🤝｜tech-support

[ornate ruin]

Lastly, with 8GB VRAM I should probably stick to single batch sizes right? Or are 2 or even 3 possible? Couldn't find a clear answer for that yet

[vocal burrow]

single batch

[severe kindle]

Funny you should have posted that @fair oxide @ornate ruin I was just testing if InstantID does a better on realism style prompt with LCM's due to both needing low guidance scale to work.
It least thats my finding InstandID see much better at photo style images with guidance scale < 3

[ornate ruin]

But from testing, the time it takes is actually quite equal tbh.. like if do two in a row or two in one batch

[fair oxide]

you don't need to bother with batch size, has no technical relevance other than automating your task which btw you could alternatively click on generate forever until you cancel

[ornate ruin]

No? I thought batch count would be just that but batch size allowing to generate 2 images at the same time so basically doubling up efficiency

[fair oxide]

doesn't affect the actual output quality at all, its just automation at the cost of memory resources

[fair oxide]

ahh interesting find, instant id for now is only limited to XL model application, incidentally i would also encourage you to look into face_id ip-adapter

[ornate ruin]

i suppose the actual image size has a major impact too, this one only took 26.3 seconds

makes only sense tho..

[fair oxide]

an ideal approach would be to generate images at 512x768 then hires fix by 2x and if needed upscale even further in the EXTRAS tab

[ornate ruin]

generate an image - then hi res fix confuses me.. you do mean to do the hires fix within the generation progress of the image tho right?

[fair oxide]

well hires fix is integral to generating image, its the extras tab that is seperate process to enlarge even bigger

[severe kindle]

The images from SDXL models are okay, that is just from theSDXL LCM base model (no refining done)
Need to see if the LCM LoRA and a better model work any better.

original image

[fair oxide]

the thing about using LCM LoRA instead of LCM based model or turbo model is that you can apply LCM lora to any models you may want to use

and LCM LoRA file size is around 300mb

[severe kindle]

Indeed, I just used my frankensteined lcm model from before the lora came out as it didn't need the code change to add a LoRA in (working with a diffusers script)

Plus technically I working and they're paying me overtime ATM 🙂

I have tried faceid and PhotoMaker. IP-ADapter Face (Plus v2) is where I worked out the colourfully light almost caricture faces for realism styles where down to guidance scale. Photomaker gives the cleanest images so far (although I've not played with the guidance scale, it's at the default value in my test script).
If you does more artistic styles they're all pretty good.

[fair oxide]

an experiment i was doing with face id ip-adapter plus v2, which came out interesting and aesthetic from the source photo i used

actually thats just the plus not v2

[wide vault]

hi

im having problems with command line arguments on my webui , could i get help here or where should i go?

[severe kindle]

So, LCM lora with SG161222/RealVisXL_V3.0 give me this, need to play with the guidance scale at bit this was at 1.0
The InstantID adapter needs to be loaded first before the LoRA or it'll fail to load it.

Photomaker

[dim egret]

Hi, bots (1-9) don't work I can't write anything c what's the problem?

[severe kindle]

https://discord.com/channels/1002292111942635562/1047610792226340935

[wooden iron]

what checkpoint you are using for this

[vague dagger]

Hi guys, i'm looking for a python developer who can help me with building stable diffusion pipelines. Anyone here, that might be interested?

[tribal wave]

I can’t type in the bots I’m confused

[tacit light]

How to add Lora to amd ui and use it

any tutorials?

also my amd ui got considerably slower for some reason

[wooden iron]

i am having multiple hands problem can somebody know how to resolve it

[vocal burrow]

#1047610792226340935

[vocal burrow]

LoRa/Lycoris quick guide :
1/ Where do I put my LoRa : stable-diffusion-webui\models\Lora, Loras are NOT stable-diffusion model but rather patches for said models.
2/ How do I use my LoRa, you can either :

• Invoke it directly in your prompt by using the following syntax <lora:filename:weight> (or <lora:alias:weight> depending of the setting in setting->extra network). eg : <lora:animeTarotCardArtStyleLora_v20Offset:1> for using the Lora animeTarotCardArtStyleLora_v20Offset.safetensors
• Use the UI "Extra Networks" / card system by doing as described in my screenshot. Clicking an associated lora card should automatically insert the lora prompt liked mentioned above. https://cdn.discordapp.com/attachments/1002602742667280404/1191779742534348870/loraHowTo17.jpg
  3/ Be careful, LoRa are trained specifically for some SD version. Lora trained for SD1.5 (and derivatives) models will not load on SD2 models and vice versa (same goes for SDXL). Also even though Lora trained on SD1.5 models should work for every SD1.5 derivatives, they'll probably work best with the model that has been used for its training.
  4/ LoRa, LoHa, LoCon, what are these ? They're all commonly refered as "Lora" but differ slightly technically :
• LoHa : Lora with hadamard decomposition
• LoCon : Lora extended to redisual blocks.
• More explanations : https://stable-diffusion-art.com/lora/, https://www.sdcompendium.com/doku.php?id=lora_0014, https://www.sdcompendium.com/doku.php?id=locon_training_0071, etc

[vocal burrow]

share your command line args, your setup, prompt settings and logs

[vocal burrow]

Ask. What problems ?
(also, https://dontasktoask.com/)

[wooden iron]

i resolve it wathing youtube

[wooden iron]

well i have one problem can you help related to confyui

[vocal burrow]

I don't use comfyui much but I for sure can't answer if I don't know the problem

Plus I'm not the only one out there, people might stop by, read about your problem and help you a couple hours down the line.

[wooden iron]

i like your attitude man

thanks ( i dont know why )

[vocal burrow]

xd. I might be a bit sassy at times but I don't bite.

Anyways, what's wrong with your hands ?

[limber bloom]

I am having trouble with the Sketch to Image function. Does anyone know a way to generate images that are more photorealistic?

[wooden iron]

https://dontasktoask.com/

[vocal burrow]

Two "questions" there, therefore two "answers".

• what's the problem with "sketch to image" ?
• Making an image more photorealistic compared to what result ? What do your outputs look like ? What settings are you using actually ?

[wooden iron]

yeah if you use right checkpoint then you dont have to give sketch to it go to civitai.com

[wooden iron]

its look too real

[limber bloom]

I'm very sorry for my behavior.

[wooden iron]

some times i am sassy but i dont byte XD

[vocal burrow]

ok let's tone down the sass.

[vocal burrow]

np, just please provide more informations like I asked.

[wooden iron]

i am sorry i am just kidding

[limber bloom]

Please wait, as I am compiling it now.Thank you.

[wooden iron]

what software you are using ( i am using confyui )

[limber bloom]

As I'm using Sketch2Image by Clipdrop API.

I am struggling to create more photorealistic images.

[vocal burrow]

and what settings are you using ?

[limber bloom]

Here you are.
https://clipdrop.co/ja/stable-doodle

[vocal burrow]

Ok so, you're not using the API then but the website.

[limber bloom]

I'm using Clipdrop API for my website.

[vocal burrow]

then what do your requests look like ?

[limber bloom]

Oh, I mistook.
https://clipdrop.co/apis/docs/sketch-to-image

[vocal burrow]

What settings are you using (in case of sketck 2 image) that would only include an img and a prompt if i remember correctly.

[limber bloom]

When generating images, I can't help but leave sketches behind. I can't generate them in photorealistic quality. How can I make them look more realistic with 3D rendering? I am searching for a way.

[vocal burrow]

I've understood that. I'm trying to help you get there. But given that sketch 2 image only use an image and a prompt, there's not much to work with.... I'm guessing that your prompt can be improved but to verify that assumption I need to know your prompt.

[limber bloom]

Oh, You are so kind. Just a moment, please. I need to be prepared for that.

I want to check my site.

[limber bloom]

I apologize for the wait. I will send you one example.

[limber bloom]

I apologize for the wait. I will send you one example.

About the prompt. It is setting like this.

[vocal burrow]

maybe try with this prompt Scandinavian style ((kitchen)), photorealistic 3d rendering, architectural visualization, high resolution

[limber bloom]

Thank you for your advice. I immediately try these prompts.

[vocal burrow]

Ok good luck with that, feel free to try out different wording for your prompt.

[wooden iron]

i am going to sleep

[wooden iron]

bye bye friend remember never ask to ask

[vocal burrow]

There's not much settings on clipdrop you can tweak to constraint the output. It might be more effective for you to use a local installation (auto1111, comfyUI, swarmUI, etc) as those have much more settings to tweak.... Although It will probably be harder (but possible) to setup if you want to use it through your website.

It's 2 am for me so I'll also go to sleep, best of luck.
二時ですから私も寝ています。頑張ってください

[limber bloom]

Thank you. Good night.

[wild tusk]

This may be a better place to ask but I posted over in #🐝｜swarm-ui because that's what I'm currently using but having issues with refining/upscaling images: #🐝｜swarm-ui message

[cold walrus]

Fixed : hi all, when i try to execute INSTANTID on comfyui i have a error (many others fixed) ComfyUI_windows_portable\ComfyUI\models\controlnet\config.json' is not a valid JSON file.

