🤝｜tech-support

[ornate elk]

nope

[kind sedge]

give me for example a good extention you use for the latest version

[ornate elk]

reactor for faceswap

[kind sedge]

ok i see, give me asec

[little gate]

ok, updated as many of these as I could find.

not sure what "adetailer" is

or model hash

[ornate elk]

adetailer is an extension to get nicer faces

the model hash ist just an identifier for the models

[little gate]

So, no matter what, this forge is actually running faster, and I can go to the size he suggested

I set the scale at 30 because I want it to not drift from the prompts.

Is there a way to ensure it follows the prompt without messing with that

?

[ornate elk]

scale?

30 is the sampling steps

and cfg is 7

[little gate]

CFG Scale is what it says

and I set that to 30

because I want it to follow the promprts

[ornate elk]

no please not

[little gate]

but thats what I am asking

[ornate elk]

you see he used 7 here:

[little gate]

but doesnt that just mean he was more open to letting the program do its own thing?

[ornate elk]

yea but anything above 12-14 will break the image

you cant use cfg 30

aswell as you cant use cfg 1 for normal models

you have to be between 7-12

[little gate]

got this, didnt follow the prompt really and its still lacking that pixel look

[kind sedge]

how do i get rid of these warnings?

[ornate elk]

you cant, they are normal and will maybe removed with upcoming updates

[chilly knot]

its normal

[ornate elk]

its pixelated

[little gate]

lol no

this is... I dont know... but not pixel

[kind sedge]

@ornate elk this is the error i was talking about

[ornate elk]

which model did you used?

[ornate elk]

try delete the venv and rerun the webui-user.bat

[kind sedge]

i tried installing the extentions from a fresh install

[kind sedge]

i mean on a fresh a1111 install

[ornate elk]

it says permission error, thats because you git cloned directly into your pc user folder and not into a self created folder

so you have to edit the webuis permission

[kind sedge]

isee

[ornate elk]

to give your user access

[kind sedge]

you are right the folder in on the user folder

so you recommend making another folder somewhere and reinstall a1111 on it

tbh i would rather just edit the permission, how would that be possible

[ornate elk]

yea like in my guides, there i say to create an Ai folder on C drive

[kind sedge]

can you link me your guides please

[ornate elk]

you can do that easily

right click the stabel-diffusion-folder, click properties, security

then on edit

then select your username

and there give it full permision (top checkmark)

[ornate elk]

okay then try without hires fix enabled

[kind sedge]

weird it's already checked

[ornate elk]

oh okay

are you on win11 ?

[kind sedge]

ye win11

[ornate elk]

oaky then you need to make an extra folder and git clone the webui in there

[kind sedge]

i see

[ornate elk]

ill link my guide

[kind sedge]

ok, like make in the documents for example or where exactly?

yes please link it

[ornate elk]

https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides
follow the nvidia Automatic1111 Webui

[little gate]

The creator specifically said it has to in order to GET the pixelated look

[kind sedge]

can i make a folder and move the stable diffusion folder to it, without reinstall?

[ornate elk]

that would break the auto update feature

[kind sedge]

ok thank you imma reinstall it

[ornate elk]

a clean install will be better

[ornate elk]

okay, i see

really bad lora then xD

[little gate]

all positive reviews

that I can see

[ornate elk]

yea but the images arent really that pixelated

just a bit

the edges mostly

but anyway its not related to an technical issue here. you have to try out different settings i guess

[little gate]

This one was better, despite it not actually following my prompt well

[main nest]

btw does anybody know how to face swap with the batch option, I dont remember quite good

[ornate elk]

looks good, but you also need a vae file for color correction

[little gate]

?

[ornate elk]

thats a file that fixes the dull colors

[little gate]

where do I find something like that?

[ornate elk]

https://civitai.com/models/23906/kl-f8-anime2-vae

put the file into the models/vae folder

then go into settings, user interface, quicksettings, there add sd_vae and click apply and reload ui

then you can selecte the vae in a dropdown

[little gate]

ok thank you

How do I make sure it follows my prompt directly, without changing the cfg scale?

[ornate elk]

you cant get a 100% prompt follow

try cfg 7 and then adjust your prompt

you can give words more attention by using (word) or (word:1.2)

[little gate]

Thats a bit weird, with midjourney I can set it to follow it at maximum and it does a rather great job

I thought stable diffusion was a bit better for things since there is more control

[ornate elk]

not prompt wise. as midjourney does really a lot of stuff in the background to get things right

[kind sedge]

@ornate elk what does this do?

-no-half-vae

[ornate elk]

its for vae compatiblity

and loads the vae in fp32

needed when using anime vae files or sdxl vae files

[kind sedge]

i have a 2060 6gb of vram

very nice guide

--medvram

i forgot to add these command on the first launch, it's fine right?

[little gate]

sooooo... how do I ensure it follows the prompts then lol

[ornate elk]

you can try using other models

for example sdxl or pony models follows prompts better

[ornate elk]

yea you can add it later after the install

[little gate]

Thank you for all your help today, you have actually made it work for me

[ornate elk]

No problem at all 🙂

[little gate]

Can you hi-res fix on img2img?

Also, can you do character references?

[wraith valley]

hey im installing comfyui but when i want install the requirments its blocks

[ornate elk]

In img2img you can use the SD upscale script

[kind sedge]

same issue

weird

[ornate elk]

Which Extension?

[kind sedge]

facechain

could this be the reason?

[ornate elk]

I checked its github and it seems broken

[kind sedge]

i see but does it have to do with the oserror

[ornate elk]

I think so

If you don't get that error with other extensions

[kind sedge]

give me an extention that you are sure that it works please

[little gate]

So... can it do character references?

Also, can SD see image links?

[ornate elk]

Boorutag autocompletion

Reactor

Controlnet

[ornate elk]

No it can't see links

References? Maybe in img2img

[kind sedge]

this is for reactor, so are these things supposed to be install manually?

[ornate elk]

Yea reactor needs a manual install of the insightface file

https://github.com/Gourieff/sd-webui-reactor?tab=readme-ov-file#insightfacebuild

[kind sedge]

look even with controlnet

same issue just with cv2

im cursed

[ornate elk]

Thats really strange

Does your user has admin permission?

[little gate]

starting gto get somewhere!

[kind sedge]

yes ofcourse

should i reinstall my whole system

that's what i feel like doing tbh

[ornate elk]

Does users have all permissiona here?

And authenticates users

[kind sedge]

yes

even for the venv

[ornate elk]

Thats the same screenshot

Click on users

[ornate elk]

Okay

And administrators?

