🤝｜tech-support

[vocal burrow]

d:\Ai 2\stable-diffusion-webui

[cursive cobalt]

It just... Idles once it reaches these 😦

[vocal burrow]

Probably because some other node is not conected/connected incorrectly. check other nodes

[pseudo prairie]

i can’t find it

[vocal burrow]

screenshot that folder for me d:\Ai 2\stable-diffusion-webui

[cursive cobalt]

Use the search function then 😄

[pseudo prairie]

i did

ur gonna hate me for it

i don’t know how to screenshot

i don’t know which one is my window key

[cursive cobalt]

The button with the windows icon on it

Also bruh. Last folder

[vocal burrow]

go from the very end "Ultimate SD Upscale" node and go backwards. check the output of every node, see if its connected, see if it s outputting everything it should, etc

[pseudo prairie]

i have a custom one

so it doesn’t have window logo

[pseudo prairie]

which one is ghis

[cursive cobalt]

Default place is on the buttom left, between CTRL and ALT. The key right in the middle

[vocal burrow]

(for next time, win+shift+s or simply push the "print screen" key on your keyboard should be on close to the top right of the board)

[pseudo prairie]

ok

so

[vocal burrow]

anyways delete that folder, run webui-user.bat and it should be good

[pseudo prairie]

omg yay

i think its working

thanks for helping and being patient with me

[cursive cobalt]

Ok. I've checked all the connections, and it looks like it is all connected correctly. Trying to gen an image now, to see if it actually wants to, or fails again.

[pseudo prairie]

The button with the windows icon on it

now what

im so lost

[pseudo prairie]

this is what im getting

[cursive cobalt]

WIndows key + Shift + S, Select Area > Ctrl + V to paste the selected area into discord.

Nop Ary. Still goes CR Upscale Image > Steps > Idles.

I'll leave this workflow in the back again, and try to...... For the 8th time, to make my own :cries;

[pseudo prairie]

its not working

its closes when i try to use any keys

at the end it says press any keys to continue

[cursive cobalt]

What I am talking about, is how to take a screenshot.

[pseudo prairie]

so when i try to take screenshots or anything its closes the thingy

[vocal burrow]

what s your gpu ?

[pseudo prairie]

i honestly don’t know

[vocal burrow]

so you downloaded the wrong stable-diffusion-webui

that s why I hate youtube video tutorial......

[pseudo prairie]

?

oh 😦

what do i do now

i need visual steps to understand

[vocal burrow]

Check the pinned messages in here for AMD tutorials

can t help you more rn, have to work, sry

[pseudo prairie]

its ok

[pseudo prairie]

im so again

lost*

[cursive cobalt]

Pardon the ping Ary. @vocal burrow

What would this box be called, and how does one add the text box?

[pseudo prairie]

Getting Zulda running bit is confusing

[vocal burrow]

duno, box should be there, try to recreate the node (name is at the top)

[cursive cobalt]

I think its an editied note title. The search function doesn't yield a result on that name

Trying to somewhat recreate the parts I need, in my own workflow, using the previous Pre-Made workflow as a template

Ah found it. Its a String Literal

[pseudo prairie]

yall i am scared to do step 3

am i doing this correctly

should i even do it

[cursive cobalt]

https://support.microsoft.com/en-us/windows/use-snipping-tool-to-capture-screenshots-00246869-1843-655f-f220-97299b865f6b#:~:text=Select the Start button%2C enter,R for a video snip.

[vocal burrow]

doing good so far

Also @ornate elk is it really %HIP_PATH%bin and not %HIP_PATH%/bin ?

(still don t have an amd card to check)

[terse lichen]

stable diffusion webui is called auto1111

[vocal burrow]

stable-diffusion-webui is the official name for it. People call it auto111, a1111 after the name of the maintainer "automatic1111" (it s also called sdwebui sometimes) because there used to be multiple forks with the same generic name.

[terse lichen]

i see

[vocal burrow]

Also, it is not an official stabilityAI project.

[cursive cobalt]

Idk why this is happening

It also ends up causing this, which i cannot cancel, and must terminate the server

Brb. Just gotta disinfect something.

[vocal burrow]

you probably created a loop somewhere

output of node A into B, into C, back into A into B into C, back into A into B into C, back etc

[cursive cobalt]

Ah yes, indeed. Had lost 2 connecting liens to a Lora loader

This makes no sense...

Why the random CLIP to nowhere?

Also keeps getting this, which just stops any subsequent attempts at genning

[terse lichen]

maybe cause your lora is disabled or something

[cursive cobalt]

They work in another workflow. Not sure why they won't work in the new I'm trying to make

[onyx solar]

The Anything Everywhere node allows you to have an empty receiving node of that type that will automatically be filled in by the transmitting node somewhere else. So if you have an empty CLIP on that workflow that needs an input, it'll just pull that in as if the noodle was connected.

[terse lichen]

is auto1111 not compatible with latest the version of python script ?

[cursive cobalt]

Oh, so its a wireless transmitter?

[onyx solar]

Yes.

[cursive cobalt]

i c

[onyx solar]

If you don't know something about a custom node, the best way to figure that out is to open the Manager, go to the Install Nodes, find the pack it's from, then click on the pack to go to its github page. There you'll typically find the information on the nodes that you need.

If you don't know what pack it is, click on Manager and add the badges to your view.

[cursive cobalt]

Oh that is whatbadge is

[onyx solar]

There are several options, including that, yes.

[terse lichen]

@vocal burrow @ornate elk

[cursive cobalt]

Now i just need to fix this 😐 Verified that they actually work in another workflow, so... Something, somewhere... Is doing something

[onyx solar]

I highly suggest reading up on ComfyUI Manager so that you can understand all of its features.

[onyx solar]

Well, considering all those nodes were purple, you're bypassing them.

[cursive cobalt]

... Purple?

[solemn moon]

Purple

[onyx solar]

In your own screenshot.

[cursive cobalt]

Oh no, that is from a screenshot of the workflow i imported

This is my Workflow

[ornate elk]

Yes

[cursive cobalt]

Soo, as u can see. it looks to be fine, but 🤷 still gives errors

[onyx solar]

Have you tried isolating which LoRA is causing the issue by disconnecting all by one at a time?

[ornate elk]

Hey do you need help with any step?

I made the guide, so feel free to ask me

You can't break anything with step 3, just show me a screenshot before saving it

[cursive cobalt]

This is with the workflow that generated an image. 2nd image is isolating Loras giving errors... But they are the same. Why is it working in 1 workflow but not the other?

[onyx solar]

Did you use the exact same 4 LoRAs in both workflows? Did you use the same model in both workflows?

[cursive cobalt]

Yes

[vocal burrow]

no, 3.10.6 to 3.10.13 if you want to be safe.

[terse lichen]

mine is 3.12 or something

[vocal burrow]

then nope. reinstall a compatible python, delete venv, relaunch sdwebui-user.bat

[onyx solar]

What lora is the 3rd one? (KIM_V2)

[cursive cobalt]

A character lora

[terse lichen]

i have not installed it yet cause i am also running comfyui latest python script

[onyx solar]

Yes, but what exactly? I am validating things.

