🤝｜tech-support

[solemn atlas]

how do i update my nvidia dribver?

driver*

[ornate elk]

Oh you have hires steps on 0.
Thats not good. Always set it to 10. thats why it crashed before.

And the denois to 0.5 for better output

[ornate elk]

To get the latest driver. Select your GPU here and then you can download it:
https://www.nvidia.com/Download/index.aspx

[solemn atlas]

i did but still doesnt work

maybe something is here wrong

[ornate elk]

Restart the webui and try again

Also show me the webui version at the bottom

[solemn atlas]

okay

[ornate elk]

Are you running anything in the background ? Like a game or Wallpaper engine?

[solemn atlas]

no nothing

maybe this is the problem

[ornate elk]

Nope the error is vram related

Try the image again but with hires fix disabled

To test if that works

[solemn atlas]

that worked somehow

[ornate elk]

And now enables hires fix and set the hires steps to 10. and the denois to 0.5 and the upscale by to 1.5

[ornate elk]

Also make sure the webui is whitelisted in any browser adblocker

[solemn atlas]

that worked too

[ornate elk]

Okay and now upscale by 2

[solemn atlas]

it didnt worked

sorry my internet broke down

[solemn atlas]

OutOfMemoryError: CUDA out of memory. Tried to allocate 2.23 GiB. GPU 0 has a total capacty of 11.99 GiB of which 4.19 GiB is free. Of the allocated memory 5.72 GiB is allocated by PyTorch, and 806.14 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
Time taken: 32.2 sec.

[ornate elk]

Okay, then you need to install the tiled vae extension

Then enable tiled vae and let its settings on default. Then it will work

[solemn atlas]

tiles vae, where do i get that?

[harsh kelp]

I run the webui stable diffusion. It used to work wonders but now it takes more than an hour to generate an image

[ornate elk]

In the extensions tab. Click on available, then on load from

Then search for tiled vae

Ans install it

[ornate elk]

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

Also which settings like resolution did you used?

[harsh kelp]

ive tried many, but small

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --disable-safe-unpickle

call webui.bat

3070ti

[ornate elk]

Remove --disable-safe-unpickle
And instead add --xformers --medvram-sdxl --no-half-vae

That should make it run much faster

[harsh kelp]

thanks a lot

worked wonders

[ornate elk]

No problem 🙂

[solemn atlas]

i somehow cant find it

[ornate elk]

Its called tiled Diffusion & tiled vae

So search for tiled and it should appear

[solemn atlas]

where can i search for it?

i somehow cant find it

[ornate elk]

Click on load from

Then click on the white empty line below the "extras"

[solemn atlas]

ah there it is

so i searched for it but i dont have that option tiled Diffusion & tiled vae

[ornate elk]

Maybe you already have It installed?

Check the installed tab

Also click there on "search for updates"

[solemn atlas]

thats what i have

[ornate elk]

Yep you already have it

Its old name is Multidiffusion

Then update the extensions

Then restart the webui

And then in txt2img you should have the Tiled VAE option

[solemn atlas]

now it worked

what was the problem, was this the vram?

thank you for your help

I hope it works from now on

[frank garden]

Hi:-) Im making a Comfyui Workflow for looped Seed Travelling, by interpolating noisy latents (Starting and ending in the same) and then using Ksampler without add_noise. Also, for low vram uses I didnt want to batch all the keyframe latents, so im using a list instead of a batch with rebatch latents. So far it is working great.

But, now i want to introduce Prompt Travelling too. The problem is that Fizznodes looks designed to work either in Batches or in consecutive prompts in queue. When i try to use it with lists, it only applies the first prompt (Because the batches are size=1).

[ornate elk]

Yes the last step is the VAE step. Which requires much more vram than the normal steps.
Tiled VAE reduces the vram usage by tiling the VAE.

[frank garden]

I was wondering if i could convert a batch of conditionings to a list of conditionings. And, in that case, if the Ksampler is going to apply them in sync with my list of batches.

[solemn atlas]

so i need to enable that all the time?

[frank garden]

Or alternatively, if the increment of a primitive worked in lists, everything would be solved...

[solemn atlas]

i'll let it as it is i guess

[ornate elk]

For upscaling yes

[solemn atlas]

okay thank you

[ornate elk]

Hey, you could ask that in #🧣｜comfy-ui
Here aren't much comfyui users in tech support. Most are in the comfyui channel

[frank garden]

Oh ok, thx!!

[magic widget]

does anyone know where I can find some good examples of character datasets for lora training?

[solemn atlas]

now i have this problem

[ornate elk]

At which settings?

[ornate elk]

Is maicha model SDXL based?

[solemn atlas]

no but it didnt worked before with normal vae too. the problem is now i cant even change vae , it stays on sdxl

[ornate elk]

Look down there is a flag set for the other vae

[solemn atlas]

what's flag set?

[reef sparrow]

Is there anyone expert in LoRA? It's not loading. Need help!

[ornate elk]

What's the issue?

[vapid sluice]

Why is the resultant image so oversaturated when using the latent to latent method in comfyui (i can’t show the workflow cuz its slightly nsfw)

The workflow by a latent place on Youtube

[vapid sluice]

I only modified a bit by adding new prompts to the second k sampler

[misty marten]

I recently reinstalled stable diffusion and it's giving me this error. Before formatting the PC and reinstalling it, it generated images normally, can anyone help me?

RuntimeError: mat1 and mat2 must have the same dtype

[ornate elk]

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

[misty marten]

i have a RX6600, @echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--opt-sub-quad-attention --disable-nan-check --skip-torch-cuda-test

call webui.bat

[ornate elk]

Oh you installed the wrong webui version.
AMD cards need a different one.
You'll find my install guide in the pinned messages of this channel or here: #🤝｜tech-support message

[tacit bone]

Brand new to this and I started with Foooocus. Does anyone know why it is utilizing my C: drive that Windows is installed on instead of the drive I actually installed it on? Is this because the temp folder being on C:? How can I change this?