[kind sedge]

it's all the same

this is the only thing that causing this issue

opencv-contrib-python

should i try and install it manually?

[ornate elk]

You can

With pip install

But you have to be in the venv

[kind sedge]

ye imma do that

in the venv

it's literally only for the this

purged the cash and still nothing

[ornate elk]

What's your Python Version?

[kind sedge]

3.10.6

i think my user stuff are messed up

wtf is account unknown

[ornate elk]

Ohh

Yea that looks bad

[kind sedge]

very off

odd

umm idk if i should delete that

yeaa

when i delete that unknown user

it starts deleting my AI folder

[ornate elk]

Lol

Really strange

Did you renamed your user maybe?

2 weeks ago we had a user with really messed up folder permissions.
But he found a tool to reset them.
#🤝｜tech-support message

[kind sedge]

i see that tool might be helpful

[ornate elk]

Yea I guess its worth a try. Also open up a CMD and run
SFC /scannow
That will fix system errors from windoes updates

[kind sedge]

im doing this for users maybe it will fix

[ornate elk]

Ok

[kind sedge]

look it's only cv2

literally

idk why

[ornate elk]

Maybe try update your Python

To 3.10.11 64bit

Then delete the venv folder

[kind sedge]

a1111 supports 3.10.11 ?

what is the latest version that is supported by a1111

[ornate elk]

3.10.11

On Linux 3.10.14

It also can work with 3.11 but some extensions wont

So best is 3.10.11

[kind sedge]

ok i see

delete the venv and open webui right?

[ornate elk]

Webui-user.bat

Dont launch webui.bat

[kind sedge]

yes it's reinstalling all of the things

https://tenor.com/view/silly-cat-silly-car-car-stare-10-thousand-yard-stare-10-thousand-yard-gif-14200271775968563996

[kind sedge]

fixed

https://media1.tenor.com/m/zXRUDvNn0hIAAAAd/been-through-war.gif

@ornate elk thank you

[stone fog]

So how do i update SD/Gitpull?

[stone fog]

my runs keep locking up and looking like this.

[little gate]

getting this warning: [Low GPU VRAM Warning] Your current GPU free memory is 1272.80 MB for this diffusion iteration.
[Low GPU VRAM Warning] This number is lower than the safe value of 1536.00 MB.
[Low GPU VRAM Warning] If you continue, you may cause NVIDIA GPU performance degradation for this diffusion process, and the speed may be extremely slow (about 10x slower).
[Low GPU VRAM Warning] To solve the problem, you can set the 'GPU Weights' (on the top of page) to a lower value.
[Low GPU VRAM Warning] If you cannot find 'GPU Weights', you can click the 'all' option in the 'UI' area on the left-top corner of the webpage.
[Low GPU VRAM Warning] If you want to take the risk of NVIDIA GPU fallback and test the 10x slower speed, you can (but are highly not recommended to) add '--disable-gpu-warning' to CMD flags to remove this warning.

[onyx solar]

This looks like you are trying to generate an image that is the wrong size for the model you are using and/or you have the wrong or no VAE.

[stone fog]

It’s being done to the correct model specifications and happens across basically every model

[onyx solar]

You'd have to share all of your parameters to be sure.

[stone fog]

Steps: 30, CFG scale: 7, Sampler: Euler a, Seed: 2421349604, Size: 512x1024, Hires steps: 10, Hires upscale: 2, Hires upscaler: R-ESRGAN 4x+ Anime6B, Denoising strength: 0.3, Clip skip: 2

[onyx solar]

model? scheduler?

[stone fog]

meinpastel and im not sure

[onyx solar]

VAE?

Scheduler makes a difference, too...so you should look at what you're using.

Also, it wouldn't hurt to paste a screenshot of your settings in case there's something you're missing that we can see.

[stone fog]

schedule is karras

all things suggested by the model

[onyx solar]

What happens when you turn off the highres stuff and switch the scheduler to normal? (I still suggest just pasting a screenshot of your settings and to mention what your VAE setting is.)

[stone fog]

hangon I might be on 1.4 still

[ruby yoke]

are you?

[stone fog]

I am.

On 1.5

[onyx solar]

Why so far back? A1111 is up to version 1.10.0 for mainstream release.

[stone fog]

AMD

[onyx solar]

The version of Automatic1111 has nothing to do with the version of the models you are using. The fact that Stable Diffusion models can be version 1.5 doesn't mean you need A1111 1.5.

[stone fog]

I mean my CMD says Im up to date

[onyx solar]

Ok...well, if you're AMD, I'm out. That introduces further complexity that I don't have to worry about since I'm on nVidia.

[stone fog]

I mean amd really isn't all that complicated

I'm pretty sure they're on the same A111

[ruby yoke]

have you read thorugh all of @ornate elk 's AMD guides pinned at the top of this channel?

[stone fog]

Yes

[ruby yoke]

why are you using SD 1.5?

[stone fog]

I'm using the download that his guide provided.

[ruby yoke]

okay but what is the reason you're using stable diffusion 1.5 instead of SDXL?

[stone fog]

I have njo idea how to check the fact I may be using XL

[ruby yoke]

when you said you were on 1.5, did you mean auto1111 or stable?

[stone fog]

AUTO111

[ruby yoke]

and you don't want to update it to the latest level?

[stone fog]

I just git-pull'd yesterday

[ruby yoke]

okay then i'm confused what the issue actually is?

[stone fog]

I was getting blue artifacting and didn't know what was causing it

But it seems to have been a issue with VAE

I am going to figure out where exactly VAE's go and move from there

[woven zealot]

Idk if that's the right channel, but I'm looking for an extension that would let me do this:
Queue up generating 10 images with the batch count
During the first image, after it's 20% done generating, cancel it and start the next one
At the end I have 10 under-rendered images and I can look through them and see which ones have potential, and just copy the seed and let them generate fully

Is there something like that?

[ornate elk]

Can you show your cmd log so I can check if anything Is outdated

[ornate elk]

There is the agent scheduler extension

There you can define the steps

[vocal burrow]

(but you can't cancel mid generation with it)

[ornate elk]

Yep so he needs to use a low step amount

[vocal burrow]

However, increasing (significantly) the number of steps will affect the composition of the output.

[lean fox]

Hi,
I'm using stable diffusion and i have RX480. Do i gain any meaningful advantage with adding RX470 i have laying around into crossfire. I did benchmark and speed goes down by a bit but can it possibly generate larger images? Curently sometimes i get an error where at the end of generation process there is no more space for tensor allocation. I'm a newbie in this so i apologise if the question is stupid.