[cursive cobalt]

Sent on DM

[onyx solar]

That's your problem.

The others are all SDXL.

You cannot mix model versions.

[cursive cobalt]

Worked fine in Auto1111 😦

[onyx solar]

No, it wouldn't have.

That LoRA wouldn't even show up to be selected if you had an XL checkpoint selected.

[cursive cobalt]

Well it did.

[onyx solar]

No, it didn't.

[cursive cobalt]

Yes it did

[onyx solar]

Do it now.

[vocal burrow]

at best they ll do very very very very little and probably not what s intended. most of the time they won t work

[onyx solar]

If you have 1.5 LoRAs showing up when an XL model is selected, you're either horribly out of date or you've got a great bug to share.

[cursive cobalt]

Lemme just boot up Auto1111 and do the same loras

[vocal burrow]

don t mix and match sd 1.5 / sd 2 / sdxl / sd whatever

[solemn moon]

Looking for a brightness model XL btw if ya’ll know of any, qr brightness control

[cursive cobalt]

Using a PonyXL SDXL merge

[onyx solar]

That's just it...if he's running a version of A1111 that's anywhere near current at all, it's not going to show up.

[cursive cobalt]

It was installed last week

[vocal burrow]

you can force them to be displayed. people often google stuff up, find the switch and then come here to complain that it does not work.

[onyx solar]

Sure...and if that is the case, it would show, but that's not typical and it would still be broken.

[vocal burrow]

99% of the time yes

[onyx solar]

Show the entire screen where both XL and 1.5 LoRAs are showing up.

[onyx solar]

Ok...now run a gen with ONLY that LoRA loaded and the XL model.

[cursive cobalt]

Ok i admit, without the Other Styles lora, it does look a bit yank, but using the Other styles lora, and it works... I don't see the compatibility issue?

[onyx solar]

Kimpossible is a common enough character that she's likely trained into most models and is absolutely in basic Pony. A1111 is not going to use the 1.5 LoRA successfully with the XL model. The geometry is completely different, which is why the two models are not compatible with each other.

[cursive cobalt]

So shouldn't it, regardless of style loras added, not be able to promt an image?

[onyx solar]

So, essentially your LoRA ain't doing shit and you're seeing Kim because she's in the model.

A1111 is lying to you because you've bypassed the filter and chosen the 1.5 model.

It's not actually using it.

[cursive cobalt]

https://tenor.com/view/ok-alright-then-fine-great-okay-gif-23114127

Good thing i recent begun naming the Lora's after their base model

[onyx solar]

There are lots of LoRAs for Kim, though...plenty for use with XL/Pony:
https://civitai.com/search/models?baseModel=Pony&baseModel=SDXL 1.0&modelType=LORA&sortBy=models_v9&query=kimpossible

[vocal burrow]

you can toss them in a folder called SD1.5, SDXL, etc... it s gonna save you some time

[onyx solar]

Exactly. Subfolders are a much better way to go.

[cursive cobalt]

Folders within the Lora folder?

[onyx solar]

Yes.

It will parse them.

[cursive cobalt]

... Did not know i could do that...
Still having issues adding a 4th path, because i ran out of space on the main drive for models

[cursive cobalt]

Right... Let me find a compatible lora, and try again

[onyx solar]

Best of luck. I'm sure you'll be able to work something out well with one of the other LoRAs that are out there for her.

[cursive cobalt]

I mean. Apparently she is thoroughly trained on Pony, soo

[onyx solar]

Yeah...popular girl character, so not surprised.

[cursive cobalt]

Ok. No errors thie time... However, I need to scrub through the workflow I'm using as a template... Cause it just stops here.

This is complicated and my poor brian can't deal with this >.<

[onyx solar]

I would skip using AE nodes until you have a fully functional flow.

[cursive cobalt]

Well. I'm missing something that's meant to recieve CLIP... Somewhere

[onyx solar]

If you run the flow without that AE node, it should highlight the error node in red when you are missing the input.

[cursive cobalt]

I just think i don't have a note thats supposed to get it 😄

Is what I'm saying

Its trying to send to [Blank]

[onyx solar]

CLIP would typically go from your LoRA into your prompt conditioning next.

[cursive cobalt]

I don't think I've added the conditioner tbh

[onyx solar]

It's somewhere, whether it's visible or not. Without that, you can't sample your text.

[cursive cobalt]

Also... WHY is it so hard to un-bypass a group?!

[terse lichen]

ae as in after effects node ?

[onyx solar]

No.

[terse lichen]

then ?

[onyx solar]

#🤝｜tech-support message

[terse lichen]

i see

[cursive cobalt]

Wireless Transmitter/Receiver

Is the TL:DR

Ok. Getting images now.
Just need to further add onto the flow to fix face etc

[cursive cobalt]

So... @onyx solar Tell me. In Auto1111, i can assign a LoRA to insert certain prompts, whenever i "add" it to the prompt box (I know it might be inefficient, but it is what worked for me).
Can i do similar in Comfy? Because I like having certain prompts assigned to Lora's, like their hairstyles, eye color etc, as a "preset". My memory ain't too good

[cinder ermine]

is there any way to remove text / speech bubble from images automatically?

would save a lot of time when going through a cleanup process

[fair oxide]

hi @ornate elk quick question, is it ok to install git and python on C drive and stable diffusion on D drive?

[wheat flare]

So this is weird. I have SDXL 1.0 running in AUTOMATIC1111, and when I run the upscaler with default settings, it gender-swaps the men. What's causing this and how do I fix it?

[vocal burrow]

Yes

[cursive cobalt]

o/ Ary. Can u assist me in a moment with a simple issue? 😄

(This) #🤝｜tech-support message

[vocal burrow]

Commuting back to home rn

Maybe later

[cursive cobalt]

Ok. Commute safely!

[fair oxide]

k thanks!

[ornate elk]

upscaler? it seems more like you used img2img

with a high denois

[wheat flare]

I clicked the upscale button. A peek at the settings shows that Denoise was set to 0.7

[ornate elk]

ah okay yea that will use the default settings of hires fix. so better put the denois to 0.35 or 0.4 and the upscaler to esrgan4x+

also hires steps to 10

[fair oxide]

@ornate elk any idea what could be causing this error ..

i see this when running the bat file, and i have git and python installed on C drive and stable diffusion on D drive

[ornate elk]

whats your gpu?

is your nvidia driver updated?

[fair oxide]

haven't updated driver yet, i installed windows 11 a fresh install

gpu rtx 3060

let me update the gpu driver

[ornate elk]

make sure you installed the latest nvidia driver

[fair oxide]

i will

4 weeks away from a1111 and i feel new

i upgraded my mobo, processor and added more ram with total 32gb

[fair oxide]

after updating the nvidia driver, things seem to be moving along fine now, thanks.

windows had old driver installed

[wheat flare]

Where's the setting to change the upscaler?

