🤝｜tech-support

[proper swallow]

I am having issues with Forge UI: error connection errored out. I cant generate anything.

I have tried to use export COMMANDLINE_ARGS="--no-gradio-queue" according to this thread: https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/9074

(!) This solved it, Make sure to reboot PC after the change: https://github.com/lllyasviel/Fooocus/blob/main/troubleshoot.md

[ornate elk]

Your following an old guide. Checkout the first link of the pinned messages for a better AMD Guide

You need to adjust the path

Idk where you want to store the models so I put in an example

[solemn atlas]

I copy paste the models folder in my ssd and delated the original folder from my webui

I did what was done in the youtube tutorial

Now I want to do what is in the tutorial https://www.youtube.com/watch?v=9UU3EuZDDVg It's 2:09

but it doesn't work. It doesn't create that "folder" in my webui

It says: You do not have sufficient permissions to perform this operation.

[hybrid kindle]

which guide should i stick to? i have rx6600 so i need zluda, but which guide is more relevant

[proper swallow]

Anyone have a good way to upscale in img2img,
using Forge UI, stable diffusion? In old stable diffusion I used latent upscaler, which I cant find in Forge UI.

[hybrid kindle]

ok now this is what im getting

[ornate elk]

Automatic1111 with ZLUDA

[hybrid kindle]

yes

[ornate elk]

Internet connection error

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

[potent oyster]

trying to do some inpainting in comfy, wth is happening here

[potent oyster]

(on the right is the image processed using lama cleaner, left inpainted) trying to use the cleaned image with controlnet to guide the inpainting

[hybrid kindle]

idk im keep getting same error every time

i'll keep searching for solution

[ornate elk]

You may need to change your DNS server then

Or sometimes a VPN can help

[hybrid kindle]

i tried with vpn

and without

proxy vpn

[ornate elk]

Try change the DNS

And no VPN

Maybe you need to git clone the repo again too

[zenith seal]

Hello ! well im here cause today my auto01 isn't responding anymore

like it was working prefectly yesterday

also when im in the ui prompt is not recognised it always show 0 /75

also can't press on the gen

my full cmdd

[viscid badge]

I tried to render something but it render me these kind of weird and bad image everytime for no reason

When I render a picture with my PC it take 15/20 second by image but here it take only 2 seconds and gave me this

and it happen with all the model I have

[onyx solar]

You are using an XL model, but are only set to 512x512. Change to 1024x1024.

[ruby yoke]

Did you get the problem solved?

[craggy epoch]

Anyone have a good systemd script so I can launch SD on reboot. Linux Mint 24, everything is in it's own directory (/ai/automatic1111 or /ai/comfyui or /ai/forge) I know I'd need a script for each, and the examples I keep seeing are running into their own issues. I can create a user specific for each instance of SD I want to run, if that helps.

It's on a headless (well, unless I plug a monitor in) machine with an nvidia card. I have it running without issue if I ssh to the box and start it with the --listen --enable-insecure-extension-access switches (baked into webui-user.sh)

[violet stump]

Hello. Whener I try to use another model other than the default one "v1-5-pruned-emaonly", it generates gray images. I already deleted venv and ran it again but it keeps doing this

[hasty orbit]

Is it a Pony model? Cause I had something similar where if you don't give it scores (i.e. score 9, score 8, score 7, etc.) like how it's supposed to be used it will generate blank images like that

[violet stump]

it's a Lora

[shadow thicket]

Hello, I'm currently planning to build a server for image generation. I'm not sure which graphics card would be more suitable, the RTX A4500 or the RTX 4090. Is there a significant performance difference between these two cards?

[zenith seal]

i have this when instaling auto

[ornate elk]

Loras goe into the models/lora folder

[ornate elk]

What python version did you installed?

[violet stump]

but there is no Lora folder here

oh nvm you are right

it's in models/Lora

but i can't find my lora models

[pale salmon]

1 — Go to the "Settings" menu.
2 — Click on the sub-menu "Extra Networks".
3 — Scroll down and click on the option "Always show all networks on the Lora page".
4 — Click on the "Apply Settings" button (at the top of the page).
5 — Go to your Extra Network tab, click the "Refresh" button

try this

[ornate elk]

Loras are not models

Loras ar on the lora tab in txt2img

Loras cant generate images on their own

They need a model

[zenith seal]

the 3.10.6

[hasty orbit]

How do I stop Stable Diffusion from naming files via the full prompt? I have the "Max prompt words" scale set to 1 but still generating files with the long names

[violet stump]

i can't find my lora yet

[pale salmon]

click here

[violet stump]

nothing appears

[pale salmon]

did u put something in ur lora local folders ?

[violet stump]

yes, the lora i downloaded

it has .safetensors extension

[pale salmon]

u put it here ?

[ornate elk]

can you show me the full cmd log?

[violet stump]

yes

[ornate elk]

they only show up when a compatible model is selected

[pale salmon]

which model is ur lora for

[ornate elk]

so if you download loras for SDXL, you also need to use and SDXL model

same for every other base model

[violet stump]

i'm using v1-5-pruned-emaonly as my model rn

should i use another one?

[pale salmon]

ur model is an SD 1.5

what is the name of ur lora

[violet stump]

it's a nsfw lora lol

[pale salmon]

i do nsfw its not a problem lol

just tell me

[ornate elk]

check the civitai site of each lora you download to see what its compatible with

[pale salmon]

"base model"

on ur lora

[hasty orbit]

Found the fix

[pale salmon]

what is urs

[violet stump]

i'm using this one a "Raven Shadbase" lora

[pale salmon]

from alan rex ?

send me the link in dm

[violet stump]

yes

[pale salmon]

a

sd 1.5 too

weird

maybe ur model is sd1.5 too, but sometime it doesnt match

check on github

[ornate elk]

whats your gpu btw? this log looks normal. when do you get the errors?

[violet stump]

i'm gonna try using Counterfeit-V3.0

[ornate elk]

have you relaunched the webui ?

after moving the loras

[violet stump]

let me try

[pale salmon]

press on refresh

webui

[violet stump]

nvm i did do that but i think it's the model i'm using

[pale salmon]

try with counterfeit

• counterfeit is better for anime lol

cartoon too

[umbral compass]

Don't remember if this been mention or posted, but should Forge | Weight of GPU be lower by 500 MB? | In my case 23,539 to 23,039?

[ornate elk]

if your using flux you can try lowering it down by like 2gb

but yea dont use max vram

[umbral compass]

Okay, I been running generation and testing it with Flux, I had to get the Prun version because using Checkpoints the one with 20GB freezes my computer and lags, I guess that's probably why.

[violet stump]

it is "working"

[ornate elk]

Yea but its not recommend to use the 23gb model anyway ^^ better use the fp8 16gb one or the nf4 flux

[zenith seal]

im on a 6800xt

i just did a full reinstall

now i have this

when loading a model

[ornate elk]

Oh and you use the directml Backend

Not recommended but works

If you want to get a much better performance checkout my AMD install guides from the link in the pinned messages

[ornate elk]

Thats one time only

[umbral compass]

Another question I have is when I first started using Flux on forge it didn't work even with the fp8 16gb model with (29be1da) -

But when I updated Forge it works, do you know why Flux doesn't work with F0017.7z (f92be1da) unless it updated?

Also, is the a difference in running the Run.bat file and "webui-user" in the webui folder?

[ornate elk]

because the old forge has no flux support

its a version from february

always start the webui-user.bat

[umbral compass]

Thank you, I was following this direction and other on youtube to do "Run.bat"

Should I do Run.bat first then webui-user.bat?