[vocal burrow]

it s impossible to use unified memory across gpus with SD.

aka you won't be able to use multiple gpu at the same time for one instance

best you can do is run one instance per gpu

[hot trout]

Yes. Sent you a dm

[royal sapphire]

Is here someone who runs stable diffusion 3 medium with confyUI and a 4090 GPU? I want to compare the timings

[plain ocean]

Hello all!!

how are you all doing?

I just have a question, is swarm ui faster on amd??

or webui forge?

im not sure which one to go with

[ornate elk]

Forge is faster or Auto1111

[plain ocean]

oh

thx mate!!

Ill go with forge web ui then, hope you have a good day!

[ornate elk]

Thx, you too!

[plain ocean]

🙂

[spiral comet]

idk that should i ask here or not - but what is now best faceswap to forge? still reactor?

or there is smth better?

[ornate elk]

Roop-unleashed or FaceFusion

[cunning robin]

hi! is this the place where i could ask for help installing stable diffusion? ive tried a couple of times and i cant get it to work

[ornate elk]

Hey yes, where do you need help with?

[cunning robin]

i keep getting a couple of errors that cause a problem in particular, when i try to use another model besides the default it hangs on loading and telling something about failed to create model

i could try installing again so i can show you the full error

or i could be patient and hear step by step, it may be a problem about how i install it

[ornate elk]

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

[cunning robin]

amd radeon 6600

and well, right now inside my webui-user.bat theres nothing, i uninstalled all in case it was something i needed to do since the begining

[ornate elk]

Ahh okay so your on AMD

Have you followed one of my guides?

[cunning robin]

no sorry, i just followed some randome youtube tutorials

where i can watch your guide?

[ornate elk]

Oh okay, best for AMD is to use one of my guides from here:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides

You need the [AMD] Automatic with ZLUDA

[cunning robin]

since its 6600 should i use zluda?

ah!

what a coincidence x3

any zluda is good? for example a1111?

[ornate elk]

Yep, best is to get A1111 with zluda

[cunning robin]

nice, on it!

[ornate elk]

If that works you can easily install the others to test if needed

Let me know how it goes

[cunning robin]

sure!

rn upgrading from python 3.10.6 to 3.10.

*3.10.11

[ornate elk]

@cunning robin in the next two days I'm gonna update my guide for older cards to use rocm 6.1 too

[cunning robin]

woah :0

[ornate elk]

So if you want to use that than dont install hip SDK 5.7 and instead 6.1

And then I can link you the correct files for your card

Related to step 3. of zluda setup

[spiral comet]

roop unlashed is same as a reactor (same engine)?

[ornate elk]

Both are based on roop

[spiral comet]

face fusion is smth diff?

cuz i think roop is not that great to sdxl

i mean, low quality

[cunning robin]

ok so i follow everything the same until step 3?

[ornate elk]

Roop-unleashed is a standalone tool so it shouldnt matter what's the input image

Roop itself was discontinued

[ornate elk]

Yep do step 1 and 2 of the zluda setup

Then let me know

[cunning robin]

Oh i need step 1 and 2? i thougt those were needed to skip since im below 6800

[ornate elk]

Thats what I meant as I'm gonna update that part

In 2 days

[cunning robin]

ohhh

nice

[spiral comet]

like i understand well: roop unleashed is the same as a reactor in result, but face fusion is what? also the same?

[cunning robin]

okay installing step 1

[ornate elk]

Idk which techniques they use but they have a discord where the devs are pretty active

[cunning robin]

what does it mean step 1 by Dont check the Pro Driver.

because i think i downloaded that

AMD-Software-PRO-Edition-24.Q3-Win10-Win11-For-HIP

is that the right i should install?

[ornate elk]

Yep

Just leave everything default and your fine

Then do step 3 and 3.3 of the zluda setup (but dont install hip 5.7)
And use these files from here:
https://github.com/likelovewant/ROCmLibs-for-gfx1103-AMD780M-APU/releases/download/v0.6.1.2/rocm.gfx1032.for.hip.sdk.6.1.2.7z
Also put the additional rocblas.dll into the rocblas folder.

[ornate elk]

I'm AFK a bit.

[cunning robin]

okay i hope you come back soon, i think im a bit lost

step 1 tells me to install amd hip, done

step 2 tells me to download zluda and put it in c:/, done

now is step 3 the part im confused}

in 3.1 should i install hip 6.1 right?

or is the 6.1.2 i installed at step 1 is good?

also, step 3.2 tells me to install zluda, didnt i do that at step 2? or is it another zluda?

also in step 3.2 tells me to download optimized roc libs, is it neccesary?

[quasi ginkgo]

In Forge, my tagcomplete extension is acting fucky. It fails to load more than half the time
Sometimes I'll have to reload the webui or refresh the page several times to get it to finally work. No errors in the cmd console. Anyone else experience that issue?

[ornate elk]

The ones I linked you here yes,
#🤝｜tech-support message

[ornate elk]

Install hip SDK 6.1

Then use the zip I linked you above

And do the steps from the 3.3

[cunning robin]

where do i find hip 6.1? i only found 6.1.2 on the guide

i cant find C:\Program Files\AMD\ROCm\5.7\bin\rocblas

i think im lost again

this step is being so hard for me, why 3:

i think i made it, turns out i was in the wrong directory

wait i think i did it wrong

[ornate elk]

6.1.2 is 6.1

Sry for the confusion

[cunning robin]

yeah what i did wrong is i took the library archives into the zip file, instead is the opposite

[ornate elk]

Ohh

xD

[cunning robin]

what do i do with that rocblas archive? the one in the zip file

oh time to restart my pc

[ornate elk]

The Library folder or the .DLL file?

[cunning robin]

the ddl

ok lets continue

step 4!

Add the C:\ZLUDA folder and %HIP_PATH%bin to your PATH in the System Variables. Like shown here:

ok im there to edith path but what do i do?

do i add a new path to C:\ZLUDA and another to %HIP_PATH%bin or its the same?

[ornate elk]

The .DLL goes into the rocblas folder

Where the Library and the old library are

[ornate elk]

You add them as two new entries into the Path entry

[cunning robin]

ok

[ornate elk]

Here is how it should look like:
https://github.com/CS1o/Stable-Diffusion-Info/issues/7#issuecomment-2408541054

[cunning robin]

yeah!

step 5!

thats it?

[ornate elk]

Yep

If everything is correct it will work ^^

Feel free to show the cmd after the webui opened

[cunning robin]

at least rn, since its still installing more stuff

[ornate elk]

Looks good

[cunning robin]

what should i do here? is it working as intended or is going to be a problem?