I found two posts with similar questions but both of them installed through the github program thing. I didn't, I just installed from the download link. and I do not want to learn what a git pull or a pip is

[ornate elk]

Hey, when installing any webui it downloads stuff in the cache for the installation. This is always on C.
To get your space back open up a cmd and type
Pip cache purge
And hit enter

That will remove the cached files

[tacit bone]

My concern is that it is incredibly slow, because my C drive is meant for simpler stuff and it's a SATA drive rather than my 4 TB nvme that I actually have focus installed on

Is there no way to change what it's writing through?

and again I don't have a github client thing

[ornate elk]

What's your GPU ?

[tacit bone]

3080S

[ornate elk]

Ah okay. Fooocus is a webui which only uses SDXL models. These models require a lot of resources. On fact you need to have 32gb RAM and 8gb vram to use it nicely.
But even if you have enough RAM it will still uses the windows pagefile.
This file is always located on the C drive by default.
What you need to do is to set it to your nvme drive.

Disable it for C and only enable it for the nvme and set it to system managed. Then make sure to have 15-20gb free space on it
Then restart the PC.

[tacit bone]

Is this guide right for the pagefile change? https://www.windowscentral.com/how-move-virtual-memory-different-drive-windows-10

[ornate elk]

Correct 👍

[tacit bone]

Alright thanks. lets see how this does. Also I meant 2080S not sure why I thought it was called 3080

shows you how often I change stuff with my PC build lol

[olive gale]

Anyone reason why I'm not seeing models in controlnet?

I downloaded the pth files and put them in the models folder

[ornate elk]

Models/controlnet

Then click the reload icon

Also you need to have compatible models.
For sdxl you need different controlnet models than for 1.5

[olive gale]

Got it, thank you so much!

[olive gale]

Still very new so I don't understand this exactly

[ornate elk]

There are 1.5 based stable-diffusion models and sdxl based stable diffusion models

If a model is 6gb its mostly an sdxl one

If its 2gb-4gb its an 1.5 one

Controlnet needs compatible model version for these two variants

[misty marten]

That was it!, I installed it using your guide and it worked, thank you very much!

[ornate elk]

Perfect, no problem 🙂

[worthy spindle]

I'm trying to test a Lora train on a model. But the model doesn't seem to accept it? Is there a reason why? Or what I should do IF I would like to train on "SDXL1.0_LoRA_Photo2Style_Pony_6"

training on any other base model seems to work.

[balmy lily]

hello friend

i have the question

.

how do i write a batch script, as in, how can i queue a batch of say 20 differently generated images, but specify how each one ought to be generated. for instance its easy to make it queue a batch where each time it changes the seed. but how can i make it so each time it switches the checkpoint to the next one, and stuff like that

[misty marten]

Does anyone know how to use two lora's characters in one image?

[acoustic saddle]

using webui, possible to have the refiner use a different VAE than the initial checkpoint?

I want to use an sdxl model and then switch to an sd1.5 model, the sdxl model's VAE won't work with the sd1.5 model.

[gritty pumice]

Hi I'm getting a 500 internal server error when I try to do image-to-image generation.

[tall vault]

Hello I purchased a stability membership but now every time I try and generate something and stable artisan it tells me I don't have a membership. Please help.

[brittle lantern]

@ornate elk I did a fac reset again for other reasons u wanna try zluda again?

[brittle lantern]

meant to be like this for 3 whole min?

[timid notch]

HI,can someone tells me how to open a new confyui in my computer? i tried but can only open 1 confyui window, how to let two windows coexist?

[lost crescent]

I like ADetailer in AUTO1111 as it deals with ALL the faces in a picture.

Is there a similar node in ComfyUI - which too will fix ALL faces?

[lost crescent]

Clipdrop SD3 is much more economical for $10/month than Artisan.

[lost crescent]

Open another tab on your browser and type Localhost:8188

[timid notch]

ok thanks i will try

[lost crescent]

U need 2 KSamplers with an SDXL VAE first; then the SD15 VAE in the second KSampler

[wet flame]

I would like to know that too wanna work together maybe we can make a script like that

Currently I'm working on a script that takes a inputs from a file a text file and reads the file what are the text from that I have written maybe it will be a Excel file instead and the prompt reader will read the what are the prompts and it will input in the generator for automatic 111

I think it should be a extension for stable diffusion

You know python?

What I'm trying to do is say I want to generate different kind of vegetables

But if I put the same text prompt it will only generate carrot not chili cauliflower etc etc

[balmy lily]

its a feature that could be created with a custom node, aswell as the ability to convert an image made in automatic1111 into a workflow of comfyui, also a feature to have the generation data copyable from images found on civitai be able to be pasted into comfyui and turned into a workflow

[wet flame]

Oh I don't like comfy I don't like anything node based

Actually dislike comfy very much

[balmy lily]

chatgpt can write most any python script

[wet flame]

Okay thanks I'll try out that

[balmy lily]

i think thats already a thing https://www.reddit.com/r/StableDiffusion/comments/17t2d83/batch_prompt_schedule_only_takes_first_prompt/, im not sure how its done in automatic1111 tho

[wet flame]

It already exists in automatic 111 you just drag as csv file

And here I was looking for a script

[lost crescent]

I'm having an error in ComfyUI

The size of tensor a (91) must match the size of tensor b (90) at non-singleton dimension 3

[ornate elk]

Yes it can take up to 40 minutes at this point

[brittle lantern]

Yh it's done that

And I finally can gen images

Ty

[ornate elk]

Perfect np 🙂

[brittle lantern]

Macbook can be used as portable or backup now

I knew it's got to do with my shitty windows installation

[ornate elk]

It was the only thing left ^^

[brittle lantern]

Yep

I was like aint no way its not this

@ornate elk last question sorry if I download any other model my first use won't also be up to 40 min?