[thorn sage]

How can I tell if I'm using SD3?
version: v1.10.1  •  python: 3.10.6  •  torch: 2.1.2+cu121  •  xformers: 0.0.23.post1  •  gradio: 3.41.2

[ornate elk]

oh sry, for forge you can run the run.bat as it directs to the webui-user.bat anyway

[ornate elk]

SD3 is a model (checkpoint). Your programm is a so called Stable Diffusion Webui, its on Version 1.10.1
they are different things.
For SD3 you would need to download the SD3 model and put it into the models/stable-diffusion folder and then select it in the webui

[thorn sage]

Gotcha, I was trying to use the update BAT assumed that would work. Thanks!

https://huggingface.co/stabilityai/stable-diffusion-3-medium

Is this it?

[ornate elk]

yes

[thorn sage]

Sorry should I go with the smallest one? or biggest

[ornate elk]

the 10.9gb one

if you have a 4080 or 4090 you can also use the big one

[urban girder]

does anybody know how to install stable diffusion? I'm trying to find the right website.

[thorn sage]

What if I have a 4060 Ti? 16gb
I appreciate you

[brittle lantern]

guys i went to open my auto1111 and its really slow on my 6600xt out of nowhere on SDXL models how could i fix this it was not this slow with zluda

[thorn sage]

I used this guide, https://stable-diffusion-art.com/install-windows/

[brittle lantern]

possibly found the issue

[ornate elk]

can you show the settings you tried to use in txt2img=?

[ornate elk]

you need to launch the webui-user.bat

[brittle lantern]

ah ok

[ornate elk]

never launch the webui.bat

or it wont use zluda

[brittle lantern]

ah got it

much faster ty

should i do a zluda update

idk if that increases speed?

[ornate elk]

checkout my Guides:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides

[ornate elk]

i would say not by much, maybe a bit better cuda support was added

but doesnt hurt to update them

[brittle lantern]

i see how would i update it without messing up anything

[ornate elk]

download the latest Zluda release build for either rocm5 or rocm6 depending if you installed hip sdk 5.7 or 6.1
Then replace all files in C:\Zluda and delete the old cusparse64_11 and cublas64_11 dll and then make new copies with the new files

[brittle lantern]

so if i have sdk 5.7 i need rocm5 build?

[ornate elk]

yep

also make sure your webui is updated

[brittle lantern]

should i do a 6.1 update also or stick with rocm5 build?

[ornate elk]

you can do that if you want. as 5.7 will be not supported in the future. (but nobody knows when)

but for the switch you also need compatible rocm files for your card.
these can be downloaded 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

you then also need to uninstall hip sdk 5.7 and install 6.1

and then also downloading the rocm6 zluda files instead of rocm5

[brittle lantern]

ok so i uninstall 5 then i download rocm6 sdk from AMD then whats after?

[ornate elk]

then you can follow my guide again at zluda setup Step 3 but with the rocm libs i linked you

[brittle lantern]

ok

[ornate elk]

the rest is the same just that you have to do it for 6.1 instead of 5.7

after all you also need to delete the .zluda folder inseide the stable-diffusion.webui folder

[steel garden]

Hi I'm trying to face swap a photo which has used a lora to create the facial expression. Every time I use Reactor to face swap it overwrites the lora and the facial expression is just bland. Does anybody know how to fix this?

[ornate elk]

hey, you would need to use controlnet + IP-Adapter Face to get expressions + faceswap working

what you still can try is to reduce the codeformer strenght in reactor first to test

[steel garden]

So it is possible to do it then?

[zenith seal]

yep so i did follow the steps for zluda and now i have this

[ornate elk]

can you show me the full cmd log?

[ornate elk]

thats not the full log

scroll up and mark everything to copy

[steel garden]

I have increased and decreased the strength and still not working

[ornate elk]

do you have a model in models/stable-diffusion?

if not place there a model

[zenith seal]

yes

[ornate elk]

have you installed hip sdk 6.1 ?

and how do your environment path settings look ?

[zenith seal]

yeah i did for next

[ornate elk]

looks okay

have you restarted the PC one time after installing hip sdk?

[zenith seal]

doing this rn

[zenith seal]

i instaled hip 1 week ago

[ornate elk]

okay then wait ^^

[zenith seal]

like i folow the pin guide but skiped the hip part

cause it was already installed

[ornate elk]

thats okay

then just wait until the cmd shows something new

the zluda process part can sometimes take 15-40 minutes

on time only

[zenith seal]

i will see thanks !!

anyway i will send a screen when finished

[ornate elk]

alright 👍

[ornate elk]

yea then i would go the IP-Adapter route maybe

[steel garden]

Okay, Do you know of any guides that would be able to help me set it up?

[ornate elk]

maybe this:
https://www.youtube.com/watch?v=_A7PraBHyg0

[steel garden]

I'll give it a try thank you

[thorn sage]

Will Flux models work with SD A1111?

[ornate elk]

currently not, but in future i guess

Flux works in Webui Forge and ComfyUI and sdNext

[thorn sage]

Which one do you personally use? Not just with Flux

[ornate elk]

I primarily use Auto1111

Best Extension support

[brittle lantern]

cmd is showing compiling in progress when generating thats fine correct

[ornate elk]

yep

[brittle lantern]

this does once or?

[ornate elk]

yep only one time and it can hapen again after a torch reinstall or gpu driver update

[brittle lantern]

ok so this is the 45 min first gen thing?

[ornate elk]

run this inside a cmd in the stable-diffusion-webui folder
venv\Scripts\activate.bat

you dont need to use cd

go into the stable-diffusion-webui folder.
Click into the file explorer bar and type cmd.
Then hit enter and in the cmd run the command

ohh okay

but the command should be the same only the you need to activate the activate.sh

instead of bat

you only have to run activate.sh in the scripts folder

[brittle lantern]

@ornate elk i have 4 to 5 it/s now on rocm6

[ornate elk]

ouh nice, what you got before?

[brittle lantern]

2 to 3

[ornate elk]

wow okay

seems like i will update my guide for 6.1 for older cards then too

just type
activate.sh or try
sh activate.sh

[zenith seal]

it's a bit long no?

[ornate elk]

for mac stuff i can refer to @karmic crown, he knows everything about mac and SD

[ornate elk]

yea it should have launched by now

try close and relaunch the webui-user.bat

[zenith seal]

i do a gen now?

nvm

[ornate elk]

yep

let me know what happens then

[zenith seal]

okey nice

[zenith seal]

i will

[ornate elk]

looks good

now just wait

[zenith seal]

okey

only the first gen gonna be slow right

[ornate elk]

sry for the wating time but it could be another 15-35mins

after that an image takes seconds 🙂

[zenith seal]

thats would be dope ahah

img to img also?

[ornate elk]

yep

[zenith seal]

and more stable them next

[ornate elk]

it always depends on model, resolution etc, but everything should be much much faster

and yea better useable than next xD

[zenith seal]

yeah next make me crash a lot

this is normal right

[ornate elk]

Is the generate button still grayed out?

And shows interrupt

[zenith seal]

yep

[ornate elk]

Looks good

[zenith seal]

well it's working but during the img to img my mouse was no moving :/

[ornate elk]

Make sure you don't run any games and Also not Wallpaper engine in the background

[zenith seal]

thats maybe why

[ornate elk]

Probably. Right click and Exit it. Pause it won't help

[zenith seal]

yeah it did exit

buut didn't know that this is causing the frezze

[ornate elk]

Wallpaper engine and SD dont like each other xD together they can even crash the PC

[zenith seal]

that whats just happen lol

working like a charms now i think

[ornate elk]

Perfect, no problem

[zenith seal]

nvm ahahh

[ornate elk]

What did you tried to do

[zenith seal]

it worked on direct

but it crash when i go at 90%

[ornate elk]

Resize by takes up a lot of vram and ia not the recommended way to upscale an image

[zenith seal]

whatss the best way

[ornate elk]

In img2img you can use the SD upscale script at the bottom.
Then set the "resize to" resolution to either 512x512 or 1024x1024 depending if you use 1.5 or sdxl model.
(Its not the end resolution).
Then set the denois slider below 0.3 and select an upscaler at the bottom

[violet stump]

how can i use xl models if i have 8gb of vram but it still crashes?

i tried setting my setting to medvram like this guide says https://rentry.co/2rq2n#-running-on-4gb-and-under-

[ornate elk]

What's in your webui-user.bat?