[ornate elk]

enable hires fix in txt2img

[safe basin]

followed the zluda guide until step 7, even redownloaded python and git again. but this time were stuck on this: even cleared cache like you said

[ornate elk]

press space to test if its stuck

[safe basin]

nothing happens

[ornate elk]

okay, then close and relaunch the webui-user.bat

alos make sure to have 15GB free space on the C drive

[safe basin]

150gb free and stuck on the same spot, text is flickering

[ornate elk]

Thats really strange

@safe basin what you can try is to open up a cmd and run
SFC /scannow
to detect and fix system errors.
Then restart the PC
then open up a cmd and run
pip cache purge
then delete the venv folder again and relaunch the webui-user.bat

[safe basin]

No luck

I've never had this many problems installing something before. It was so easy to setup on nvidia

[ornate elk]

what problem did you get now? still the freeze?

at the download

[safe basin]

now the download doesnt even appear

[vocal burrow]

it s installing stuff

let it work

[safe basin]

But I got it to work yesterday, how is the entire download broken today.

Like sure It crashed when I tried rendering hires but It worked atleast

[vocal burrow]

I don't see anything broken here. It's installing stuff.

It just does not show a cool pretty ascii bar, that's normal.

how long it takes depends of your hardware and connection speed

[ornate elk]

yes its doing stuff, let it load (can take up to 45 minutes)

[safe basin]

If you say so, but Im pretty sure its not downloading. It took me about 2 minutes max yesterday when It worked

[vocal burrow]

make sure it has the focus then

click on it then press space

[ornate elk]

@safe basin has it moved know?

[safe basin]

nope

[ornate elk]

okay then please close and relaunch the webui-user.bat again

it should at least download torch correcly

also if your from asia you need to use a VPN

and if your not from asia you need to disable any VPN and third party antivirus

[subtle quarry]

is anyone having trouble with civitai download speeds? OOKLA's speed test says my internet is fine, but civitai is only downloading at 1-2MBps

i'm not using a VPN or anything

[safe basin]

same spot, from eu. no vpn and antivirus off

Ill just walkthrough every error and step so far. first I tried with directml and got this error: RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check

So I cloned it again and got stuck on Installing open_clip after I followed some guide on how to fix it. I then fixed it by running webui-user.bat as admin. Then It said that my gpu wasnt supported. the reason was because my integrated graphic drivers wasnt disabled. got it to work but I could only generate images if I started it as admin otherwise I got this error:

[safe basin]

but when I started getting this torch error is when I tried another tutorial with miniconda and I think it broke my python somehow because after I reinstalled python again and restarted. I tried directml again I got stuck on this

[white stirrup]

anyone here know dynamic prompts?

lets say i did this {X1|X2|X3|X4} is there a way to make it so the first image gets X1, second X2 etc?

[ornate elk]

please dont follow multiple guides to fix directml, you dont want the directml webui xD

[vocal burrow]

You ran it (whatever it is as it seems you followed multiple tutorials) once as an admin so now permissions for the whole directory are messed up. git consider that those files are owned by admin. And therefore it's preventing you from touching files that you (with your non admi account) don't own. That's why you don't run stuf as admin.
The proper fix would be to take ownership back of everything in there
takeown /f c:\SD-Zluda\ /r /d y should work.
The not so so proper fix is to run the command it's telling you to run (cf screenshot), it will basically ignore ownership of files.
The "please don't do this, why" fix is to keep running it as admin and praying that you don't run into any malicious file.

Also direct-ml and zluda are two different """"backend""""/engine.

directml is easier to install but way slower than zluda.

[safe basin]

So do I run it with admin? and its the folder where "stable-diffusion-webui-directml" is located?

[ornate elk]

no never run it in admin mode, it will cause permission errors after that

[safe basin]

Cant I just delete everything I've downloaded somehow?

[ornate elk]

yep, delete every folder you created,
open up a cmd and run
pip cache purge
then start again with the zluda guide

better restart the PC before starting the guide

[vocal burrow]

yeah better start from scratch

[safe basin]

pip cache purge deletes all downloads ive done?

I meant like is there a way to get my pc back to the state it was in before I tried installing stable diff?

[long parcel]

not sure if this belongs in #sd3 or in tech-support but the website is broken I'm trying to join the waitlist

hey guys I was trying to join this list wait:

In this page: https://stability.ai/news/stable-diffusion-3
it says this:

You can sign up to join the waitlist here.
http://stability.ai/stablediffusion3
But the link gives me a 404, is the waitlist closed or is there another way to join?

[ornate elk]

yes, deleting folders + pip cache purge will d oit

[safe basin]

the folders are in the python directory or where?

[ornate elk]

no, you only need to delete the stable-diffusion-webui-directml folder

and when you run pip cache purge, it will delete the cache files of python

[safe basin]

I think I already did this but Ill restart, clone it and try again.

[ornate elk]

sure 👍

also run the
sfc /scannow
as admin, to fix any system errors, on windows 11 that helps alot xD

[safe basin]

ok the fancy bar appeared atleast, lets see

but the text is back to flickering like it was before

[white stirrup]

?

[ornate elk]

hmm okay, give it a bit of time

[safe basin]

ill wait 15 min but I still think its bugged

I waited like an hour before

[maiden egret]

what's a recomended upscaler for comfy ui (ignore the loaded model)

[subtle quarry]

i believe the wait list is closed?

[subtle quarry]

I use 4x-ldsir I think it's called

yeah, 4x-ldsir

[safe basin]

nothing so far

[subtle quarry]

just saw the edit. I think Ultimate SD Upscale is the one everyone uses, I haven't seen anything else personally

but i'm also like... not the most knowledgable. All I know is that USDU works really well for me so I just stick to it

[ornate elk]

never saw anything like that. your internet connection is so good, but it doesn place the file inside the cached folder it seems

[safe basin]

yeah idk, I must have broke something along the installation. do you think upgrading to w11 could help or would I have to do a fresh install.

[ornate elk]

im running on windows 10 with no problems

have you made the sfc /scannow ?

[safe basin]

yeah there were no errors

[ornate elk]

a fresh install wil fix any issues but sure is the last option

[safe basin]

Yeah but I think since It only worked with admin access I might have ran some pip commands through there. maybe thats why its not working now

the privileges might be messed up. this is one of the outputs I got yesterday when I tried running webui-user.bat without admin priv

[ornate elk]

yea you shouldnt need to install any pip stuff by yourself

ohh i have an idea

go into C:/users/klask/appdata/local/pip/
then right click the pip folder and under options, security, enable the read and write permissions for your User and other software, and programms

[safe basin]

I think it worked

ah im back at this, its what happened when I didnt run it as admin before

but everything else worked

do I just give that file permissions aswell?

[ornate elk]

yep

i hope you didnt launched the webui-user.bat as admin

[safe basin]

haha no, learnt from my mistake

new error new file without permission

[ornate elk]

new error?

[safe basin]

I mean not error but new file that doesnt have permission

[ornate elk]

give your Users folder permission for your User

[safe basin]

Is it thanks to me running stuff in admin cmd thats causing these issues or why dont these files have access?

