfyui model install

[vestal plank]

Same

So strange... If I load image and make img2img from file, not from node, its work

May be im out of VRAM?

[ornate elk]

Nope, but for Comfyui models just put then in the models folder

[soft fiber]

Have you renamed the symlink file from its initial form of .example? It looks like you set it to D:/ in the yaml but it's still trying to pull from C:/ in the terminal

For clarification, that yaml is initially downloaded as extra_model_paths.yaml.example and we're supposed to rename it (simply delete the .example)

[hazy wigeon]

yup deleted the .example

do i have to do some mklink stuff

[hazy wigeon]

oh comfyui is installed in C:/. i tried moving it to D:/ as well but same error

[karmic crown]

has anyone tried to move models to an external drive? that should only affect model load time, right?

[karmic crown]

do not symlink anything, just put proper location(s) in yaml file

if all models loras and stuff are in deafault place in a1111, just add correct a1111 location

if not, change the other things that you need

for example path to models if they are in some other place

[hazy wigeon]

so add the full file path?

[karmic crown]

yes

just check if you need / at the end or not

i do not have comfy installed at this moment

base_path: path/to/stable-diffusion-webui/

with

if your models are in default location do not touch this
checkpoints: models/Stable-diffusion

if there are for example in d:/aimodels
than change it to that location

[solemn moon]

OMG lol

came here to help but bruh

[sinful halo]

anyone knows of a way to use amd gpu ? i dont wanna convert to onnx tho

[limpid monolith]

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

You can run A1111 on an AMD 🙂

[ornate elk]

Sure what's your gpu?

[sinful halo]

Rx 5700 XT

[ornate elk]

Ive made 2 guides

[sinful halo]

thanks, ill take a look

[ornate elk]

You find the guides in the pinned messages of this channel

They cover every step

[sinful halo]

thank you so much

[sinful halo]

Sorry to bother you, i only have one question;
in this part, my card is 5000 series, where do i put the no half arg ? i substitute the --no-half-vae with it ?

Edit the webui-user.bat (right click), At the line COMMANDLINE_ARGS= You add: --use-directml --skip-ort --medvram --opt-sub-quad-attention --opt-split-attention-v1 --no-half-vae --upcast-sampling
If your GPU has only 8GB Vram and you run into out of memory, you may need to change --medvram to --lowvram.
GPUs with less than 8gb Vram need: --lowvram
RX 5xxx cards additional need: --no-half

or it goes at the end of everything ?

[ornate elk]

You change --no-half-vae to --no-half

[sinful halo]

thank you, i apreciate it

[ornate elk]

No problem

[chilly knot]

cs10

you can help do a good deed

a man was banned for posting images of girls with piercings and tattoos, they were fully clothed

is it possible to unban him he won't post this anymore

[fierce hollow]

Help

[true sphinx]

yo im having a weird issue every time i try to generate an image i get absolute horrors that make 0 sense it used to work last night then i loaded a new SDXL model then when i swapped back to a sd 1.5 i get werid pics like this

any idea on how to fix

or when its not this nonsense

its like a weird grid

[onyx solar]

What software? What configuration settings? Need more information.

[true sphinx]

automatic 1111

i reset to default

gen settings are 40 steps

[onyx solar]

Show a screenshot.

[true sphinx]

ok

gen settings

[onyx solar]

Turn off hires for moment and what happens?

[true sphinx]

settings for sd 1.5

trying

now

[onyx solar]

Also, turn off the Tiling checkbox on the 2nd screenshot.

That's your real issue.

[true sphinx]

thats the issue

idk how that got turned on

wtf

[onyx solar]

Shit happens. 🤷‍♂️

[true sphinx]

thanks homes

[solid locust]

I'm frying my brain here x_x

I'm trying to connect ForgeUI to AI Diffusion in Krita, and I believe I do that with Cyanic SD, but everytime I try to click generate, it just shows the percent... and then goes back to generate. I thought it might have something to do with it having no checkpoints & loras, so I try to search for it, but I just can't find it. And I am so stumped, I don't even know where to begin or go. I might even become a himbo without the body.