You only should use:
--xformers --medvram-sdxl --no-half-vae
And nothing more in the commandline_args=

Then sdxl models should work.
Also if you only have 16gb system memory you need to increase your Windows Pagefile

[violet stump]

i'm using linux

this is my webui-user.sh file

[ornate elk]

Yea then same thing add only the three commands I mentioned there

Then save and relaunch

[violet stump]

got it, let's hope it helps 🙏

[zenith seal]

damn it worked !!!!

tysm your are the best ❤️

[ornate elk]

Linux and SD need a big swap file for sdxl (32gb)

[ornate elk]

Thx, no problem 🙂

[violet stump]

yes, i'm noticing my swap file is around 4gb

how can i make it bigger

[ornate elk]

Idk sry

[violet stump]

it's all good

[ornate elk]

I'm off for now gn

[quaint idol]

Hi Im getting an error with cuda 11.6 im using visual studio 2022 and an RTX 2080ti im new to using cuda so am I using the correct version 11.6 for SD or should I use 12.0?

2024-09-22 16:04:58.1913869 [E:onnxruntime:Default, provider_bridge_ort.cc:1992 onnxruntime::TryGetProviderInfo_CUDA] D:\a_work\1\s\onnxruntime\core\session\provider_bridge_ort.cc:1637 onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : LoadLibrary failed with error 126 "" when trying to load "C:\Users\Shaun\Desktop\AI Program\stable-diffusion-webui\venv\lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll"

I tried putting a set path to the dlls to the cuda toolkit but webui says python cant load

[quaint idol]

...is pytorch needed for cuda on win 10?

[ornate elk]

Hey you need python 3.10.11 64bit and cuda 11.6 is to old.
At least 11.8 is needed. But you don't need any manual cuda installation. The webui will install the right pytorch version with the cuda version compatible with your GPU driver.
So make sure that's updated too

[steel garden]

Hi I'm having issues face swapping two people using Reactor. I have set the source image as 0, 1 and target as 0, 1 but Reactor always says ReActor - STATUS - Swapping Source into Target
08:34:43 - ReActor - STATUS - Detecting Source Face, Index = 1
08:34:43 - ReActor - ERROR - Age Detection: No face with index = 1 was found
08:34:43 - ReActor - ERROR - Gender Detection: No face with index = 1 was found

[charred rover]

Hi! I have a 2080ti. How much more inference speed would I get with 2x 2080ti (nvlink)? how much with 4x 2080ti (2x nvlink)?

[ornate elk]

Stable Diffusion doesn't support nvlink

[charred rover]

then why they recommend it on forums

[ornate elk]

Where and who?

[charred rover]

I can give many links and chatbots also say so

[ornate elk]

Didnt found anyone that got nvlink working with stable diffusion

And a chatbot isnt a prove for aynthing

What you can use are two instances of a webui.
For each card one instance to generate images simultaneously

And LLMs work with nvlink. What also doesn't mean stable diffusion works with it

[charred rover]

'You can use swarmUI and queue up a second generation with the other gpu at the same time'

[ornate elk]

Yea that's not nvlink

Thats using 2 GPUs separately

[charred rover]

sorry, so it is reasonable to buy 4x 2080ti right?

[ornate elk]

Swarm can use multiple backends and is good for multi GPU systems

[ornate elk]

As a 2080ti only has 11gb vram and Vram is the most important stuff I dont see a benefit here

[charred rover]

I dont see how vram is so important. this is two times vram difference (inference speed - puget systems)

[ornate elk]

Vram is the most important stuff when it comes to ai tasks.
For example run a 23gb Flux model with a 24gb 3090 or a 10gb 3080 and it will outperform it by a lot

More vram = bigger models work faster

[charred rover]

but thats for flux, I dont use it

[ornate elk]

Yea that was just an example

An other thing is that more vram can create higher native resolution

Or upscaling too

[charred rover]

so for sdxl I think I can do better resolution? but it ll still take more time right so..

I can just upscale it on 11gb right

[ornate elk]

Yes you can upscale on 11gb

[charred rover]

so no need for 24 gb vram

I m buying 4x 2080ti I guess

[ornate elk]

Depends on your use case. Why do sou want 4x 2080ti ?

[charred rover]

to have 4 images generated in time of 1

[ornate elk]

You would need to test how long a specific image takes on a 2080ti because with more vram you can also use batch size which means "generate images at the same time"

[charred rover]

just like swarm?

[ornate elk]

Swarm can do that too

[charred rover]

it s 2.15 it\s for me currently

but I want to install tensorRT

[ornate elk]

That can improve the performance a lot but also limits the extension compatibility

[charred rover]

can you explain a bit please

[ornate elk]

Here is also a benchmark for 1.5:

[ornate elk]

Some Custom Nodes may not work when using TensorRT as they are coded for the base cuda stuff.
Also TensorRT needs manual model conversion

[charred rover]

I see thank you. maybe I ll consider 4070ti 16 gb instead of 2080ti

oh no they are expensive nvm

[ornate elk]

Yea 4070ti is crazy expensive

[karmic crown]

if you still have issues let me know, but i need to go to the hospital tomorrow, so i will probably not be able to answer in the next 2 days

[mystic heath]

who is a good python developer svlbto

[charred rover]

I dont get it, is it a solution? https://github.com/comfyanonymous/ComfyUI/issues/162#issuecomment-2241298064 what you think?

[ornate elk]

Its a barebone code to make it work. It would need to be implemented in comfyui first to make it compatible with nodes too

[charred rover]

I see thank you so much

[charred rover]

I ll also need to switch intel to amd for more lanes. comfy does work with amd efficiently right?

[ornate elk]

Yep works flawlessly

[thorny dragon]

is 6gb vram enough for sdxl? 🙃

[uncut totem]

any tips for hanging, takes a very long time to start gen when I switch model then after first gen they fly

[ornate elk]

Yes, can work

[ornate elk]

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

[uncut totem]

4090 Trio X

@echo off

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

@REM Uncomment following code to reference an existing A1111 checkout.
@REM set A1111_HOME=Your A1111 checkout dir
@REM
@REM set VENV_DIR=%A1111_HOME%/venv
@REM set COMMANDLINE_ARGS=%COMMANDLINE_ARGS% ^
@REM --ckpt-dir %A1111_HOME%/models/Stable-diffusion ^
@REM --hypernetwork-dir %A1111_HOME%/models/hypernetworks ^
@REM --embeddings-dir %A1111_HOME%/embeddings ^
@REM --lora-dir %A1111_HOME%/models/Lora

call webui.bat

Im using forge if that makes a diff

[thorny dragon]

i am getting 9.14s/it speed on sdxl....is this normal? it took 4mins to gen a 1024x1024 img..i see ppl saying that it only takes them 30 secs for a img like that...

[ornate elk]

What's in your webui-user.bat?

[ornate elk]

And it hangs after the generation is done or only when switching models?

[uncut totem]

when switching

like I just ran one now and it fired off

[thorny dragon]

i am currently using invoke ai via stability matrix...so i dont know how to access my webui file in here...i tried a1111 before but it was slow like this too....

[ornate elk]

Can't help much with invoke but with 6gb vram you should use Auto1111 or Forge webui

In Auto1111 I can give you some tips to improve the performance

[thorny dragon]

forge works quite fine...but it has issues with some extensions...thats why i was looking for any other option...gotta stick with forge i guess....

[ornate elk]

Check if you can enable fp8 mode for sdxl in the settings

[ornate elk]

How much ram and which model do you use?

