🤝｜tech-support

[karmic crown]

it should be like this when works correctly

something is wrong with display of geneated images

[wind sun]

And when there are more pictures separated they are all just the same

[karmic crown]

but they should be in output folder

[noble belfry]

are you looking in grids folder?

[karmic crown]

with -1, they are diffrent

[inland seal]

they are all the same image, there should be 2 different images

[noble belfry]

actually, im going to take five, im getting frustrated now. We're all trying to help 🙂

[wind sun]

They are together in one photo

[storm vector]

@karmic crown look at the 2 images In this

[wind sun]

I know I appreciate it so much

[karmic crown]

forget the ui

find output folder

[wind sun]

We are not fighting you guys just we are trying to explain

[storm vector]

Its the same as well

[karmic crown]

ui is obviously broken

[wind sun]

Ye

But if we reinstall everything we will lose so much progress and stuff

[vocal burrow]

Whenever you generate for multiple pictures it will create a grid aggregating all the output and display this grid first. The next entries in the gallery should be your individual output. Maybe it's just a gradio bug with the thumnail. Try to click on them and see what full res image is displayed. Also try restarting your sdwebui instance

(and yeah in last effort try checking your output folders directly as they're telling you to)

[wind sun]

They are also not saving in the output folder xD

[vocal burrow]

show us your output folder

[wind sun]

That’s it

[river flame]

didnt help

[wind sun]

It’s on the @inland seal PC, we are just all talking in discord trying to fix his issues

[storm vector]

@halcyon zodiac have you ever had this problem??.

[vocal burrow]

check what values you have for the output directories (default should look like this)

[vocal burrow]

try a full reinstall then :/

[wind sun]

Everything is the same

[river flame]

the updater now says

[vocal burrow]

you didn't delete the files from the previous installation ?

[river flame]

I just copied and replaced

except Models and Extensions

[vocal burrow]

well don't, do a proper clean install. if you want you can keep your usual models / embeddings / etc for sure in a seperate folder

but run it from a clean empty folder

[karmic crown]

rename the old forder, do not delete
reinstall
move models from old to new install

[vocal burrow]

ok.... restart sdwebui, do one generation with batch count 2 then copy paste the full console log in there please

[karmic crown]

than delete the old folder

[river flame]

thanks

[loud jasper]

Hi, im installing SDXL so where should I put the refiner model?
In the same folder as models or a different one?

Im using A1111

[vocal burrow]

same folder

[river flame]

I renamed the Forge folder, installed in a new one, updater ran well, ran, still same error

I ll try deleting everything

[vocal burrow]

if you installed it in a different folder it should be the same

no need to delete anything

[vocal burrow]

what's your hardware ?

[river flame]

12900k, 64gb ram, 2080ti (just changed)

[vocal burrow]

oh

yeah

[river flame]

to be fair, reddit says not only to delete VENV

[vocal burrow]

that has nothing to do with your error

however because you just changed your gpu I would :
1/ completely remove any trace of your video drive using Display Driver Uninstaller (DDU)
2/ run pip cache purge
3/ try to reinstall again in a clean folder

there might be some leftover from your previous driver confusing Torch

maybe ....

[river flame]

wait should I reinstall the main Nvidia driver

[vocal burrow]

yes, cf https://www.tomshardware.com/how-to/uninstall-nvidia-amd-intel-gpu-drivers

[river flame]

thank you

[cursive spire]

have a strange question, for my upscale i want the image to be 1600x900 but for some reason it always comes out to 1600x896 i was wondering if there was a reason for that

[vocal burrow]

900 isn't dividable by 8

so it's rounding to 896

[cursive spire]

ohh i see didnt know it had to be divide by 8 makes sense ty

[noble belfry]

when possible use the little arrows to change the values. i find an aspect ratio extension is super helpful too

[river flame]

fixed everything, thank you so much!!

[vocal burrow]

that was an usual one :p

[solemn atlas]

Hi I have a question, i downloaded latent couple but cant find it

[manic snow]

Is it in "scripts"?

[vocal burrow]

it's the last one

you need to enable it

[solemn atlas]

I thought its LatentCoupleHelper

thank you so much

[solemn atlas]

thank you!!

And another question, I downloaded ComfyUi

should I put it in the venv folder?

[vocal burrow]

put what in which venv ?

[solemn atlas]

The comfyui folder

[vocal burrow]

in auto1111 venv ?

why ?

comfyUI is its own thing.

don't mix and match it with Auto1111

[solemn atlas]

Ahh sorry it was said to not mix them.
But what should I do with it then. I looked at it but there is not .bat

sorry I'm a little bit lost with these things

[vocal burrow]

it's ok

[karmic crown]

python main.py
to start comfy

[vocal burrow]

(or simply double click main.py, if you followed the setup instructions it should be enough)

[solemn atlas]

Maybe I'm blind but I cant find it

[vocal burrow]

windows is hiding file extensions by default, I'd suggest you bring them back https://www.howtogeek.com/205086/beginner-how-to-make-windows-show-file-extensions/

(it should be this one)

[solemn atlas]

I tried to open it before but it doesn't open

[vocal burrow]

what did it say ?

[paper crag]

general question about inpaiting.
I generate a image that looks relatively good, but has some parts i want to improve. I send the image to inpaint and select the area, now, do you guys keep the prompt unchanged, or change it so all the prompt commands are ones that focus on the area you want to improve.
Example, i have a image with bad looking face. I select the face area. Do i remove from prompt all stuff that doesnt have to do with the face area and replaced them with more detailed prompts for face? Maybe change the lora and add ones that would help out the face?

[solemn atlas]

If i click on it, it seems a new window is opening but then it closes before I cant see what it is

[vocal burrow]

I'd change the prompt so it only describe the masked area (and overall tone of the picture).

[noble belfry]

.py files are just text files btw, you can open them in notepad

[dense sandal]

Actually I start comfyUI with the runGPU.bat file or something

[paper crag]

So if i have for instance off putting eyes, i remove all prompts for body, clothes, etc and add ones that would effect that eyes.

[vocal burrow]

let's say you have a girl walking a dog in dark futuristic city => inpainting the girl face => girl face, dark futuristic city background

[paper crag]

got it

[river flame]

I m testing generation. Both Forge and A1111 give me 10-40 mins ETA for a genaration that took me 2 mins on 1080 ti

[solemn atlas]

is there a folder for that

[vocal burrow]

Get in the ComfyUI folder, click in the explorer bar (where's the folder path is) type cmd then press enter. It should open a console in that folder.
Then run python main.py in it.
What does it say ?

[loud jasper]

can someone tell me how to check the SD version?
I want to check if its SDXL or SD1.5

[vocal burrow]

there should be one next to those (a run_nvidia_gpu.bat)

[vocal burrow]

like so

[noble belfry]

what program are you using? normally the file name should be a giveaway

[loud jasper]

A1111

[noble belfry]

top left, it should show the model name under "stable diffusion checkpoint"

if its an XL model it will likely have XL at the end of it. if not, you might be able to see from the cmd window

[loud jasper]

so it means I can use both XL models and 1.5 models at a same time?

[solemn atlas]

i only have these 3

[vocal burrow]

check if your antivirus deleted them.
There are still in the archive .zip I just double checked

[hollow pike]

you can switch between them at any time yes

[loud jasper]

ohhk thanks

[vocal burrow]

sdwebui version != sd model version

[solemn atlas]

sorry, how do i check this?

[vocal burrow]

I don't know it depends of your antivirus

where did you even get your archive ?

[loud jasper]

?
how to check that?

[solemn atlas]

should i delate and install it again

[noble belfry]

yes. theres differences but yeah you can. its like how a record player can play LPs and the little 45s.

i just realised thats such an archaic metaphor its useless lol

[vocal burrow]

check what ? the version of sdwebui you're using ? It should be displayed at the very top of the console log or in the footer of the webpage

[noble belfry]

people are now more confused

[loud jasper]

Im more confused rn

[vocal burrow]

stable-diffusion-webui (also known as auto1111) is a program... so like all programs it comes with different versions. Its version number gets bumped with each update.
With stable-diffusion-webui you load and use stable-diffusion models. Models that come with different architecture (sd1.5, sd2.1, sdxl, sd-cascade, etc)

As long as you get a recent version of auto1111 you can use sd1.5, sd2.1, sdxl, etc models with it.

[vocal burrow]

v1.8.0

you're up-to-date

[loud jasper]

Ohkay
I thought SDXL had a different client or something

[vocal burrow]

where did you get the comfyUI zip from ?

[noble belfry]

nope 😄

[river flame]

the ETA was long cause I ran out of RAM restarting SD

[loud jasper]

Thanks for helping

[vocal burrow]

nope, stable-diffusion-webui had to be updated back then to support the new sdxl architecture at the time. But that's about it.