[safe basin]

gave users full control but still getting that error

[ornate elk]

it could be yes

something is very wrong with your file permissions

thats why it cant place and install/download stuff on C

[safe basin]

well I gave "huggingface" access and now im stuck here instead

nvm I was too quick to calling it stuck

well I got the ui to work but like the last time I got zluda to work I cant generate images

[ornate elk]

so your in the webui now?

in browser?

[safe basin]

yes

[ornate elk]

okay perfect, then type something in the prompt box and press "Generate"

the first image will take (no joke) 15-40 minutes

the cmd will show nothing in this time

the Generate button will be grayed out

after pressing it

then its just waiting, watch a youtube video or seomthing

[safe basin]

why does it take so long, on the other engine it was like instant

[ornate elk]

its only for the first image.
it builds and prepares a file needed for Zluda, in the background. this file takes that long to be created

the other webui version uses an other technique which doesnt neeed this first time build up

[safe basin]

alright, whats the odds that my vram usages spikes up to 16gb and crashes like it did on the other engine?

I was doing hires 25 steps, it shouldnt need that much no? it never went down unless I closed it completely. I remember a while back I was messing around with stable diff on a 3070 8gb and I had no problems.

heading off but thanks a lot for the help. knowing my luck there will def be another problem tomorrow.

[ornate elk]

That was in the other version

It is not optimised in vram and performance

The one you installed now will be not using as much vram and you can easily use hires fix and upscaling

[pseudo prairie]

what do i do now

how can i add C:\ZLUDA folder anf %HIP_PATH%bin

everything is so confusing

also

the only edit option was notepad

[ornate elk]

yep with notepad is okay

[fair oxide]

@ornate elk any idea why im getting this error?

[ornate elk]

it should look like this:

add these two lines

[ornate elk]

hmm strange, it should work

[pseudo prairie]

oh so i just copy paste those files here

perf

and thats it

or any other steps

[ornate elk]

not copy
you create new entrys
in the environment path

[fair oxide]

i even updated the pip as you can see but not working

[pseudo prairie]

oh ok

[fair oxide]

following this https://github.com/Gourieff/sd-webui-reactor?tab=readme-ov-file#installation but getting that invalid error

i accidentally ran this command before insightface-0.7.3-cp312-cp312-win_amd64.whl

312 wasn't compatible with my platform

could that be causing the invalid error?

this is the link to those whl files. https://github.com/Gourieff/Assets/tree/main/Insightface

i was supposed to run 310

if that caused this issue, any way i can reverse it?

[pseudo prairie]

do i press okay now

[ornate elk]

yep

[ornate elk]

yep, have you deletet the venv after changing python?

[fair oxide]

i havent changed python

it was insightface installation

[ornate elk]

@fair oxide ah okay

you can open up a cmd and type
pip cache purge

then try again

[pseudo prairie]

i think my laptop has RX 6800 GPU

do i still need to follow step 4

[fair oxide]

ok going to try that in a bit

[pseudo prairie]

do i do those steps

[ornate elk]

is it a 6800 or a 6800M ?

[pseudo prairie]

M

6800M

is that bad

[ornate elk]

nope, but let me check

[pseudo prairie]

ok

that’s the specs i think

[ornate elk]

okay you need to do step 4 and the others

a M variant is below an 6800

[pseudo prairie]

oh sad

[ornate elk]

so you need to download the ROCmLibs.7z file and do the steps

[pseudo prairie]

okkk

i will do that after i come back home

[pseudo prairie]

thanks

[ornate elk]

sure

[ornate elk]

no problem 🙂
let me know if you have any questions again

[fair oxide]

@ornate elk didnt work after pip cache purge, im going to delete venv and rebuild

[fair oxide]

@ornate elk that error still showed up when i tried to install insightface ... whl file, but surprisingly insight face faceID controlnet is working fine

[molten creek]

using auto1111 checkpoint merger, i had set a checkpoint for model c but cant seem to remove it so that model c is blank. before i was just reloading ui, but i guess at some point i applied defaults so now theres always a checkpoint in model c. any ideas how to remove it?

[pseudo prairie]

Go into C:\Program Files\AMD\ROCm\5.7\bin\rocblas folder.
There copy and rename the library folder to old_library
Open the .zip file and drag and drop all files of the library folder into the library folder. Not into the old_library folder.
Important Restart your PC after this Step before proceeding.

for that step

idk how to do it\

i downloaded everything on D:

and i don't have program file in D:\

as for C:\ there is program file but no ROCm

also without the last step ⬆️ im still running into this issue

i am pretty sure i picked that path option check box thingy

[pseudo prairie]

this is what it says without step 4 & 4.5

i feel like im close to finishing it off

but yet so far

idk what to do

im guessing i have to update pip to 24.0 but idk what pip is

[ornate elk]

Wherever you have rocm installed

Go to its folder like the path said

You don't have programm files on D. But you should find the AMD/ROCm folder

[ornate elk]

Yes your only missing the rocm Library step

[pseudo prairie]

wait

i am confused

[pseudo prairie]

i can’t find it im so dumb

[ornate elk]

By default its in C:/Programm files/AMD/

[pseudo prairie]

those are my options

[ornate elk]

Maybe you didn't installed HIP sdk

[pseudo prairie]

how do i install HIP

[ornate elk]

The link is in the guide.
Step 1 of the Zluda instructions

[pseudo prairie]

which one

[ornate elk]

5.7.1 for windows 10/11

[pseudo prairie]

i think i fid

did*

[ornate elk]

Nope

Thats zluda

[pseudo prairie]

ok

so i am downloading it

whats after that

[ornate elk]

Installing it but make sure the option
AMD "Pro Driver" is disabled or not checked

[pseudo prairie]

how do i make sure of that

after i download

or before

[ornate elk]

After

[pseudo prairie]

so after that’s download i click on it?

[ornate elk]

When you double click it to install

Then you get options

[pseudo prairie]

thats what i get after i double click

@ornate elk

oops sorry

[ornate elk]

Yes

[pseudo prairie]

i press yes?

[ornate elk]

Yes

[pseudo prairie]

okkk

i don’t see pro drivers

[ornate elk]

At the bottom. Its not enabled

Thats good

You can click install

Don't change anything

[pseudo prairie]

am i close

im scared

[ornate elk]

Nothing to be scared of

[pseudo prairie]

im guessing accept

im soo dumb with it comes to tech

like its not even funny its just embarrassing

[ornate elk]

Thats the browser site. You can close that

[pseudo prairie]

done

i pressed finish

now what

[ornate elk]

Click skip this and then close

[pseudo prairie]

done

now AMD?