[glad fiber]

how can i make the a1111 version not to open a browser tab?

[keen edge]

hey, I'm having an issue with controlnet, It won't stay ticked after applying, and no models appear in controlnet even when there are several in folder

[karmic crown]

try to add
set SD_WEBUI_RESTARTING=0
to your webui-user.bat

[glad fiber]

will try

[karmic crown]

let me know if it work or not

put it in this part here

below commadline_args for example

[solid locust]

help x_x

[karmic crown]

have you setup a1111 correctly and does it work by itself?

[solid locust]

no it connects to Comfy instead.

also let me try if Forge works.

[karmic crown]

it should be similar as one for photoshop

you need to set proper port

it also might have a1111/comfy radio buttons

or something like that

[sinful halo]

Hi, Im getting this error on auto1111;
RuntimeError: don't know how to restore data location of torch.storage.UntypedStorage (tagged with privateuseone:0)
only using hires fix

[solid locust]

proper port?

Oh, it's setting up.

[karmic crown]

7860 vs 8188 if i remember correctly

i do not have comfy currently installed

check settings of the plugin

the one for krita

[solid locust]

well here we are.

[karmic crown]

leave it open

and open krita now

and try again

[solid locust]

ok

[karmic crown]

your webui/comfy needs to be open, so that plugin can connect

[solid locust]

And now the generate button doesen't work at all Q_Q

[karmic crown]

see this

[solid locust]

yeah?

I mean no

[solid locust]

Wait, now I do when I click the AI diffusion docker.

[karmic crown]

click that button

[solid locust]

sorry. ^^''

[karmic crown]

as far as i can see you can use current installation, so you need to use connect for that one

or to let krita plugin install it, when you need to specifiy the first option and directory

[karmic crown]

try this one

but open your comfy first

[solid locust]

ok

comfy & forge both booted up.

@karmic crown And then I connect the custom server now that they both are open, right?

[karmic crown]

please screenshot config screen when you select that option

so i can see what you have there

basicly you can use standalon comfy or you can let plugin install comfy for you in desired location
i am not sure if you can use forge/a1111

[karmic crown]

ok

[solid locust]

weird... This guy in this ponydiffusionTURBO said he changed to ForgeUI and it fixed the issue with it being extremely slow.

[karmic crown]

is there a new version of that plugin?

i do not see forge or a1111 there

on your screenshots

[solid locust]

I think so.

I mean that I updated ForgeUI

[karmic crown]

wait

https://github.com/Interpause/auto-sd-paint-ext

this one is for a1111

https://github.com/Acly/krita-ai-diffusion
and this one is for comfy

as far as i can see at least

[solid locust]

what about Forge? o.o

[karmic crown]

forge is a1111

[solid locust]

oooooooohhhhh...

[karmic crown]

you cant use comfy plugin with forge

i thougt it is similar as with photoshop plugin, it can be user for both comfy and a1111/forge

but looks like krita has two difrent plugins

[solid locust]

now to figure out how to remove the comfy plugin and replace it with a1111 plugin.

[karmic crown]

it should be easy

check where plugin is, remove it and restart krita

i havent seen krita for at least 10 years to be honest

maybe even more

[solid locust]

Just remove the comfy file?

[karmic crown]

not that

%APPDATA%\krita\pykrita

find plugin here and remove it

you are on winows if i see correcly

[solid locust]

windows, yeah.

[karmic crown]

i am on laptop and images are small

[karmic crown]

i need to open image in new tab and to zoom to see 🙂

remove those two

and restart krita

that should be it

[solid locust]

oki :3

[karmic crown]

then you can remove comfy from the other drive

[solid locust]

entirerly or?

[karmic crown]

this

you do not need it if you do not plan to use it

[solid locust]

I mean... it is where Krita is in.

I just chose to name it that.

[karmic crown]

wiaht

[karmic crown]

leave krit as it is

[karmic crown]

ok, do not touch that

[solid locust]

😛

thought so.