[noble belfry]

all the clients updated so they can run the new models 🙂

[solemn atlas]

https://github.com/comfyanonymous/ComfyUI/ this one

[vocal burrow]

from here ?

[solemn atlas]

yes

[vocal burrow]

then yup that's the good one. Either your antivirus deleted the .bat or something went wrong with your download

[solemn atlas]

I delated the other version and download it again

till it gets downloaded i have another question. I have now Latent Couple but where do I find LatentCoupleHelper or is this it?

[vocal burrow]

latent-couple-helper is yet another (old) extension

[solemn atlas]

i have it but it doesnt show

[hasty orbit]

Is there a log file anywhere in stable diffusion that keeps track of previous prompts?

[solemn atlas]

Latent Regional Helper , is this better?

[vocal burrow]

From what I can see (https://github.com/Zuntan03/LatentCoupleHelper?tab=readme-ov-file) it's not even an extension... It's just an external tool that you run along your auto1111

[solemn atlas]

Oh okay, so i need to download it from url code

[vocal burrow]

never used it either

[vocal burrow]

by default the prompts and some settings should be saved in your images's metadata.

[hasty orbit]

Thank You!

[vocal burrow]

You can see them either by using the PNG Info tab or simply by draging an image into txt2img's prompt text area (and if you click the blue arrow under generate it will automatically apply every setting to match the ones used for that image)

Time for me to log off and try to get at least one good night of sleep before work.

[noble belfry]

also, if you save the images into the log folder it will create a csv of all previous prompts

but almost nobody uses that as it by default saves files anywaying into output

[small zenith]

Hey, i hope this is the right place to write. I just downloaded SD (one click installation for win10), but i think it doesnt use my gpu (3070) fully enough. Where could i increase the vram allocated to SD? Do i also need to allocate more RAM or is that irrelevant?

[noble belfry]

by default it absolutely should

you'd actually have to go out of your way to tell it not to use all your gpu in fact

[hollow pike]

Make larger images and you'll notice it

[noble belfry]

what makes you think it isnt? there are lots of little speed up tips and tricks, but out of the box it should try its hardest

[small zenith]

btw, how could i make the image more clearer, with no weird artifacts? increase sampling steps?

[paper crag]

say guys, i keep trying to change her expression buts not working

do i have the wrong settings on?

[ornate elk]

Are you on AMD?

[noble belfry]

are you masking it correctly?

[paper crag]

yeah

[noble belfry]

try it without the refiner

[ornate elk]

If you use the directml webui and not zluda To get inpainting and adetailer working you need to add --no-half to the webui-user.bat

[paper crag]

i already do that

[ornate elk]

Okay then also disable refiner

[paper crag]

also did that, still no change

[ornate elk]

Ans change the resolution to a square

[paper crag]

changed to a square, still no change

[ornate elk]

Can you restart and only set
Inpaint area to mask only.
Then set the resolution to 512x512
And denois to 0.8

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

And also make sure to run the webui-user.bat and not the webui.bat

[paper crag]

huh, it working now

thanks for the help

[ornate elk]

No problem

[vivid viper]

i cant get ReActor for Stable Diffusion extension to work even after installing the extension, should I fully uninstall the Stable Diffusion and start again

[odd nymph]

I need help running Animated Diff, it keeps thowing an error and i have no clue how to fix

[vivid viper]

nevermind I got it to work now

[azure magnet]

Sorry, how to update torch and xformeres

[azure magnet]

please someone

[dense sandal]

Look for my messages

@azure magnet

[azure magnet]

DM?

[dense sandal]

--reinstall-torch --reinstall-xformers

[opaque viper]

I'm not sure what happened. Been running Forge for a while now with no issues. Today, whenever I use SDXL models, I get this error: TypeError: 'NoneType' object is not iterable

Nothing really changed, other than a small update for the webui

1.5 models work fine

[fierce nimbus]

I have a 1.5 model loaded. The lora tab doesn't appear near "Checkpoints" and "Hypernetworks"

[fair oxide]

How Do I Fix This

[ornate elk]

Wrong python version.
Please follow my install guide from the Pinned Messages in this channel

You need to uninstall python 3.12 first

[hoary lodge]

I'm truly struggling with this error trying to run a simple SDV Node into a KSampler,... I keep getting the same error. I've updated my nVida Drivers, and downgraded them with no avail, I've also uninstalled my CUDA Toolkit, (12.3) and reinstalled it. I've uninstalled PyTorch, and Reinstalled it. I have no idea how to fix this error. Could it be that I need to reinstall everything (ComfyUI) and start from sctratch?
Error occurred when executing KSampler:

CUDA error: invalid configuration argument
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.

[gentle anchor]

im unsure of how to fix this, i just got it set up and its giving me this error prompt

[ornate elk]

Hey, what's your python version?

[gentle anchor]

10.9@ornate elk

[ornate elk]

3.10.9 ? Okay, also do you have any third party antivirus ?

Also when do you get this error?

[gentle anchor]

i dont think so no

and i get it upon trying image generation

[urban current]

you can delete the venv folder and launch the bat file, it may take a while to redownload but that should fix it

[gentle anchor]

sweet

[ornate elk]

Ohh that error comes from the wrong settings

No need to delete the venv

[gentle anchor]

oh

[ornate elk]

Disable refiner and hires fix or use them the correct way xD

[gentle anchor]

so no refiners or hires fi?

fix*

[ornate elk]

Refiner is for sdxl so that won't work with 1.5 models

[gentle anchor]

i see

stilll not generating, same error prompt

[urban current]

how did I miss that

[ornate elk]

Hires fix is good, but the "default" settings are incorrect

[ornate elk]

Have you restarted and check the cmd for any errors?

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

Also what's the model your trying to use

Also set cfg to 7, 25 is way to high

[gentle anchor]

i am using a amd radeon 6600

config reset to 7, will restart program

[ornate elk]

Then your using the wrong webui version in general

AMD needs a different one

@gentle anchor
You have two options with that card. You can find boath of my AMD guides in the pinned messages of this channel.
The Zluda version is much faster than the Directml one, but requires more steps to get it to work.

[gentle anchor]

yeah i guess im just gonna have to start from scratch tomorrow

[ornate elk]

Yep, checkout the guides and for any questions feel free to ask

[gentle anchor]

thanks for your help, and will do!

[ornate elk]

No problem 🙂

[rain wind]

my sd hangs when its out of vram but doesnt do an error, how do i stop it from going into this state? i have to kill the cmd to get it back

usually when i ask a resolution too high

it will max out the vram but not stop, so it becomes stuck and the interrupt button isnt doing anything in auto1111

[ornate elk]

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

[rain wind]

2070 super with 8GB VRAM with --medvram --no-half-vae

[ornate elk]

Ah okay, you need to add --xformers and change --medvram to --medvram-sdxl
Then it should be much faster and probably don't freeze

[rain wind]

ok ill try that, thanks!

[ornate elk]

@fair oxide delete the venv folder and relaunch

[fair oxide]

um

[ornate elk]

It uses the wrong python

You can't use the windows store python

[fair oxide]

uninstall ?

[ornate elk]

Uninstall every python you dont need and install the 3.10.11 64bit with the installer

[fair oxide]

alr

[ornate elk]

Then delete the venv folder again before launching

[fair oxide]

should I delete the python launcher?

[ornate elk]

If its from the windows store yes

[fair oxide]

is this from the windows store?

[ornate elk]

Yes

Its from the 3.12 install from before

[fair oxide]

ok

Ive downloaded it however

[ornate elk]

Did you checked add python to path?

[fair oxide]

how do I do that

[ornate elk]

Run the python installer again, click on modify, click next, select "add python to system variables"
Then it should work

[fair oxide]

enviromental variables?

[zealous leaf]

Are there any good reliable SD alternative APIs, that don't use StabilityAI's servers? I need those services to use as a fallback model when the StabilityAI is down.

I have found Clipdrop, that just acquired by Jasper, they use models developed by StabilityAI, and the company was formerly under StabilityAI, so I'm not sure they don't use the same servers.

There is also https://stablediffusionapi.com/ but they don't have any kind of status page, to understand do they rely on StabilityAI or not.

[ornate elk]

Do you have any third party antivirus programm installed?

[fair oxide]

mcafee?

[ornate elk]

Yes xD

Stuff like mcaffe breaks the python install

[fair oxide]

should I turn off my firewall?

[ornate elk]

For the python install yes

[fair oxide]

ok

[ornate elk]

Also mcaffe isnt a good antivirus programm, so your better of with the default windows defender

[fair oxide]

oh

it just came with my pc when I bought it

it still has the same error msg

Ill try fix it tmmrw

[valid gorge]

Something odd going on with Comfy atm. Yesterday evening images were generated just fine, not almost every image is overexposured. Haven't done any changes to anything.