[tacit light]

I copied both a Lora file and LyCoris to the folders but they did not show up on my Ui, am I missing the models etc? All I did was download them from Civitai

also my Ui is back to the speed i had yesteday

does it have anything to do with the time?

or did the restart resolve the speed lag

[ancient plover]

This is a stupid question, but what is the preferred way to start SD (auto 11 webui for amd) once its installed. Do I keep running the webui-user.bat in command line, like I did for the first install, or is there another way? I am brand new to this stuff, but I tried to run the launch.py in CL and I got errors. Thanks in advance.

[rancid wagon]

i'm having trouble generating images on automatic1111's web ui. I followed all of the steps on github and when i try to generate anything, all i get is just a black image, sometimes even "NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type."
also i'm new to ai so i might not know some tech words or etc so please hold my hand guide me through 🙂

[ornate elk]

Hey, did you followed my install guide for the AMD webui?
Its in the pinned messages of this channel. Check it out for the best performance settings.

Also you always have to start the webui-user.bat to launch SD.
You can copy that to your desktop if you want

[ornate elk]

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

[rancid wagon]

i got the message nans exception....... so i searched for guides and got told to set the commandlineargs as above

and after i did that it just generated a black image

[ornate elk]

Ahh okay a 1660.
Remove all the commands you added.
Then only add:

--xformers --medvram --no-half

Then save and relaunch the webui-user.bat

[rancid wagon]

okay ill try it

oh ok it worked

tysm

[ornate elk]

No problem 🙂

[wooden iron]

good morning everyone

i am doing iterative upscaling by image and latent both trying one by one they work awesome
but it again giving me gpu error

for god sake if anyone know how to solve out of gpu memeory error in confyui please let me know

i am using amd gpu

[ornate elk]

Can't help with comfyui, but could help with auto1111

Your limited in your actions what you can do with 12gb vram.
Better use the ultimate SD upscaler extension to upscale images. With that you won't get out of vram errors

[worn frigate]

Luffy

[wooden iron]

monkey d luffy

are these really matter , select other will give me some perfomance or just increase quality of picture

[ornate elk]

some are faster than others, some require less steps, some give not as good outputs as others

here is an explanation on what they do different:
https://stable-diffusion-art.com/samplers/#Samplers_overview

[vocal burrow]

Are the loras you have compatible with the model you're using ?

[wooden iron]

good morning friend

[wooden iron]

thankyou luffy

[tacit light]

where can i check for that

[vocal burrow]

Check the description of the loras. Loras are made for a specific SD base model. So if you're tring to use a SD1.5 lora with a SDXL model it won't show up in the list.

[tacit light]

ok

ty

[tacit light]

How to upscale final image in the automatic1111 ui?

[lost crescent]

Load Image node not functioning - shows no image. Workflow png

[ornate elk]

With the SD upscale script in img2img

Set the
Denois below 0.3
And resolution to 512x512

[tacit light]

okay

What resolution will this upscale to?

[frank mortar]

Hey, could someone help me?
I want to change people's facial expressions and I found some kind of wildcards thingy and I tried instsalling dyanimic prompts but it doesnt show up on the stable diffusion interface, I also installed wildcards and it also doesnt show up please help

[wooden iron]

bro what promt should i use to avoid this ugly and taboo result

i think resolution is wide enough to create error bcz when i decrease width it showing good result atleast not mutant

oh by the way i updated my graphic drivers get good it speed

[fair oxide]

SD base models are trained on specific resolutions, and all other community models are built on that settings, if you break away from that with random sizes you will likely get anomalies

[wooden iron]

yes you are right

what if i say 1 girl or 1 man in promt should it resolve that

let me try it

[fair oxide]

those are not related to resolutions

an ideal resolution would be in these values 512x512 / 512x768 / 512x912 / 768x1024

[ornate elk]

It will upscale by 2

[fair oxide]

@wooden iron besides setting those resolutions you would also always use hires fix by 2x which is not only to enlarge image but also accentuate the detail of the image.

[wooden iron]

you are right

i just found something

it also depend on checkpoints some checkpoints give nice result

and some give very bad result

[fair oxide]

they may, but most checkpoints that are properly trained would be ok with the resolutions i mentioned, which are part of the base model specifics.

[wooden iron]

https://tenor.com/view/takagi-gif-19085008

i am so intelligent now

[fair oxide]

@wooden iron i wouldn't suggest struggling with your prompt on models that don't follow those conventions

[wild lodge]

im trying to set up but my python isnt working. I have python installed and even have Path env var checked when isntalling py. Whats going on?

[wooden iron]

yo

[wild lodge]

sup

[wooden iron]

installing comfyui

?

[wild lodge]

what?

[wooden iron]

you instaling what

[wild lodge]

python

[wooden iron]

why python

[wild lodge]

i cant launch the webui-user

[wooden iron]

ohh

[wooden iron]

i also intalled python for comfy ui

[wild lodge]

gives me this

[wooden iron]

do this in you cmd

type python and let me see the version

[wild lodge]

i did it but only py works

and python and python3 just bring me to the windows store

[wooden iron]

let me see you environment path

[wooden iron]

show me this list inside path

[wild lodge]

oh i fixed it

lemme try launching stable

bruh

why is it not a git repository

do i have to use git to install it?

i just downloaded the zip and moved it to where i wanted it

[wooden iron]

you can do that to i mean moving file

[wild lodge]

is it because its not updated?

[wooden iron]

yes it showing this

[wild lodge]

E:\Stable Diffusion\stable-diffusion-webui-master\venv\Scripts\python.exe -m pip install --upgrade pip

so just copy and paste this in cmd?

or in git?

[wooden iron]

this one

[wild lodge]

it doesnt work

[wooden iron]

you can also go to th respective folder from file explorere
and then type cmd at path of folder

[wild lodge]

so right here?

and type this?

python.exe -m pip install --upgrade pip

[ornate elk]

first thing first, your python version is the wrong one

only 3.10.6 up to 3.10.11 is supported

second thing, dont download the webui as .zip file

Please uninstall python 3.12 and delete the stable-diffusion-webui-master folder.
Then follow my install guide. Its in the Pinned Messages of this Channel.

[wild lodge]

the webui-user folder should look like this? I have 24 GB of GPU ram so it should be +11GB or 24GB?

[ornate elk]

remove the +11gb

[wild lodge]

all good?

[ornate elk]

all good

[wild lodge]

i got this when launching the bat file

[wild lodge]

its the same problem i had the last time

[ornate elk]

that error means your mostly using a third party antivirus that blocked the proper installation of Python

That happens if you have any of the following Antivirus programms running:
MCAfee, Avira, avast, AVG, Norton, Bitdefender, Eset, 360security, kaspersky, panda

[wild lodge]

i dont have any antivirus though

just windows defender

[ornate elk]

okay thats strange then

[wild lodge]

and looking at its protection history theres nothing from today

[ornate elk]

then uninstall all python versions

restart the PC

and install 3.10.11 64bit with "add python to path" checked

[wild lodge]

how do you uninstall all python versions through cmd?

[ornate elk]

nope trough system control

or the windows settings under apps

[wild lodge]

those 2 at the bottom?

[ornate elk]

nope

[wild lodge]

uninstall both?

[ornate elk]

its okay if you uninstall 3.10.11 only

[wild lodge]

aight gonna restart

this is what i did last time

same thing

[lost crescent]

LoadImage in ComfyUI broken - I can see a menu of images - just there is no access to them!!! ComfyUI. Win 10. 64Gb mem. RTX 2070 8Gb VRAM

[wooden iron]

what is clip skip

[exotic oak]

New to SD, are there any merits in sharing models folder between Koyha and auto1111/comfy etc? I'm setting up some new docker images and have serious decision paralysis in how to structure things! I've just never trained any models so I am not entirely sure on the best route to take.

[ornate elk]

delete the venv folder and relaunch

[wild lodge]

oh yea that worked

and its working now

[ornate elk]

perfect 🙂

[wild lodge]

i saw that on a forum somewhere

[hollow parrot]

Nsns

[wooden iron]

hey hey

https://tenor.com/view/pat-pat-anime-pat-funny-mandakh-pat-skrunkly-gif-24287712

what is clip skip

[fair oxide]

I have not used SD/Automatic1111 in a while and before using it i did a git pull and when generating an image i am getting this error:

[Previous line repeated 5 more times]
File "/x/stable-diffusion-webui/venv/lib/python3.10/site-packages/torch/nn/modules/module.py", line 833, in _apply
param_applied = fn(param)
File "/x/stable-diffusion-webui/venv/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1158, in convert
return t.to(device, dtype if t.is_floating_point() or t.is_complex() else None, non_blocking)
NotImplementedError: Cannot copy out of meta tensor; no data!