[ornate elk]

Nope. Thats a first time only thing.
It can happen again when the zluda files get updated

[brittle lantern]

Ok ty and what's some good extensions to get for QoL on SD?

[ornate elk]

An important one is Tiled Diffusion & Tiled VAE.
Its needed for upscaling to larger resolution without getting out of memory error. Therefore only activate Tiled VAE and your fine.

Boorutag autocomplete extension for tags and prompting.

Controlnet is very powerfull and has so many features. Needs additional files to work.

[brittle lantern]

Ok ty and I've installed 4x ultrasharp upscale and put it in the correct earth I've reloaded ui and reopened the bat even but it doesn't show up why?

Downloaded all 3 mentioned

[ornate elk]

It needs to be in models/ESRGAN

Then it will show up when you use any upscaling features like hires fix in txt2img or SD upscale script in img2img

[brittle lantern]

It is

As I've done it before on my mac

But it's not showing

[ornate elk]

Have you restarted the webui too?

[brittle lantern]

Like mention before done UI and relaunched the Bat

[ornate elk]

Ans when you select hires fix its not there ?

In upscalers

And you got the upscaler from here?
https://openmodeldb.info/models/4x-UltraSharp

[brittle lantern]

From hugging face

[brittle lantern]

It's not no

[lost crescent]

The size of tensor a (91) must match the size of tensor b (90) at non-singleton dimension 3
problem in ComfyUI

[ornate elk]

That can be done with Tiled Diffusion extension. It has regional prompt control feature

[ornate elk]

Thats not possible but you can merge the VAE into the model with the checkpoint merger tab

[ornate elk]

Hey, you have to link the account to discord to generate I think

[ornate elk]

@wet flame This extension should be able to do that:
https://github.com/ArtVentureX/sd-webui-agent-scheduler

[brittle lantern]

Ok so I've deleted the 4× and reinstalled it from Hugging face put it back in did the same thing

It didn't show

So idk what's going on

[ornate elk]

Try the one from my link

[brittle lantern]

Ok I'm just head into the garden rn it's 9:15 AM I always go outside for some morning sun if that's ok my fren

[ornate elk]

Sure np

[brittle lantern]

Unf recently I stay home quite often so I need to leave the house for some air (u prolly don't care I just wanted to explain why)

[dry geyser]

Hey, newbie +1 here. I git clone comfyUI local and started playing with img2vid model. my gpu is 3060ti and the loading speed of the KSampler seems super slow to me. it takes about 150~160 seconds in average. is it normal? I want to know if the workflow is using cpu or gpu and if it is using cpu, i'd like to change to gpu. HELP PLZ!!!!!

[brittle lantern]

@ornate elk yep that link works

It's actually so much faster than my macbook with Zluda on my 6600 xt

Thank f*ck bro

[pliant cipher]

Hey, will a 4090 power consumption max out while using inference or training? I want to know if using a 750w PSU with a 85% undervolt is ok for this usecase

[ornate elk]

No problem 🙂 great to hear it performs that good
Maybe you can check the it/s you get on an
512x512, 30 step, DPM++ 2M
Image with only Cat as prompt and blurry in negative

Without upscaling

[rocky bear]

Hi.. How do I get started making video and images? Is this all public? Or can I create a private space ?

[ornate elk]

Hey, it depends on what you use.
If you use the paid credit service Stable Artisan, then you can use it in the channels for it or add it to your own discord server.

If you want to generate images and videos for free then you need to install a local webui on your PC.

Guides for installing a local webui to get started using Stable diffusion localy, can be found in the Pinned messages of this channel

[rocky bear]

@ornate elk Thank you -- i've already signed up with Stable Aritisan and I have created a server. I cant find an app called stable artisan .. how do i add this ?

[ornate elk]

I asked the team about the discord server integration. Will reply when I know more. For now checkout theses channels:

FAQ: #artisan-faq
Image and video generation: #1237459938901491852

[mortal compass]

I need your infinite wisdom @ornate elk 😩 You mind taking a look at a text file of my SD boot?

[ornate elk]

Hey, sure

[solemn moon]

let me try too

[mortal compass]

Thanks! Okay so I'll try to remember what I did exactly.
I was installing some extensions, (sd_civitai_extension, BrushNet). I was also reinstalling IF_prompt_MKR because I thought it might not be working as intended. After I did that, all of these errors started to show up, and now some extensions are not loading properly.

lol not sure why the text file looks like that xd it's readable if you download it

[solemn moon]

no idea but a prompt is missing

[whole grove]

Hey there! I'm starting to use SDXL1.0 by API, is there any supported SDK to use?

[mortal compass]

Can you be more specific? I'm not sure what that means, are you talking about arguments?

[solemn moon]

positive what you want to see, negative what you dont want to see

maybe you can post your webui-user.bat file

open with text editor

[mortal compass]

Oh lol right, yeah well I just started up the UI, the errors are showing in the terminal and my problem is that some extensions stopped working even though I didn't touch them (sd-webui-prompt-all-in-one)

Sure

[solemn moon]

need @ornate elk the hero

[mortal compass]

My only hope

https://tenor.com/view/helpme-saveme-starwars-obiwan-obiwankenobi-gif-5693087

[mortal compass]

Actually I had another question you might be able to help me with! I've seen the argument --api pop up here and there. What does it do and do I need it?

[solemn moon]

if i was in yr position, with no immediate response, i would disable any new extensions and delete venv folder and then rerun .bat file

https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/API

[mortal compass]

I'll take that into consideration! Gonna wait for CS1 and see what he sais, and if he doesn't know I'll delete venv

[solemn moon]

yea it juist replaces it

[mortal compass]

I was here 20 min ago haha. Didn't really understand it. I'm not a programmer, might be why

[solemn moon]

same bruh 😅

[mortal compass]

Thanks for trying 😊

[solemn moon]

what graphics card you got?

my .bat file looks way simpler

@echo off

set PYTHON="C:\Users**\AppData\Local\Programs\Python\Python310\python.exe"
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --xformers --no-half --medvram --api
git pull
call webui.bat

[mortal compass]

NVIDIA GeForce RTX 3070 Ti

I'm also using SD Forge, that fork doesn't use medvram and no-half vae I believe

[ornate elk]

Okay it seems like you installed Brushnet but it isnt a webui extension. Its not compatible.
So it needs to be deleted.