[ornate elk]

Have you reinstalled python?

Best way, uninstall python, and McAfee, reboot the PC, and install python again while checking "add python to path"

[fair oxide]

damn I had 3 vpns on my pc ??? 💀

[ornate elk]

Wtf xD uninstall them too if you didn't installed them

[fair oxide]

I have 4 anti viruses that I havev no idea of

[ornate elk]

You said you bought the PC new ? xD wow

4 is a new record

[fair oxide]

ill msg this channel tmmrw if it works cz i need to sleep

[ornate elk]

Uninstall them all lol

[fair oxide]

will do

[ornate elk]

Sure!

Gn

[fair oxide]

gn

[brisk heath]

while i was updating torch and xformers, this shows up. i went to the Scripts folder and updated pip but the -rotobuf warning and pip update notice still reappear. how do i fix this?

if i run the .bat file normally (without torch and xformers update), this shows up. please lmk if it's related to the above or requires a different solution

i've restarted the Terminal, and i've restarted the PC. idk what server it's asking me to restart

[ornate elk]

Thats a message from the Reactor Extension

The rotobuf not.
You can update torch and xformers.
Then update your extensions

[royal plaza]

How to solve this error while using controlnet in comfy

[hot spindle]

Hey, trying to update to A1111 1 .8 git pull gives me the following error:

Tried unsuccessfully to delete offender file and redo, after git merge --abort ... I'm clueless, I really dont wanna go trought a fresh install , and all the extentions,etc...

Any idea of what to try?

[fair oxide]

Two questions:

1. Is it possible to programattically/automatically switch models while running batch jobs? I want to test prompts across multiple models, but don't know of a way to have A111 switch models

2. How can I find the trigger words automatically for LORAs? I have 100s and didn't bother to document all the trigger words for the LORAs listed on CivitAI

[ornate elk]

for 1, you can use the X/Y/Z script in txt2img, with that tool you can select multiple checkpoints that will be used with the same settings

for 2. there is the civitai browser+ plugin, which should list them

[fair oxide]

would that be "checkpoint name"?

[ornate elk]

yes

[fair oxide]

hell yes, thank you!

[ornate elk]

No problem 🙂

With x/y/z script you can compare everything with everything

[fair oxide]

yeah, that's great

I use it regularly for prompting, didn't know it could do checkpoints too

very nice

[solemn atlas]

Hi guys, yesteray I download comfyui again

but i dont know how to start it

so this is what happens

it says "An error has occurred. Do you want to view the WINZip log file?"

I said yes

and this apperard

I tried with CMD and put this in "python main.py" but it says there is no such things as this file

[river flame]

can I get generation settings of the images in Output folder? really need it

[manic snow]

assuming you're using A1111/Forge, you can! just go to the "PNG info" tab and upload your image

[river flame]

thank you so much

[spiral roost]

Hey I need help to make my local stable diffusion work again, I have an error concerning xformers when lauching the webui, yet my pytorch version is the right one and I tried to reinstall it multiple times. It keeps telling me I have the 2.0.1+cu118 version no matter the version I installed. It was working fine last time I used it and I didn't change anything. Thanks

[river flame]

I get this error constantly

[tardy sparrow]

Hello, how to add ControlNet models for "webui_forge_cu121_torch21" ?
the only one i got is this :

[river flame]

I m getting either yellow\red cast or a white clown face

also this: "FaceSwapLab - ERROR - Failed to swap face in postprocess method : No reference face found" I do have a face in FaceSwapLab

[lunar bloom]

So we have Differential Diffusion now, which is awesome (https://differential-diffusion.github.io/) and it works with ComfyUI now for example (the Node you need is under _for_testing), but the way the example works changes the whole picture, because it Vae Encodes the Original with the "InpaintModelConditioning" Node. Is there a way to have get the DD changes without changing the black areas in the mask at all (similar to the DetailerDebug approach from the ImpactPack) ?

[karmic crown]

git restore webui-macos-env.sh
git pull

if that does not work, backup your command line from webui-user.sh
run

sh a1111-setup.sh -f all

and restore the command line params

that question is more for #📝｜prompting-help

this one is for technical issues related to installation, and errors

[valid gorge]

Not directly related to prompting but I'll try maybe general chat, Thanks

I think it's more like inpainting problem

[karmic crown]

#📝｜prompting-help is not strictly limited to prompting, you might get the answer there even faster than on #💬｜general-chat

[valid gorge]

Ok great, thanks!

[warped imp]

Hey guys, I followed two tutorials and got no luck

Just wanted to ask if it's possible to get stable diffusion on a Microsoft Surface Pro 9?

[manic snow]

it doesn't have a dedicated GPU, does it? because if that's the case, you're not going to have a good time running it

[karmic crown]

it does not even have Intel ARC (as far as I know), so I guess it will not work

only CPU mode will work

and that will be painfully slow

[manic snow]

yeah, i wouldn't recommend CPU mode to my worst enemy

[warped imp]

Oh man that's rough

Well that's a lil miserable any work arounds this?

[manic snow]

running it in the cloud is an option

[warped imp]

Only really got my college laptop right now haha

[karmic crown]

i tried on 10 CPU VPS with 60GB of RAM, and it was painfully slow

[warped imp]

Sorry, im new to this stuff, what does that entail?

[manic snow]

you pretty much rent a computer in like a datacenter or something, and you can use it to generate images

[karmic crown]

or some website like civitai

[warped imp]

so like going to a library and using their pc

[manic snow]

kind of, except this pc is powerful and you pay depending on how much you use it

[warped imp]

I see

Alright not necessarily trying to go that route,

Are there any other programs I could use? Like websites?

CivitAI i just saw from viking

[karmic crown]

https://civitai.com/

[warped imp]

But basically, what im looking for

Free program, or a one time purchase

And the ability to replicate art styles

[manic snow]

well, imho there aren't any good free websites, and a one time purchase isnt really a thing in this industry

[karmic crown]

https://www.liblib.ai/ is fine, but i do not recommend unless you do not speak chinese or have chinese friends to help you

[warped imp]

ahh yeah credit systems suck

Yeah that's a little annoyin

[karmic crown]

my chinese frind use liblib more than his nvidia 3070, since it is faster, and not so limited as civitai

[warped imp]

Basically trying to replicate anime characters into scenes that allow nsfw as gore is sort of part of the thing we're running but

[warped imp]

is this another site

[karmic crown]

sorry, i accidentally pressed enter before i finished the message

[warped imp]

all good haha

[manic snow]

i guess the closest thing you can get to a one time purchase is commissioning someone to do it for you, but i don't know how to feel about that tbh

[warped imp]

Haha yeah

definitely looking to do it myself

do you guys know what specs i should look for in a laptop to run stable? if i buy a new one

minimums would be cool trying to get a lower end laptop than this microsoft pro cuz it was expensive

[karmic crown]

nvidia 3xxxx or better would be the best

[warped imp]

got it

[manic snow]

definitely a dedicated gpu with 16gb or more VRAM. so not regular RAM, VRAM. that should keep you future proof for at least a while

[karmic crown]

laptop with 3070 works fine

[warped imp]

Is that a nvidia graphics card

[karmic crown]

also, what YellowSoul said

[manic snow]

yes, that's another thing. make sure to get an Nvidia card

[karmic crown]

more VRAM, the better

[warped imp]

got it

alrighty, hope that applies to other software too

ive been having hell to deal with after effects

it runs, but it runs like shit

[manic snow]

a dedicated GPU should solve that issue i'm pretty sure

[warped imp]

got it

@karmic crown question, what does civit use for credits?

noticed buzz didnt go down on prompt

[karmic crown]

you can earn some, or buy some

[solemn atlas]

I dont want to annoy someone but I hope someone can help me

[warped imp]

what are they for?

[karmic crown]

buzz = credit

10000 credits = $10

[warped imp]

mine dont seem to go down with the prompts though?

like im generating pics and nothing is happening

[river flame]

reinstalled everything again, could not install faceswaplab extension (which I really need)

permission denied

I tried typing this:

no permission

[karmic crown]

do you have some 3rd party antivirus like McAfee?

[past rampart]

Hi, does anyone know if Stable Cascade will be added to the Dev Platform?

[warped imp]

if this is to me then no

[karmic crown]

it is for @river flame since he has my "favorite" error

[river flame]

no, I dont. Trying to reinstall the extension with system defense off, but how do I reinstall it?

I deleted the folder but it says the extension is alreasy installed

[karmic crown]

you are trying to install forge protable?