[ornate elk]

Download a model and put it into the models/stable-diffusion folder

[cunning robin]

there is already one, but do i download more?

[ornate elk]

Relaunch the webui-user.bat
Then select it in the dropdown.
And then try to generate and image.
The first time it will take up to 40

[cunning robin]

what do you mean by select it in the dropdown?

i still have the OS error

OSError: None is not a local folder and is not a valid model identifier listed on 'https://huggingface.co/models'
If this is a private repository, make sure to pass a token having permission to this repo either by logging in with huggingface-cli login or by passing token=<your_token>

[ornate elk]

That can be ignored

[cunning robin]

ah alright

[ornate elk]

It will get fixed in the next big update

[cunning robin]

is it normal to be stuck at Compiling in progress. Please wait...

?

[ornate elk]

Now just let it run for 40 minutes.
If you don't get an image then.
Close and relaunch the webui-user.bat
Then try again and it should take seconds

[cunning robin]

like that? glazed donuts

[ornate elk]

Yep just let it compile. Its a one time only thing

[cunning robin]

oh! something moved

is that alright?

[ornate elk]

Yes thats good

[desert tree]

hi guys, is there any guide out there explaining in more detail how to write extensions for a1111? the official guide doesn't explain much

[cunning robin]

Yummers! thaks cs1!

[desert tree]

in particular I'm looking for explanations on the differences between p.prompt, p.prompts, p.all_prompts etc, there seem to be a ton of different ones and idk which are used when

[ornate elk]

No problem 🙂
Now you can download more models and stuff from Civitai.com

[cunning robin]

thanks!

[cunning robin]

is this normal trying to switch model?

its just keep loading

ah there!

[vague trail]

can someone help me figure out how to get a model for controlnet 🥺

im trying to get multiple characters generated and in all the yt videos they just have a model but never explain how to get it

and i have no idea

[ornate elk]

For sdxl and pony you can add --medvram-sdxl to the webui-user.bat
That can make it faster

[ornate elk]

For 1.5 you can get them here:
https://civitai.com/models/38784/controlnet-11-models

[vague trail]

wdym by 1.5

ok well doesnt matter ig

i downloaded the thing you just sent

can you tell me where to put it

🥺

wait @ornate elk do you know how to use multiple character loras in 1 image

because ive been looking through tutorials for HOURS

id genuinely let you ram me from behind if you showed me

like full on sloppy toppy

🙏

[vague trail]

@ornate elk 👉 👈

🥺

[ornate elk]

What model do you use at the top?

Is it 1.5, sdxl or pony?

[vague trail]

uhhh

hold on let me launch it to check

[ornate elk]

Because you need the compatible Controlnet models

Controlnet models go into models/controlnet

[vague trail]

dude you tlel me which model to get i will get it

i have these

[ornate elk]

Meinamix is 1.5 based

So you can use the Controlnet models I linked you

[vague trail]

ok let me go put them in

ok

uhh

can you give me a prompt

to just test it

like fucking

uhh

can i do

garden background
AND 1 bald man
AND 1 blonde woman

would that owrk

[ornate elk]

Idk what you try to do

[vague trail]

ok listen man

my end goal here is

im trying to use

multiple character loras

into 1 image

and people say to use latentcouple

or controlnet

and im trying to figure out how to dothat

if you could tell me step by step id literally let you violate me raw

[ornate elk]

You should use regional prompter

[vague trail]

i have these

[ornate elk]

That works good with multiple character loras

[vague trail]

can you teach me how to use it

🥺

[ornate elk]

Yea but need to get to the PC first

[vague trail]

ok!

thank you

my hero 😍

we have the same time zone btw

im up all night

just lmk when you're free

[ornate elk]

In 10 mins I can give you an example

[knotty turret]

@ornate elk hey cs1o you asked me if using AMD and i not

[vague trail]

hero..

[knotty turret]

so what should i do?

[vague trail]

no rush

[knotty turret]

.

[cunning robin]

I haven't done that but to clarify

Ponyv6 was working fine, but switching to sdxl froze my pc

Why is this?

[ornate elk]

thats because you have only 16gb of RAM. to fix the freezing issue you have to increae your Windows Pagefile

[cunning robin]

how

[ornate elk]

https://www.tomshardware.com/news/how-to-manage-virtual-memory-pagefile-windows-10,36929.html

Make sure its only enabled for C drive. and set it to 16000 Min and 24000 Max
then restart the PC

[vague trail]

@ornate elk you're so smart..

[ornate elk]

have you tried any other flux model?

[fossil nacelle]

Hi! Can anyone help me? I used my trial credits but I cannot upgrade my plan, there is no such option on the billing page. Only "Cancel subsription"

[knotty turret]

i will try that one to see

first try to different one

but i think that one its not quantized

[ornate elk]

how much vram does your gpu have?

[knotty turret]

8

[vague trail]

cs once you're done helping these guys lmk if you can help me with regional prompter ❤️

[ornate elk]

okay then you should use the gguf models if your in comfyui. or using forge webui as its better for 8gb vram

[knotty turret]

there github with that? gguf models

[knotty turret]

i using comfyui use that one?

[ornate elk]

you can find them here:

https://huggingface.co/city96/FLUX.1-dev-gguf/tree/main

try the gguf Q4_0

[knotty turret]

q8?

[knotty turret]

Alright

[ornate elk]

and try this workflow for it

[knotty turret]

trying to use inpaint

with flux

[ornate elk]

you also need the clip l and t5 tex encoders and vae

[knotty turret]

do u have workflow with inpaintg?

[ornate elk]

nope

[knotty turret]

already have it

[cunning robin]

this was the result

[ornate elk]

have you restarted the PC after changing the pagefile?

[cunning robin]

yes

[ornate elk]

hmm

can you show me your RAM in taskmanager?

[knotty turret]

not showing here for some reason

tried checkpoint and unet folder

[vague trail]

im so hard rn

[ornate elk]

try updating your comfyui

[knotty turret]

first and third file?

[ornate elk]

yes

then also update via the manager ui

[knotty turret]

trying to select but nothing

[ornate elk]

Here is my example:

[vague trail]

ok wait hold on

let me go through this step by step

so first you got the model

can i have the name

ill get the same one

[ornate elk]

and when you click on the right arrow?

[knotty turret]

start with null

then click on right arrow

and stay undefined

maybe the name?