Also your reactor extension has an error. Hopefully you installed it via its github guide. If not you need to reinstall it with that.

Also mov2mov has a known bug that has to be fixed manually by following this here:
https://github.com/Scholar01/sd-webui-mov2mov/issues/146#issuecomment-2054157728

[ornate elk]

--API is needed for third party programms like Blender and Photoshop to get access to the webui. Also for server stuff like discord bots

[mortal compass]

Well that makes sense. I wasn't sure if brushnet was for the sdUI or if it was a stand-alone 😅
I believe I did install it that way. Might aswell just reinstall it and see if that fixes reactor.
Okey I'll just delete it, wasn't keen on downloading 30 gb or whatever

[mortal compass]

Do I need an actual api key in the argument, or is just --api needed? Is it like an activation argument, if that makes sense?

I know you need to put api keys into settings if you want civit to work in webui, not sure if there's other important steps to it

[ornate elk]

No key needed as its your local webui

Its just an argument to open the webui up for other programms to access it

[mortal compass]

Oh, of course

[mortal compass]

Right, then I understand! Thanks! Gonna follow your steps and see where it leads:)

Almost forgot @ornate elk, I noticed a line in the terminal "Device: cuda:0 NVIDIA GeForce RTX 3070 Ti : cudaMallocAsync". I don't know shit about computer language, but device: cuda:0 leads me to believe that my cuda is inactive. How wrong am I?

Everything works now! My hero ❤️

[solemn moon]

https://tenor.com/view/clapping-yay-excited-soexcited-greatnews-gif-8845875809066863059

[ornate elk]

No its active. If you look at the taskmanager under the performance tab you will See that your GPU is GPU0 sometimes you have also GPU1 if a CPU has an integrated GPU.
Cuda:0 refers to the first GPU in your system that has cuda support

[mortal compass]

Ah fantastic. Bit strange imo that they set it to 0, in my mind, 0 = null. But what do I know 😄

Have a great day, thanks again

[ornate elk]

Thx, you too 🙂

[wet flame]

how do I include a negative prompt in my text to prompt?

[solemn atlas]

what does that mean

[ornate elk]

What's the model you try to use?

[solemn atlas]

is something with this wrong? My generations are also a bit slow than usual

[solemn atlas]

oh it was a normal one and the other was xl sorry

[ornate elk]

What's your GPU again?

Maybe you can provide the full cmd output too

[solemn atlas]

4070S

I want to send you but my computer freezed because of Stable diffusion. I need to wait a a few minutes

that happens a lot these past days

it also once happend that my computer shut down

[ornate elk]

I see no error in the log

[solemn atlas]

maybe something here is wrong? should i change there something?

[ornate elk]

No thats good

[solemn atlas]

There must be some problem because SD always crashes

[tiny vale]

Help Needed. Trying to lauch a feature to a National client today and seeing period 504s with some APIs.

[pastel kettle]

Hi, I can't seem to get xformers working on 1.5, I keep getting this error

    PyTorch 2.1.2+cu121 with CUDA 1201 (you have 2.0.1+cu118)
    Python  3.10.11 (you have 3.10.6)
  Please reinstall xformers (see https://github.com/facebookresearch/xformers#installing-xformers)
  Memory-efficient attention, SwiGLU, sparse and more won't be available.
  Set XFORMERS_MORE_DETAILS=1 for more details```

[magic widget]

does anyone know why I am getting ValueError: too many values to unpack (expected 2)?

I reckon it has something to do with optimizer arguments, but I have no idea what, I believe they are properly entered

[ornate elk]

How much ram do you have?

[ornate elk]

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

Your webui version might be outdated

[solemn atlas]

32,0 GB installed RAM

[ornate elk]

okay then check your windows pagefile settings.
Make sure its only enabled for the C drive and disabled for other drives. also it has to bet set to system managed

[pastel kettle]

3060m

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --no-half-vae --autolaunch --xformers --update-check
git pull
call webui.bat```

[magic widget]

I found the issue and fixed it by editing some python code, it seems it doesnt correctly parse given arguments, it just takes it as a list with one string in it so I had to take that string and split it by commas, and then reconnect parts where comma was inbetween brackets to fix it

[ornate elk]

then also add --medvram
then delete the venv folder and relaunch the webui-user.bat

[ornate elk]

then xformers and torch should be self updating

[solemn atlas]

it says its C

[ornate elk]

go in the pagefile settings

[solemn atlas]

here

[ornate elk]

ah wait

[ornate elk]

your pagefile is to large

its 98gb wtf

[solemn atlas]

oh .. the system did it automaticlly but why

[ornate elk]

yea it shouldnt go that high

mine is big and its 32gb

never seen 98gb before

[ornate elk]

okay first check that your have the latest windows updates.
then open up a cmd as admin and run
sfc /Scannow
that will repair corrupt os files.
Then reboot and check the pagefile again

[solemn atlas]

my system is really weird., i bought my pc in february and SD always has issues every now and then

[graceful loom]

Hey i need to composite my video to look futuristic is their a way i can do that...it will be like 2-5min clip

[solemn atlas]

it's the latest windows update. It automaticlly updates

[solemn atlas]

what are OC files?

[ornate elk]

OS = Operating System

[solemn atlas]

I'll also have the option to unable but should i leave it as it is? the cmd is running right know

[solemn atlas]

thank you

[solemn atlas]

it said that it repair corrupt files

this is the update

[ornate elk]

okay and you restarted the pc?