[river flame]

I m trying to reinstall a extension in Forge portable

[karmic crown]

https://www.geeksforgeeks.org/how-to-take-ownership-of-a-folder-in-windows-10/

make sure folder d:\forge is yours

do not runt anything as admin

take ownership, and then try again

i guess you run something as admin

and that messed the privileges

[river flame]

but how do I run again, it says the extension is already installed although I deleted the folder as reddit said

[karmic crown]

this is your problem

[river flame]

can it be because it s the portable Forge?

[karmic crown]

nope

check the link i gave you, and fix your permisions

and do not run anything related to stable difusion as admin

[river flame]

my folder properties do not have such tabs

I m on windows 11, I googled for it, it says press this

but I dont have this button either

oh I see I should install it first

[karmic crown]

that is windows option

i am on mac, so i cant make a screenshot

[river flame]

ok I m taking ownership

[river flame]

no, the same error of permission again

it worked perfectly before I switched GPU

and reinstalled everting

[river flame]

can it be because I have both Python 3.10 and 3.12 installed?

[karmic crown]

if you are using portable forge you do not need any of them

[empty cipher]

I'm having a terrible time training on this style. I build dataset of 50 decent images. Good resolution, some headshots, some body shots, some action shots with multiple people - even a handful of "detailed background" images

And the style is consistent

And I captioned them as best I could

I've tried a whole bunch of different configurations, but the results are inconsistent and rather poor

Any advice?

[karmic crown]

ask in #📝｜prompting-help or #💬｜general-chat
this channel is mainly for technical issues relating to installation, bugs and similar technical issues

[empty cipher]

Oh, thanks!

[vocal burrow]

#🔧｜finetune would probably be the best place for that actually

[karmic crown]

is that new one?

[vocal burrow]

nope but it's far down the list so very few people ever see it :p

(I go back into my cave)

[karmic crown]

@river flame "I just changed permissions of the problematic folder from READ to FULL CONTROL. It is in the folder's proprties, SECURITY tab." - try that

[river flame]

Oh I have to not only install the extension but also a .onnx model

but still there should be no error while sinstalling, that s not right

[karmic crown]

you have issues with privileges, so installing more things will not help

you need to fix privilegies first

folder (and things inside) must be yours, with full control

it is so easy on linux and mac

[river flame]

I dont have a security tab there let me google

[karmic crown]

windows does not have chown and chmod 🙈

[vocal burrow]

did you try to restart your computer (could be something in the background holding it hostage) ?
Also did you try to run d:\Forge\system\python\python.exe -m pip cache purge ?

[karmic crown]

he has some access errors, so i do not thing purge will help

this

[river flame]

btw, should I do this?

Hint: your device supports --pin-shared-memory for potential speed improvements.
Hint: your device supports --cuda-malloc for potential speed improvements.
Hint: your device supports --cuda-stream for potential speed improvements.

[karmic crown]

ignore hints until you fix errors

[vocal burrow]

probably but later (it's all explained here https://github.com/lllyasviel/stable-diffusion-webui-forge?tab=readme-ov-file#forge-backend), focus on your error first

[river flame]

ok thanks

[karmic crown]

i just found this, this should be equivalent of chown and chmod on linux/mac

ICACLS "D:\forge" /setowner "username"
ICACLS "D:\forge" /grant:r "username:(F)" /C

just replace username with your actual username
and run this from admin command prompt

@vocal burrow have you used that ever?

[vocal burrow]

if you run the install command yourself in a cmd, do you have more informations ?

[river flame]

I missed the last line: You can try to install dependencies manually by activating venv and installing requirements.txt or requirements-gpu.txt

[karmic crown]

if you are using portable forge you do not have venv

[vocal burrow]

the hinted stuff or forge ? To both, yes, it's working great for me. The speed and loading improvments are nowhere close to the numbers they promise, but it still works a little better and the vram consumption is significantly improved. Currently I'm using set COMMANDLINE_ARGS=--cuda-stream --cuda-malloc --pin-shared-memory --unet-in-fp8-e5m2 trying to see if I can squeeze a whole 1024x1024 sdxl generation within my 8gb (nope but very close)

[river flame]

in the secont line do I change only the username?

[vocal burrow]

in both lines I'd say

[river flame]

it says syntax error of file name or folder

yes in both

[karmic crown]

i was asking about ICACLS, sorry for confusion

[vocal burrow]

oh. nope

[karmic crown]

if that works, that would be great for fixing privileges issues

since clicking through 7 levels of hell in folder properties might be confusing 🙈

[river flame]

the second line didnt work, same error with extenstion again

I can try this

[vocal burrow]

please try to run the install command manually before tho

#🤝｜tech-support message

maybe it will tell us more

(cause of all the @echo off stuff)

[river flame]

permission denied

[vocal burrow]

https://tenor.com/view/frodo-keep-your-secrets-alright-then-lotr-gif-12907121

[karmic crown]

does it work like this (wihtout /C)

ICACLS "D:\forge" /grant:r "username:(F)"

with your username

[river flame]

yes it worked

[karmic crown]

try this now

[vocal burrow]

stupid question. You didn't have forge running this whole time somewhere else ?

[river flame]

no. Tried installing the extension again, not working

I ll reboot

nope

[vocal burrow]

what antivirus do you have ?

and wher is it installed ?

[river flame]

just the windows defence

[vocal burrow]

ok

[river flame]

I added a Security tab in reg, gave the rights, still dont work

[vocal burrow]

well I don't like it but try "d:\Forge\system\python\python.exe" -m pip --user onnxruntime-gpu=>1.15.0 --prefer-binary

let's see where it goes

[river flame]

I should change "user"?

no error reinstalling the extension in A1111

[vocal burrow]

no

run it as is

[vocal burrow]

wait I' m dumb

"d:\Forge\system\python\python.exe" -m pip install --user onnxruntime-gpu=>1.15.0 --prefer-binary

forgot install :p

[vocal burrow]

rooooh cmon

I'm still sick and it shows

"d:\Forge\system\python\python.exe" -m pip install --user onnxruntime-gpu>=1.15.0 --prefer-binary

[river flame]

oh I should install Python again

[vocal burrow]

no ?

why ask ?

what did it said ?

[river flame]

A1111 gives me this with FaceSwapLab on (juggernaut)

[vocal burrow]

probably incorrect vae

[river flame]

it was actually, but the default VAE gives the same result

[vocal burrow]

meh I don't mind this, ignore it

or

[river flame]

ok

[vocal burrow]

why is it showing in appdata

not in the path... ok.... in appdata tho ?

[river flame]

where should it be?

[vocal burrow]

1/ open a cmd in your forge install folder.
2/ run call environment.bat
3/ run "d:\Forge\system\python\python.exe" -m pip install --user onnxruntime-gpu>=1.15.0 --prefer-binary

[river flame]

oh I works with a different checkpoint - DarkSun

[river flame]

I did, it said nothing

[vocal burrow]

good then

it must have worked

did it ?

Ok then I'm guessing it made your pc explode then :p

[small zenith]

i created a good image with SD, how can i "save" it and make the AI know i want pictures like these?

[river flame]

it really worked, thank you SO MUCH!!

you are a genius

[karmic crown]

https://github.com/deforum-art/sd-webui-deforum/pull/952

this will finally fix deforum without need to install missing library on new clean a1111 installations

[vocal burrow]

funny I've seen someone complain about the file loader a couple days ago too

anyways wait and see if it gets accepted :p

[karmic crown]

second PR for deforum in 2 days, all i am doing is fixing deforum on 1.8.0 🙂

first one that actually solves the main issue was merged

https://github.com/deforum-art/sd-webui-deforum/pull/950

[vocal burrow]

I don't think there is anyway to nudge the AI locally toward "something that you like". It's too vague.

[karmic crown]

a1111 1.8.0 changed some things that authors of deforum plugin was not aware of

[vocal burrow]

lots of stuff happened between deforum extension and auto1111 lol

[karmic crown]

https://github.com/deforum-art/sd-webui-deforum/pull/950#issuecomment-1977041729

see this

i made new version of my script, but i am waiting to see what will happen with PRs

Instruct A1111 to use latest stable version of torch...
Add recommended command line parameters...
Fix cannot convert a MPS Tensor to float64 dtype error...
Fix contolnet and deforum requirements hell...