.gguf

[ornate elk]

looks normal

[vague trail]

is this hte one you got

https://civitai.com/models/684007/ebara-ponyxl-pony-sdxl

or can i just use one i already have

https://cdn.discordapp.com/attachments/1002602742667280404/1296937610022752417/image.png?ex=67141ac6&is=6712c946&hm=29daacb4135f24139efbd9aff8e30ea684591118e843521ae7d781ebbd374921&

[ornate elk]

nope

but i used a wrong one xD it was for realism and not for anime

[vague trail]

https://media.discordapp.net/attachments/1002602742667280404/1296946154646540428/image.png?ex=671422bb&is=6712d13b&hm=9a0f10de15ce3412f56c32f3790e23f5479b622759ba0f114c61080846dd08d2&=&format=webp&quality=lossless&width=960&height=271

[ornate elk]

https://civitai.com/models/502370/rev-engine-ponyxl

[vague trail]

how do i get

sd vae

to show up there

is it an extension

[ornate elk]

nope, go into Settings, User Interface, Quicksettings, there add: sd_vae
then hit apply and reload ui

[noble belfry]

not helpful, but revengine sounds like a metal gear solid game name

[vague trail]

ok ok @ornate elk so i have a model and i got a vae

so when i go to regional prompter

what do id o

[ornate elk]

set the resolution to the same as the top resolution

and then enable "use common prompt" and disable the "use base prompt"

[vague trail]

done

what now

ty btw, i just get it easier if i do it step by step

alr i got it to look somewhat like yours

the colors different tho

does that matter

[ornate elk]

you have to set it Active

[ornate elk]

nope

[vague trail]

ok i pressed active

now

for the prompts

do i just put in the loras

do i have to write "BREAK" like you did

[ornate elk]

yep

and the first line is the overall description

like a scenery

[vague trail]

can you

give me the prompt youused

i wanna try generating that one

also, can i use hiresfix?

i usually do this

[ornate elk]

yes if your card is good enough

[ornate elk]

oh damn xD

[vague trail]

i make these in 2 minutes

[ornate elk]

dont use 150 hires steps

[vague trail]

why

[ornate elk]

not needed at all

[vague trail]

alr

[ornate elk]

also 150 normal steps is to much

[vague trail]

dude

i want the best quality possible

[ornate elk]

more steps doesnt mean better quality xD

[vague trail]

oh

[ornate elk]

steps only change the image

[vague trail]

alr well ill experiment with that later

right now

can you give me

the prompt you used

or just

make a new one thats simple

i have these 2 character loras

[ornate elk]

`2girls, score_9, score_8_up, score_7_up, source_anime, masterpiece, best quality, in korean city,.BREAK score_9, score_8_up, score_7_up, source_anime, masterpiece, best quality, 1girl, dark magician girl, lora:dark-magician-girl-anime-ponyxl-lora-nochekaiser:0.6,.BREAK score_9, score_8_up, score_7_up, source_anime, masterpiece, best quality, 1girl, lucy, cyberpunk, lora:cyberpunk-lucy-s1-ponyxl-lora-nochekaiser:0.6,

Negative prompt: 3d, realistic, score_6_up, score_5, score_4_up, blurry, deformed, worst quality, ugly, source_pony,`

[vague trail]

alr ill do that and

ill replace your loras

with mine

ill lyk how it turns out in a sec

@ornate elk

in your prompt you used "source_pony"

do i remove that

since im using

[ornate elk]

snowpony could be a pony based model

so its okay to keep that in

[vague trail]

ok

also

to confirm

i should not use 150 steps right

ill try 100 for now

oh my god its working

[ornate elk]

even 100 is to much, your fine with 30-35 xD

upscaling does the rest

[vague trail]

dude

id legit

let you hit raw

rn

my hero..

also

also

can regional prompter

use these

its alright if it cant

just asking

[ornate elk]

thats for controlnet

[vague trail]

yea

assumed

[ornate elk]

you can use both together

but i didnt tested it

[vague trail]

why is she disintegrating 😭

[ornate elk]

your denois was 0.7

thats to high

0.5 should be max

[vague trail]

oh

i dont even touch that

its 0.7 by default

[ornate elk]

yea its wrong by default

[vague trail]

what do i put it at

[ornate elk]

0.5

[vague trail]

alr

do i lower the steps on anything

[ornate elk]

yea use max 20 hires steps

[vague trail]

alr let me interrupt this one

[ornate elk]

and 35 normal steps

[vague trail]

itssshe broke apart

@ornate elk

it finished the image

but at the end

it just showed this

[ornate elk]

yea because you used a to high upscale resolution and to much upscale steps

[vague trail]

these are my

things

[ornate elk]

set the upscale by to 2

and set the upscaler to RESRGAN4x+

and then you may need to lower the resolution or install the tiled vae extensions

[vague trail]

its not there

[ornate elk]

and you also need the fp16 vae

[ornate elk]

click into the upscaler

there it should be listed

[vague trail]

found it

ty

alr

i lowered upscale

to 2

ill try again

[ornate elk]

do you use xformers ?

[vague trail]

no idea what that is

also

this happened again

[ornate elk]

okay out of vram error

whats in your webui-user.bat?

[vague trail]

im on windows 11 and i think thats why my file isnt called .bat

its just webui-user

[ornate elk]

you can enable that at the top under View > Show > File Name Extensions

[vague trail]

done

what do i do now

[ornate elk]

right click, more options, edit

[vague trail]

i dont see edit

oh you mean on

the bat file

shit mb

[ornate elk]

you need to add --xformers

[vague trail]

after --theme dark

?

like this?

alr ill save

do i reload the ui

@ornate elk

[ornate elk]

yep

[vague trail]

ok

[ornate elk]

and you also need the fp16 vae

[vague trail]

can you send me that

[ornate elk]

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

[vague trail]

let me get that

[ornate elk]

the sdxl.vae.safetensor

[vague trail]

theres 2

@ornate elk wake up pookie

[main nest]

quick question, what purpose does controlnet has

[ornate elk]

i dont answer on questions when the answer is one message above

[vague trail]

oh

shit

,y bad

i didnt see t

sirry

[ornate elk]

np

[vague trail]

do i put it in

[ornate elk]

with controlnet you can control different aspects of the image generation. for example the depth of images or how persons should be placed in the image

[vague trail]

the vae folder

[main nest]

ohh yea i remembered

[vague trail]

ok now let me reload ui and reassemble everything

and gen the image

[main nest]

well I got into an aciddent and suffered a memory loss so now imma ask everything from the start

so now the big question is how do i install Face Fusion to SD

[ornate elk]

the extension of facefusion seems outdated but its a standalone tool.
But would recommend Roop-unleashed over it.
You can try install the facefusion extension but idk if it will work