[fair oxide]

ok after running @karmic crown script it works

[sonic ferry]

ComfyUI, anyone know a save node or how to save to subfolders? /output/folder, so I can do scenes, people, anime, etc. instead of everything in /output/ together. 😛

[vocal burrow]

What is clip_skip ?

Clip_Skip changes the way CLIP interprets your prompt. Basically CLIP itself works by iterating over your prompt (or image but in this case it's the prompt), it will loop 1,2,3,4,etc times.... until it's "done" and spit out some "tokens" to feed to StableDiffusion. So tldr it translates your prompt into "tokens"
What clip_skip does is telling CLIP "hey remember the result of the last xxx iteration ? let's use that instead of the final one"

[wooden iron]

i see thats why no one replying to my messages bcz its too hard to understand

[worldly rain]

I am getting this error only when use hires fix. why am I getting this error? i have 12gb vram and 60gb empty space on my c drive

[wraith bronze]

You exceed your capacity, if you have a scaling of 2, lower it to 1, if it finishes, you can raise it little by little until you find the limit.

[worldly rain]

alright, will do

but if i set scaling to 1, isn't it gonna be the same image?

[wraith bronze]

Just the size will remain the same. But if you apply any other settings in Hires they should affect the image.

[ornate elk]

whats in your webui-user.bat?

[worldly rain]

lemme see rq

[ornate elk]

also when using hires fix, set hires steps to 10-15 max

[worldly rain]

xformers, no half vae, dark theme

[ornate elk]

okay thats good

[worldly rain]

yes it was 15

[pearl pelican]

Model is not in Diffusers format, this makes loading slower due to conversion. For a speedup, convert it to a Diffusers model.

[worldly rain]

but my point was to upscale it

[ornate elk]

what was the resolution?

[pearl pelican]

how do i fix this

[worldly rain]

1080x1350, upscaler 2, 15 steps, denoising 0,3

[ornate elk]

yea thats to high, you would need to install Tiled Diffusion to upscale that high. When installed only enable Tiled Vae and leave it on default

then it will work

[worldly rain]

bet, thank u

[plush flax]

trying to figure out how to get this working ;w; anytime i use my own models i get this

[sonic ferry]

Hmmmm, I think it's a boy. But seriously, I usually get that in ComfyUI when I have strength set too high on something or ClipSkip is wrong for a LoRA.

[ornate elk]

whats the webui your using?

[pearl pelican]

im using windows gui

[ornate elk]

which one?

[pearl pelican]

1.12.0

SDGUI

[ornate elk]

ah the NMKD webui, i would recommend you something more updated

most people here use Automatic1111 as local Stable Diffusion webui

you find my install guide for Auto1111 in the Pinned Messages of this Channel

[pearl pelican]

ty will check it out

[ornate elk]

np, if you have any question feel free to ask 🙂

[wind sun]

any idea ?

after installing controlnet

[ornate elk]

is your webui updated?

[boreal bison]

@ornate elk Hello

I'm been playing with comfy ui recently a lot. Now I wanna try automatic1111 too.

[ornate elk]

Hey, ah so you started with comfyui before auto1111

[boreal bison]

no, I've played a bit with a1111 than comfy

now I wanna try instant id and I found a tutorial with automatic1111

[ornate elk]

Ahh

[boreal bison]

I wanna make sure I'm doing everything right again

can you help me with that? 🙂

[ornate elk]

Sure

I have instant Id working on my auto1111

[boreal bison]

Is it possible I send you a private message?

so we can keep focused please

[ornate elk]

Sure

[boreal bison]

sent you a message

[sonic ice]

trying to download stable difsuoion and i go thtis error

RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check

whats causing it? how do i fix it? if i just need to add the thing it says to add where do i add it?

im using an nvdea 4080.. so i do in fact have a graphics card

[river tartan]

hi guys. i cant use LDSR Upscaler can you help me?

[lost crescent]

ComfyUI loads up - then when calling Localhost:8188 - I get a blank screen (after updating!)

[lost crescent]

Edit your user.bat file in notepad - you should see a line KWARGS= - and all entries there start with a --

So put --skip-torch-cuda-test there and save

[wooden iron]

good morning furands

[sonic ice]

and that will allow stable diffusion to use my graphics card it didnt read before?

Also I did it and it still gives different error... RuntimeError: "LayerNormKernelImpl" not implemented for 'Half'

[lost crescent]

Take out no-half-vae

[brisk heath]

how do i update my python 3.10.13? they only put up installer up to 3.10.11

[ornate elk]

Can you show an full screenshot of that error in cmd?

Don't add --skip-torch-cuda-test or it will run on CPU only mode

[ornate elk]

You can't since there is no installer

[spare umbra]

I am not able to find where i can put my prompt and create or generate image

[ornate elk]

The bots on this discord are under maintenance right now

[left iron]

How can I run Automatic1111 as HTTPS instead of HTTP?

[lofty belfry]

😋😋

[lost crescent]

What does this mean at all? https_key OK: B:\ComfyUI\custom_nodes\comfyui-mixlab-nodes\https\certificate.crt B:\ComfyUI\custom_nodes\comfyui-mixlab-nodes\https\private.key

[ornate elk]

whats in your webui-user.bat and whats your GPU?

[ornate elk]

okay at the line Commandline_ARGS=
you add: --xformers --medvram --no-half-vae
then save and relaunch

[native loom]

okay ty)

[valid plank]

hey there, maybe someone can help me with this problem.
I had Stable Diffusion installed on my old PC and saved all the files on One Drive. Now that I have a new PC, it doesn´t work anymore. I already reinstalled python and Git, but when I try to launch webui.bat I get this notification in CMD:

venv "C:\Users\kayju\OneDrive\Desktop\StableDiffusion\stable-diffusion-webui\venv\Scripts\Python.exe"
No Python at '"C:\Users\kay\AppData\Local\Programs\Python\Python310\python.exe'

(if there´s anyone who can walk my through what I gotta do in german, that might be easier for me but english will do too)

[ornate elk]

hey, to copy the webui to another PC isnt the right way to get it to work.
you need a clean install of the webui by following my Install Guide. Its in the Pinned Messages of this channel.
After the install you can copy your models folder over and it will work.

for German help, dm me

[sonic ice]

bruh.. i got told i had to..

[ornate elk]

okay, you installed the webui the wrong way, by downloading it as zip file.
Please follow my complete Install guide to get it work.
You find it in the Pinned Messages of this Channel.

[plain axle]

Hello!

I just got a 2T External Solid State Drive that I would like to install Stable Diffusion on.

Can someone help walk me through the process?

⚠️ Warning: I am computer illiterate and it took all my courage to post this so please be kind.

Thank you!

[ornate elk]

hey, you can checkout my install guide in the Pinned Messages of this channel. With that you can install it on any drive you want. But Python and Git needs to be installed on the C: drive

[sonic ice]

no i didnt? i did it the "proper" way.. using git clone and cmd as it said too

[plain axle]

Thank you. However, I don't know about you, but when I'm trying to learn something that is waaay outside of my wheelhouse, sometimes I need a pro assuring me that I'm doing it right. I've delayed 2 weeks because I'm afraid I'm going to fubar my computer by doing this wrong.

[ornate elk]

no, because your folder is called stable-diffusion-webui-master
and thats only if you download the webui as zip file

if you would git clone the repo it would never get that name -master at the end

even in your screenshot it says no git repo

so thats not the right folder

[dusty garnet]

Belp

Help*

[lost crescent]

What does this mean at all? https_key OK: B:\ComfyUI\custom_nodes\comfyui-mixlab-nodes\https\certificate.crt B:\ComfyUI\custom_nodes\comfyui-mixlab-nodes\https\private.key

[plain axle]

soft sigh Looks like u have ur hands full @ornate elk

I'll try again later.

[ornate elk]

no problem, for any questions or guideance you can ask here, or dm me and youll get help

[dusty garnet]

Stuck here forever

AMD gpu

[ornate elk]

AMD GPU cant train on windows

[dusty garnet]

So I literally can't?

[ornate elk]

only on CPU mode which would take hours or days, or if you install Linux and then train on that

Linux is and was always free

whats your gpu btw?

and whats in your webui-user.bat?

[dusty garnet]

Mb I read something else

[dusty garnet]

Amd ryzen 5 5500u

[ornate elk]

thats the cpu not gpu

[sonic ice]

now i get just an error code when trying to instal..