[ornate elk]

Then do the 4.5 step

Of the guide

[pseudo prairie]

just to clarify

i duplicate the library folder

or do i just rename
the already existing folder

[ornate elk]

You duplicate it and rename one of them

One is called library and one is called old_library

Then you move all the files from the ROCm libs zip you downloaded into the library folder

Not the zip itself. Only the files that are in it

[pseudo prairie]

whats the ROC’ libs

oh wait

i found it

so i copy paste all of that

into the library one

[ornate elk]

Yess

And if it asks to replace some files then click yes

[pseudo prairie]

how do i copy

when i select all of that and right click i don’t get the copy option also control c doesn’t work

[ornate elk]

Bad zip Programm then.
Just close it and right click the rocmlibs zip file.
Select Extract as rocmlibs

Then you should have a folder called rocmlibs

Or install 7zip and open the zip file with that

[pseudo prairie]

what should i do

[ornate elk]

Continue

[pseudo prairie]

also does it matter where it is. Cause all my file is in D but this Amd thing is in C

[ornate elk]

Thats fine

That AMD thing has to be on C

The rest doesn't matter

[pseudo prairie]

ok done

its that it

mow i restart?

[ornate elk]

Can you show me first

The Library folder

Or the files inside it

[pseudo prairie]

i put those on the original libary

no the duplicate old_libary

[ornate elk]

Okay perfect

[pseudo prairie]

am i done?

[ornate elk]

Then yes restart is the next step

[pseudo prairie]

fingers crossed

if it doesn’t work i cry

[ornate elk]

Nah we can fix it if it doesn't work

[pseudo prairie]

i open it through the .user.bat thingy

[ornate elk]

Webui-user.bat

Always use that to launch stable diffusion

[pseudo prairie]

okk

i think its working so far

is is running of this this command

im just waiting for it to end

[ornate elk]

If its done it should open in your browser

[pseudo prairie]

ok perfect i will let u know

its stuck on installing requirements

[ornate elk]

Its not Stuck

Its doing stuff in the background

[pseudo prairie]

ur right

now its doing something else

[safe basin]

Bad news after letting the first image render for about 20 minutes my pc just shuts down. I can’t start it back up until I click off/on on my PSU.

Happened twice now

My ram sticks still had their lights on but I couldn’t start the pc.

[pseudo prairie]

its working

expect is it local?

[ornate elk]

Yes its local

[pseudo prairie]

yay

im so excited

[ornate elk]

@pseudo prairie the first image you generate will take 15-40 minutes.
Just let it load

[pseudo prairie]

thank u so much

[pseudo prairie]

okkk

[ornate elk]

Its only one time

After that it will work normal and fast

[pseudo prairie]

wait how do ik its loading

cause i pressed generate

[ornate elk]

The generate button will be grey.

Until its done

[pseudo prairie]

it did

for a second

but it went back to orange

[ornate elk]

Oh then check the cmd

[pseudo prairie]

thats why im confident

[pseudo prairie]

what cmd

😦

[ornate elk]

Ohhh you closed it?

[pseudo prairie]

no

i didn’t close anything i think

[ornate elk]

After you launch the webui-user.bat it needs to stay open

That black window is a cmd

[pseudo prairie]

nope its still open

ik that much cause i modded other games begore

is it because i have no model

or

[ornate elk]

Okay it seems like there is something wrong with the Rocmlibs thing

[pseudo prairie]

oh no 😢

[ornate elk]

You should have a model in the top right

Left*

[pseudo prairie]

what should i do now

[ornate elk]

You need place the rocm lib files again inside the library folder

[pseudo prairie]

and how i do that

[ornate elk]

Then restart the PC again

Just like before

You still have the rocmlibs zip file

[pseudo prairie]

ok so

step one

i go back to the libary

[ornate elk]

Whats your PSU wattage power?

[pseudo prairie]

21 files

right

do i do it for all 21 files

all current otems

items*

[ornate elk]

Oh wait

Yea its fine

[safe basin]

I think its 650w

[pseudo prairie]

i opened the folder copyied all the files

now im pasting it

[ornate elk]

And what's your GPU?

[pseudo prairie]

so continue?

[safe basin]

6950xt

[ornate elk]

Yea that could be to less power dor that GPU

[pseudo prairie]

but i did this before how is this going to fix my issue

[ornate elk]

Now delete the venv folder inside the stable-diffusion-webui-directml folder.

Then restart the PC. Then relaunch the webui-user.bat

[pseudo prairie]

that one right?

[safe basin]

I mean recomended is 750w but I havent had any problems since I bought the pc. My last build had a 3070 and the recommendation for that is 650w and I used a 500w for stable diff with no issues.

[pseudo prairie]

wait

the venv thing came backe

even tho i deleted it

[ornate elk]

Yep, but that card needs more power than a 3070.
AMD GPUs need more than nvidia cards in general.
So 750 or 850 would be the best case.

[pseudo prairie]

it happened before too

[ornate elk]

Thats intended

[pseudo prairie]

oh okk

fingers crossed

[safe basin]

Yeah I probably could have went for a better PSU but during the first image render my GPU is under 10% utilization. If its not even using my gpu that much how come it crashes?

When It worked on directml version it was on 100% but it never crashed, except it failing to render the image.

[ornate elk]

The first image of the Zluda install makes your PC generate stuff in the background that is needed for the zluda/cuda Backend.
These file creation or convertion uses GPU power, RAM, and CPU, and at the last step tries to generate the image.
I think that this will draw more than on directml which only uses ram, and GPU separately after another. Which draws less power at peak

[pseudo prairie]

what do i do now

😦

i was so close now it feels far away

[ornate elk]

What's in your webui-user.bat

Your missing the launch args

[pseudo prairie]

how do i check

[ornate elk]

Close the cmd.
Right click the webui-user.bat

And click edit

[pseudo prairie]

edit in notepad?

wait

it launched

the stable diffusion started itself

do i try to see if it works

or close it and edit notepad

[ornate elk]

Show me what's inside ^^

[pseudo prairie]

i was expecting a lot more

[ornate elk]

@safe basin so yea the only way is to upgrade the PSU to use Stable-diffusion with Max performance.
In the meanwhile you can install the Directml Webui seperatly and use that until you got the 750w or 850w

[ornate elk]

You should have edit it like stated in the guide xD

Your missing the essential launch arguments

[pseudo prairie]

no i did

i swear

[ornate elk]

Maybe you didn't saved it

[pseudo prairie]

how do i save lol

[ornate elk]

Click on file at the top left and then on save

[pseudo prairie]

i add that?

copy paste?