[solemn atlas]

no, I'll do it now

[ornate elk]

please click restart and not shutdown

so it does a clean reboot

[solemn atlas]

I did

[ornate elk]

looks good now

[solemn atlas]

will it stay like this now? what did it do with what I did now?

and should I always enable tiled vae

[ornate elk]

only enabled it if you also use hires fix or upscale

[solemn atlas]

i always use it, so i should use it too, but what is the diffrence to tiled diffusion?

[ornate elk]

tiled diffusion is like generating an image in tiles, while tiled vae only tiles the vae at the last step

the vae step is the step where it can crash if the resolution is to high

using tiled vae doesnt mean that you can now generate like 8k images with hires fix

[solemn atlas]

okay so I need tiled vae but not diffusion at this point

[ornate elk]

yes

[solemn atlas]

thank you so much, i hope now my SD works

[ornate elk]

no problem 🙂

[drifting pendant]

Wise CS1o, it it's definetly a bad motherboard problem, I will by a new one next month 😉

[ornate elk]

ohhh okay

how did you found out?

[fair oxide]

Hi, I don't know where to ask this but, how do I fix this issue?
I'm trying to make app that generate images using my basic knowledge of c#

[fierce tartan]

Is this server still alive?

[ornate elk]

Yes it is

[drifting pendant]

Just by using it at factory settings, weird behavior, I cannot explain it fully, sometimes it's impossible to get out of sleep, I need to reset the BIOS to boot, very frequently, to factory settings, but I was in factory settings before... weird stutters at low CPU usage in games
My undervolt / overclock is more stable than the factory, that's why I was in trouble to diag this... And above all, none of the symptoms I've described before...
I've tried "reasonable" high overclock too, Stable at all levels BUT if I shutdown my rig, sometimes, Impossible to boot to any system, black screen, fan @ max level, I have like 14 years of exp in this, never seen this kind of sh*t, and sometimes I was "kinda" extreme with my rigs
And losing the drive by generating stuff etc ? Dahek bro

[ornate elk]

Have you made a BIOS update?

[drifting pendant]

Yeh, all of them, when they are released 😉

Same for my drivers etc... Except Windows I read updates before, and sometimes wait for review of peeps

[ornate elk]

oh okay

what brand is it?

[drifting pendant]

https://www.gigabyte.com/fr/Motherboard/B550-AORUS-PRO-rev-10

[ornate elk]

ah well gigabyte... xD

[drifting pendant]

I agree

[magic widget]

I am trying to train a lora and am wondering what does 2390 mean here?

is that a total amount of steps?

I have 18 images, repeats 12, and epoch 10, and batch size 4, while both max train epoch and max train steps are on 0

so how did we get that number?

My goal is to make epochs smaller to get sample updates quicker

So I don't know what I should change

[magic widget]

by sample updates i mean sample images, i want to get them more frequently, once per epoch so my aim is to make epochs have less steps each

[magic widget]

nvm I am just dum, I was changing the repeat count which created multiple subset folders which in turn gave me a ton of training images

why do I have a feeling that when training epochs are getting progressively slower? first epoch is done relatively quickly but the others are getting slower and slower

and my ETA is slowly increasing

[sinful needle]

Question: Is score_9 only for pony diffusion or is it for any model?

[ornate elk]

its only for pony based models

[zenith sail]

How would I convert a string to prompt for the combitorial node?

[unique bluff]

Hello again

I’ve tried generating an image but it just became static

Is there any way to fix that?

[wet flame]

how can i turn all my image more anime like btw in batch process?

any suggestion?

i can do it in single image prompt but that is not what i want i want to scan all my image in the folder and turn into anime style

currently it creates broken images when i do that

also is there way to read the image and generate description from it?

and maybe use that in the promp generation text input to generate the images

[wet flame]

is there any small app that just reads the image and describes its style etc by reading an image

[half island]

Hoi, i'm at a corner with comfy atm, i almost got it fully working on steam deck lol, and it has 1GB pre-shared in bios and several GB free in ram which it can expand upon to use more, but comfy for some reason can't see past 1GB. Games can see past it just fine, just comfy that can't. Any way to mitigatemake it use more ram as video memory?

File "/home/deck/anaconda3/envs/comfy/lib/python3.12/site-packages/torch/nn/modules/conv.py", line 456, in _conv_forward
return F.conv2d(input, weight, bias, self.stride,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
torch.OutOfMemoryError: HIP out of memory. Tried to allocate 540.00 MiB. GPU 0 has a total capacity of 1024.00 MiB of which 358.00 MiB is free. Of the allocated memory 379.17 MiB is allocated by PyTorch, and 54.83 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_HIP_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)

[dreamy saffron]

Excuse me, where can I check my seeds, everyone？

[sudden hawk]

hello, excuse me, I have again the grey images but don't understand cause I have no error message in my webui, and it's happen only with two checkpoint

[wet flame]

nvm figured out a way to do this

just try training a lora but don't build you get the tags

[ornate elk]

What checkpoints are they?

[sudden hawk]

animeIllustdiffusion/ ariesFP16/ pavomixFp16NoEma/ subtoonBf16NoEmaNoVae

[ornate elk]

Very strange.
You can try add --upcast-sampling to the webui-user.bat

[exotic garden]

Hi, Is there anyone available for a short discussion of integrating tripoSR API with node js app

[sudden hawk]

ok I did, it has work for 1 images on each checkpoint and now it doesnt work anymore I have this error mesage in my webui /
Error running process: C:\A1111\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\controlnet.py
Traceback (most recent call last):
File "C:\A1111\stable-diffusion-webui\modules\scripts.py", line 784, in process
script.process(p, *script_args)
File "C:\A1111\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\controlnet.py", line 1245, in process
self.controlnet_hack(p)
File "C:\A1111\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\controlnet.py", line 1230, in controlnet_hack
self.controlnet_main_entry(p)
File "C:\A1111\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\controlnet.py", line 901, in controlnet_main_entry
model_net, control_model_type = Script.load_control_model(p, unet, unit.model)
File "C:\A1111\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\controlnet.py", line 369, in load_control_model
control_model = Script.build_control_model(p, unet, model)
File "C:\A1111\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\controlnet.py", line 379, in build_control_model
raise RuntimeError("You have not selected any ControlNet Model.")
RuntimeError: You have not selected any ControlNet Model.