[karmic crown]

but i guess you do not need this

now install that other plugin

and lauch forge

[solid locust]

hmmm...

[karmic crown]

that is the comfy that plugin isntalled

remove that

[solid locust]

all of them or just the comfyUI file?

[karmic crown]

all this was installed by that plugin

remove the server folder

[solid locust]

aaaah...

[karmic crown]

and if there is nothing else in ai_diffusion left after that

remove it as well

[solid locust]

https://tenor.com/view/haha-bongo-cat-press-the-button-explosion-gif-14344779

[karmic crown]

can Apple make anything like other brands? 🙈

[solid locust]

Right, the whole AI_diffusion is deleted.

So uhm... I start up Krita, and keep the ForgeUI open only, and close Comfy, right?

[karmic crown]

yes, but install krita a1111 plugin

[solid locust]

oh! right!

[karmic crown]

@vocal burrow if i put models on sdxc it should only affect model loading speed, right?

and before you ask it is 1TB 🤣

[solid locust]

Extension or Plugin? ... Plugin, right?

extension is not necessary, right?

[karmic crown]

both

plugin on kirta

and extension in forge

[solid locust]

ah

Ok, my brain is acting like a potato again.

with cloning, do they mean to pack up the ZIP file into this extension folder?

[karmic crown]

clone = using git clone

but you can download zip

and extract in folder

[solid locust]

ah ok, I suck at CMD stuff.

[karmic crown]

kids those days 🙂

we used to launch windows from dos 🙂

i used windows 3.11 🙂

[solid locust]

eh, born 1996 myself. 😛

[karmic crown]

do not forget this part

[solid locust]

before or after?

[fair oxide]

these*

[karmic crown]

english is my 3rd language, and i am exhausted

i am usually aware of my grammar mistakes, but not at this moment 🙂

[solid locust]

https://tenor.com/view/loading-wtf-ruocgiff-gif-25631449

I click on... ForgeUI folder, and then stable diffusion webui depository...

and then find a file that says commandline_args inside webui user bat, or the sh version.

[fair oxide]

Oops fair enough

[solid locust]

Did I get it right?

[karmic crown]

in forge

find webui-user.sh

and add --api to command line args

[solid locust]

let's see... .sh is translated to Shell Script file, right?

[karmic crown]

sh is for mac and linux

you need to edit .bat

[solid locust]

ah, bat.

[karmic crown]

webui-user.bat
or without extension if your extensions are hidden

and launch that one

[solid locust]

?

[karmic crown]

post your webui-user.bat here

[karmic crown]

set COMMANDLINE_ARGS=--api

change it too look like this

save it

close ALL running a1111/forge apps

and double click the webui-user.bat

or click (however it works on your windows)

[solid locust]

look like what?

[karmic crown]

set COMMANDLINE_ARGS=--api

instead of
set COMMANDLINE_ARGS=

here

[solid locust]

OH in the bat where I originally copied, right?

[karmic crown]

webui-user.bat in your forge folder

[karmic crown]

is there anything above that in cmd window?

before exit code 1

[solid locust]

nope

nothing

just exit code

exit code 1

[karmic crown]

do you have python in that folder?

what version is it?

[solid locust]

oh, I do ^^''

[karmic crown]

open cmd and type
python --version

what you get?

Python 3.10.13

something like this

[solid locust]

3.12.2

[karmic crown]

thats the problem

you need 3.10 for forge

do you have nvidia?

[solid locust]

I kind of have AMD and Nvidia.

hold on

[karmic crown]

https://github.com/lllyasviel/stable-diffusion-webui-forge

and scroll down and find

Or you can just use this one-click installation package (with git and python included).

Click Here to Download One-Click Package<<<

that is forge with python

[solid locust]

ah ok ^^

[karmic crown]

it is fastest

and just change webui-user.bat

[solid locust]

ok

[karmic crown]

since you have nvidia that should be enough

[solid locust]

ok

wait... I already installed that.

[karmic crown]

run that

it should work

[solid locust]

ok

I mean, it already worked before.