[main nest]

well I can do it with Roop

[desert tree]

I'm going absolutely insane trying to write an a1111 extension, any ideas where I could get help? even on the official github repo I didn't get any answers when I created a discussion

[main nest]

and what was the best preprocessor with Controlnet for realistic pictures

[ornate elk]

maybe you can ask on the discord of Automatic1111

but i dont have a link to share

[ornate elk]

controlnet doesnt enhance the quality

[desert tree]

they have one? where

[ornate elk]

its for stuff like poses, depth, or lineart, converting stuff into images etc

[vague trail]

bro @ornate elk i tried genning it and my pc crashed

its updating by itself now

[main nest]

i meaaan for poses

[vague trail]

it got to 100% then it just went to the blue screen of death

[main nest]

found it

[ornate elk]

blue screen 😮

[vague trail]

what do i do pookie 😭

[ornate elk]

you need the tiled vae extension then

i used it too

[vague trail]

send

from the screen to the ring to the pen to the king!

[ornate elk]

click on extension tab, then on "available", then on "load from" then you get a list

there search for tiled

[vague trail]

alr

let me juat wait for my pc to updare

[ornate elk]

it seems your already in autos discord

[desert tree]

just found it and joined

literally this second

[vague trail]

ok @ornate elk pc back up let me get the extension and ill try genning it again

i got this

[ornate elk]

enable only the tiled vae function and keep its settings on default

[vague trail]

how do i

do that

[ornate elk]

have you restarted the webui ?

[vague trail]

yes

[ornate elk]

then you see at the bottom of txt2img the setting

tiled vae

[vague trail]

i do not see tiled vae

hold on

let me

restart the ui again

i mightve not

ok NOW its ther

e

my bad

ok @ornate elk what do ido

pookiiieee pleeeaseee 😭

[ornate elk]

only enable tiled vae

thats all

[vague trail]

do i touch anythng here

if not ill start genning the image

[ornate elk]

let them default

[vague trail]

alr

IT WORKS

https://tenor.com/view/whitebeard-marineford-goat-gif-21001348

[ornate elk]

nice !

[vague trail]

ok

uhh

@ornate elk

question

can i use

like

style loras

artstyle loras

[ornate elk]

yes

[vague trail]

how do i

do that

do i just add

BREAK (lora)

[ornate elk]

nope

add the style lore to the top line

then play with the strenght of it

[vague trail]

u mean

denoising strength?

[ornate elk]

nope, the lora strenght

its the :1

at the end of lora:loraname:1

reduce the lora strenght

for example

or 0.7

Its not easy to get it right

You have to test a lot

try full body or describe their shoes, skirt etc

or wide angle

[main nest]

btw how was I supposted to face swap using the batch option

[stone fog]

The CMD log when I launch webUI user?

[cunning robin]

what can i do?

[cunning robin]

pony works fine but i cant get to use sdxl

[ornate elk]

The char searcher extension and the artist study extension gave errors.
I would delete them and after that delete the venv folder.
Then relaunch the webui-user.bat

[ornate elk]

Have you added --medvram-sdxl to the webui-user.bat?

[manic quarry]

Hey guys, I just installed Stability Matrix and the SD WebUI reForge. I tried different models but all give the same error: an image is generated and it looks good until 95% or so, but then in the very end, it looks like this and I have no idea how to fix this. Can anyone help?

[hallow latch]

It's not a tech support in say that is a problem but.. Is there a way to enable to install extenstion and update stuff even if you have --listen or --public flag when you start it so I can avoid "AssertionError: extension access disabled because of command line flags" ??

[manic quarry]

I just installed another UI (Forge instead of reForge) and the problem is not there. Maybe this helps anyone ¯_(ツ)_/¯

[ornate elk]

--update-all-extensions

And git clone extensions manually still works, but other than that you can't install while --listen or share is active

[hallow latch]

Was hoping to be able to run the update from the extenstion tab and be able to download stuff with --listen active :/

[ornate elk]

Thats not possible

[manic quarry]

Where can I find a manual for all the settings (in Stability Matrix, using SD UI Forge)?

[blazing warren]

hi, I wanted to create some pixel art for my game for fun and I'm using this checkpoint - image1 and a Lora - image2 but when I try to generate image it gives me weird results - image3, can anyone help me with this?

[vocal burrow]

sdxl + 512x512 resolution == this

sdxl is trained for 1024x1024 (with some margins)

[blazing warren]

why is is so slow? I have rtx 3070 with 8gb ram and it takes like a minute now somehow. When I used comfy ui in past, it only took around 10 seconds for 1024x1024 image, why is that?

great is got bsod

[vocal burrow]

what command line args are you using ?

[blazing warren]

where is it?

[vocal burrow]

what's in your webui-user.bat ?

[blazing warren]

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=

call webui.bat

looks empty

[vocal burrow]

change that set COMMANDLINE_ARGS= line to
set COMMANDLINE_ARGS=--xformers --no-half-vae --medvram-sdxl

restart and try again.

You can also enable fp8 for sdxl model in the webui's settings

[blazing warren]

oh now it's fast

thank you!

[blazing warren]

what does that do?

[vocal burrow]

--xformers Enable xformers for cross attention layers. aka use some math library to optimize computation at the cost of some very very very minor change in the output and making SD not deterministic from device to device (different hardware == different output)
--no-half-vae Do not switch the VAE model to 16-bit floats. , don't use small precisions values for the VAE part of SD pipeline, will make SD consume slightly more VRAM but will minimize the risk of getting "nan (not a number) errors" due to lack of precision.
--medvram-sdxl enable --medvram optimization just for SDXL models, will chop the SDXL model into multiple part and then proceed to load (and unload) and work with those parts as needed. Needed if you don't have enough vram for sdxl (8gb is borderline not enough), otherwise SD will fill your gpu's vram (fast) and spill out in your ram (slow)

[chrome prism]

I'm new in image generation via ai and wanted to make a cartoon style of some game assets that i bought so the style is consistent. I played already a bit around but its really overwhelming 😄 Are there some good resources for the use case i have?

[warped canyon]

anyone knows the solution to this issue? https://github.com/lllyasviel/stable-diffusion-webui-forge/issues/2009

[manic quarry]

Where can I find a manual for all the settings (in Stability Matrix, using SD UI Forge)?