[ornate elk]

Check your controlnet settings

It says no Controlnet model selected

[sudden hawk]

I have this msg now, should I force ?
2024-05-14 10:18:07,281 - ControlNet - INFO - ControlNet UI callback registered.
C:\A1111\stable-diffusion-webui\venv\lib\site-packages\huggingface_hub\file_download.py:1132: FutureWarning: resume_download is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use force_download=True.

[mortal compass]

Mornin! A new day, and a new oppertunity for me to screw something up in stable diffusion 😎 Got this bad boy of a message today. No idea what this could be, not much on google regarding it.

But I suspect it has something to do with me installing extensions lol

[ornate elk]

That message can be ignored

[ornate elk]

What extension did you installed?

[mortal compass]

I've lost count at this point. There's too many damn extensions that looks fun 😖

[ornate elk]

Does your webui still boot up?

[mortal compass]

When I first booted it, the UI had time to launch, but I could only see the UI's "skeleton" if that makes sense. No text, no buttons to press. Loaded it up again, and now it seems to be working? Except that I got the exact same error as yesterday with the "prompt all in one" API and UI unable to load

Did not get an error msg in terminal either

[ornate elk]

You shouldn't install multiple extensions at once ^^

[mortal compass]

Whaaaaaat

[ornate elk]

Haven't seen this error before. So its mostly from an extension I don't know

But you have 2 civitai extension's installed. Thats not good. Only the Civitai helper plus should be used

[mortal compass]

Hmm, right.. I think I installed that one when I started using SD, and I think I didn't like it or something. This is the one you are referring to, right? https://github.com/BlafKing/sd-civitai-browser-plus
Looking at it right now

I'll give it a go

[mortal compass]

eh voila, everything as good as new! Saint CS1 strikes again 😇

Just curious, no need to answer if you don't want to: what background do you have regarding computers?

[ornate elk]

Np, im working as a System Administrator and also my Hobby is mostly PC stuff ^^

[crimson bridge]

Good morning,
I tried to install reactor on stable diffusion but it doesn't work, can someone help me please?

[ornate elk]

Hey, Unchecked the "disable all extensions" button.
Also what's your GPU?

[crimson bridge]

I deactivated and reactivated it doesn't work? the extension does not appear on my terminal either.
it's a mac M1

[ornate elk]

Have you read the Install guide of reactor? It needs additional steps to get it working

[crimson bridge]

I used a tutorial on youtube, do you have another installation guide link on mac please ?

[crimson bridge]

I used a tutorial on youtube, do you have another installation guide link on mac please ?

[ornate elk]

Can you show the complete terminal log?

Here is the install guide of reactor:
https://github.com/Gourieff/sd-webui-reactor?tab=readme-ov-file#installation

It should work for Mac I guess

[crimson bridge]

@ornate elk terminal
Thanks

[ornate elk]

Yep its missing insightface

So you need to do the install step with vscode

Or by manually install the whl

[crimson bridge]

Do you have a tutorial link for this please? Thank You

[ornate elk]

Nope but these steps would be the easiest:
https://github.com/Gourieff/sd-webui-reactor?tab=readme-ov-file#viii-for-windows-users-if-you-still-cannot-build-insightface-for-some-reasons-or-just-dont-want-to-install-visual-studio-or-vs-c-build-tools---do-the-following

[grizzled zealot]

Hi Guys, I‘m currently working on a project to put my customised a1111 with a simplified UI on my own website. I don’t have any knowledge in python or similar, just the basics in html, css and JavaScript. I just want to run my local a1111 on my public website. Can anyone help me getting started ?

[ornate elk]

Hey, therefore you need to put --API in the webui-user.bat
So that you can use the API on your website

[grizzled zealot]

Ok so I ran it with this command. Should there now be a API key or something like that? Sorry I‘m really a complete beginner

[ornate elk]

Can't help much with API sry.
But here is information on how to use the API:
https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/API

@grizzled zealot you said you want to customise the ui. So you would have to use the API I guess

If you just want to let friends access the webui. You can do that by using --share --gradio-auth username:password

Then you get a link that works 24h

[grizzled zealot]

Ok thank you very much, I‘ll have a look

[mortal compass]

Anyone here familiar with Dynamic Prompts? I can't seem to get the combinatorial sampler to work. My prompt is "A {cat|dog|mouse} sitting on a {chair|branch|skateboard} and eating,{apple|banana|hamburger|fish}". Supposedly, this will generate all the possible combinations for these synthaxes, but I'm only getting one image 😫 Got all the settings enabled, now I'm out of ideas

[lost crescent]

Pixart Sigma - How do you make a T5 Text Prompt with a String (as opposed to Conditioning) as output?

[lost crescent]

Batch count needs to be a multiple of all possibilities - and you have 3 x 3 x 3 possibilities - so 27 should cycle you thru all eventualities

0 means 0

[zenith sail]

Anyone know how I would convert the string to a "Prompt" in the combinatorial propts node I have there?

[simple crow]

Can I get a great tutorial to animate a video? I am able to do great animation with an inage but I wish to do animation on a video and it looks really complicated to play around with ComfyUI, it's really not comfortable as the name suggest 😂😂

[mortal compass]

Oh really?? I mean now that you say it, it feels obvious. I'm one of those who take Readme's very literally, I feel like they kinda missed that part 😄

Not sure if I understand it fully, here it sais "will ignore batch count". I see it as "put this at zero if you want to run all the possible outcomes" 🤔

[iron hound]

keep on getting this error

whenever installing AUTOMATIC1111

[polar crater]

when i run SD webui it don't want to load some loras and none of embeddings.