List of modified files:
  modules/sd_disable_initialization.py
  requirements_versions.txt
  webui-user.sh

see the 4th line

🙈

deforum is asking for basicsr, which was removed from a1111 1.8.0

controlnet cannot install insightface, and asks for manual installation

manual instalation install protobuf 4.25.3, which is conflicts with some other libraries

☠️

[vocal burrow]

classic deforum stuff

they've been battling since day 1 on how to handle dependencies

I haven't seen much issues with controlnet tho

[karmic crown]

i fixed deforum, but controlnet is now the problem

my script add this to requirements to fix the issue, but that is just a workaround

protobuf==3.20.3
insightface==0.7.3

i will probably release 0.0.6 in next two days, i will just wait if i can at lest remove basicsr from requirements (and leave those two for deforum)

btw. the same error happens with clean install of forge and "native" a1111

it is easy to overlook that when you have existing venv

[vocal burrow]

good to know. Although I don't know why someone would use the "native" controlnet extension when forge comes with its own

[sour ferry]

Hello, for a month I've been using only the turbo and lighting checkpoints, I have about 30 of them. Is there a way, when loading one, to automatically change the cfg and the sampling methods, and sampling steps with the best choices for the loaded checkpoint? For example, because when I create a prompt then I activate x/y to ask it to make an image with each of my checkpoints, to see which is the best choice for a prompt, I would like it to be able to change the cfg, and step, for the best choice.

[vocal burrow]

I don't think so, there's stuff such as https://github.com/rifeWithKaiju/model_preset_manager?tab=readme-ov-file and https://github.com/Zyin055/Config-Presets but for me it was cumbersome to use and I ended up reverting back to my usual model workflow of having a .txt holding notes and some .jpg for each checkpoint that I can check in with model preview https://github.com/CurtisDS/sd-model-preview-xd

[spiral roost]

Hey I need help to make my local stable diffusion work again, I have an error concerning xformers when lauching the webui, yet my pytorch version is the right one and I tried to reinstall it multiple times. It keeps telling me I have the 2.0.1+cu118 version no matter the version I installed. I also tried to reinstall xformers. It was working fine last time I used it and I didn't change anything. Thanks

[crisp prawn]

I cant get forge layerdiffusion to work if anyone can help me.

[vocal burrow]

1/ you're using an incompatible python 3.11.7, go back to some python between 3.10.6 and 3.10.11
2/ in the second screenshot you're using your system's pip not your venv's one

[spiral roost]

Thanks but I saw people saying that python 3.11 was working for them and like I said it was working for me before today. So what can I do to force the environment to select the correct version ?

[vocal burrow]

Some people also drive their car without seat belt

I can't tell you more, especially without seeing a full log

I strongly advise you to just get a known compatible python, delete your venv folder and try again

[hot spindle]

Tks..I did manage to solve with "git rm webui-macos-env.sh" at least I think so, it runs , updated to latest nightly torch and kept all extensions working

[spiral roost]

thanks, somehow just deleting the folder solved the problem

[normal quarry]

how do i know if stable diffusion is using my gpu? i have a good gpu and the photos dont look that good

[vocal burrow]

image quality doesn't depend of your gpu.

only speed is affected

[crisp prawn]

I keep getting this error msg trying to run layerdiffusion: *** Error running process_before_every_sampling: D:\Ai\stable-diffusion-webui-forge-main\extensions\sd-forge-layerdiffusion\scripts\forge_layerdiffusion.py
Traceback (most recent call last):
File "D:\Ai\stable-diffusion-webui-forge-main\modules\scripts.py", line 835, in process_before_every_sampling
script.process_before_every_sampling(p, *script_args, **kwargs)
File "D:\Ai\stable-diffusion-webui-forge-main\extensions\sd-forge-layerdiffusion\scripts\forge_layerdiffusion.py", line 130, in process_before_every_sampling
vae = p.sd_model.forge_objects.vae.clone()
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: 'VAE' object has no attribute 'clone'

if anyone can help I'd appreciate it TIA

[old parrot]

Is there a way to install SD Forge on an AMD GPU?

[ornate elk]

What's your GPU?

Normaly you can install forge and run it via --directml

[old parrot]

@ornate elk Can you link me a guide please? It's for a friend

Not sure what GPU off hand

[agile nimbus]

This is a dumb question probably, and please anyone can get to it when they can, becuse I feel like im interupting other discussions. But I dont really know exactly what type of SD I am using, other than I go into a folder and click a file that says "webui-user.bat".

My question is, how do I use SBXL models? I have tried putting them in my models folder and when I try to load it, it never works. I know I am doing something wrong, any help would be very appreciated.

[noble belfry]

You're never interrupting conversations, we can thread them or people can respond whenever.

[ornate elk]

He would be better of using one of my two amd guides.
You'll find them in the Pinned Messages of this channel.

I tried forge ones with amd and it didn't worked :/

If he has a new GPU gen then he should use zluda

If he has a older card then directml

To get forge running you need to install it like stated on their github and then add --directml to the webui-user.bat
But I doubt that it works good

[gloomy wave]

Upgrading automatic1111 to forge bumps up speed in generation greatly but does it have a negative effect going up so fast, like for example worse quality?

[old parrot]

Thank you! I will tell him. What variation of Stable diffusion do you use?

[old parrot]

Nope quality is the same in my experience, love Forge!

[noble belfry]

Not in quality, but it has some drawbacks. like models are slower to swap out

[gloomy wave]

I mean it sounded "too good to be true" at first but I will try it perhaps unless there any risky issues.

[ornate elk]

Im have directml and zluda installed. And only using Zluda now as its much faster than directml.
Zluda is the best for AMD on Windows right now. But requires a 6000 or newer GPU series

[gloomy wave]

I can live with that. I often use same model.

[noble belfry]

its not risky or anything. its just slow to do some things, faster to do others. a1111 is one of the contributors

[agile nimbus]

thanks, I was just trying to wait my turn hahaha

[noble belfry]

oh then you'll never get one lol

[agile nimbus]

true true

[noble belfry]

do you know which client you're using? a111/comfy?

[agile nimbus]

how do i find that?

[ornate elk]

Your mostly using Automatic1111 as webui.
But you need to check the version. Its at the bottom of the browser when the webui is open

[agile nimbus]

v1.8.0

[ornate elk]

Okay and what's your GPU?

[old parrot]

Thank you very much!

[agile nimbus]

3090

[noble belfry]

should run sdxl rightly

[ornate elk]

Then you shouldn't have any issues using sdxl models

[noble belfry]

are you wondering if the models are sd1.5 or sdxl?

[agile nimbus]

I guess I am not worried about running them, its just that when I try to load the checkpoint model in the top of the UI, it never actually loads, it stops the process after a little, but it goes back to the previous model i was using

[noble belfry]

thats ood. any errror in the cmd window? (black bg white text)

[agile nimbus]

let me try, give me a sec

"AssertionError: We do not support vanilla attention in 1.13.1+cu117 anymore, as it is too expensive. Please install xformers via e.g. 'pip install xformers==0.0.16' "

[ornate elk]

Ohh do you have dreambooth installed?

[agile nimbus]

i have no idea haha

[noble belfry]

i thn thats from the old zip installer

[agile nimbus]

should i try a different way?

[ornate elk]

You should go into the Extensions tab and show us a screenshot

[noble belfry]

your hardware is fine, your software is old

[gaunt mortar]

hi guys, I am running stable cascade on ws2l. after finishing up what I am doing, I use "wsl --shutdown" in windows and it kills it, I also manually close docker, I no longer see the .exe associated with wsl, but my system RAM usage remains high, its like the memory is still allocated to linux even when the processes are closed

[ornate elk]

Okay then its easy, your webui is updated but not the core components. Like torch and xformers.
You only need to delete the venv folder and relaunch the webui-user.bat

[agile nimbus]

oh really? that is all?

[noble belfry]

yeah youre golden

[agile nimbus]

wow, you guys are smart, I will report back after giving that a try

[noble belfry]

CS1o might be smart, i just keep having errors and seeing other people the same errors.

now you can answer the same question 😄

one upside: i realised i spend my evenings helping people online with tech problems so finally enrolled in university. 🙂

[fair oxide]

new version of torch and xformers

torch==2.1.2

[wispy wasp]

Quick question, is the directml branch supposed to gobble vram like a starving hyena? I don't know if it's my settings but I get 2.5gb usage on the main branch with xformers, while 7gb on the directml branch.

Also I had medvram on for the 7gb

[vocal burrow]

yup, but why would you use the directml fork if you can use the main one ?

[worldly rain]

guys quick question. I am using inpaint anything. I see the inpainting models here like realistic vision. I installed a few inpainting models. which folder do u put these models so I can use them in inpaint anything?