So what now? @karmic crown

[karmic crown]

add a1111 extension to that one

add --api to commald line params
and that should be it

scroll back to see how, if you forgot

i need to go

i am very tired

[solid locust]

ok, good night ^^

[whole skiff]

Hello, what does this mean? Tried to find but nothing

[ornate elk]

I think it could be from this extension:
https://github.com/wcde/sd-webui-kohya-hiresfix

[whole skiff]

Yes it is! Thank you very much! It is kinda hard to study and obtain new information sometimes 😅

[whole skiff]

And NGMS:3 it's Negative Guidance minimum sigma that can be found in setting--->Optimization

[fair oxide]

idk if this is the place to ask for help but does anyone know why this isnt working

was watching a tutorial and it worked fine for him

[vocal burrow]

you didn't install git

[fair oxide]

oh I thought I did

will try again

thx

[vocal burrow]

cf https://git-scm.com/download/win

[fair oxide]

while installing the new one it said I already had it

should I worry about that

[vocal burrow]

weird, but nothing to worry about

[fair oxide]

still doesnt work

[vocal burrow]

also, don't clone in a onedrive / any kind of sync/cloud folder

and avoid spaces/non latin characters in path

what happens if you run git --version ? (in a cmd)

[fair oxide]

idk why it says that I nuked onedrive a long time ago

[vocal burrow]

did you make sure that the "add to path" checkbox is checked when installing git ?

(it should be by default)

[fair oxide]

I dont remeber

[vocal burrow]

might as well reinstall then :p

[fair oxide]

ok

I didnt see any options saying add to path

we try again ig

[vocal burrow]

this

[fair oxide]

yep I did that

[vocal burrow]

and where did you install it ?

[vocal burrow]

that's the install .exe, not the installation folder. Where did you install git ?

[vocal burrow]

run echo %PATH% in cmd and screenshot the result

[vocal burrow]

don't run it as an administrator

use the standard cmd

[fair oxide]

its automatically at administrator

[vocal burrow]

that's not normal

[fair oxide]

is that hard to fix

or idk

[vocal burrow]

did you turn off UAC ?

[fair oxide]

is that the thing where it asks you first?

[vocal burrow]

yes among other things

[fair oxide]

I dont think so

I might have tho and dont remember because about a year ago my pc got bitlockered and I didnt have the code and after I fixed that it kept saying I didnt have premison to open files I downloaded

so how do I check

[vocal burrow]

run
REG QUERY HKEY_LOCAL_MACHINE\Software\Microsoft\Windows\CurrentVersion\Policies\System\ /v EnableLUA and screenshot the result

[vocal burrow]

it's turned off

[fair oxide]

sorry

how do I turn it back on

[vocal burrow]

follow those instructions https://learn.microsoft.com/en-us/mem/intune/user-help/you-need-to-enable-uac-windows

(set it to the default value)

[fair oxide]

that worked thank you

[vocal burrow]

I'd still recommend using a different folder tho.

seeing that onedrive in the path is making me anxious

there's a tutorial in the pinned messages if you want @fair oxide

[fair oxide]

ok I will look at it

[rose solar]

Hey there
Hope you are doing well
I am a senior full stack developer who has full experience in web and AI development
So if you have some projects, please let me know
thank you

[fair oxide]

new issue

[fair oxide]

are you talking to me?

[fair oxide]

ive been trying to get this to work all day and I would really appreciate any help

[earnest jackal]

Having issues merging models how do I fix this?

[pallid kettle]

same having trouble w this as well (i wanna merge some models)

[lime falcon]

What a coincidence...

Error loading/saving model file
    Traceback (most recent call last):
      File "E:\SD\stable-diffusion-webui\modules\ui_checkpoint_merger.py", line 21, in modelmerger
        results = extras.run_modelmerger(*args)
      File "E:\SD\stable-diffusion-webui\modules\extras.py", line 213, in run_modelmerger
        theta_0[key] = theta_func2(a, b, multiplier)
      File "E:\SD\stable-diffusion-webui\modules\extras.py", line 97, in weighted_sum
        return ((1 - alpha) * theta0) + (alpha * theta1)
    RuntimeError: [enforce fail at ..\c10\core\impl\alloc_cpu.cpp:72] data. DefaultCPUAllocator: not enough memory: you tried to allocate 335544320000 bytes.