Cloning into 'C:\Users\Micha\Music\AI\stable-diffusion-webui\repositories\stable-diffusion-stability-ai'...
remote: Enumerating objects: 580, done.
remote: Counting objects: 100% (357/357), done.
remote: Compressing objects: 100% (128/128), done.
error: RPC failed; curl 92 HTTP/2 stream 5 was not closed cleanly: CANCEL (err 8)
error: 6332 bytes of body are still expected
fetch-pack: unexpected disconnect while reading sideband packet
fatal: early EOF
fatal: fetch-pack: invalid index-pack output
Traceback (most recent call last):
File "C:\Users\Micha\Music\AI\stable-diffusion-webui\launch.py", line 48, in <module>
main()
File "C:\Users\Micha\Music\AI\stable-diffusion-webui\launch.py", line 39, in main
prepare_environment()
File "C:\Users\Micha\Music\AI\stable-diffusion-webui\modules\launch_utils.py", line 408, in prepare_environment
git_clone(stable_diffusion_repo, repo_dir('stable-diffusion-stability-ai'), "Stable Diffusion", stable_diffusion_commit_hash)
File "C:\Users\Micha\Music\AI\stable-diffusion-webui\modules\launch_utils.py", line 192, in git_clone
run(f'"{git}" clone "{url}" "{dir}"', f"Cloning {name} into {dir}...", f"Couldn't clone {name}", live=True)
File "C:\Users\Micha\Music\AI\stable-diffusion-webui\modules\launch_utils.py", line 116, in run
raise RuntimeError("\n".join(error_bits))
RuntimeError: Couldn't clone Stable Diffusion.
Command: "git" clone "https://github.com/Stability-AI/stablediffusion.git" "C:\Users\Micha\Music\AI\stable-diffusion-webui\repositories\stable-diffusion-stability-ai"
Error code: 128
Press any key to continue . . .

[ornate elk]

that means you had a connection error when trying to git clone, also you tried to git lcone it into the Music folder, thats a windows default folder ant isnt supported because of Permission conflicts

[dusty garnet]

nvidia geforce gtx 1650

[sonic ice]

i thought a folder was a foler and it didnt matter wehre it was..

[ornate elk]

so thats not an AMD GPU, and you should be able to train.
what you need to do is to edit the webui-user.bat
at the line COMMANDLINE_ARGS=
you add: --xformers --medvram --no-half-vae
then save and relaunch thr webui-user.bat

[ornate elk]

nope not for SD

thats why in my guide i say to create a new one on C directly

[sonic ice]

would it work to put it on an external harddrive connected to my pc?

[dusty garnet]

On the same line or the next line?

[native loom]

how can i fix the git pull?

[ornate elk]

same line

[unique glacier]

https://tenor.com/view/hug-virtual-hug-hug-sent-sending-virtual-hugs-hang-in-there-gif-17500745

@ornate elk for help 3 people at once. gotta be hard!

[ornate elk]

do that:

Update Fix
To fix this go into the stable-diffusion-webui folder and click in the File Explorer bar (not searchbar) and type cmd and hit enter.

Then use the following commands:
git init git remote add origin https://github.com/AUTOMATIC1111/stable-diffusion-webui.git git fetch --all git reset --hard origin/master git branch --set-upstream-to=origin/master master git pull
Then launch the webui-user.bat

[dusty garnet]

This guy is a tech genius wtf

[ornate elk]

thank you ^^

[dusty garnet]

Were you created with stable diffusion bro

Thanks tho

Imma try and tell u

[ornate elk]

yes that would work, if the drive is formated to ntfs

you can also install stable diffusion on a PC and only store the large models on the external drive to safe space

[sonic ice]

:/

I don't understand

I keep getting that same connection error. But everything else works just fine. Like no lag for downloads. Speed tests work fine. Connectivity tests say no issue

[ornate elk]

did you made a new folder to git clone into?

[sonic ice]

Yes.

i.. i think i got it to work?

ty for help and being patient with me 😛

[ornate elk]

no problem ^^

[sonic ice]

For sure if it's just a different version of automatic that I was using before. Put in the top right corner where I have the generate button, it used to have the thing for Loras there, but now it's like it's own tab.

[dusty garnet]

Hey man it's still getting stucked! Is my gpu not good enough?

Also it's saying style database not found

[placid torrent]

Can I use 3.10

[ornate elk]

it needs to be 3.10.6 or 3.10.11 or anything between them

[placid torrent]

bruh any ai use a different python version

I just downloaded 3.10

[ornate elk]

i dont know any webui who needs 3.10.0

maybe you installed 3.10.6

that would be ok

[placid torrent]

Like it would be good if you can just use 3.10 3.11 or 3.12

for any web ui

or like other ai tools use another python version

[ornate elk]

any webui will work with 3.10.11

i run multiple ai programms with only that

[placid torrent]

mhh okay but like gpt pilot and gpt engineer im not sure

[ornate elk]

ok, maybe these require an other version

[placid torrent]

bruh

i just installed it and add to path too

[placid torrent]

I downloaded it now

it says python can't be found etc

[ornate elk]

can you show the full cmd ?

[placid torrent]

i started the web ui bat

[ornate elk]

you need to start the webui-user.bat

[placid torrent]

Yeah i mean that sry

[ornate elk]

open up a cmd and type
where python

[ornate elk]

it seems you also have python 3.9 installed

[placid torrent]

Actually I removed it but something is messed up

idk why

[ornate elk]

can you check in the control pannel if you have them installed still?

[placid torrent]

seems not

[ornate elk]

okay

then go into the python 310 folder

and copy the path to the python.exe

then edit the webui-user.bat

at the line PYTHON= you paste the path

[dusty garnet]

.

[dusty garnet]

Have you checked on the set python to path checkbox while installing python

[ornate elk]

then your GPU is maybe to weak :/

[placid torrent]

yeah always

[dusty garnet]

Like I only used like 20 pics

😦

[ornate elk]

does the 1650 has 4 or 6gb vram?

[dusty garnet]

4GB i think

[ornate elk]

that would be to less for training anything

[dusty garnet]

...

Ohh

[ornate elk]

6gb is the lowest with that you can train

for training you would need a google collab training thing or asking someone with a better gpu

or rent a cloud gpu

[dusty garnet]

google collab will work?

[placid torrent]

I cleaned all python version with a powershell script and then installed python again but it still not working

[dusty garnet]

What's yours stable diffusion version?

I'm using python 3.10.6

I also had to keep the file path to the default one recommended by python

[ornate elk]

Make sure you delete the venv folder before Launching the webui-user.bat again

[placid torrent]

exit code: 1

stderr:
C:\Program Files\Python310\python.exe: No module named pip

Launch unsuccessful. Exiting.
Drücken Sie eine beliebige Taste . . .

[ornate elk]

Do you have any third party antivirus installed?

AVG, Avira, Avast, McAffe, Norton, etc?

[scarlet pebble]

So most likely this issue been seen and someoen can easaly fix it but.. how can I fix this UI issue ?
If I do add --Share I cant downlaod the zoom function inside the image2&image

\I tried ot use diffrent browsers same issue

[placid torrent]

I did something and that is why

its working now

idk how but I used some script and add it to the path and it seems working

i also installed manually but i think it did not work because of this

Disable App Execution Alias for Python:

Go to Settings > Apps > Apps & features.
Scroll down and click on App execution aliases.
Find any aliases for Python and turn them off. These aliases can intercept the command to run Python from the command line and might redirect it to the Microsoft Store or do nothing if the alias points to an uninstalled app version.

how to use random seeds

what should i choose

[placid torrent]

why are they no sharp

[scarlet pebble]

Anyone met this problem ?

[grand fable]

hi, could you help me with this error in ComfyUI? It appears only when I choose the "alpha channel", doesn't appear with red/blue/green

[ornate ruin]

@ornate elk good day to you:)

Trying to figure out heavy inconsistencies why my generation timers.

I just had everything working well, had stuff generate within 50 seconds. Now today it takes far over 5 minutes for single images. No other applications are opened on my PC, only web.ui after a complete PC restart

[vocal burrow]

try to restart your sdwebui

[vocal burrow]

what's your hardware ?

[scarlet pebble]

fixed it

needed to edit reg files

[ornate elk]

What's the issue exactly?

[scarlet pebble]

it is even i nthe troubleshoot inside GITBUG

but it was that the Register files

of the CSS was appernalty changed most likely though some of my programs

[ornate ruin]

RTX 4060 Mobile
8GB VRAM
64GB RAM
I7-13700K

[scarlet pebble]

to text/something..

and I needed to retur it to text/css

https://www.bilibili.com/read/cv19519519/

[ornate elk]

Ah okay

[scarlet pebble]

in chinese with translation it worked

[vocal burrow]

what command line args are you using ?

[marsh flax]

hello! i am interested in downloading the stable diffusion ai thing, but im trying to figure out computer requirments, so i dont accidentaly yknow. lag out my pc. can anyone help me out with that?

[vocal burrow]

ok then what does console log says when after one of those 5 minutes generation ?

[vocal burrow]

Any nvidia gpu above gt700 should be able to run it abeit slowly, anything above gt1000 should run it well enough.

it all depends of your workflow

[marsh flax]

... i have no idea what those mean

i have, like. 4 gb of memory. and
let me check task manager rq

[vocal burrow]

were you maybe running on battery ? energy economy mode ?

[marsh flax]

1.8 gpu memory?

gb's

[ornate ruin]

Can you even read that? Why is that screenshot so pixeled for me?