[vocal burrow]

inpainting stable-diffusion models are still stable-diffusion models, therefore they go in models\Stable-diffusion

[worldly rain]

oh, alright. thanks a lot 🤝

just put it there but it still doesnt appear in the models section in inpaint anything

[noble belfry]

theres a little refresh button. looks like the recycling icon but blue

[worldly rain]

yes I see it in the stable diffusion checkpoints window but I was talking about here in inpaint anytinhg

[wispy wasp]

amd gpu

[crisp prawn]

Here is the error msg i get when I try to run Layerdiffusion, any help is appreciated: *** Error running process_before_every_sampling: D:\Ai\stable-diffusion-webui-forge-main\extensions\sd-forge-layerdiffusion\scripts\forge_layerdiffusion.py
Traceback (most recent call last):
File "D:\Ai\stable-diffusion-webui-forge-main\modules\scripts.py", line 835, in process_before_every_sampling
script.process_before_every_sampling(p, *script_args, **kwargs)
File "D:\Ai\stable-diffusion-webui-forge-main\extensions\sd-forge-layerdiffusion\scripts\forge_layerdiffusion.py", line 130, in process_before_every_sampling
vae = p.sd_model.forge_objects.vae.clone()
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: 'VAE' object has no attribute 'clone'

[karmic crown]

forge has exactly the same problem with clean install 🙈 i actually noticed the issue with forge first 🙃

[neat saffron]

question, a lora from civid ai told me to use a adetailer, however, i dont seem to have a folder for that in my stable diffusion folder, is it probably cause of what version i have? i am not too skilled with it so im not sure if my version is out of date

[vocal burrow]

try to put "inpainting" in its name if you're that it's an inpainting model

[worldly rain]

file's name is "chilloutmix_NiInpainting.safetensors"

[hoary plaza]

Quick question how would i go about installing the ohter ones like waifu ddiffusion etc

[vocal burrow]

adetailer is an extension for auto1111, you have to install it additionaly. Use the "extensions"->available tab for that, click the "Load From" button and from there adetailer should appear in the list along with an "install" button

[vocal burrow]

NiInpainting

so there's no "inpainting" in the name

rename it

[worldly rain]

yessir

[vocal burrow]

it's quite stupid but there is no way to recognize an inmainting model from a regular one..... that's why most code just look for "inpainting" in the name of the model.

Just look at that ....

• https://github.com/AUTOMATIC1111/stable-diffusion-webui/blob/311354c0bb8930ea939d6aa6b3edd50c69301320/modules/sd_hijack_inpainting.py#L322-L325
• https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#inpainting-model

[neat saffron]

Thanks

[fresh phoenix]

hi, not sure if it's right channel but we paid $20 for commercial membership (twice!) but don't see the same reflecting in our account. It's still the basic 25 credits account. Stability have specified their support emails but getting no response at all (been 3-4 days). Would anyone have an idea or faced something similar? Is there a way to subscribe from within your account (membership payment was done via https://stability.ai/professional)

[worldly rain]

yeah they can make it easier for people, idk how it makes sense

[odd nymph]

Is Animatediff broken? it never works on my system

out puts an error

[dense obsidian]

I also found I had an issue with xformers today. What was that I saw before, the solution is to delete the venv folder and stable diffusion will update it by itself?

[twin linden]

When I save an lora's settings like activation text, preferred weight and preview picture it doesn't update immediately like it used to. Now I need to restart the webui for it to be applied. Any fix to this?

[vocal burrow]

you probably want to contact them directly then. https://stability.ai/contact

[vocal burrow]

The person who originally named the model is also partially to blame here.... It's a well known fact among advanced users / people training models.

[worldly rain]

I completely agree

it is what it is thankfully we got people like u to help us out 🤝

[vocal burrow]

deleteing the venv is one way to reinstall xformers for sure yes

[vocal burrow]

I don't think so.

[vocal burrow]

not sure what you mean exactly here.

[odd nymph]

huh, i cant get it funtional, is it linux only maybe?

[vocal burrow]

nope should work on windows too

what's the full log ?

[odd nymph]

Dang, idk what im doing wrong then

[odd nymph]

ill send one just a sec!

[vocal burrow]

you can take your time, I have to commute to work anyways :p

I'll be afk for a while pretty soon

[vocal burrow]

the full log please

there's stuff cut before that

[odd nymph]

oop ok, one sec again, ill need a clean terminal then im sorry

Clean full terminal output

[vocal burrow]

looks like it's installed correctly, so you're probably "just" using it wrong

would need to see your settings to get what's wrong

(time for me to go)

[odd nymph]

figured also have a good day!

[twin linden]

When you save a lora's settings by editing metadata and click on the image it will apply what you saved but when I click on lora it's defaults to weight of 1 with no other activating text that I saved with. It creates an .json and png but doesn't update in real time which is what it suppose to do. I have to restart SD for it show up.

[vocal burrow]

you can always screenshot your webpage so we can check what might be wrong

[odd nymph]

ill have to do that in a bit

not able to right now

[vocal burrow]

np

[vocal burrow]

cant check stuff like that while I m at work maybe someone else could.

[gloomy wave]

I am trying to install Forge Ui but it is acting weird and gets stopped by windows protection in install phase, should i be worried or just install anyway?
Also if i already have Automatic1111 installed do i need to do a clean install instead?

[vocal burrow]

what did you use to install it ?

[gloomy wave]

I have only tried to install but stopped. I used the 1-klick option from here: https://github.com/lllyasviel/stable-diffusion-webui-forge

The file folder also gets locked by vpn making so the downloaded file cant be deleted unless i turn off vpn.

[nocturne topaz]

Clean install recommended

[river flame]

yesterday everything was good, now it s

for both Forge and A1111, most of RAM is free

it was like this before only when RAM was close to full

[autumn void]

Hi, is there a way to set up Focus to open in a specific browser without having to have that browser as default in windows? How, if it's possible?

[river flame]

and the result is this again, I m gonna go crazy soon

[ornate elk]

Thats an VAE issue

[river flame]

THANK YOU SO MUCH. it switched it by itself for some reason

[ornate elk]

Np ^^

[fair oxide]

hi @ornate elk have you updated A1111 in the last few days?

[ornate elk]

Only the Directml zluda version

My nvidia Version is still 1.7.0

[fair oxide]

they released newer version of torch and xformers

[ornate elk]

Yes

Based on cuda 12.1

[fair oxide]

yeah

[rose solar]

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

[dense obsidian]

If I do delete the venv folder, do I need to put anything back into it? can't remember if there is anything important that I had to manually add in

[vocal burrow]

no don t mess with the venv, it will reconstruct by itself once it s deleted

[dense obsidian]

got it, alright, deleting venv then

[autumn void]

Hi, is there a way to set up Focus to open in a specific browser without having to have that browser as default in windows? How, if it's possible?I've found this but I dunno where to put that line

--disable-in-browser

looks like something I'd put in the webui-user batchfile if it were A1111 but can't find anything like that in Focus

[river flame]

I do have an image in IP-adapter. ETA 30 mins. VAE is automatic

[autumn void]

solved:
edit the run.bat file and add this in the first line
set BROWSER=C:\Program Files\Google\Chrome\Application\chrome.exe
(or whatever your browser's path is)

[river flame]

solved long ETA with Low VRAM (I have 12 Gb), but it still doesn`t see the face in IP-adapter

why do I have different errors every day, am I cursed

[ornate elk]

Make sure the webui is whitelisted in any browser adblocker

The error says no input image given

So maybe wrong file type

Should be .jpg or .png

[gusty herald]

does anyone know how to make the skin tone match after img2img inpaint generation?

[river flame]

turned off all adblockers, IP-adapter now sees the image, but faceswaplab still not

[ornate elk]

would suggest using Reactor instead of faceswaplab

since faceswaplab dont receive updates anymore

[crisp prawn]

Hi I'm trying to figure out a Layerdiffusion error in Forge: *** Error running process_before_every_sampling: D:\Ai\stable-diffusion-webui-forge-main\extensions\sd-forge-layerdiffusion\scripts\forge_layerdiffusion.py
Traceback (most recent call last):
File "D:\Ai\stable-diffusion-webui-forge-main\modules\scripts.py", line 835, in process_before_every_sampling
script.process_before_every_sampling(p, *script_args, **kwargs)
File "D:\Ai\stable-diffusion-webui-forge-main\extensions\sd-forge-layerdiffusion\scripts\forge_layerdiffusion.py", line 130, in process_before_every_sampling
AttributeError: 'VAE' object has no attribute 'clone'

[serene hatch]

can anyone help me figure out how to fx this????? please.