[uncut totem]

pc ram or vram?

PC= 128GB VRAM= 24GB

[ornate elk]

Okay and the model you tried?

[uncut totem]

flux1-dev-bnb-nf4

[ornate elk]

Looks like it doesn't trys to store models in the RAM, instead it loads them directly into the vram

[uncut totem]

could that have to do with swap location

cpu vs shared

[ornate elk]

Also make sure the models are on an SSD

[uncut totem]

they are on m2

ahh I think I got it I put gpu weight on max

[ornate elk]

Yep thats bad too

For flux keep it 1-2gb below max

[uncut totem]

should I have to install this every time I start??

Legacy Preprocessor init warning: Unable to install insightface automatically. Please try run pip install insightface manually.

[hybrid kindle]

how much time it will take? already like an hour

[ornate elk]

if you want to use faceswap stuff only

then oyu need to install it one time

[ornate elk]

relaunch and then try to generate again

[uncut totem]

Everything i launch it tells me that

[hybrid kindle]

without deleting venv folder?

[ornate elk]

yea dont delete the venv

[ornate elk]

yea because its not installed

if you install it one time into the venv correctly, it wont complain anymore

[uncut totem]

Ok I install in terminal says successful then asks every launch I'll try to see if it's just installing to the container instance or something

[ornate elk]

you need to activate the venv and then install it

[fair oxide]

what is this slowdown in iopaint? It seems that if I close it and reopen, time per iteration drops from 10 ~ 11 sec to 2s

[uncut totem]

kk ill try that ty

[hybrid kindle]

and i think it started compiling endlessly

OSError: None is not a local folder and is not a valid model identifier listed on 'https://huggingface.co/models'

FutureWarning: clean_up_tokenization_spaces was not set. It will be set to True by default. This behavior will be depracted in transformers v4.45, and will be then set to False by default.

i tried to generate something and it repeats Compiling in progress. Please wait...

[edgy dust]

how do i fix it?

[ornate elk]

that can be ignored

[ornate elk]

switch to a compatible model so they will show up

[uncut totem]

should I add the pip install insightface to my webbat?

[ornate elk]

no

[uncut totem]

sorry kinda a noob

[ornate elk]

your using forge right

[uncut totem]

yes

in docker

[ornate elk]

in docker

why

[uncut totem]

I have everything in one place

[ornate elk]

on linux or windows?

[uncut totem]

windows

docker desktop

[ornate elk]

okay wouldnt recommend that for nvidia stuff but okay ^^

then you have to run pip install insightface in your dockers terminal for the forge instance

[hybrid kindle]

what if i try use any other model? i have no clue what to do about infinite and repeating compiling, ill keep searching

[ornate elk]

whats your gpu?

[hybrid kindle]

rx6600

[ornate elk]

is it still compiling?

[hybrid kindle]

it repeats every time i click soomething in webui

if i try generate it sends me this

[ornate elk]

you have to generate one image first and dont click on other stuff meanwhile

so if the generate button is grayed out and showing interrupt, its fine, let it load

[hybrid kindle]

after some time it just repeats Compiling in progress. Please wait... but nothing

[edgy dust]

where do i change the model

sorry i'm a beginner

I got it here

I had to restart

thankyou

[uncut totem]

can forge spaces be integrated into the windows version?

[hybrid kindle]

omg it worked

[ornate elk]

Perfect 🙂

[hybrid kindle]

i just downloaded diffrent model and noticed lag spikes because of my ramn usage is 16gb out of 16

[ornate elk]

If you use sdxl models you need to increase your Windows Pagefile

[hybrid kindle]

wait can you tell me how do i know what im using rn, sdxl, sd3 or something else

[ornate elk]

Always check the models info box in Civitai.com
But you can also tell it at the file size.
2-4gb is mostly 1.5
6gb is sdxl/pony

[hybrid kindle]

so this one i downloaded is most probably sdxl

[ornate elk]

Yea and they need more ram or a bigger pagefile

To not lag

[edgy dust]

I need to create about 20 images for 1 that I like

[hybrid kindle]

it takes a looot of time to generate pic, i think my pc is too weak for that

[hybrid kindle]

what will happen if i add --medvram or --lowvram in the line?

[ornate elk]

can you shwo the txt2img settings you tried?

[ornate elk]

you can try add --medvram-sdxl
it speeds up the sdxl models
dont add --medvram, its for 6gb or less

[hybrid kindle]

i will never set it to 1920x1080

maybe lower resolutions

[ornate elk]

yea dont go to high, better generate a bit lower and then use upscaling to get a higher resolution

but first add --medvram-sdxl to the webui-user.bat

[hybrid kindle]

almost at 95% !!

[ornate elk]

Yea to high native resolution

[ornate elk]

If you increase the windows pagefile, it also would be faster

[hybrid kindle]

yes it works much faster now!

thank you

[ornate elk]

No problem!

[stone orchid]

any python devs here want top help me with adding control net to to my script

[ruby yoke]

probably find better help in #🧣｜comfy-ui channel than here

[thorny iron]

Is there a way to identify a model

So I can find the actual model file and make my own pictures.

[zenith seal]

Yo

well idk why but all my first gen are blurry

[ornate elk]

do you have an example?

[zenith seal]

yeah let me find

[ornate elk]

looks like you used a latent upscaler with a denois above 0.5

[zenith seal]

oh thats maybe why

let me try without

[ornate elk]

try esrgan4x-anime6b with a 0.5 denois

and 10 hires steps

[zenith seal]

in high res fix or imgg to img?

[ornate elk]

in highres fix. but if your in img2img then just lower the denois below 0.3 and selecet Esrgan4x Anime6b as upscaler

[zenith seal]

i try in gen rn

well it's not perfect but i still need to get use too this workflow

[ornate elk]

looks good

[zenith seal]

thanks for the help

[ornate elk]

no problem

[lethal sorrel]

Hello I am new with the stable diffusion. Trying to make "inpaint" for a product so I followed a tutorial. For inpaint people suggested foooocus so Im using that. For the outputs there is pixel changes on the area I drew at the inpaint tab but the product is not there. Tried to increase the weight but no visible change. Anyone has an idea ?

I dont even know what Im trying to do is feasible or not tbh

[azure magnet]

An help with comfyui

[granite flare]

bet

Tech support

I need help finishing the download. I got to run but it keeps failing . Its like it will go all the way then boom can't do nothing. UI is up and all but doesn't want to work

I will send a few screen shots of the code

My system is amd

[balmy lily]

on the comfyui package i cant seem to use flux. while im using this exact workflow with all the models and nodes installed and everything, it just crashes when it reaches the load diffusion model node when i have a flux model in there

this seems to be the only thing that the console has to say from the time i click queue prompt to when it crashes

Using pytorch attention in VAE
Using pytorch attention in VAE

[granite flare]

I will be back to check this later for answer hopefully.

[loud pond]

Did you manage to solve it?

im having the same issue icl

[loud pond]

anyone know how to fix this? the inpaint keeps giving me a weird result, i wanted the person to have a big toothy smile but instead their mouth is yellow?

these are the settings im using

[urban girder]

is there any chance somebody could get on voice chat and help me with installing stable diffusion? I'm trying to download it but there's so many codes and steps and so many different websites that I am lost and call me a dumbass but I really need someone to walk me through the download procedure cause I am lost as fuck.