[ornate ruin]

No I'm plugged in on default energy mode

[vocal burrow]

yes I can

[vocal burrow]

and what settings in txt2img were you using for those ?

[ornate ruin]

You don't mean those tho do you?

[fallen zephyr]

I followed the steps in here, but somehow reactor is not showing up in automatic1111

[vocal burrow]

yes, are those the one you were using ?

[ornate ruin]

Yes

Nothing changed, did run it two times then restarted my PC and tried it again, always the same long time

[vocal burrow]

that doesn't mean much. run dxdiag and screenshot "display 1" tab

[vocal burrow]

I don't know then.... were you using the same model before when you had those "fast results" ?

[marsh flax]

where do i put dxdiag in?

[vocal burrow]

start menu

[marsh flax]

i dont have python or anything installed yet. im waiting till

like. i figure if my pc can handle it or not

[marsh flax]

okay

do you need system info?

or somthing else

oh.

[vocal burrow]

"display 1"

[marsh flax]

distplay

[vocal burrow]

I suppose there's no display 2 ?

[marsh flax]

nope

[vocal burrow]

then you have no proper gpu

[marsh flax]

oh my.

[vocal burrow]

Meaning it will run but depending of your cpu it will range from very slow, to extremly slow.

[marsh flax]

i see.

[ornate ruin]

I mean, I always used Realistic Vision 5.1 but I could try it with some others too to make sure. Could also try other samplers again..

[marsh flax]

this is a laptop.

[vocal burrow]

Yes I can see that. It's using intel UHD chipset.

so it's basically a low power gpu included inside your cpu.

[marsh flax]

ah. okay.

[plain axle]

Is anyont running stable diffusion on their EXTERNAL SSD? I kinda need help

[vocal burrow]

it shouldn't matter where it's installed

[plain axle]

I need assistance with the external SSD

[vocal burrow]

Sure, I don't have any idea right now about what could be causing the issue. Maybe @ornate elk will.

[plain axle]

Never had one and just plugged it in, but nothing popped up to install it.

[vocal burrow]

that's because you need to format it using "Disk management"

[plain axle]

Ok, I'm as computer illiterate as they come.

Talk to me like I'm 5

uwu

[vocal burrow]

https://learn.microsoft.com/windows-server/storage/disk-management/initialize-new-disks

[plain axle]

thanks!

[vocal burrow]

(make sure to use NTFS and not fat32 when it asks you for it)

And make sure to select the proper disk of course. you don't want to format another disk.

[plain axle]

Ok, so apparently this is a used disk and it says i can only reformat it, but doesn't tell me how.

[ornate elk]

Did you followed the install guide carefully?

[vocal burrow]

right click on it and delete its existing partition and reformat it ? screenshot what you have

[plain axle]

It's ok, I'll figure it out. Thank you tho ❤️

[ornate ruin]

Maybe. But I have refreshed the Textual Inversions and that seems to have fixed it somehow.. don't know why but it's now sub 60 seconds again...

(it didn't show my Textual Inversions, that's why I refreshed it)

[fallen zephyr]

Yes, but look like I get failures during the start

the react extensions folder also has only 5mb in size

[umbral plover]

When using ultimateSDupscale tile size should be compatible with model or with image that's being upscaled?

[vocal burrow]

use model's dataset size

SD1.5 -> 512x512
SD2 -> 768x768
SDXL -> 1024x1024

[ornate elk]

No insightface found

Which of the 3 methods did you used?

Visual studio,
visual build tools c++,
manually

[cerulean mortar]

why its so bad ??

my all photos are so bad

[ornate elk]

because you use the old 1.5 default model

try using community made models like this:
https://civitai.com/models/4384/dreamshaper

[cerulean mortar]

oh thanks

[ornate elk]

then put the file into the models/stable-diffusion folder

[cerulean mortar]

👌

[ornate elk]

then reload the models and select it in the dropdown

[umbral plover]

Can someone explain or recomennd node for comfyUI that can save batch of images and not mess up their resolutions. Im using Image Save from WAS Suit and it's great but when i give it a batch of images (all of different resolutions) it saves it all as the lowest res of the three, whats wrong?

[cerulean mortar]

where can i download

i cant find

[placid torrent]

someone know why i cant load turbo

in invoke ai

[placid torrent]

help

[placid torrent]

why does this normal turbo mode looks so bad

[vocal burrow]

Wrong resolution (sdxl_turbo is trained for 512x512 it's more akin to SD1.5 than SDXL in some ways), it's also meant to work with a CFG <= 1 and with 1 or 2 steps.

[solar karma]

I haven't used SD in a while, and I go to open it now and get

exit code: 1
stderr:
The system cannot find the drive specified.```

I just uninstalled and reinstalled python, and I can't find anyone else having the same error

[vocal burrow]

Make sure you have python in your path / have checked the "add to PATH" tickbox during python installation.

If you moved your sd install or python, delete the venv folder in SD

[solar karma]

Both are already done, yeah

And I haven't moved either python or SD

[vocal burrow]

What do where python, python --version and py --list-paths say ? (run those in cmd)

[solar karma]

py --list-paths should be python, not py right?

Er, actually I have no result for the third one

[vocal burrow]

no

py --list-paths

[solar karma]

py isn't recognized as a command

[vocal burrow]

what about the two others ?

[solar karma]

C:\Users\DXHHH101\AppData\Local\Programs\Python\Python310\python.exe

C:\Users\DXHHH101>python --version
Python 3.10.6```

[vocal burrow]

ok where is installed your sdwebui ?

[solar karma]

I tried copying it somewhere else to see if it fixed it. So currently it's in two places

It was in C:\Users\DXHHH101\Desktop\stable-diffusion-webui

And now it's also in E:\stable-diffusion-webui

Oh
Deleting the tmp folder fixed it. Weird, there's only a couple files in there, and neither of them have anything to do with anything

[patent meadow]

Stable-diffusionCan't run without a checkpoint. Find and place a .ckpt or .safetensors file into any of those locations.

what too doo?

[ornate elk]

Download a model from civitai.com like Dreamshaper v8.
Then put it into the models/Stable-diffusion folder.

https://civitai.com/models/4384/dreamshaper

[hushed niche]

hello all. I have Stable Studio running at home, I'd like to be able to access it across my network via browsers on other machines, is there a setting to accept connections from non-localhost ?

[vocal burrow]

--listen

[patent meadow]

do i have to add the default v1.5 too?

[vocal burrow]

no

[patent meadow]

because this did not work

[vocal burrow]

share the full log from console then please

[hushed niche]

thanks for the reply, I'm a bit new to this, where do I implement in the --listen command?

[vocal burrow]

in webui-user.bat (assuming you're using windows) there should be a set COMMANDLINE_ARGS= for you to put ... well command line arguments

[hushed niche]

ok thanks I'll go poke around

[vocal burrow]

here's the list of command line arguments https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Command-Line-Arguments-and-Settings#all-command-line-arguments

[vocal burrow]

You can't ignore errors and expect everything to go fine. SD is telling you to update torch at the very beginning of the log.

==============================================================================
You are running torch 1.13.1+cpu.
The program is tested to work with torch 2.0.0.
To reinstall the desired version, run with commandline flag --reinstall-torch.
Beware that this will cause a lot of large files to be downloaded, as well as
there are reports of issues with training tab on the latest version.

Use --skip-version-check commandline argument to disable this check.
==============================================================================

when errors start happening you fix them in their order of apparition.

[hushed niche]

well I may be havinga senior moment (highly likely) but I cannot find a webui-user.bat file anywhere in the StableSTudio install folder

[vocal burrow]

oh you're using stablestudio... my bad I was also assuming you were using automatic1111's stable-diffusion-webui.

[hushed niche]

yep stablestudio

[vocal burrow]

Sorry I don't know about it.

To be honest, I see that the last commit on it date back to 9 months ago....

It might not be the best choice to pick this one.

If you want to pick a "stability-AI" official UI I'd recommend https://github.com/Stability-AI/StableSwarmUI instead

[hushed niche]

that's ok I did actually have an older install of automatic1111 that I can still useinstead

[vocal burrow]

well then now you know what command line arguments to use for this one

And with that I'll go to sleep, good night discord.

[patent meadow]

i did that twice same error

[hushed niche]

I tried to update my install of automatic1111, then it wanted me to update torch, I've done that but now webUI wont launch, here is the error info...

[solemn moon]

Maybe try delete VENV folder, then run bat file again. That’s my “go to trick” for errors 😅

[hushed niche]

Installing torch and torchvision
Looking in indexes: https://pypi.org/simple, https://download.pytorch.org/whl/cu118
ERROR: Could not find a version that satisfies the requirement torch==2.0.1 (from versions: 2.2.0, 2.2.0+cu118)
ERROR: No matching distribution found for torch==2.0.1

earlier it was complaining that my version of torch wasn't up to date, so I installed 2.2.0 and then it complained it was too new lol, apparently it wants 2.0.1

[solemn moon]

OMG lol

[hushed niche]

i'm just going to nuke it and start from scratch

Yeah that fixed it

[old warren]

Why don't I have the permission to send messages in the drawing channel

[old warren]

What should I do

[ornate elk]

Bots are under maintenance.
#1047610792226340935

[valid gorge]

How do I create something similar in ComfyUI as the PNG Info in A1111?