[hardy garnet]

Hello, I've been getting some images that are hosed...here's an example. I've lost about 10 images now (I get this instead of my picture), so I gave up. It will intermittently do a good picture, then do one of these

[carmine ibex]

I installed a pack of custom nodes for comfyui. Only a few popular ones. When I restarted it, I noticed in the startup terminal, it said something like uninstalling torch or cuda or something. Now, neither one of my comfyui's open at all.

[loud pond]

whats the difference between a negative and a embedding and where do i put them?

[cosmic furnace]

@ornate elk Ahoi). R u here ?

Installed forge... Considering auto on zluda running fine. Skiped that part. And... Forge not making even 1 simplest image... Just error. -_- is there some kind of error-log file?

Optional: Move one 1.5 based model from your other existing webui into the models/stable-diffusion folder. Or skip this if you dont have one. - skipped this and it didnt download any base model...weird but w/e i dont rly need any i guess

and copying the one 7 gb didnt do any difference any way

https://civitai.com/models/618692/flux - tried using def pruned fp8

and tried this (even tho not sure what error had) but anyway no result

5.1 If you get a missing GGUF module error, do this:
Go into Forges folder. Then click into the File Explorer bar (not searchbar).
type cmd and hit enter
then run the following two commands one by one:
venv\Scripts\activate.bat
pip install gguf

[cunning robin]

will that help me stop running out of memory?

[ornate elk]

For vram not memory

For the memory error you could try increasing the pagefile to 32000 instead of 24000

[ornate elk]

What error do you get in forge when using a normal model?

1.5 or sdxl

[cunning robin]

ahh ok ok, my first problem is that when i try to change to sdxl it keeps loading until it throws me an error and returns to pony

[ornate elk]

Yea then add the --medvram-sdxl and also increase the pagefile

Then restart the PC

[cunning robin]

ok!

[cosmic furnace]

i didnt tried >_<

only that only pruned flux popular one >_>

btw why pro version isnt available for download but i see images generated with it? Is it locked behind buying or whuht? o.0

[ornate elk]

Pro is to big for consumer

And they want to make money

[ornate elk]

Try that and then I can link you models that should work

[cunning robin]

i got that message since i installed reactor

still opens tho

[heavy blade]

Hey, Is this only support for Stable diffusion or Flux as well?

[loud pond]

@ornate elk ,
whats the difference between a negative and a embedding and where do i put them? (also i found the error to the generation software, i needed to have a negative prompt lora and the images faces came out a lot better)

[cunning robin]

this was the result

[ornate elk]

That can be ignored, but in reactor settings you have to set it to CPU mode

[ornate elk]

Both

[ornate elk]

Negative embeddings go into the negative prompt box

[loud pond]

do i just put them all in the lora file?

[ornate elk]

Embeddings go into the embeddings folder

[cosmic furnace]

No error using pony model o.0 everything works...

Do i not know something about using forge+flux ? -_-

[loud pond]

found it thank you!

[cosmic furnace]

or flux will work now? hm

nope flux ends error on try -_-

[heavy blade]

Could someone share their config.json for character lora, I'm confused with settings learning rates and unet learning rates. I'm trying to generate Flux Loras.
I've used FluxGym , it worked great (but only allows 75 images with captions). But I cannot replicate these settings with Kohya UI (I want to use 200+ images).

[hallow latch]

I just need --listen and --share to access it from outside my network and inside my network right?

set COMMANDLINE_ARGS=--no-half-vae --medvram-sdxl --theme=dark --xformers --listen --share

[stone vortex]

We are a commercial company, and we currently have a paid project requirement. We need assistance with editing some images by replacing the heads of different people into specified positions. Could you please let us know if there are professionals available to help us with this task?

[ornate elk]

use Flux gguf Q4_0 or Q8_0

https://huggingface.co/city96/FLUX.1-dev-gguf/tree/main

and the t5 encoder gguf Q8_0
https://huggingface.co/city96/t5-v1_1-xxl-encoder-gguf/tree/main

[cosmic furnace]

recently friend asked how will flux do on a 12gb card ?

this flux Q8 as good as the one dev pruned fp8 16gb?)

hm. Q8 + encoder weight above same gbs or even more) hope so -_-

[ornate elk]

will work good when a model with less than 16gb gets used

gguf works great

[bitter idol]

Can someone tell me 1-2 image to video converter ai ?

[cunning robin]

wouldnt that make it run with my cpu instead of my graphic?

[ornate elk]

only for the faceswap

which is still very fast

[cunning robin]

ah good

where is the reactor settings?

[ornate elk]

in txt2img at the reactor tab

then on the left

[cunning robin]

thanks!

[warped canyon]

bumping

[cosmic furnace]

where do i put encoder?)

[ornate elk]

models/text_encoder

[ornate elk]

tested it rn, got the same issue the first time, now on the second run and not on the forge tab again i get the sound

now i opened up a new tab switched to it, minimized the browser, opened up a different programm and still got the notification sound

[gleaming prism]

i can delete the miniconda folder for space?
or it's used by stable diffusion?

[high drum]

Anybody have any idea why FLUX SCHNELL creates very pixelated on my machine? I’ve tried all samplers and every other model seems to work fine. I’m on M1 Apple generating at 1K resolution

[onyx solar]

You'd need to show the output and your settings. There are too many variables and not enough information supplied to be able to help.

[high drum]

I removed the model a little while ago hence I haven’t got any more images but it looks like a zoomed in version of a 128 x 128 image. I tried DPM++ 2M Kara’s and Euler A as well as many other samplers at 1024 x 1024 on FLUX schnell with everything between 1 to 4 Steps. It seemed to get better quality as I increased the resolution further. But my machine won’t realistically generate 2048 x 2048 sadly

Karras^

[onyx solar]

Well, you're listing off some of your issues right there. For Flux, you should probably be starting with just Euler for the Sampler and Simple for the Scheduler. There are issues with ancestral samplers and with karras schedulers. It'll give you output like these:

However, the same combination with Euler & Simple:

(I'm not using schnell, but the same rules apply.)

Flux doesn't like certain things.

[warped canyon]

What did you do differently the second time?

[onyx solar]

Grabbed schnell to run a couple quick examples for you. Here is with Euler/Simple:

Euler/Karras seems to be fine:

Euler a/Simple, however:

And Euler a/Karras just to see....back to working:

So maybe try going back to it and give some of the working stuff a shot.

[high drum]

I’ll try it again in the morning. I saw some people had problems with blur but my generations literally looked pixelated like a low resolution image zoomed in. Thanks for your help

[onyx solar]

Make sure you've got the right VAE as well.

[main rune]

Anyone knows what the deal is with Forge and the Regional Prompter extension?