[ornate elk]

Yep

[pseudo prairie]

or type

[ornate elk]

You can copy it

[pseudo prairie]

space between =— or = —

[ornate elk]

Make one space

[pseudo prairie]

now restart again

or just open

[ornate elk]

Relaunch the webui-user.bat

[pseudo prairie]

still not working

the generate thing still goes to orange

doesn’t generate

[pseudo prairie]

it opened but back to square one

[ornate elk]

What does the cmd shows

[pseudo prairie]

invalid argument to memory_allocate

[ornate elk]

Can you show me the whole text

When you scroll up

[pseudo prairie]

can i copy paste

cause their is a lot

[ornate elk]

Sure

[pseudo prairie]

would that be easier

[ornate elk]

No don't do that

[pseudo prairie]

i didn’t

dw

[ornate elk]

I may found the issue

[pseudo prairie]

yay

is it fixable

[ornate elk]

The webui tries to run on your igpu not the 6800m

[pseudo prairie]

what is the fix

[ornate elk]

You have to open up the Device Manager.
Then go to Display Adapters there check if you have two entry's there.
One of them should be called:
AMD Radeon Graphics TM
The other one 6800M
Then you right click the
Radeon TM one and disable it

[pseudo prairie]

i do actually

so i disable ghe TM

would that cause any issues tho

which one do i diasble

[ornate elk]

Nope as you have the 6800M it should use that instead

[ornate elk]

TM graphics

[pseudo prairie]

done

now?

i restart

or

[ornate elk]

Now relaunch the webui-user.bat

If you get the same error again you may need to restart the PC

[pseudo prairie]

hopefully i don’t

cause last night i spent like 3 hours trying to follow this tutorial clearly didn’t work

[safe basin]

@ornate elk I tried one more time and apparently as soon as the first image loads it crashes. I don’t think it has anything to do with the psu.

[pseudo prairie]

today another 3-4 hours hopefully it works

[safe basin]

Maybe I’ll just try a fresh install.

[ornate elk]

Okay, yea that could help too.
But a PC crash is most likely PSU, RAM or Driver related.
But still worth a try. One guy fixed his problems with a reinstall

[pseudo prairie]

ITS WORKING

I THIO

THIS TIME ITS GRAYYYY

im sooo happy

[ornate elk]

Thats good 🙂

Now its just waiting

Be patient. 15-40 minutes

[pseudo prairie]

okkk

patient is keyy

[ornate elk]

Also the cmd shouldn't show any errors.
In fact it will show not much the whole time while waiting.

But its still doing stuff

[pseudo prairie]

so pretty much the first image didn’t load and the generator went back to orange but i pressed again but that time it worked and it worked faster

now what i wanna know is how can i generate better

this are giving me weird images

also what are models

[ornate elk]

Models are used to generate images.
The default model you already have isnt good.
Go on Civitai.com
And there for example download Dreamshaper v8
It can generate better images.
Put the .safetensor file into stable-diffusion-webui-directml/models/stable-diffusion folder.
Here is the link to the model:
https://civitai.com/models/4384/dreamshaper

[pseudo prairie]

what is the safetensor file

[ornate elk]

thats the model

[pseudo prairie]

im guessing that

into the model file

[ornate elk]

into SD-Zluda/stable-diffusion-webui-directml/models/stable-diffusion folder

[pseudo prairie]

perfect

so models are the artstyle

and i can use that artstyle to create anything

so if there is women model can i create men from it

[ornate elk]

yep

[pseudo prairie]

same for the other way around

[ornate elk]

you can create everything with any model

some are better than othersd

[pseudo prairie]

oh

[ornate elk]

some models can do better realistic stuff, and some are made for anime

or cartoon

[pseudo prairie]

i got stable diffusion for reference cause i do art

• i also got it for fun and silly things

[ornate elk]

😄

you can do a lot with it

from amazing art to silly meme stuff

[pseudo prairie]

i tried using different models but each time it created nothing

it didn’t create anything

[ornate elk]

what does the cmd shows?

[pseudo prairie]

im not sure i have guests over

i will look at it tomorrow

thanks for all the help tho

[ornate elk]

sure, let me know then

no problem

[primal glen]

I'm having trouble getting inpainting to work. Someone tell me what I'm doing wrong? Here's the drawing I start with...

[tacit maple]

Anyone else keep getting 'NoneType' object has no attribute 'rsplit' randomly?
It happens with img to img, usually (but not always) after changing the cfg scale or denoising strength.
Closing and re-entering the image sometimes helps, relaunching the program always helps, it's pretty annoying, and it only started happening with xl models.
I'm using A1111, with PonyXL

[primal glen]

then I go to the "inpaint" tab, draw over the portion I want to change, type in the prompt, click "generate".

[tacit maple]

show us your cfg scale (should be at the bottom)

[primal glen]

cfg scale is 7

[tacit maple]

you mean 0,7?

oh wait

what am I saying

[tacit maple]

denoising strength

there we go

set that higher

to 0.95

[primal glen]

okay done. I'm on my slow computer so it'll be a bit to see if that fixed it.

[wild estuary]

so i'm trying to train a character lora for PonyXL. The issue I am having is that after the first epoch is trained and saved, the loss will become nan and all remaining epoch files will require me to enable no half, at which point they will take upwards of 7 times my normal sdxl speeds and output random noise with obsolutely no coherent concepts(not even like when you leave the prompt blank and get a perplexing image, this is like what i imagine you would see right before going blind).

any help would be much appreciated. (training colab i used: hollowstrawberry/kohya-colab/blob/main/Lora_Trainer_XL.ipynb)

[safe basin]

@ornate elk did a fresh install of windows 10 and downloaded with no problem this time from start but when I try generating an image I get this

[ornate elk]

whats in your webui-user.bat?

and what are your txt2img settings

[safe basin]

wait I think I fixed it, I deleted a character in the bat file by mistake

[wind roost]

hi guys

got a predicament

dpm ++3M SDE karras, or any of the karras ones are not showing up

[vapid sluice]

How can i install packages INSIDE comfyui’s python embedded? I can’t find anything about it online, and i don’t really understand how it works

[vapid sluice]

I know what im doing though

Also has anyone here successfully installed musepose node for comfyui?

[ornate elk]

Hey, in Auto you have to select Karas at the scheduler

Or leave it on automatic

[ornate elk]

Did it worked now?

[safe basin]

Well I got it to generate the first image now. But as soon as I switched to hires 25 steps It crashed again. Installing the directml version and im gonna see if I can find a way for it to not use 100% of my gpu.

[fair oxide]

i havent had this much crappy errors with insightface ever

[ornate elk]

dont use 25 hires steps lol. thats to much

use 10

and then upscale by 2

you also need to install tiled vae extension to upscale large images without getting an out of memory error

[safe basin]

where do I find that

And It just crashed again. It didn’t even let me generate a simple image this time.

[ornate elk]

click on extensions tab
then on available
then click on "load from"
then you get a list, there search for tiled

[ornate elk]

pc crash?

or cmd error?

[safe basin]

Is it really the PSU. I’ve never had any problems until now

Pc

Just shuts off no warning. Doesn’t even say that my gpu is being used.

[ornate elk]

you said you reinstalled windows, so i guess it can only be the psu as a 6950xt is a highend gpu

needs a lot of power, and SD can draw more power than games

maybe do a benchmark test

[safe basin]

But shouldn’t it crash because it’s drawing out more power than I have access to. My task manager says my gpu is on like 10% before it crashes

Yeah I’ll run a benchmark

[safe basin]

But like what kind of benchmark and what should I look for?

[ornate elk]

a heavy one for GPU

which draws a lot of power

[cinder ermine]

do i need to caption regularization images

[fair oxide]

@ornate elk if i'm installing visual studio 2022 which are the feature should be installing for insightface?

just the one that says Desktop Development with C++ ?