[rancid wagon]

any possible reasons why the checkpoint takes so long too load? and i think its not gonna stop. using the auto1111 rn

[autumn void]

Hey, I don't know if this is the right channel but here goes, I've been trying to make seamless tile patterns but it turns out it's very visible where they connect. Is there a way to correct that? Here's a screenshot of what I mean

[lyric vector]

Are you using HDD for your stable diffusion?

[rancid wagon]

ye

should i not be?

[ornate elk]

How big is the model file?

[rancid wagon]

1.98gb

[valid plank]

I reinstalled SD and dropped my old Model files and embeddings into the folder and now it doesn´t work anymore. can anyone help?

[ornate elk]

Did you placed them where the default folders where?

[valid plank]

and replaced them, yes

[lyric vector]

putting it on ssd or nvme should speed that up a bit

[ornate elk]

So the models are now in stable-diffusion-webui/models/Stable-diffusion ?
And the embeddings in
Stable-diffusion-webui/embeddings ?

[valid plank]

including a controlnet folder, might that be causing the issue, since I didn´t get the extension yet?

[ornate elk]

The controlnet folder needs to be in the models folder

But yea maybe try to cut it out and restart

[valid plank]

so not in that folder:
D:\Stable Diffusion Ai\stable-diffusion-webui\models\Stable-diffusion

[ornate elk]

Yes not in there

[valid plank]

https://tenor.com/view/that70s-show-eric-foreman-light-bulb-idea-gif-4680281

[ornate elk]

Stable-diffusion-webui/Models/controlnet

[scarlet hull]

Hey guys, im having an issue with unclip. I think ive set it up right but the images look nothing like the blog posts im following. They look cartoony for some reason. Any ideas?

[placid torrent]

What is the best UI?

I think Invoke AI has more features

[ornate elk]

Automatic1111 as it has the biggest support for extensions and models.
Comfyui for people who want the Full Control of everything.
Invoke has a nice design but for auto1111 there is the Lobe Theme extension which is quite nice too

[placid torrent]

Okay and what about fooocus

[ornate elk]

Oh right, Fooocus is the best in terms of simplicity.
Its for the people who used Midjourney before and just want a artistically good output with less fiddling

But its not good for lower end PCs, because Fooocus relys on SDXL models primarily

Then there is also Stableswarm-ui
Its a combination of settings like seen in auto1111 but with the Backend of comfyui

[valid gorge]

I've tested A1111 and Comfy so far and to me A1111 is better and less hassle free. I'm also getting better results with A1111 and I'm able to work faster. At this point I don't see any reason why to use Comfy in my current workflow. SDXL Turbo will change things for sure, it's great with Comfy.

[jolly pawn]

how should I write this

[vocal burrow]

replace + with (a space)

[jolly pawn]

still no work

[placid torrent]

okay thx

[vocal burrow]

what s the console log ?

[placid torrent]

I think Invoke AI is not so confusing than A1111

like holy

[jolly pawn]

ah xformers

done

[wooden iron]

i resolve this problem if anybody wanna know then ask

[wide sonnet]

Hello SD community, I load someone comfyui workflow with UltralyticsDetectorProvider bbox/Eyes.pt, I don't understand where I can find it, google doesn't know about it

[placid torrent]

Hey is there an AI which can help me build prompts?

[rugged latch]

me has quesitionnnnnnnnnnnnnn

;^;

how do i get that to happen

i dunt rememebr mai actions

and im trying to help a friend achive the same

[elder turtle]

What GPU are you using?

[rugged latch]

AMD 5600 apu

also known as

Radeon

is the memory monitor just automatic when not using an Nivida GPU ovO

im just tryna replicate the actions i did to disable memory monitoring

was it the --precision full cmd?, or the -nohalf, or maybe its the cuda test, i dunno

[vocal burrow]

not happening, there s no memory monitoring for AMD hardware on windows

[elder turtle]

Seems like you installed Nvidia related software or drivers?

Wait no, it says its a lack of nvidia drivers

[elder turtle]

You can check the A1111 wiki to see what which flag does

[karmic crown]

i am extremely busy at work, trying to finish the project on time, so i do not have time to be here, but i see that you manged to solve your problem
i am glad i could help

[fair oxide]

hi, is there any way to limit my gpu usuage on sd? i got a pretty okay computer and id like to generate in the background without dedicating all my gpu to it

[ornate elk]

You installed the wrong webui version. You need the one for AMD
Checkout the Pinned Messages of this channel for the AMD Guide

[vocal burrow]

also that yes

cf pinned messages

[rugged latch]

nu...

das not wut me asking

<v>

im not having any issues ;v;..

[ornate elk]

Your running it on CPU only mode so I'll guess you have slow speeds

[rugged latch]

im asking, how to get the memory montior diaabled message to appear for my friend, so they dont have an memory runtime error

[vocal burrow]

that not how it works....
"memory monitor disabled" => sdwebui has no way to monitor the vram usage.
Disabling the monitoring will not fix memory runtime error

[rugged latch]

oh

[vocal burrow]

Just because you re closing your eyes while driving at a wall doesnt mean you will go through it

[rugged latch]

well

i thought it meant it wasnt keeping track of memory usage, and that it would continue to genarate, and not error out, whethert it has the ram or not

u-u

[vocal burrow]

Not that I know of. You can underclock your gpu or set a timer to pause between each rendering but it will always use everything that it can.

[fair oxide]

ouff, that's a shame, but thank you for the reply!

[rugged latch]

ya ty

[vocal burrow]

error happen for a reason, trying to force any program despite it throwing errors is asking for disaster

[rugged latch]

well

its like

my friend has nivida, and stable by defualt is desgined for exactly the type of PC they have, and they get a runtime error for insuffient

vram

but, i, , a pc with no graphics card

all AMD apu, dont get the runtime error

when i try to make an image at 2048x2048 res

[vocal burrow]

Probably because they re using it wrong. I cant fix your friend's problem without their log

[rugged latch]

where would that log be

i can try to ask for it

from them

is it just printed on the termianal

im asking them to replicate the error, and to ss the terminal

[vocal burrow]

it s in the terminal

having them post in here directly would be easier for everyone involved

[ornate elk]

Because your running it on CPU only mode

And what's your friends GPU ?

He may just need some commandline args in the webui-user.bat

[rugged latch]

,_, im (not) having issues X3

also she*

also i think itds nivida 30 sum

[ornate elk]

The vram amount would be good to know

[sonic ferry]

ComfyUI question, I have a 'Upscale Latent By' followed by a 'Latent Noise Injection' going into my second KSampler, what exactly does the 'Latent Noise Injection' do? Does it add onto the Denoise I have set in the KSampler? I actually followed a guide which had that after a ControlNet for hands, don't understand the reasoning for it and there wasn't much of an explanation as to what it adds to it.

[rugged latch]

im assuming, based on context, thats 8GB vram on their card

[ornate elk]

Yes thats right.
Therefore she needs to add:

--xformers --medvram-sdxl --no-half-vae
To her webui-user.bat

Then relaunch the webui-user.bat and it should work

[rugged latch]

oh oki , thx chu :3

[ornate elk]

no problem 🙂

[tepid crypt]

well, hello, im looking for help to optimize and generate images faster, it usually takes a few mins to generate a small image (512x512) with euler, im using a model called realistic vision, hope someone knows what to do to make it better, these are the parameters im using:

also im using opera gx

[ornate elk]

hey, whats your gpu?

and commit hash: none means your webui is maybe not updated, but in anyway cant be autoupdated

[tepid crypt]

how do i do that?

also is

also

[ornate elk]

okay, is SD installed on C drive or on D, E ?

[tepid crypt]

E

[ornate elk]

okay, is it a Laptop or PC ?

[tepid crypt]

Pc

[ornate elk]

okay, and how long does an 512x512, 30 steps, euler a image takes for you ?

[tepid crypt]

like 3 or 4 mins

[ornate elk]

okay, thats very long, your using the realistic vision 2gb model or the 6gb xl model?

[tepid crypt]

realisticVisionV60B1_v51VAE

thast the name

it neccesarily has to do with the model? if i try another ones would go faster?

[ornate elk]

smaller models are faster on gpus with low vram like yours. so dont use sdxl models or any models that are larger than 6gb

but realistic vision is 2gb, so thats ok

[tepid crypt]

i see, so besides the model, any other way to generate faster? any config or something?

[ornate elk]

yes, you should reinstall SD based on my install guide on your SSD, (because yours isnt able to update anymore, (you later can link your models from the hdd) so that your ssd space doesnt get filled up.

you also need to make sure to have enough free space on the C drive, best would be 15-20gb

then the last step is to check the windows pagefile settings, these are important for the overall system performance,

[tepid crypt]

what does it have to do?