[cosmic pulsar]

Hi, can anyone tell me if this error occurs when trying to use openpose, with canny everything is fine.
urllib.error.URLError: <urlopen error [SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1007)>

[civic nova]

OK I have an older free use version of Stable diffusion. Went in today to make some character pics for my D&D Gaming toons and it does not work? says GIT is an unrecognized command or some such. I am not a programmer, last time I did anything was with DOS 6.0. Yes, I am over 60 and still play games, LOL is there any "easy Idiot proof way" to update this program that does not require manually writing in code lines? Running Windows 10 and Google Chrome. Any assistance is most welcomed. It has been a long time since I used it, did I need to put in a specific URL into chrome maybe?? I did take a marker and wrote ANY KEY on my return key but that did not seem to work??? HDG US Army Ret. VFW

@echo off

if not defined PYTHON (set PYTHON=python)
if not defined VENV_DIR (set "VENV_DIR=%~dp0%venv")

set SD_WEBUI_RESTART=tmp/restart
set ERROR_REPORTING=FALSE

mkdir tmp 2>NUL

%PYTHON% -c "" >tmp/stdout.txt 2>tmp/stderr.txt
if %ERRORLEVEL% == 0 goto :check_pip
echo Couldn't launch python
goto :show_stdout_stderr

:check_pip
%PYTHON% -mpip --help >tmp/stdout.txt 2>tmp/stderr.txt
if %ERRORLEVEL% == 0 goto :start_venv
if "%PIP_INSTALLER_LOCATION%" == "" goto :show_stdout_stderr
%PYTHON% "%PIP_INSTALLER_LOCATION%" >tmp/stdout.txt 2>tmp/stderr.txt
if %ERRORLEVEL% == 0 goto :start_venv
echo Couldn't install pip
goto :show_stdout_stderr

:start_venv
if ["%VENV_DIR%"] == ["-"] goto :skip_venv
if ["%SKIP_VENV%"] == ["1"] goto :skip_venv

dir "%VENV_DIR%\Scripts\Python.exe" >tmp/stdout.txt 2>tmp/stderr.txt
if %ERRORLEVEL% == 0 goto :activate_venv

for /f "delims=" %%i in ('CALL %PYTHON% -c "import sys; print(sys.executable)"') do set PYTHON_FULLNAME="%%i"
echo Creating venv in directory %VENV_DIR% using python %PYTHON_FULLNAME%
%PYTHON_FULLNAME% -m venv "%VENV_DIR%" >tmp/stdout.txt 2>tmp/stderr.txt
if %ERRORLEVEL% == 0 goto :activate_venv
echo Unable to create venv in directory "%VENV_DIR%"
goto :show_stdout_stderr

:activate_venv
set PYTHON="%VENV_DIR%\Scripts\Python.exe"
echo venv %PYTHON%

:skip_venv
if [%ACCELERATE%] == ["True"] goto :accelerate
goto :launch

:accelerate
echo Checking for accelerate
set ACCELERATE="%VENV_DIR%\Scripts\accelerate.exe"
if EXIST %ACCELERATE% goto :accelerate_launch

:launch
%PYTHON% launch.py %*
if EXIST tmp/restart goto :skip_venv
pause
exit /b

:accelerate_launch
echo Accelerating
%ACCELERATE% launch --num_cpu_threads_per_process=6 launch.py
if EXIST tmp/restart goto :skip_venv
pause
exit /b

:show_stdout_stderr

echo.
echo exit code: %errorlevel%

for /f %%i in ("tmp\stdout.txt") do set size=%%~zi
if %size% equ 0 goto :show_stderr
echo.
echo stdout:
type tmp\stdout.txt

:show_stderr
for /f %%i in ("tmp\stderr.txt") do set size=%%~zi
if %size% equ 0 goto :show_stderr
echo.
echo stderr:
type tmp\stderr.txt

:endofscript

echo.
echo Launch unsuccessful. Exiting.
pause

Sorry that is about the extent of my programming knowledge these days. I hope at least some of that was helpful.

[main rune]

@civic nova maybe you somehow uninstalled GIT

[ornate elk]

You have to install Git
But also your using a very old webui version.
Maybe a clean install is better at this point.
Easy install guides can be found in the pinned messages of this channel.

[civic nova]

Humm do not rememeber ever uninstalling GIT will try a reinstall of that first just to check to be sure. Do I install that into the same folder or a stand alone folder on the same drive?

OK I reinstalled GIT it was listed as a seperate folder on the same drive. But that seems to have fixed it. Thanks. NOW where are these "Pinned" install instructions you spoke off please.

[ornate elk]

You can find the guides here:
The latest Stable Diffusion install Guides for Nvidia and AMD:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides

[civic nova]

thanks again..

[mystic zodiac]

what if Lora discription on civitai does not clarify use model? does not seem to work on any

should i watch videos on webui lora setups? or is there a certant way it should be done

[main rune]

weird, do you have a link?

[cosmic furnace]

flux load model every launch ? o.0 Auto loads it only when u switch ... If u used it previously it doesnt load -_-

Anyway i tried generating and speed is very slow , as if it uses cpu . 1024-1024 image 11 mins 😄

oh and final quality is blah

btw had to download & install clip module "clip_l.safetensors" cuz it was giving clip error without it
also installed vae "ae.safetnzor" just incase

[gritty delta]

Hey, I'm having issues downloading and running stable diffusion.

If anyone could DM and help me get it sorted that would be much appreciated.

[ornate elk]

For me it takes 1 minute

You have to set the GPU weight to 19000 in forge

So it dont use the slower pagefile

[ornate elk]

Hey did you used the guide from the pinned messages?

[gritty delta]

yeah

but it was when adding the roop extension when i started to get issues

[ornate elk]

Yea roop extension is deprecated and breaks the webui

To fix it delete roop from the extensions folder and then delete the venv folder and relaunch

[gritty delta]

ahh

i went the long of of juust deleting the whole thing

and re downloading

[ornate elk]

For faceswap use Reactor extension

[gritty delta]

in the process of trying that now 😅

[cosmic furnace]

but 1 minute is slower than in auto same image... and ive read forge shud be faster... its faster with pony-sdxl but not flux? or whuht? o.0

[gritty delta]

i got it

just had to be more patient xd

[ornate elk]

Flux is always slow

1 minute is fast for flux

[cosmic furnace]

damn 😦

and it always loads model every launch? even if same used as previous launch? which means kinda ssd dependant?

[ornate elk]

Normaly not

Make sure you lower the GPU weight

To 19000 or 20000 at the slider