*** Error loading script: reactor_faceswap.py
Traceback (most recent call last):
File "D:\AI\Stable Diffusion\stable-diffusion-webui\modules\scripts.py", line 469, in load_scripts
script_module = script_loading.load_module(scriptfile.path)
File "D:\AI\Stable Diffusion\stable-diffusion-webui\modules\script_loading.py", line 10, in load_module
module_spec.loader.exec_module(module)
File "<frozen importlib._bootstrap_external>", line 883, in exec_module
File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
File "D:\AI\Stable Diffusion\stable-diffusion-webui\extensions\comfyui-reactor-node\scripts\reactor_faceswap.py", line 13, in <module>
from scripts.reactor_logger import logger
ModuleNotFoundError: No module named 'scripts.reactor_logger'

---

*** Error loading script: reactor_swapper.py
Traceback (most recent call last):
File "D:\AI\Stable Diffusion\stable-diffusion-webui\modules\scripts.py", line 469, in load_scripts
script_module = script_loading.load_module(scriptfile.path)
File "D:\AI\Stable Diffusion\stable-diffusion-webui\modules\script_loading.py", line 10, in load_module
module_spec.loader.exec_module(module)
File "<frozen importlib._bootstrap_external>", line 883, in exec_module
File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
File "D:\AI\Stable Diffusion\stable-diffusion-webui\extensions\comfyui-reactor-node\scripts\reactor_swapper.py", line 18, in <module>
from scripts.reactor_logger import logger
ModuleNotFoundError: No module named 'scripts.reactor_logger'

[ornate elk]

You have the Reactor comfyui extension installed in auto1111

That wont work

You need the normal Reactor Extension

[serene hatch]

how to delete without messing up anything?

[ornate elk]

Delete the Extension from the Extensions folder

That should be enough

And then read the install guide for Reactor

On github

[sand dawn]

CS1o you are a treasure to this community, in case no one has mentioned that. Every once in awhile I find myself here and you're always helping someone

[ornate elk]

Thanks for the kind words 🙂 always here to help if I can ^^

[serene hatch]

@ornate elk you are a lifeline. appreciated.

[sand dawn]

Blasting out a question regarding using ControlNet via A1111's API.
When using the Web UI, when user clicks one of the "Control Type" buttons it selects an appropriate default Preprocessor rather than "None".
Is there a way to get an appropriate default preprocessor for a given control type / Model selection via API?
Checking before I go making feature enhancement suggestion...

I've tried sending a typical payload without a 'module' key but it simply errors, would be sweet if it would just default to something

[river flame]

Reactor gives me bad results. FaceSwapLab doesnt see the face bit somehow makes the result non-clowny, which I want. After restart ControlNet doesnt see the image again. I noticed that it recommended to reinstall torch for 2.1, I did it and now I get this

and now controlnet works

[sand dawn]

idk if you switched to/from the dev branch (as I did for 1.7RC), but I had other random errors saying to reinstall torch, reinstall xformers, which I did, but I still had bugs until I said screw it and just reinstalled A1111

Otherwise I've never had problems with Reactor/Controlnet

[ornate elk]

Did you installed insightface like stated in reactors guide?

[river flame]

no, it only says to do it for SD Next

I tried making a checkpoint face in Faceswaplab, no errors this time, but the result is the same as without faceswaplab

[ornate elk]

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

thats the install instructions of reactor for auto1111

[river flame]

yes, I did that, but it doesnt say anything about insightface

[ornate elk]

the steps are a preparation so that isightface can be build

you can also install it manually by following this guide:
https://github.com/Gourieff/sd-webui-reactor?tab=readme-ov-file#installation

you may also need to disable faceswaplab to get reactor to work

[river flame]

it works, I just dont like the result

[ornate elk]

reactor or faceswaplab?

[river flame]

faceswablab sees the image perfectly with realisticVision checkpoint

[river flame]

reactor

[ornate elk]

ah okay

[ornate elk]

you could also try controlnet with ip-adapter for faceswap

[river flame]

I always use it, without it faceswaplab is bad

[ornate elk]

do you use face id or instant id or facefull?

[river flame]

face-id

[sand dawn]

ended up making that c-net feature enhancement suggestion I mentioned earlier https://github.com/Mikubill/sd-webui-controlnet/issues/2679

[ornate elk]

face id plus v2 works good

[ornate gorge]

any way to fix black screen withouth using -no half command on broken models?

[languid tinsel]

is img2img broken? I'm getting so many issues

it worked fine and now almost all my models spit out black images or garbage

ok screw this I'm rolling back...

[ornate elk]

make sure the models are larger than 2gb, otherwise they are not models

[ornate gorge]

6.46 gb, re downloading it currently with yaml files as second file (if i correctly remember yaml goes to same folder as model itself for auto?)

[ornate elk]

yes

whats your gpu?

[ornate gorge]

4080 super

16g vram

[ornate elk]

do you use --xformers --no-half-vae?

[ornate gorge]

no half slows the shit out of generation so im avoiding it i do use xformers command tho

[ornate elk]

i mean --no-half-vae

--no-half is not needed for your gpu

[ornate gorge]

vae will not impact generation speed? ill let u know once i got model and try without and give feedback if still does not works i will use this command line, thanks for trying to help

[ornate elk]

make sure your webui is updated too

[ornate gorge]

should be up to date, i did reinstalled it like few weeks ago

[ornate elk]

also normaly sdxl models dont need a yaml file, if they do, they are likely bad merged models

[woven canopy]

I have 12gb vram but I still get a CUDA: out of memory error when I try to run the clip interrogator on "best." On top of that, I'm able to run the interrogator on "fast," but after that I get an out of memory error in txt2img. Does clip interrogator really use that much memory, or is there some kind of glitch?

[glad fiber]

how can i convert this to a single safetensors file?

[ornate elk]

hey, whats in your webui-user.bat?

[rotund zinc]

this is happening with every gen, any ideas?

[woven canopy]

@echo off

set PYTHON=C:\Users\pmax\AppData\Local\Programs\Python\Python310\python.exe
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--listen

call webui.bat

[vocal burrow]

Looks to me like you download the whole repository of whatever that is instead of getting just the trained model.

[vocal burrow]

incorrect VAE most likely

[ornate elk]

ah okay, you should add --xformers

then it should work

[glad fiber]

the problem is that there is no final trained model

only the repository

[vocal burrow]

most likely there is

what's the repo

[upbeat crane]

@ornate elk gunna get a "what's in your webui-user.bat?" tattoo

[vocal burrow]

yeah we should get a macro on or keyboard ready for that....

Or "what's the full console log ?"

[upbeat crane]

ikr? lol

[glad fiber]

https://huggingface.co/monadical-labs/minecraft-skin-generator

[ornate gorge]

--no-half-vae fixed it for me, thanks straw hat

[ornate elk]

perfect, np 🙂

[river flame]

I really like InsightFace+CLIP-H (IPAdapter) preprocessor, but I dont have it in A1111, only in Forge, with doesn`t work for me. Cant understand how to copy it to A1111

the controlnet folder is the same

oh I see, there s no ControlNetPreprocessor folder in A1111

can I just copy the folder?

[ornate elk]

its in the extensions folder of controlnet

[river flame]

what you mean? there are no folders in the controlnet folder

[ornate elk]

in auto1111 the preprocessor are stored here:

[vocal burrow]

oh yeah those are diffusers

you need to convert them

[glad fiber]

how

[vocal burrow]

I'll check if there's a tool for it when I get back home soon. Otherwise it will be command line stuff that I don't remember at the top of my head

[river flame]

I moved it, reloaded A1111, but cant see it. Did I move it wrong? From here to here?

[ornate elk]

make a folder and call it clipvision

then put it in there

cant guarantee that will work, but worth a try

because i have this folder

[glad fiber]

this script did the trick https://github.com/huggingface/diffusers/blob/main/scripts/convert_diffusers_to_original_stable_diffusion.py

[vocal burrow]

lol just finished testing with this https://arcturusforge.itch.io/diff-to-ckpt

[glad fiber]

and i am on linux

so it would not work for me

[vocal burrow]

for sure

there's this one too just in case https://gist.github.com/jachiam/8a5c0b607e38fcc585168b90c686eb05

[glad fiber]

it is literally the same i used lol

but thanks, anyway

[river flame]

no it doesnt work. I shouldnt rename those to .pth right?

[vocal burrow]

lol looks like someone stole it then, # Written by jachiam

they're not the same, 235 vs 353 lines

[ornate elk]

these diffusion_pytorch_model files wont work,
if you tried to use Instant-ID you have to rename them correctly

[river flame]

I dont know what are the correct names

[ornate elk]

source: https://github.com/Mikubill/sd-webui-controlnet/discussions/2589

[river flame]

thanks

[river flame]

I did everything like there, it doesnt work. whatever

[ornate elk]

you need to switch the units

[river flame]

no ,same error

[ornate elk]

an yo show your settings?

[ornate elk]

like i said you need to switch the units

Unit 0:

Unit 1:

[river flame]

that s what I had when you said to switch the first time)

[ornate elk]

okay then switch back again and for unit 2 select None

[river flame]

same

[ornate elk]

did you renamed the files correctly?