[bleak mirage]

@ornate elk Sorry for the long wait but here is my issue 🙂 if more is needed I will get it 🙂

[rugged latch]

boop

i got a video i can send that helped me

basically its prety straight forward tho

u type, git and then the stable diff repository into cmd, (provided u have git installed first)

then u just let it clone into C:, or whereever, once its done, u go to that directory, double click the file called "webui-user.bat" and then let it finish downloading dependencies, if u dont already, u will also want to install python

specifically python 3.10.6

i believe any higher than that and it wont run

and thats pretty much it =3

i forget what u have to do for AMD cpus, but, i think its relatively the same setup

[urban girder]

I'm sorry but I am too dumb to grt all that

[rugged latch]

D;

[ornate elk]

Hey, check the pinned messages for install instructions

[ornate elk]

Try delete the venv folder and relaunch the webui-user.bat

[ornate elk]

What was your GPU? Nvidia or AMD?

[ornate elk]

You followed an old tutorial it seams.
For the AMD setup please follow my install Guide from the first link of the pinned messages.

[hybrid kindle]

i can move my stable diffusion folder to somewhere else without any problem?

i want to move it on another ssd, do i need to move it with zluda folder or only stable diffusion folder

[ornate elk]

You can only move the models folder and then link it in the webui-user.bat by adding --models-dir "D:\ExamplePathToAnyhwere\models"

[hybrid kindle]

ooo ok thanks

😅

[bleak mirage]

Thanks a bunch that fixed it. Though I have a new problem involving Aria2. It won't start properly. its tied to an addon I have called "CivitAIbrowser-plus" with my A1111😬

[ornate elk]

Try updating your extensions

[bleak mirage]

I checked for updates but nothing changed. Could I just delete the extension and reinstall it 🤔

[high smelt]

on civitai i saw there are clip skip for the img, what extension do i need to download to add that to my ui?

and guidance is cfg right?

[loud pond]

AMD

[ornate elk]

Sure

[ornate elk]

No need for an extension.
Go into the settings, User interface, Quicksettings, there add Clip_stop_at_last_layers
Then hit apply and relaunch

[ornate elk]

Your using directml then?

If so you have to add --no-half to the webui-user.bat to fix the Inpainting.

[lethal sorrel]

Anybody have an idea for this ?

[ornate elk]

Thats not how it works and its really difficult to do that

[median geode]

i need help i tag "git pull" to update my webui, and then something get wrong. it can run well before that.

[warm jackal]

Hello! I need some help here. I am trying to change the location of both the Checkpoint and lora files, but whenever I use the command we are supposed to use, it only loads one or the other, Not both. What am I doing wrong here?

[loud pond]

Not sure, I followed ur pinned link on how to install for AMD tho

[loud pond]

Pretty sure I did already, ill check again tho

[vocal burrow]

you ve modified launch.py, you re not supposed to do that. Revert its modification by running git checkout --theirs launch.py from sdwebui folder

[vocal burrow]

what does the log says ?

[warm jackal]

where are the logs?

[vocal burrow]

in the console

white text in the black window thingy

[warm jackal]

it just says

and if lora is first it only loads that

[vocal burrow]

try --ckpt-dir "c:/AI/Models" --lora-dir "C:/AI/Loras"

[median geode]

i downloaded webui 1.10.0 , double clicked webui user.bat, and it says like this. it took so much time to download pytorch and then failed to make it. whats wrong with it? can anybody help me?

[warm jackal]

same

[median geode]

mine have lots of red words. looks like so many errors occurred.

[ornate elk]

Follow my install guide from the pinned messages

Idk where you downloaded the webui from. But you have to Git clone it

[median geode]

from automatic111

[ornate elk]

Okay, if your in Asia you need a VPN for the first install

[median geode]

yeah im using VPN now, but still don't work

[ornate elk]

Did the git clone worked without any errors?

[ornate elk]

How do the folders look like?

[median geode]

yes it works

[warm jackal]

What do you mean? they are just normal directores

[ornate elk]

Do they have the same name as shown in the log?
aiLoras ?

Because in your webui-user.bat is ai\loras

[ornate elk]

Change the first c to C

And use \ instead of /

[warm jackal]

and there is nothing in those folders presently, yet when I get into SD all of the checkints are there, but the loras are gone

[median geode]

did i make anything wrong with the edit?

[ornate elk]

Looks fine

[median geode]

ok i'll try again

[warm jackal]

I was just using what someone posted, didnt look to see it was wrong

[ornate elk]

Try open up the cmd and type
Pip cache purge
Then delete the venv folder and relaunch the webui-user.bat

[warm jackal]

the lora folder wont load, its stuck in a loop

[ornate elk]

Whitelist the webui in any browser adblocker

Lora list sometimes gets blocked

[median geode]

sorry for disturbing, but what does the "At the line COMMANDLINE_ARGS= You add: --xformers --no-half-vae" work for? i didn't added it because i didn't see it in github

just curious

[ornate elk]

Its for performance and compatibility

Xformers boosts performance and needs less vram.
And no half vae is for VAE compatibility

[warm jackal]

Lora works, but the checkpoint folder does not switch

[ornate elk]

do you see them in the dropdown or not?

[warm jackal]

yes, but i shouldnt, because the directoy i told it to look in has nothing in it, so the list should be empty

[ornate elk]

and where does it get the models from?

where are they located?

[warm jackal]

the models that are showing in the list are in the old Sd directory. the inital one

[ornate elk]

ah okay

but you need put at least 1 model into the new directory

or you wont see if it works or not

[warm jackal]

well, its actually already got 1

[ornate elk]

and does it show up after a relaunch?

[loud pond]

i added no-half-vae already

[ornate elk]

what was your gpu again?

[loud pond]

amd

[ornate elk]

i mean the exact name

[loud pond]

oh 1 sec

[ornate elk]

inpaint with directml works only with --no-half, not with --no-half-vae

[loud pond]

radeon 480

[warm jackal]

So its showing both directories, which is annoying, but ok

[ornate elk]

it will always show the default one, if there is a model in it

[loud pond]

oh so delete '--no-half-vae' and add '--no-half'?

[loud pond]

ok will delete that and add the other

[ornate elk]

yes then it will work, but also that increases the VRAM usage, so maybe make two different bats one with --no-half and one with --no-half-vae

[median geode]

thank you so much btw, what if i don't add anything besides "--xformers --no-half-vae" , even though my card is GTX 16XX and 4GB? will anything get wrong?

[loud pond]

oh alright so how do i create two different bats?

[ornate elk]

copy the webui-user.bat, then rename it for example to Inpaint-webui.bat and there use the --no-half

and for the other keep --no-half-vae

[ornate elk]

a GTX16XX always needs --no-half or it wil get problems with some models
also it needs --medvram as it only has 4gb vram

then your ready

[loud pond]

alright done ill give it a try and see how it goes, thanks!

[median geode]

what if my webui is already launched without modifying it? should i add them and relaunch webui?

[ornate elk]

you can do that after its done installing and opened

[median geode]

thank you sooo much

for staying up so late and help solving my problem

[raw maple]

For some reason my WebUI won't load. I get a "Invalid HTTP request received" in console, and ERR_SSL something in my browser. Has anyone else experienced this? Tried running it on different ports etc..

ComfyUI works fine?

[vocal burrow]

what are inside those folders ?

(commuting back from work, gonna be a while till I answer)

[loud pond]

hey so the images are now a lot better its giving me images which are no longer yellow, but the out put is still clunky any way i can sort this out?

[median geode]

OMG

still don't work

another error

[vocal burrow]

I didn't follow everything, but if you're from china you might have to use a vpn.
Also cleaning pip cache might work (run pip cache purge)

[median geode]

run pip cache purge in where？🥲 webui folder cmd？sry i don't know that much about coding.

[vocal burrow]

doesn't matter, open cmd and run it

it's gonna clean pip's cache globally

[median geode]

cmd in webui folder or just cmd

[vocal burrow]

doesn't matter

[median geode]

done. trying installing again.

[solemn atlas]

Hey guys, where do i find my Stable Diffusion install folder?? I had some problems back then and i delated everything and installed it again but i feel stupid because i dont know what is meant by that.

this is what i want to do

I copy&paste my models folder to my ssd

now i need to make this command so my webui still can use my models folder. BUT I'm stupid,. i just don't know how

[median geode]

what's the problem? you can't find your Stable Diffusion folder?