[ornate elk]

the windows pagefile is needed for caching and is like an "virtual ram"
its usually stored on the C drive, but sometimes windows changes it to the other drives. and thats your HDD. if the pagefile is on an HDD it will slow done every process that requires an good amount of RAM

and because you only have 8gb of ram that could be the case here

[tepid crypt]

i see, i think i changed that a long time ago, thats why it says i have 31 gb

[ornate elk]

yea its good if you can check that again, so your sure its only enabled for the C drive and not for D or E

[valid gorge]

Cool, maybe I’ll try it out. 🙂 As far I know Invoke can be sligtly slower compared to A1111 or Comfy so I haven’t bothered.

Comfy can be a bit comfusing (sorry, pun inteded lol) at first but you’ll get hang of it quite fast as you just dive in and start testing things out your self. So far I haven’t seen any benefit why to use it but I might change opinnion later when I need something more inctricate.

[upper scaffold]

Okay, I'm having a world of a time trying to get Kohya's GUI installed. For some reason, my python version isn't changing? I think that's the problem. I've downloaded 3.10.9, changed the environment varible path to 3.10, and I still get an error on the setup.bin. Any ideas of what I'm doing wrong?

[placid torrent]

What is the best sd fine tuned model?

[ornate elk]

You maybe need to delete khoyas venv folder

Then it will get rebuild with the new python version

[upper scaffold]

I've deleted the entire folder a few times and started fresh with the same issue. But I'll try again!

Is there a way to force Windows to use a certain version of python within Command Prompt?

[ornate elk]

i dont think so, do you have multiple versions installed?

[upper scaffold]