[white stirrup]

anyone know how to fix this error?

i tried to restart the UI after installing an extension and this happened

i then tried deleting it in my files and same thing happened

[south vigil]

hi, where do i put pth files typically inside?

its some sort of an "addon" for image2image in stable diffusion

[dense rune]

hi, im new in comfyui and i have a problem with this, basicly i did a txt2img+hiresfix with ultrasharp(here everything is fine), but then i wanted to do another upscale but using latent upscale but the output looks like this

the edges looks blurry

[livid lark]

ELLA has no contorlnet support right?

[abstract fractal]

really annoying that the api returns blury images almost always if its a woman character. Like this prompt gets blured out, for "woman", but if I swap it with man, its not blurred out.

"Elara, a young woman with grease-stained fingers and worried eyes, stands in her workshop surrounded by anatomical sketches, glowing crystals, and intricate clockwork gears, holding a still-beating artificial heart in her hand."

if i remove the word young, it still gets blurred.

[ornate elk]

Delete the venv folder too and relaunch the webui-user.bat

[ornate elk]

Yep

[ornate elk]

Hey, they go into models/ESRGAN

[white stirrup]

I did that 2 hours ago i just forgot to say it

[ornate elk]

Have you delete the extension too?

[white stirrup]

Yeah

It works

I dont have the issue anymore

[ornate elk]

Perfect

[fair oxide]

did that, i get bunch of red texted errors, but FACEID seems to be working despite that, my OCD is bugging me

[pseudo prairie]

so

whenever i try to generate any image

it goes to that

and after that

it goes back to this

im using models btw

this is just some of the models

venv "D:\SD-Zluda\stable-diffusion-webui-directml\venv\Scripts\Python.exe"
WARNING: ZLUDA works best with SD.Next. Please consider migrating to SD.Next.
Python 3.10.11 (tags/v3.10.11:7d4cc5a, Apr 5 2023, 00:38:17) [MSC v.1929 64 bit (AMD64)]
Version: v1.9.3-amd-23-gd23a1724
Commit hash: d23a1724921e58d6ba92c40236cbf8dd139c130b
Using ZLUDA in D:\SD-Zluda\stable-diffusion-webui-directml.zluda
Skipping onnxruntime installation.

| You are not up to date with the most recent release. |

Consider running git pull to update.

no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
D:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: pytorch_lightning.utilities.distributed.rank_zero_only has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from pytorch_lightning.utilities instead.
rank_zero_deprecation(
Launching Web UI with arguments: --use-zluda --medvram-sdxl --update-check --skip-ort
Loading weights [c667408232] from D:\SD-Zluda\stable-diffusion-webui-directml\models\Stable-diffusion\Goro_XL_V10_OPT.safetensors
Creating model from config: D:\SD-Zluda\stable-diffusion-webui-directml\configs\v1-inference.yaml
Running on local URL: http://127.0.0.1:7860

To create a public link, set share=True in launch().
D:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\huggingface_hub\file_download.py:1132: FutureWarning: resume_download is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use force_download=True.
warnings.warn(
Startup time: 10.8s (prepare environment: 12.3s, initialize shared: 2.3s, other imports: 0.4s, load scripts: 0.4s, create ui: 0.6s, gradio launch: 0.2s).
Applying attention optimization: Doggettx... done.
Model loaded in 3.0s (load weights from disk: 0.7s, create model: 1.2s, apply weights to model: 0.4s, load textual inversion embeddings: 0.2s, calculate empty prompt: 0.4s).
100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [00:11<00:00, 1.70it/s]
Total progress: 100%|██████████████████████████████████████████████████████████████████| 20/20 [00:11<00:00, 1.75it/s]
Total progress: 100%|██████████████████████████████████████████████████████████████████| 20/20 [00:11<00:00, 1.80it/s]

thats what my cmd say

why do u have that

🙄

caught u red handed

[steel geyser]

Generating images with SDXL crashes my computer, not even a blue screen, just turns off immediately. SD 1-2 works fine.

Using easy diffusion with a GTX 1080Ti

[formal trout]

Have to try adding flags like --medvram-sdx

[lost crescent]

At pytorch.org - I request version 2.3.0 - but it persists in giving me 2.0.0!!!

[ripe tendon]

i typed in ahegao girl and this is what i got

any reccomendations for settings so they arent all morphing or botched

[lost crescent]

The dimensions of your output seems to be too large for the data on which your checkpoint was trained?

[ripe tendon]

ill look into it thanks, ima try to fix it

any idea how to fix the faces?

the faces were looking better before(exept for that cursed image i send before this one)

[ripe tendon]

how do i have bigger dimensions without it bugging out?

[ornate latch]

i wante to try the stable artisan but when i finish the checkot it leeds me to stable assitante . when i enter the discord server of the stable artisan it say that i dont have live subsecreption

hello , i need help with this please

[lost crescent]

Tiled VAE, or Tiled Upscaling

[lost crescent]

Reactor Restore Face - or Face Detailer

[crystal nacelle]

hey good people. ive been using Zluda on my AMD for ahile now , and its excellent

it told me to downlad a git pull update so i did that. now its just stuck on this screen

im gna try closing and re-launching

relaunching worked. but generation not creating anything

[ornate elk]

Let it load

It will take a while

It's like the first image after doing my guide.
It can take 15-40 minutes for the image.
Then it will be normal again

[crystal nacelle]

okay so this is normal? thank you. ill try again later

[ornate elk]

Can you try an other model?
My guess is that Goro XL isnt a model but a lora. They need to be in an other folder.

[ornate elk]

Yep its normal. It can happen after a webui update and after GPU driver updates.

And also whem updating zluda

[ornate elk]

Hey, you need to connect your stability account with your discord account.

[ornate latch]

how to do so

[ornate elk]

Do you have made an account + subscription on https://stability.ai/stable-artisan ?

Then also checkout #artisan-faq

[ornate latch]

yes i have entred my credit card detailes and i supposed to be on free trial now

but when i entred i founde myself on the stabe assitante not the artisan ok

?

after i clicked on the discord of stable artisan it said that i have no live subsecreption

i think it is some issue with the connection between discord and the stabilty ai account

[ornate elk]

Yea mostly, can you check your stability ai profile? Maybe there you can link your discord account

[ornate latch]

the thing is why when i enter the credit card details it leads me to the assistant and not the discord

i didi try it 2 times and the same promblem after the checkout page

do you use it ?

[ornate elk]

Nope haven't tried it yet

@ornate latch What you can do is ask in
#🗣｜artisan-support-feedback

[ornate latch]

i did , they take too muchn time to responde .

[ornate elk]

Ohh

[ornate latch]

thank you for ur help

[ornate elk]

No problem. What you can also do is to create a ticket at the Stability ai site

@ornate latch can you login here?
https://assistant.stability.ai/

With your discord

[ornate latch]

yes

im in

[ornate elk]

Okay and there you have an active trial?