[ornate elk]

Thats because you didn't git cloned the webui and instead downloaded it as zip

[iron hound]

fatal: could not create work tree dir 'stable-diffusion-webui.git.': Invalid argument```

oh i am severly neurodivergent

i forgot to remove the dot at the end

[ornate elk]

Yes

[iron hound]

after installing:

ModuleNotFoundError:
No module named 'jsonmerge'
Press any key to continue . . .```

[dapper sigil]

what ai model is best for creating/training a hyper realistic person/model?
im new to sd
any tips appreciated

[wet flame]

hey how good is zluda

nvidia cards are so expensive i can't afford that

but i see that 16 gigs of gpu for about 450 dollars

[barren shard]

Helllo can anyone help me out with stable diffusion? I try to generate a pic but it's showing a blink picture

[last oyster]

hallo everyone!When using Stable Diffusion's Deforum, clicking 'Generate' has no response?

[ornate elk]

Checkout my install guide in the pinned messages of this channel.

[ornate elk]

Zluda is very good. The performance is not as good as it would be on Linux with rocm, but with newer GPUs its very nice.

Still for ai stuff nvidia leads. But the prices are insane

[wet flame]

yeah i think i will settle on the 4060 ti

cost like 700 CAD 😦

i do need an upgrade from my old 4 gb amd

[ornate elk]

How much for a 7900xt or 7900 GRE ?

[wet flame]

7900 1k

gre 746

I think since i am going to mainly use it for AI stuff and all that nvidia is better

I am learning AI developer stuff like auomation and etc really interested in that

lot of nice job opportunites too i think for it

like 80k+ per year for sure

[ornate elk]

Yep if ai is important then go for nvidia

[ornate elk]

Crazy prices aswell

[gleaming steeple]

I generated an AI image and I want to change it into realistic photo. How can I do that??

[drifting pendant]

Hello there !
Does anyone knows about any good documentation, I want to learn more about all the technical stuff and AI / Stable Diffusion un general
And above all, I want to stop asking dumb questions

[edgy heath]

really dumb question but comfyui automatic1111 fooocus whats the difference?

a.k.a which best

[ornate elk]

What's your GPU ?

[ornate elk]

Depends on your use case and your hardware.
Fooocus is for people who came from midjourney and don't want to mess around with to much settings to just get good images.
But it requires a GPU with 6gb vram or more. Better 8gb.

Comfyui is a node based webui. With that you can make your own workflows for image generation.
Its very complex but you can do anything in it. But its not for beginners.

I would highly recommend starting with Automatic1111, as its the most used webui, supports low to high end GPUs. And the usage can be simple and more advanced the more you want to try. Also most yt tutorials are for auto1111 or comfyui. Also most extensions are made for auto1111

[ornate elk]

Hey, that can be done in Img2img or with controlnet IP-Adapter

[edgy heath]

So comfyui is like a stick shift and automatic 1111 is well.. automatic gear shift?

so comfyui is in theory, better, but for the average person, automatic1111 is better and faster?

[gleaming steeple]

Can you provide me some more detail??

[ornate elk]

Yes, exactly. But comfyui and Auto1111 have on average the same performance.

You can also install multiple webuis and compare them to your liking

Some people use comfyui for specific workflows of video generation. And then for the image based stuff they use Auto1111 (both can do video stuff with extensions)

[ornate elk]

Are you using a local webui? Or an other service?

[fierce hollow]

How does it work? Why is there only one video and I get an image?(
https://civitai.com/models/416775/faceswap-albedobase?modelVersionId=464339

[gleaming steeple]

I am using local web

[fierce hollow]

help(

[manic bridge]

I am new to Stable diffusion and trying the image-to-image API from Postman but it is giving the following error {
"id": "2b42fc09d44e633d5c409b12df78279f",
"message": "can't decode multipart body",
"name": "bad_request"
}

I am following this format - https://platform.stability.ai/docs/api-reference#tag/Image-to-Image/operation/imageToImage

[mortal compass]

Impossible to know when there's no information on the model's webpage

[fierce hollow]

is there a way to create a realistic video in SD? That is, in order to distinguish from reality, it was impossible not to be a professional

[mortal compass]

I've heard about AnimateDiff and mov2mov, but that stuff is kinda difficult to do if you are a beginner

[fierce hollow]

Animated gif is complete nonsense, sorry! He is not able to make a realistic image, each new frame is drawn a little differently.

[mortal compass]

Well okay, I'm only saying that I've heard of it, not if its good or bad. I'm only into image generation

[acoustic wind]

Does anyone know if it is possible to get the "live preview" images in A1111 to save as either frames or a .gif?

[manic bridge]

Help(

[ornate elk]

Yes there is an extension that can do this

But I don't know its name anymore :/

[ornate elk]

Found it, its called Steps Animation

[lost crescent]

How do I defeat "the undefeatable"?!

WAS Nodes will under no circumstances load up in my ComfyUI

[frail bobcat]

Anyone know how to solve this error?
modules.devices.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.

Never had this error before.
Did recently install Kohya_ss, did this do something?
Error occurs when using adetailer.
I'm at a loss.

--disable-nan-check gives black pics

the --no-half commandline argument did not help either

[ornate elk]

Hey, what's your GPU?

[frail bobcat]

4090

NVIDIA GeForce RTX 4090, sorry

always have to check what it exactly is

[ornate elk]

Then remove --disable-nan-check and --no-half
Only add: --xformers
Then check the model file size. If its smaller than 2gb its not a model

Oh you said it only appears with Adetailer enabled

Make sure your extensions are updates via the extensions tab

Also try an other model

If you still get the error, you can share a screenshot of your txt2img settings

[frail bobcat]

Alright, and need to update controlnet I see

[barren shard]

I have a Nvidia and I've put the --medvram thing in the notepad just to make sure

[frail bobcat]

Also, in 3rd pic at the bottom, what is this progress?

Updating controlnet didn't solve the problem.

Forgot to try on another model, used an golden oldie but same problem.