I don't know if I can fit 330 gigs of ram on my board...

[sinful halo]

Can someone tell me why everyone render an image in low res and after upscale it ? why not render in high res from the beginning ?

[hazy wigeon]

if you have a beefy pc sure. but most people dont, so they render lowres first.

[sinful halo]

no i dont, but the entire process of sampler, hi res fix, sampler, upscaling could take more or less the same time, maybe theres another reason why people render in low res, i see youtubers with beef pcs rendering in 512:512

[fair oxide]

hires fix has a different method of correcting an image that direct render does not

[sinful halo]

Thank you so much, unfortunately i cant use hires fix, my auto111 doenst give any error, simply does nothing when is time to do the hires part /-/

[fair oxide]

when using sdxl i dont use hires fix, but artifacts with sd 1.5 models were common at least used to be

[sinful halo]

oh

what resolution do you usually render ?

[fair oxide]

you have technical issue then, im not quite sure what could be that cause, i suggest talking to the tech help here

for sdxl i generally go with 896x1152 could also go higher

[sinful halo]

i see, thank you

[fair oxide]

i dont have any need for hires fix with sdxl as far as correcting goes

[sinful halo]

thats nice to know, ill start using sdxl

[fair oxide]

and i could go into extra tab to upscale with minimum resources

[sinful halo]

thats true

[ornate elk]

Run the python 3.10.11 installer again, click modify, next, and check "add python to system variables"

[hazy wigeon]

Hi I've been trying to install the comfyui krita plugin, i've got all the dependencies except controlnet preprocessors. but comfyui manager gives me this warning, should i download it anyway?

[pseudo nebula]

return F.conv2d(input, weight, bias, self.stride,
RuntimeError: Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same

[pseudo nebula]

https://huggingface.co/lllyasviel/control_v11p_sd15_openpose/tree/main

how do i download all of this? there isnt a link and i dont have git

[fair oxide]

you'd have to download them manually, also look into https://huggingface.co/lllyasviel/ControlNet-v1-1/tree/main and https://huggingface.co/webui/ControlNet-modules-safetensors/tree/main

[pseudo nebula]

idk seems pretty busted, i think ill give up on openpose

[fair oxide]

your input image has to be good quality

if you are trying to replicate poses, a combination of open pose and depth work great

[pseudo nebula]

its like 700 pixels

700 by 1200

[solid locust]

@karmic crown Man, I am SO close, so close. but suddenly after re-downloading and re-installing everything, and heck, I even found out how to change to --api in the bat file, but somehow the webui-user.bat doesen't work, because it can't find any python file in the directory.

[pseudo nebula]

i think u gotta reinstall python

https://www.python.org/downloads/release/python-3106/?_fsi=ORWqMLBj

3.10.6

[solid locust]

translation: "Python was not found, run without arguements in order to install from microsoft store or inactivate this shortcut from settings > change alias for running of program.

[solid locust]

wait a minute...

[pseudo nebula]

??

[solid locust]

viking told me to install Python 3.10.13.

[pseudo nebula]

well get one of em

[solid locust]

13 or 6?

because I am using 13 now.

could it be that I tried installing these two togheter?

[pseudo nebula]

i couldnt figure out how to install the python with just the bianaries and no installer like what ur looking at

if you go to the 3.10.6

its got a windows installer in that list

[solid locust]

ooooooh ~

and this works for krita, right?

I'm trying to use ForgeUI with Krita.

[pseudo nebula]

if it uses python, probably

[solid locust]

cool cool 😄

I'll try number 6 then.

[solid locust]

no module named pip... WELL, at least we're getting somewhere.

[pseudo nebula]

dont run the -user

just webui

[solid locust]

no?

ok

[solid locust]

nope, still the same.

now where could I find a pip?

[pseudo nebula]

python -m ensurepip --upgrade