Then go into #1237459938901491852 and test if you can use the dream command

[ornate latch]

yes it worked

thank you man

i apreciat that

[pseudo prairie]

i think its a lora XL whatever u call that

how does that work

[ornate elk]

No problem 🙂

[pseudo prairie]

what do i do

[ornate elk]

Loras are not models.
They need to go into models/lora folder.
They are used together with a model to help the model create a specific character or style

[pseudo prairie]

ohhh

cooolll

so i have bunch of lora models

lol

[ornate elk]

If the lora is made for XL you also need to use an XL model to use it

If a lora is made for 1.5 you need an 1.5 based model

[pseudo prairie]

whatt

[ornate elk]

You can find more information on the model and loras page on Civitai.com
Every model and lora has a little Info box there

[pseudo prairie]

im lost

Life is confusing

[ornate elk]

Here for example is the Dreamshaper v8 model.
Its based on 1.5 and its a model (checkpoint)
You can see that information when you go on its page and scroll down.
https://civitai.com/models/4384/dreamshaper

[pseudo prairie]

oh kk

[ornate elk]

Here important infos:

[pseudo prairie]

oh okk

[ornate elk]

Loras are type lora

And base is either 1.5 or sdxl

[pseudo prairie]

k let me checked

check*

what things i have

ok

so almost all are Lora

and the base model is Pony

although the name says pony XL

[ornate elk]

Then you need a Pony model

Its a variant of sdxl

And move the loras into models/lora

[pseudo prairie]

how do i get pony model

[ornate elk]

Here is one:
https://civitai.com/models/311817/anime-confetti-comrade-mix?modelVersionId=395263

[pseudo prairie]

oh thanks

that goes into model right?

which one should i get

[fair oxide]

@vocal burrow https://suno.com/explore

[vocal burrow]

yes ?

[fair oxide]

wanted to share the suno link

its no doubt going to be the future of music entertainment

even as a free user you can create 10 tracks daily

[vocal burrow]

Yeah I know about it. Is it gonna be "the future of music" ? I hope not. It it gonna be part of it ? Maybe, probably. To what extent tho, impossible to predict for sure. But that's out of the scope of #🤝｜tech-support :p

[ornate elk]

The pruned model goes into models/stable-diffusion
The VAE goes into models/vae

[pseudo prairie]

yay my guess was right

[river flame]

I heard you can use Midjourney with InstantID, but cant find any info, can anyone help please? Dont know if its okay to ask it here

and also, what is better for portraits with real faces - Midjourney --cref or ReActor or InstantID?

[solemn atlas]

hey I have a question, what should I do if i have no GB left

[jolly crag]

Hello, (first of all sorry for the pronunciation, English is not my native language), can you help me with this problem, I do not know why before I generated the images well but a problem happened and I had to reset my pc, I re-install stable diffusion and now all the images gives me so, someone can help me?

[dense rune]

where is this option?

[primal glen]

I can't get inpainting to work. Any idea what's going wrong?

[ornate elk]

on a drive? or on gpu memory ? or RAM ?

[ornate elk]

hey, whats your gpu and whats inside your wbeui-user.bat?

[south vigil]

hey guys i have a picture 9:16 and want to convert it to 16:9 (let the ai add the rest of the picture on the side) till now i only get disgusting pictures out of it, is it even possible to do it right? if yes you got any tipps? :3

[ornate elk]

set the inpaint mode to "mask only"
then set the resolution to 512x512 and the denois to 1.0

[ornate elk]

its poissble with so called outpainting

it can be done in img2img with the outpaint script (but that worksd only on ai generated images)
or it can be done with controlnet outpainting

[south vigil]

alright then im gonna do a copy of the image 9:16 and then try to outpaint it

[south vigil]

can i ask you, I get often very bad faces from img2img and cant fix it even with inpaint, how do you correct bad/ugly faces when the rest looks good in a picture?

[ornate elk]

thats easily fixable with inpainting the face or using the Adetailer (after detailer) extension that will inpaint and fix the face automaticly

[south vigil]

Wow, Huge help, thanks G!
ADetailer is super strong

[south vigil]

Sorry i dont want to bomb you but do you also have a tip for weird hands?

[ornate elk]

sdxl models can do better hands, but other than that there is not as much you c an do about it

[stable dirge]

Any idea how to make ReActor run faster? Feels like the setting up takes a lot, not the swapping itself.
It can be felt especially when using mov2mov for "deepfake" style video. one frame takes about 2-3seconds, and that's all the info I get each frame:

[fair oxide]

IMPORTANT QUESTION:

When training a character (in my case, celebrity) LoRA, it is said that one must not include too many pictures of one particular attendance of an event (Golden Globes, BAFTAs) as this will result in bias within the LoRA. However, if I included 10 pictures of each event, for every event, wouldn't that get rid of the bias as it is now evenly distributed? I'm asking because, there is sometimes great variety in camera angles, composition and facial expressions in photographs of one event, but I am not "allowed" to use all of them, because the maximum is one for each event, greatly reducing the potential of the LoRA.

[zenith sail]

Anyone know the change to model.load_device?

                              ^^^^^^^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'load_device'```

[fair oxide]

does anyone else have problems with euler a? lol

also, does anyone know how to increase VRAM usage for stable diffusion alone? I have 12 gigs and it used to not be a problem for me, but now I can only generate a few before i get the CUDA error

[brave frigate]

For some reason my VAEs don't do anything it seems

[vocal burrow]

no why ?

[vocal burrow]

So you want to DEcrease vram usage then.

copy pasta incoming

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. To alleviate Vram usage you can also try to turn on fp8 mode in the 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,

[brave frigate]

With some it just crashes right at the end, others its lots of yellow noise... One even froze my computer for 3min. I have installed them correctly, put them in the right folder and selected the sd_vae in the Quicksettings.. but for some reason nothing changes or it just brakes the generation. Does someone know a solution? The two images attached is one with a vae and the other with no vae.

[vocal burrow]

most likely you're using mixing 1.5 vae with sdxl model

can't tell much more without seeing every setting used for those.

(also please keep it classy)

[brave frigate]

mb want me to delete the images?

[fair oxide]

yo chef mine look like the left one too

it gets so grainy out of nowhere

[brave frigate]

yea

idk how to fix it

[fair oxide]

i have been trying to google it but no luck

imma make a reddit post

[vocal burrow]

well are you using sd 1.5 vae with sdxl models ?

same thing, let me see your settings

[fair oxide]

no, i use some sort of VAE with euler a

i just reset my stable diffusion to use the medram code thing

thank you for that btw

[vocal burrow]

that's not answering my question tho

[fair oxide]

i saw that on another post but your epxplanation was a lot better

[vocal burrow]

"some sort of VAE" is not telling much and I don't know what model you're using with it

screenshot me those webpage

[fair oxide]

oh yes, i use 1.5 i believe

okay

a blank page?

[vocal burrow]

as long as it's the same setting you've used for your generation.

[fair oxide]

okay

let me replicate it

[vocal burrow]

the whole page

it's not showing the model you're using (and extensions if you're using any)