I should only have 3.10.9 installed, but this is what i see

 -3.11-64 *
 -3.10-64```
and then
```>python --version
Python 3.8.9```

[ornate elk]

check in the control pannel or under settings apps

if you find any other python versions to uninstall

[upper scaffold]

only ones

Wait! I did a search and found 3.8.9 in an old app. Weird

[pallid geode]

Hi, I unfortunately always get this error when trying to install sd

updating pips doesn't work

and I have python 3.10.6

[upper scaffold]

OKAY! So, an app called Inkscape had python within it's folders. I removed that, restarted, repaired the 3.10.9 version of python, and now it's working.

[vocal burrow]

run pip cache purge in cmd

[pallid geode]

done, still getting the same error

[vocal burrow]

what's the full log then ?

[ornate elk]

try delete the venv folder

[pallid geode]

I do that every time before I run the webui file

[vocal burrow]

that's weird it seems like it stops the download by itself very early and consider it 100% downloaded...

just to be sure how much free space do you have on C: ?

[pallid geode]

116GB

[vocal burrow]

what's in your C:\Users\USERID\AppData\Local\pip\cache ?

search for torch-2.0.1 in there

[pallid geode]

holy shit, I think I got it to work after months

idk why updating pips suddenly worked now when it didn't then but I'm glad

thank you guys for your help

[vocal burrow]

well glad you got it sorted out

[pallid geode]

yeah I was gonna resort to google colab but Im glad I didn;t

[tardy sparrow]

Hi everyone, I discovered Stable DIffusion today and I did pretty well (i mean, for a beginner) but I have a few questions if you know how to answer them.
I've figured out (more or less) how to know the keywords of LORA(s), but I can't find any explanation on how to know the keywords of MODEL(s) (and the ones I find are incomprehensible because they're too technical for me)
Would you know how to know these keywords?

Another question: I've read that the order of keywords and lora calls in the prompt is of no importance, because everything ends up being read in a row
Is this true?

[onyx solar]

Models don't have keywords, really. They are the larger chunks of data that are inferred from to generate the content you want. You just draft your prompt with the thing(s) you wish to see.

[tardy sparrow]

Okay, thank you

[vocal burrow]

You can't really guess "trigger words" be it for Loras or Models. Some have them, some don't. Whatever you're using read their description/readme before installing them if there is a "trigger word" it should be listed in there.
And yes some models have trigger words but they're still models in the end so you don't necessarily need to use them.
Does the order matter .... tough question ....

Let's take an example. two shots, same settings, same seed, same everything... But only one prompt with two parts switch around for each output. So detailed anime cute cat, cloudy skyin one case and cloudy sky, detailed anime cute cat in the another

Without looking at the metadata can you guess which is which ?

It definitely had an impact on the output. But the overall "idea" of the prompt remains in both case.

If you want to emphasize some aspect of your prompt, you're probably better off using increasing/decreasing the weight of some words/loras, rephrasing some parts of it than switching its words. At least this way you'll have some control on the output.

[tardy sparrow]

Thank u very much for this explanation

To increase the weights of some words, u have to put them into () right ?

like ((((sky))))

[bitter quartz]

Why won’t it let me use the bot channels? It just keeps saying “check out tech support”

[tardy sparrow]

or ((((cloudy sky))))

[vocal burrow]

check #1047610792226340935

(the big red dot kinda gives it away)

[vocal burrow]

it's all explained in the wiki https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#attentionemphasis

[tardy sparrow]

perfect

thats exactly what i was looking for

thanks

[tardy sparrow]

Where are theses "readme" or "description" ? all i have is safetensors file which i have to open in notepad to know the keywords

it's kinda a pain in the ass, i'm sure there's a easiest and fastest way

[vocal burrow]

it should be on the webpage you downloaded them from

[late tree]

So just curious, with AMD on Windows, will RocM just work (WHEN it's actually supported), or will you need to add --use-rocm?

[fair oxide]

@ornate elk this video explains basic comfyui functions so well im feeling tempted to dive into comfyui https://youtu.be/aL4xrbg4NjU

[fair oxide]

fair to say these two resources can help learn and explore comfyui from basic to advanced levels

1. https://www.youtube.com/@c0nsumption/videos
2. https://openart.ai/workflows/academy

[unique glacier]

Join us. ComfyUI is love. ComfyUI is live.

[unique glacier]

It is pretty simple once you understand how the node work. Making a basic txt2img and img2img is easy. It only complicated once you add on a bunch of shit like control net, face swap, detailer and all kind of other stuff.

[fair oxide]

im trying it out now lol, but as for controlnet that's a necessary for a lot of my workflow, i will also need something like adetailer

[unique anchor]

Can anyone recommend a good Google Colab Notebook for Kohya SS LoRA Training?

I've tried the one from kohya github page but it is too complicated for me to understand

[tribal spoke]

Hi. Am getting "Failed to extract snapshots for the custom node" when I try to upgrade all in comfyui, anyone get that ?
I have it on two puters, the other one has conda and made and env for it and it's running ok. This one doesn't have a conda env and is acting up.

[sonic ferry]

I got an issue with my upscales in ComfyUI, it's morphing, I got first KSampler using Euler_Ancestral+Normal Scheduler, denoise 1, then sent through a 'Latent Upscale By 1.5x+Noise Injector' set to 0.1, then KSampler #2 set to dpmpp_2m+Karras denoise 0.5, and last the Ultimate SD Upscaler, which I had set same as second KSampler set to dpmpp_3m+karras, denoise 0.3...am I unknowingly setting myself up for mutations? The background looks fine, face looks fine, hands look fine, is it my LoRA maybe?

[tribal spoke]

I searched but the only hits that come up are from the comfy git repository

[unique glacier]

Nice, i am typically using workflow with facedetailer also.

I have my own server where we share several comfyui workflow. If you want. I can invite you.

[junior pilot]

Why i cant use the bots

[fair oxide]

that's cool, but im very new to comfyui so i need some time on my own to get familiar with the basics

i mean really the basics 🙂

[vague compass]

how can i disable the queue in --api, i have already added --nowebui flag

[arctic zinc]

I'm using mac and I get the error UnboundLocalError: local variable 'h' referenced before assignment

[ornate elk]

--no-gradio-queue

[zenith widget]

Lora Files Not Loading

Good Morning, ive only recently downloaded stable diffusion web ui and today is my first day using it. im trying to use loras however i get this error. Does anyone know how to solve this issue?

[zenith widget]

seems to be fixed now, i removed the command --use-cpu all and the lora shows

[ornate elk]

Hey, What's your GPU ?

[zenith widget]

NVIDIA GeForce GTX 1650 Ti

[jolly pawn]

my controlnet should have been latest version but it's still missing some control types

[vocal burrow]

restart your UI or check console logs for proper error message

[ornate elk]

Then you need only
--medvram --xformers --no-half
In your webui-user.bat

[ornate elk]

It says 18.06.23 in your Screenshot.
So its outdated by half a year.
You need to delete the controlnet extension and reinstall it

[zenith widget]

alright sweet ill take a look and try those out, i appreciate the help CS1o and Aryetis ❤️

[lost crescent]

/how-to-resolve-torch-not-compiled-with-cuda-enabled

🙂

[ornate elk]

Depends on the full error log

[lost crescent]

Where would I find that please?

[ornate elk]

Just the full cmd output

There is no log file, but the full cmd output would help much more

[ornate elk]

And your using an nvidia gpu?

[lost crescent]

RTX 2070

[ornate elk]

You can run the update-torch bat inside the comfyui folder

[lost crescent]

I think I need torch2.1.0 cuda118

[ornate elk]

It should be automatically installed when launching the start bat

Also make sure your on the latest nvidia driver

[lost crescent]

I do not see torch.bat in Comfy folder

I installed Cuda Toolkit 12-3

Search found no torch.bat

[ornate elk]

Can you show the comfyui folder contents?

Also cuda toolkit 12.3 is not supported

11.8 would be the best cuda toolkit to have

But none of them is needed for comfyui or auto1111

[lost crescent]

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

I'm installing now ...

OK, I will rollback to CUDA-TOOLKIT 11.8

I need to set TRANSFORMERS CACHE as an Environment Variable?

[ornate elk]

No clue on that error

[lost crescent]

OK but I thin it's here https://stackoverflow.com/questions/63312859/how-to-change-huggingface-transformers-default-cache-directory

[sweet moat]

https://github.com/THUDM/ImageReward?tab=readme-ov-file#integration-into-stable-diffusion-web-ui

how do i install this

i did paste the file

it doesnt seem to show up

[ornate elk]

It says exactly that you have to copy the script inside the scripts folder

And not into extensions

Then restart and it will show up in txt2img under scripts at the bottom

[sweet moat]

oh damn

mb

thank you

[ornate elk]

Np

[sweet moat]

do you know why the image reward score is not being shown?

it did load in scripts in img2txt

oh wait

it doesnt even work while generating images

[ornate elk]

It says that it needs to load stuff before the first run

So maybe try again

Or check the CMD Console

[glossy sage]

Why don't I have premmission to use the bots?

[vocal burrow]

#1047610792226340935

[unreal valley]

can anyone help me with this issue?

would like if there is a video for the fix with this

[ornate elk]

What extension do you want to install?

Reactor ?

Did you followed these steps?
https://github.com/Gourieff/comfyui-reactor-node?tab=readme-ov-file#installation

[unreal valley]

yeah

[lost crescent]

ComfyUI InstantID - AttributeError: 'StableDiffusionXLInstantIDPipeline' object has no attribute 'dtype'

[autumn void]

Hi, for some reason running Juggernaut XL causes my PC to bluescreen and reset. Has anyone got a solution for that?

[ornate elk]

Hey, how much RAM, and Vram do you have?

[coral tangle]

why is my img2img generation in comfyui bleeding?

txt2img generation works perfectly fine btw

I've been playing around with cfg, steps, samplers and schedulers for over an hour and I can't seem to fully fix the bleeding

also it worked few generations before the first bleed

[autumn void]

16GB ram and 12GB Vram

[ornate elk]

Are your drivers updated and your windows?
Also check if the power supply has enough power for the GPU.
And then make a photo of the bluescreen error so that mostly gives a hint on the Problem

Also open up a cmd and run:
SFC /scannow

[tacit light]

yo, questions so the command arguements we use to run the amd version. What are the limitations that we're working with for that? --use-directml --medvram --opt-sub-quad-attention --opt-split-attention-v1 --no-half-vae --upcast-sampling

[autumn void]

Drivers are up to date, PSU has overhead and I'll try updating windows again (it's a fresh install)

It happened before the cleanup and happens now as well. I'm guessing it's not a hardware issue then?

[ornate elk]

The limitations depend on your vram amount.
But if you have more than 12gb you can create up to FullHD images in txt2img with hires fix. And upscale to 4k with img2img

[autumn void]

btw it's only with this particular checkpoint

[ornate elk]

So other sdxl models work?

[autumn void]

yessir

[ornate elk]

Is your webui updated?

And what's in your webui-user.bat?

[autumn void]

think so it's a fresh install too, from yesterday

I followed the pinned tutorian in here

tutorial*

[tacit light]

I have 16gb how do I use hires fix and also I meant limitations as in is the generation limited in other ways then just resolution? like the way it works with models etc

[autumn void]

so it's got --xformers --no-half-vae in it

[ornate elk]

Ah okay, sdxl models will work but without hires fix. Otherwise it could error out

There is no other limitation, every model and Extension I tried works on AMD

[autumn void]

here's the SFC results

[tacit light]

Yeah I am using SDXL however it takes like 100 seconds for 30 steps with 1024x1024

[ornate elk]

Okay so it repaired something

[tacit light]

okay

[autumn void]

Seems so, should I try running juggernaut xl in the ui again?

[tacit light]

how do I use hires fix with sd 1.5, Is it better than waiting 2 mins per image for SDXL models?

[ornate elk]

Hires fix is slow so it will take 2 minutes maybe too.
But the result is cleaner than the sdxl output

You can try with these settings:
Resolution: 512x768 or 768x512 or 512x512
Steps:30
Hires fix settings:
Hires steps: 10
Denois: 0.5
Upscaler: Resrgsn4x+
Upscale by: 2

[tacit light]

okay

Also the method you told me to upscale with, Does it only work if I'm on Sd 1.5 as well?

like the sd upscale script

[ornate elk]

You can try, but if only that model has problems, I would redownload it

[ornate elk]

SD upscale script works with sdxl images too

But therefore you need to set the resolution in img2img to 768x768 or 1024x1024

[autumn void]

Will do and get back here then

[sonic ferry]

omg, my cfg has been too high!!! I'm loosing detail in the background now using 2cdg but it's hard to test like this. Hell, id take a 5 minute gen. lol

I was going up, 3, 3.5, 4, wasn't seeing a big enough difference to know to go back the other way....how was I doing 7cfg before , 20 steps vs 12? is that the reason? is it a ratio?

[tacit light]

it works yeah but i need to be on 1.5 for the script to work otherwise it fills the image with random tiled images lol

[autumn void]

MEMORY_MANAGEMENT bluescreen just occurred

I think I've solved it... by disabling XMP in BIOS

It's loaded and I generated something with it so I guess crisis averted?

[ornate elk]

You need to set the denois value below 0.3

[ornate elk]

XMP is something that boosts your PC performance by letting the RAM use its full potential

So if its only on model that makes errors I would rather have XMP enabled

Better is to check your Pagefile settings

[autumn void]

The paging file size is 5993MB

Think that's the default

[ornate elk]

Check it in the pagefile settings, best is System Controlled for C. And disabled for all other drives

If you have enough free space on C (15-20gb).
You can set the minimum size to 16000 and max to 20000
And then restart the PC.

[autumn void]

I'll see if this changes anything with XMP back on. Thanks for the tips boss.

[lost crescent]

ComfyUI InstantID - AttributeError: 'StableDiffusionXLInstantIDPipeline' object has no attribute 'dtype'

[gilded merlin]

Hey friends. I’m a noob trying to run ComfyUI on my M1 MacBook. Every time I reach the KSampler it is killing python. Can some instruct me on what to do to fix this… or, nicely tell me that my computer won’t be able to pull it off. Haha

[unreal valley]

got the reactor node to work and was on my way to face swap but now I have this

[severe kindle]

We'll need an error message to have clue. But my advice is try another sampler there are one of two that doesn't work with MPS in certain situations (step count below 30 IIRC).

[unreal valley]

I changed the detect gender to no instead of female and it worked

[karmic agate]

Hey guys. I have been using an RTX 3060 card (6GB) in my laptop. I'm going to migrate to a desktop and can choose between an RTX 3080 (24GB) or an RTX 4080 (16GB). The greater amount of VRAM on the 3080 seems interesting to me as it allows me to do more things, even if a little more slowly. What is your opinion?

[tardy sparrow]

Hello

My face restoration button is missing from A1111

i swear i had it yesterday

can someone help ?

[tepid crypt]

does a1111 have an option like pixai that lets you enhance and upscale a generated image?

[tardy sparrow]

@tepid crypt i think so

pretty sure thats this option

[tardy sparrow]

does anyone knows if the prompt in hi rez fix activate options ?

like this for example : a master jedi anthropomorph dog in star wars with a lightsaber, Hires upscale: 1.6, Hires upscaler: R-ESRGAN General 4xV3, Face restoration: GFPGAN, Denoising strength: 0.3, wearing a jedi cloak hood (dog: 1.3), battlefield background

" Hires upscale: 1.6, Hires upscaler: R-ESRGAN General 4xV3, Face restoration: GFPGAN, Denoising strength: 0.3"

[umbral plover]

Hi, i have a problem with faceDetailer in comfyui. It doesn't change anything on face, hair on the other hand changes like crazy, did i setup something wrong?

[umbral plover]

Another problem face mesh doesn't detect faces ;d

[dense sandal]

does someone know how to fix this?

What am doing wrong?

oh, is it a comma after the first prompt?

[ornate elk]

3080 has only 10gb,
a 3090 has 24

More vram is better for any ai stuff

But 4080 would be faster

[ornate elk]

Yes its called Hires fix

In auto1111