[river flame]

.

[wide musk]

im trying to train my model more. i did it already. but it doesnt let me. it says my model is dead. its actually exactly what it says. i train on dreambooth. it was working perfectly and i dont know.... it not lets me continue train it

[wide musk]

never mind i found a solution

BRUH NO

]WARNING:dreambooth.train_dreambooth:Loss is NaN, your model is dead. Cancelling training.

HELP WITH THIS ALREADY

[hollow quartz]

Hi, Stable Diffusion noob here, lets say I want to create 20 different images based on 20 different prompts. Is there a way to queue these 20 different images, so that 20 different images are created one after the other automatically, without having to input extra stuff, or can I only input a new prompt after I have finished creating a single image?

[nocturne topaz]

depends on your user interface

comfyUI can do it, a1111 cant

[ornate elk]

Yes that can be done with an extension:
https://github.com/ArtVentureX/sd-webui-agent-scheduler

[random pagoda]

is there any way to set higher res than 2048?

i wann use kohya highres fix, and i cant select more than 2048

on the sliders

[gloomy crown]

Is this the right place to ask about
DOCKER BUILDING
if anyone do know about how to build one i have few questions
i want to build an image for my paperspace, it uses comfyui, it installs some packages and stuff
when building the image, should i be on the linux machine or my local machine ?
SINCE, Uhm, i cant build images using the paperspace machine cuz of some fuckery they thought it was genius (not being able to restart the machine)
so if anyone kind enough to help, i'd appreciate

---

TL;DR
HELP !!! Docker image build for paperspace (to run ComfyUI)

sorry for the

https://tenor.com/view/imostergolfra-gif-25274972

[fair oxide]

im using runpod and control net wont show ipadapter

its on v1.1.313

[woven canopy]

thanks for your response! I actually do have xformers activated, I just use a separate program to keep a1111 updated and I guess it's stored in a different file:

[fair oxide]

also when i upload the ipadapter models in the file, they dont show in the stable diffusion page

[woven canopy]

that wasnt a reply for some reason

[abstract badge]

hoping someone can help me. ControlNet not working as intended- no sketch output picture, and it doesnt give me the same posture. what am i doing wrong?

[dry lintel]

Guys, I can no longer use the stability platform. I tried everything other browser, log in again, restart browser, delete cookies, turn off ad blocker... but nothing helps, I just get the message failed to fetch.

Any solutions?

[fair oxide]

Waiting, I guess
https://stabilityai.instatus.com/cltf4c3hd204800b6ob2n1f33kn

[lost chasm]

my pip cache folder in appdata now takes up 25.9GB on disk, anyone know if I can delete this shit? Also huggingface folder is now 89GB

[dry lintel]

Thank you

[fair oxide]

controlnet doesnt show any updates in the extension tab

how do i update it?

[upbeat crane]

make sure you are putting a check in the ALLOW PREVIEW option next to PIXEL PERFECT if the sketch preview not showing

and try the option below CONTROLNET is more IMPORTANT, instead of using the balanced option

[abstract badge]

im getting the 'sketch' but its just a replica of my sample image which did nothig for me

im sorry i am new to this

[hollow quartz]

Thanks a lot for the answer, I'll be sure to try that out! 🙂

[upbeat crane]

sorry i was giggling at the picture on the right, ur good bro

[abstract badge]

np, i also utilized chatgpt to help me understand the cmd server log and it mentioned 'basicsr' package wasnt installed and so i got that installed and i believe because of that im getting the second image but it still isnt a sketch

also i have followed all of the youtube guides to a T but its as if im missing something because they are getting these results and im not

[upbeat crane]

have u tried any other controlnets?

[abstract badge]

i havent, i figured i should probably figure out this basic one before i can even concieve of how they work

[upbeat crane]

give canny a shot

canny works with the edges of the image

[abstract badge]

maybe the problem is in my files and im not seeing it. i didnt get the 'cldm_v15 & 21' during any of the installs so i manually installed it but idk where to place it

[upbeat crane]

https://stable-diffusion-art.com/controlnet/

[abstract badge]

ok ill take a look. thank you

[silent cedar]

In the park, a scene of spring, a simple and elegant Chinese woman, aged 35, is taking a walk in the park

[abstract badge]

oh yes canny works with the edges, thats great ill see how this pans out. but the route im trying to go to will be needing to have a 3d pose tool that i can mose around all the joints to create the specific poses of my characters

[upbeat crane]

nice yeah worth a shot you can run multiple controlnets as well so worth trying canny and openpose if canny doesnt help out

but where this chinese woman is going....is the question 🕵🏻

[young bridge]

Hey guys

I need help on how to install IP Adapter control net

[dusk condor]

Why I can't start my webUI

The logs (I only captrured the error part):

logs

[gentle anchor]

im on this step of the AMD installation guide, im on a 6600 and i have no idea where to find %Hip_PATH%bin

[ornate elk]

Its C:/Programms/AMD/Hip5.7/Rocm/bin I guess

[ornate elk]

Did you installed HIP SDK?

[gentle anchor]

yes

wait one moment

[ornate elk]

You can also open up a cmd and type:
where hipinfo.exe

[gentle anchor]

im a little special

[gentle anchor]

thanks

[ornate elk]

Np, did it worked?

[drowsy wigeon]

I'm having and issue with the Trimesh module and tried laready 100 things, but whatever I try

it doesnt work

https://www.reddit.com/r/StableDiffusion/comments/1b72jh5/comment/kti4raq/?utm_source=share&utm_medium=web3x&utm_name=web3xcss&utm_term=1&utm_content=share_button

It keeps saying that it can't find the module Trimesh when running comfyui nvidea

I worked with an conda env

If I do pip list or run python and do an import, the module can be found. I'm working on a windows11 environment

It happens when it runs this file and trying to do an import trimesh

So I'm out of ideas :/

I just found out that comfyuio is using this python: (myenv) E:\ComfyUI_windows_portable\python_embeded>python.exe and this one cant find trimesh. Is there a way to fix this?

[dense obsidian]

trying out animatediff for automatic1111, wondering if anyone has practice with it they could help me try and figure out how to get it started? I keep getting:

AttributeError: 'NoneType' object has no attribute 'mode'

[worn frigate]

cs1go

bro

@ornate elk

i need your help urgently, i am literally shaking and about to cry

I have no contact with developer and i do not understand much. I installed something called keygen and when I press remove on windows defender, nothing happening

i know its not related to Stable diffusion but only a developer can help me. i was trying to install and it said i need to run patch

when i click remove or quarantine nothing happens 😭

please help me

[manic snow]

a keygen? what do you need a keygen for?

[worn frigate]

it came with software

it says It will patch, and it did patch software

and it worked like it promised

also I deleted it from pc as well

but it says this temp/dup2patcher.dll thing is not going anywher

[manic snow]

hate to break it to ya, but that's illegal. keygens are designed to "crack" software. where did you get this program?

[worn frigate]

getintopc

😭

it did nothing, it did not harm

it worked too

but its not going

it says this everytime i run quick scan

then i go to protecttion history

[worn frigate]

and it says this

pls help i am literally crying here sitting on my chair

i really can not afford another pc'

i had saved so much to buy it

[manic snow]

i have one recommendation, and if that doesn't work you're on your own.

besides, you don't need to buy another pc if this doesn't work, you can just reset windows

[worn frigate]

what i do

pls help

[manic snow]

there's a script called "TronScript". do some googling and you'll find it. the main resource is on reddit. it should remove the malicious file, along with things that aren't necessary such as OneDrive if it isn't in use and other bloatware

[worn frigate]

but does it require knowledge of codes?

what it is i have opened its reddit

[manic snow]

it's pretty much just a batch file that you run. it can take a couple of hours to complete tho

[worn frigate]

will it wipe my pc?

[manic snow]

NO

[worn frigate]

i have many projects saved

of after effects

[manic snow]

it will simply detect things that aren't supposed to be on your pc, e.g. that keygen, and remove it

[worn frigate]

hm

it seems it will download many things first

like i do not have torrent

i really do not understand how to perform it

[ornate elk]

I would start with ADWcleaner by malwarebytes

[manic snow]

right, completely forgot about that

[worn frigate]

so i download adwcleaner

then run it okr ight?

is it paid

[ornate elk]

Its free

[worn frigate]

i downloaded setup and it says smart sreen run dont run

[ornate elk]

Run it and everything it finds can be deletet or put into carantine

[worn frigate]

ok i am trying rn

[ornate elk]

You can also screenshot what it finds

Then we can take a look

[dense obsidian]

wasn't aware you guys were knowledgeable on other pc techy stuff. Gonna have to remember this stuff if I need assistance