[vocal burrow]

it should be wherever you ran the git clone etcetcetc command

[vocal burrow]

what's the full log ?

[solemn atlas]

I don't know where that was

[solemn atlas]

yeah, the install folder

[vocal burrow]

can't help you with that ^^""""

search for stable-diffusion-webui folder

[solemn atlas]

i mean i have that one but i dont think it is this

[vocal burrow]

looks like it

[median geode]

looks like no model folder

[vocal burrow]

cause they moved it

[solemn atlas]

I moved it to my ssd

now i need to create a connection but this is what i get.

[median geode]

why not just move back😂

[solemn atlas]

You do not have sufficient permissions to perform this operation.

[solemn atlas]

because its over 500GB and I have no space left?

[vocal burrow]

need to run cmd as admin

[ruby yoke]

try running that command as administrator

open your command line window as admin, then you'll be admin

[vocal burrow]

search for cmd in start menu, right click "run as admin"

[solemn atlas]

Thats what I also tried to do. I copy&paste it and this is what I got:

C:\Windows\System32>D:\Users\moonshine\stable-diffusion-webui>mklink /D models "C:\Stable diffusion"
The command "D:\Users\moonshine\stable-diffusion-webui" is either misspelled or
could not be found.

[ruby yoke]

cd into that directory and see if it exists

[vocal burrow]

because there is no "models" folder in your current directory (c:\windows\system32)

also you're trying to run "d:\users\moonshine\stable-diffusion-webui>mklink"

instead of just mklink

[solemn atlas]

the models folder is now in C

[solemn atlas]

mklink /D models "C:\Stable diffusion"

where to put that cd?

[ruby yoke]

cd is the command for change directories. you don't put it with the mklink command. when i said cd into that directory, what i was saying was for you to try to be in the directory and see the files in it

[vocal burrow]

mklink /D x:\whatever\someLinkToSomething x:\TheSource\OfThelink

[ruby yoke]

do not copy and paste that, btw - that's just an example of how you write the command

you need to use your actual directories, not the x:\whatever... text

[vocal burrow]

lol I doubt they'd have enough drives to add up to x: and not know how to use cmd. But fair enough I guess.

[solemn atlas]

Yeah I got that one haha,. I'm trying to find out what to put there,.

[solemn atlas]

i swear,. i have no clue what to do.. I don't understand anything of these things haha

[ruby yoke]

okay. what you are doing is creating what is basically an instruction to your system. you're telling it "when something wants you to make or read files here, they will actually be there" - iyou are creating what is called a 'symbolic link' from one place to the other

to do that, you use the make link command which is mklink

and then you tell it 'this is where you think they are, but this is where they really are'

IF you are using comfyUI - you can just add a statement in the extra path's file for where the models are

IF you are doing something else, or using something else, obviously that option is not avaiable

[solemn atlas]

I use automatic 1111

[ornate elk]

you just need to add --models-dir "Path to models folder" into the webui-user.bat

nothing more

[solemn atlas]

will there also be created a symbolic folder?

[ornate elk]

maybe i dont get what you are trying to do, but the command links the models folder (no symlink) but it will work normaly

[ruby yoke]

no. go into your auto1111 filder and find the webui-user.bat file

edit it

add the line:

--models-dir <where yoru models are>

[solemn atlas]

in this?

[ruby yoke]

and replace <where your models are> with the actual directory like d:\models\mymodels

[ornate elk]

yes there

[ruby yoke]

yes.

where ARE your models?

[solemn atlas]

C:\Stable diffusion

in my C drive

i created a new folder in my new ssd and put the models folder there

[ornate elk]

but C is the syste mdrive

do you have 2 folders for models now?

[solemn atlas]

no, only one,. i dont know what happend when i tried to move my whole C drive,. somehow c and d got switched up. 😅

[ornate elk]

D is your new drive i guess then. maybe you need to move your models first xD

[solemn atlas]

I moved them haha, thats why its now gone from D to C. at least its not red anymore

[ornate elk]

and your webui is on C too or D ?

[solemn atlas]

3TB is my new SSD. I moved my models folder to C

on D

[ornate elk]

you installed windows on a 4tb ssd ?

why? ^^

[solemn atlas]

No i installed windows on D

but when I tried to move my whole stuff to my ssd everything went there expect for stable diffusion,. what i actually wanted to move.. i have no clue what i did

[ornate elk]

the windows logo says , Windows is currently on C

really strange what you have done xD

but its not that great

[solemn atlas]

and the letters also got switched up. C was D and D was C when I bought my ssd.

Yes windows also moved to my ssd when I tried to move everything to my new ssd which was actually (D:)

Now it's (C:) i dont know why.

[median geode]

😂 complex situation

[solemn atlas]

really why? I used a program to move everything. (dont remeber which one it was) Everything moved expect for stable diffusion. that was why i bought my ssd in first place..

[solemn atlas]

i swear, i don't know what I did,... but I know something is not ok with my computer,. it's slower since i moved everything to my ssd and generating with SD also got very slow ....

[ornate elk]

because the best case is to have Windows installed on the smallest SSD drive on the system and use the big drives to store data.
The Drive where Windows gets installed is used more frequently and does more read/write cycles than any other drive, which wears the drives health out over the time.
Also for easy reinstalling its easier to wipe the whole windows drive than haveing to fear losing 2tb additional data on it

[solemn atlas]

can't i just move windows to my, now, (D:) drive?

[ornate elk]

is your new drive an M2 SSd or a Sata SSD? because there is a huge speed difference between them

[solemn atlas]

uhm wait

i bought this one https://www.amazon.de/dp/B0CY37V52M?ref=ppx_yo2ov_dt_b_fed_asin_title

it says Sata SSD

[ornate elk]

yea thats a sata ssd, good for storing files and games

but slow for windows and big games

[ornate elk]

you said you used a tool at first. but idk if you just could move it again.

[solemn atlas]

maybe thats why my windows got slower after moving it. Can't i just move my windows back to my (D:)?

Whats also weird is, that (D:) was my SSD
and (C:) was my computer drive

[solemn atlas]

it needed over 1 full day to move everything.

[ruby yoke]

did you turn the machine off, move the drives around, and then turn it back on?

[solemn atlas]

If i rememer right, I used aomei backupper

[solemn atlas]

the program did it. it started moving everything to my ssd and it needed over 1 full day

[ruby yoke]

you need to start by putting everything back on C:

[solemn atlas]

i still have it on my computer so it was def. that

[solemn atlas]

but how? my models folder has already over 500GB. I dont think that will work or can i left that out?

[ruby yoke]

and then juist make a directory on your SSD and hand copy the model files over to it

[ruby yoke]

i dunno. can your backup software be told NOT to back up some directories?

[ornate elk]

yea your windows is now slower on the sata ssd.
What i would do is the following:
Check whats on your D drive and move everything over to C in a BACKUP folder.
Then you can Shutdown the PC.
Unplug the new Sata SSD (C drive) and keep it unpluged.
Then reinstall Windows on your old SSD with an USB Stick.
Then replug the Sata SSD and then you may need to give the drive a new Drive Letter (dont format it)

or Backup everything on an external drive if you have one. then reinstall windows

[ruby yoke]

windows should automatically give it a new drive letter if C: is in use with system files on it

[solemn atlas]

I really don't know. Actually I'm afraid using tha again. back then i used it and my whole computer is slow now. I want to do that but thats how this looks. I would like to change the language to show you,. but there is no option to change

[ruby yoke]

but yeah, that's what I would do so. just shut down, remove that sata drive, reinstall windows. you have a reall mess there.

and do NOT use that backup software - ever again

backup software isn't desgined to just move folders and files around, anyway, which is what you needed to do with the models to move them from C: to D:

[solemn atlas]

The problem is, i never plugged it in., my brother did that for me,.
and its inside of my computer,. I need to completely remove the ssd?

[ornate elk]

you dont need to remove it but if you want to be 100% sure you dont accidently format the drive (and lose all data) when reinstalling windows, then yes better remove it

[solemn atlas]

oh,. that will be a mess,. since I don't even know where he put that in my pc...

i mean i try to do that but... it's a lot of stuff to do.

i need to buy a usb stick first..😅

[ornate elk]

maybe ask him for help when you do it

[ruby yoke]

get him to help you. was that backup software his idea too?

[solemn atlas]

he won't. he lives in another city now.
he only put my ssd in my computer.

[ornate elk]

ah okay, then write him and ask where he put it in xD

[solemn atlas]

that's not the problem haha,

i need to find someone who can help me do all of that since i'm afraid i could loose everything

but thank you all for your help. I think i have a bigger problem thatn just moving my models folder 😅

and about that.. should i just delate it?

[ornate elk]

yea definitly get some help then ^^ and no problem, also feel free to ask us if you have any questions when your doing it

[ornate elk]

you still can use it but you need to change the path to "C:/Stable diffusion/models"

without < or >

[solemn atlas]

on cmd?

[ornate elk]

no no

remove this:

and add there "C:/Stable diffusion/models"

[solemn atlas]

like this

okay I try to open SD if it works

sadly it didnt worked, it states

The command "--models-dir" is either misspelled or could not be found.

[loud pond]

any advice?

the outlined area's colour is different to the actual skin colour of the image

[ornate elk]

that could mean your webui is outdated

[loud pond]

its a visible line where the inpaint was used

[solemn atlas]

is there a way to update it?

[ornate elk]

because the command is pretty new

[ornate elk]

yea with git pull

[solemn atlas]

wait thats weird,. now i have this folder but the models are not in it . odes it have something to do with --models-dir

[solemn atlas]

i just use git pull on my cmd on webui right

[ornate elk]

it gets rebuild but is empty

[solemn atlas]

so i used git pull and it did some updates. now i tried to open webui again but it says the same thing.

The command "--models-dir" is either misspelled or could not be found.

[ruby yoke]

can you post a couple of the images so we can see them?

[loud pond]

alright lemme make somenew ones tho cuz i deleted the error ones

[ornate elk]

what does the cmd shows when you git pulled

or share the whole cmd log when launching the webui-user.bat

so we can see the version

[solemn atlas]

it says: Already up to date.

[ornate elk]

looks updated. try edit the webui-user.bat again and move the --models-dir + path before the --xformers

make sure all is seperated by one space

[solemn atlas]

like this?

now it worked,. i also tried to generate something,. its very slow,. if i generate something,. i cant do anything else expect to wait for it to finish because nothing works

[ornate elk]

yea mostly because windows and the pagefile are on the slower drive

[solemn atlas]

Yes thats the reason,. i need to somehow move everything back

[ornate elk]

set the inpaint area to "mask only" and the resolution to 768x768 (better for lower vram)
also dont use a realism model to inpaint an anime image. best use the same model as uses for the image

[loud pond]

alright

[ruby yoke]

yes. always use the model you generated with to inpaint with - or you'll get very unpredictable results

[loud pond]

what if the image isnt generated and is a stock image?

that image above is a stock image

[ornate elk]

thenm ist more difficult. best is to use an inpaint model then

but its hard to not get the outlines

[ruby yoke]

if it's a stock image, then try to find a model that will produce a look that's fairly close to it - you'd want a model that did good photographic images to use to inpaint a stock photo. or a model that did good cartooning images for a vector cartoon, etc

[loud pond]

oh ok thanks then, is there anyway i can just use the head as a reference and then generate a new image from that?

[urban girder]

I'm sorry dude I know I sound annoying but I checked the pinned message for installed instructions and I went to the webpage and I am overwhelmed as shit like there's so many steps and code words that I do not know of.

[ornate elk]

oh now problem, so at which step your stuck?

For an easy install guide use the Automatic1111 one

[urban girder]

I installed python and git but then when I go to the automatic1111 step it's telling me to create a new file and create a new code then go to another website that shows at least a billion automatic1111 files and next thing I know I'm like "Wait what the fuck there's more than one? Which one is the right one??"

[ornate elk]

oh so your at step 3?

[urban girder]

In a sense yes

[ornate elk]

yea thats not a link to go to

you have to create a new folder first+

on C drive or any other drive

call it AI for example

C:\AI\

[urban girder]

Would OC (C:) count?

[ornate elk]

yes

go into the folder where your windows folder is

there create a new folder

[urban girder]

Ok so created the new folder and labeled it AI

[ornate elk]

Then go into the folder you created in this case Ai, then click in the File Explorer bar (not searchbar) and type cmd then press enter.

[loud pond]

also the 'mask only' tip worked thank you a ton, its now only focusing on that area and its giving a better result!

[urban girder]

ok so I don't see a file explorer bar yet.

[ornate elk]

no problem, yea mask only is very important for inpainting and is faster

[ornate elk]

at the top

where you see the path

This PC - C - Ai

there click

[urban girder]

and that's where I type cmd?

Ok so now I'm here

[ornate elk]

now copy the command from the guide or here:
git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git

and paste it into the cmd

then hit enter

[urban girder]

all of it or just where it starts at https?

[ornate elk]

all

[urban girder]

What now?

[ornate elk]

Close the cmd

And then do step 5

[urban girder]

ok so hitting the x button now

[urban girder]

how do I launch the run_nvidia_gpu.bat?

[ornate elk]

You don't. Don't mix up the guides

Your on the Auto1111 guide

Not comfyui

[urban girder]

oh.....ok then I gotta scroll down some more

[ornate elk]

No xD

The Auto1111 guide is the first one

At the top of all others

[urban girder]

oop, sorry. Ok I'm there and it's saying to edit the web user, do I go back to files?

[ornate elk]

Yea edit the webui-user.bat

[urban girder]

it's saying no results match the search

or that the folder is empty

[ornate elk]

go into the ai folder

is there a stable-diffusion-webui folder?

[urban girder]

found it,

ok so with this do I change the name of that?

[ornate elk]

no right click and edit it

[urban girder]

ok so I did a right click and it's showing me these options. Now just wanna confirm with you do i click the button that says open?

[ornate elk]

nope

click show more options

[vocal burrow]

Show more options

[ornate elk]

then click edit

[urban girder]

This is what it's showing me so far I don't see a button that says edit

[ornate elk]

not the folder

go into the folder

and scroll down

to the webui-user.bat

then right click that file

hit show mor options

then edit

[urban girder]

ok so i click on the stabl-diffusion-webui folder then scroll down the list till I find the webui-user.bat?

ok is this it?

[ornate elk]

thats it

[urban girder]

It's still not showing me an edit button

[vocal burrow]

you're not clicking on the file

[ornate elk]

you havent selectetd it

[urban girder]

ok I'm not clicking on the file instead I'm selecting it? Like click hold left and drag the blue box towards it?

[vocal burrow]

no. We mean, in the last screenshot you are not clicking the file

[urban girder]

?

[vocal burrow]

you are clicking next to it

you should right click the file

[urban girder]

ok right clicked

[ornate elk]

click "show more options", then "edit"

[urban girder]

I thought I clicked right the first time, guess not.

ok so I'm on the editing part of it, do i type COMMANDLINE_ARGS= under or above webui?

[vocal burrow]

it's already there

[urban girder]

oh.....so what do I do now?

wait hold on do I add --xformers --no-half-vae now?

[azure magnet]

An help please

[urban girder]

like right after the equal sign?

[ornate elk]

--xformers --no-half-vae

How much vram does your GPU has?

[urban girder]

I dunno, lets see.

so I type the code in and hit enter?

[vocal burrow]

context please

[ornate elk]

First check the GPU vram. Because you may need to add an additional command

[urban girder]

how do I do that?

[ornate elk]

Open the taskmanager, click on the performance tab and then on the GPU

There you see "dedicated memory"

[azure magnet]

Generation with comfyui

[vocal burrow]

probably gonna need to see the workflow then, but if I had to take a blind shot I'd say you're missmatching lora version and model version (eg : lora sd 1.5 with some stable diffusion xl model)

[vocal burrow]

click here

[urban girder]

ok so it says i have 6 gb of dedicated gpu memory storage.

[azure magnet]

Probabily you have right. Where can i see in comfyui if a model is made for 1.5 or XL?

[vocal burrow]

a model is either 1.5, 2.1, xl, 3.0, etc.
a lora is made for a certain type of (sd) model.
And sadly you can't really tell what a lora is for, you have to be organized and name or sort them correctly. So make sure to read lora's descriptions when downloading them.

[ruby yoke]

probably not, but you could generate an image and then do a face swap from the head to the face of the generated image

[azure magnet]

Fuck i dowloaded a bunch of lora

[loud pond]

ah ok thanks anyways

[urban girder]

is 6 gb good enough?

[ruby yoke]

just make sure you have all the base models and then you can use those loras - just sort them out so you know what base model they are each for

[ornate elk]

You you need --medvram in the webui-user.bat too

Then save the file

[urban girder]

Wait do I backtrack?

[ornate elk]

?

[urban girder]

I'm confused on what you mean I need medvam

[azure magnet]

Last question how could i enable tiled decoding

[vocal burrow]

there's a specific node for that "VAE Decode Tiled" I think

[urban girder]

like how do I get medvram on the webui? Cause I didn't know what that is

[ornate elk]

You have to edit the file again

Like you did before with xformers

[urban girder]

Edit? Wait I don't understand where am I supposed to edit it?

[vocal burrow]

there

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

[urban girder]

Ok so I type --medvram in the webui-user first?

[vocal burrow]

And if your gpu is a GTX16xxxxxx something you also need to put --no-half in there, that's also why I asked for a screenshot of your GPU info in task manager

[urban girder]

Ok

[vocal burrow]

Ok so in the webui-user.bat file, you find the set COMMANDLINE_ARGS= line and turn it into a set COMMANDLINE_ARGS=--xformers --no-half --medvram

(--no-half includes the functionalities of--no-half-vae so there's no need for both)

[urban girder]

--no-half includes the functionalities of --no-half-? arent those the same thing?

[vocal burrow]

--no-half
--no-half-vae

-vae

[urban girder]

oh......ok so I don't need to type the one that says vae.

[vocal burrow]

not the same things

[vocal burrow]

yes, cause it's included with --no-half

[urban girder]

ok i copied and pasted now do I hit enter or another button?

[vocal burrow]

save it, double click it

and voila

[granite flare]

welp time fix this mess I made

[urban girder]

ok so, wait double click it? double click the code I just pasted in?

[vocal burrow]

double click the the webui-user.bat

just like you would to start any program

[urban girder]

wait double click it in files?

[vocal burrow]

double click this

[urban girder]

ok I did and then it showed the black screen for a sec then exited out.

[vocal burrow]

go back there, insert a new line at the end saying pause .
Then try to run it again

[urban girder]

under set COMMANDLINE_ARGS=--xformers --no-half --medvram?

[vocal burrow]

no, at the very end

[urban girder]

Like this?

[vocal burrow]

ok my bad I guess.... no dot after pause

but yes

[urban girder]

do I exit out of it now?

[vocal burrow]

save and run it again yes

[vocal burrow]

where is that webui-user.bat you just ran

screenshot its folder

[vocal burrow]

that's the result view of your search for it

I want to see what's next to this file

[urban girder]

is it possible to hop onto voice chat to screen share?

[vocal burrow]

not right now, I'm just about to go to sleep.

[urban girder]

ok so you wanna see what's next to this file, like above, below, right, left? cause I don't know what you mean by next to this file.

[vocal burrow]

as in "what's in the same folder"

[urban girder]

there's just the one thing in the folder

[vocal burrow]

there should be a bunch of stuf (can't screenshot everything)

[vocal burrow]

ok it's probably windows being windows then.
run the second last one. It will be the same file as the one you edited. But running it outside of this weird "search result" context should work fine.

[urban girder]

Ok so it's showing me this

[ornate elk]

Looks good

It should open when its finished

[urban girder]

like stable diffusion? The software? Like the actual one where everyone can type in their prompts to generate an image?

[vocal burrow]

stable-diffusion-webui the sofware written by Automatic1111 used to generate images using stable-diffusion models

so tldr, yes

and with that I'm off, bye

[urban girder]

thank you so much, I'm no tech wizard so you saved my ass bigtime

[ornate elk]

To start the software you always have to run the webui-user.bat.
It will also start much faster after this first install process

The webui then opens in your default browser

[urban girder]

ok so stable diffusion is called webui

[vocal burrow]

different things.

your car run on gasoline (or diesel, whatever)

same situation here

[urban girder]

?

[vocal burrow]

you're using stable-diffusion-webui (you car) which is using stable diffusion models (gasoline)

[urban girder]

ahhhhhhh

[ornate elk]

The webui is the interface to control everything

[urban girder]

Ok I think I'm in

Now it's showing me these

[ornate elk]

You need to update your nvidia graphics driver probably

[urban girder]

how do I do that?

and when I do, would I need to start the code all over again?

[ornate elk]

Nope

You only need to restart the PC and the webui-user.bat

But first download the latest driver from Nvidia for your gpu

[urban girder]

ok so I should head to the Nvidia website or do I have to go through a third party website?

[ornate elk]

You have a GTX 1660 ti

No please only use nvidias official site

Like here:
https://www.nvidia.com/en-us/drivers/

[urban girder]

Which one?

[ornate elk]

The geforce experience

[urban girder]

alright it's downloaded, do I open the file now?

[ornate elk]

Yea install it

[urban girder]

ok so I installed it, now what?

[ornate elk]

Restart the PC and then relaunch the webui-user.bat

[steady thistle]

I don't think this is hurting anything but there's an error somewhere and I'd like to clear it up. Can you tell me where this issue is found to be repaired? :
To see the GUI go to: http://127.0.0.1:8188
FETCH DATA from: D:\Ai\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager\extension-node-map.json [DONE]
[ERROR] An error occurred while retrieving information for the 'Florence2ModelLoader' node.
Traceback (most recent call last):
File "D:\Ai\ComfyUI_windows_portable\ComfyUI\server.py", line 564, in get_object_info
out[x] = node_info(x)
File "D:\Ai\ComfyUI_windows_portable\ComfyUI\server.py", line 531, in node_info
info['input'] = obj_class.INPUT_TYPES()
File "D:\Ai\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Florence2\nodes.py", line 148, in INPUT_TYPES
"model": ([item.name for item in Path(folder_paths.models_dir, "LLM").iterdir() if item.is_dir()], {"tooltip": "models are expected to be in Comfyui/models/LLM folder"}),
File "D:\Ai\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Florence2\nodes.py", line 148, in <listcomp>
"model": ([item.name for item in Path(folder_paths.models_dir, "LLM").iterdir() if item.is_dir()], {"tooltip": "models are expected to be in Comfyui/models/LLM folder"}),
File "C:\Users\alton\AppData\Local\Programs\Python\Python310\lib\pathlib.py", line 1017, in iterdir
for name in self._accessor.listdir(self):
FileNotFoundError: [WinError 3] The system cannot find the path specified: 'D:\Ai\ComfyUI_windows_portable\ComfyUI\models\LLM'

[urban girder]

Ok doing that and it says this

and fuck me sideways it actually works.

[ornate elk]

Nice ^^