[frail bobcat]

So, updating controlnet didn't work, using another model produced the same problem, removed the --disable-nan-check and --no-half, screenshots shared (I think you can disregard the 3rd one-removed it)

[acoustic wind]

Have you tried different base checkpoints?

Sometimes I have encountered also faulty VAEs.

[frail bobcat]

Base checkpoints? Sorry, I'm a beginner so the term checkpoint does ring a bell but its far far away.

I did try another base model with the same error as result.

[ornate elk]

How much vram does it have?

And what's the GPU model

Because some cards need additional commands

[ornate elk]

And when you set the batch count to 1 ?

[frail bobcat]

Got this from Steam system info:

Video Card:
Driver: NVIDIA GeForce RTX 4090
DirectX Driver Name: nvldumd.dll
Driver Version: 31.0.15.5244
DirectX Driver Version: 31.0.15.5244
Driver Date: 5 7 2024
OpenGL Version: 4.6
Desktop Color Depth: 32 bits per pixel
Monitor Refresh Rate: 59 Hz
DirectX Card: NVIDIA GeForce RTX 4090
VendorID: 0x10de
DeviceID: 0x2684
Revision: 0xa1
Number of Monitors: 1
Number of Logical Video Cards: 1
No SLI or Crossfire Detected
Primary Display Resolution: 3840 x 2160
Desktop Resolution: 3840 x 2160
Primary Display Size: 27.44" x 15.43" (31.46" diag), 69.7cm x 39.2cm (79.9cm diag)
Primary Bus: PCI Express 16x
Primary VRAM: 24563 MB
Supported MSAA Modes: 2x 4x 8x

[ornate elk]

Thats okay. The question above was to the other guy.

But please try again with batch count 1

I think adetailer might have an issue together with Controlnet when the batch count is set to more

[frail bobcat]

batch count 1 works

[ornate elk]

okay, a user a week ago had the same problem

there seems no fix for this strange issue right now, but its the same bug for everyone

no fix other than using batch count 1, or disabling either controlnet or adetailer

[frail bobcat]

Alright, thanks for the assistance.

But at least the use of 1 works and I didn't realize that, so also thanks for that!

[ornate elk]

no problem 🙂

[noble juniper]

After updating ComfyUI I am unable to input prompt

I have tried to use incognito mode, other web browser. I've found on GitHub that supposedly (https://github.com/comfyanonymous/ComfyUI/issues/3279turning) disabling "Anything Everywhere highlight connected nodes", in the options should fix that but I cannot find that option anywhere

Changing workflows, restoring workflow to default didn't help

[barren shard]

Really? Last time I did it (been a year) it worked fine. Not sure if I gotta update python or git hub sine I haven't used it in awhile, but I'll check once I'm finish

[misty marten]

Anyone know a good VAE to improve eyes and faces?

[noble juniper]

It's really frustrating coz it generates

But there's no way to preview the prompt

Or edit it

[azure marsh]

im trying to use kohya_ss and torch is not detecting my gpu at all i have an amd rx 5700 what do i do im on windows 10

[ornate elk]

What's your GPU then? ^^ I can tell pretty quick what arguments you need

In the pinned messages there is also a guide for Nvidia amd and which specify what to use

[ornate elk]

You can ask that in #🧣｜comfy-ui
Maybe someone there knows how to fix it

[barren shard]

I'm checking now turning on my laptop owo

[noble juniper]

I restored backup with previous version and it's working

[ornate elk]

Hey, with that GPU you can't use Khoya SS. But you can use Onetrainer. It has AMD Zluda Support.
But to make that work you first have to follow my AMD Zluda webui guide. Its in the pinned messages of this channel.
With the zluda webui version of automatic1111 you can generate images pretty fast even with amd. Not like the slow directml variant.

[azure marsh]

thank you so much. it has training right?

[barren shard]

@ornate elk hopefully I got it right..but it says RTX 3050 laptop GPU and the vram is 8

[ornate elk]

No the webui is for generating images. But when you have set it up you can install OneTrainer and it will use the same Zluda files. So that it works with training.

You could also skip the webui part and only follow the zluda steps. Then install OneTrainer.

[barren shard]

Funny thing is in uninstall python and git hub last night cuz it wasn't working

[azure marsh]

what would you reccomend? i need to train my own style and generate images using it, right now i have normal stable diffusion and kohya_ss but ill probably remove both of them for what you are talking about.

[ornate elk]

@azure marsh Khoya won't work with AMD cards at all :/
Also the directml webui is slow and not optimised in vram usage.
I would recommend to use my guide for the zluda webui version.
Ans then installing OneTrainer, to make a lora of your style.
But I can't promis that OneTrainer works (but it should) as I haven't used it for myself. (Still planing to do)

[ornate elk]

Oh xD then you need to install.them again

[azure marsh]

okay cool thanks, ill set it up and let you know what happens

[barren shard]

Oh yeah definitely do it later x.x If it doesn't work then oh well

[ornate elk]

a 3050 8gb vram needs --xformers --medvram-sdxl --no-half-vae

[barren shard]

Sdxl? Huh never put that in

[ornate elk]

its needed for 8gb cards to gen faster with sdxl models

[azure marsh]

how do i install onetrainer? i followed the instructions from the pinned message, and ive been waiting a bit for it a generate an image. id like to get ahead a bit and start install onetrainer to get it over with

i made SD-Zluda on a different hard drive from my OS

[ornate elk]

you need to go into your SD-Zluda folder (not into the stable-diffusion-webui-directml)
there click into the explorer bar (not searchbar)
then type cmd and hit enter.
then copy and paste this command:

git clone https://github.com/Nerogar/OneTrainer.git

Then run the install.bat
if its finished, run the start-ui.bat

And here is the OneTrainer site, in the wiki is a usage guide:
https://github.com/Nerogar/OneTrainer

[azure marsh]

okay thank you

[ornate elk]

no problem, let me know if it worked

[azure marsh]

i got it all installed and it appears to be working but i have no clue what to do. i have around 40 images already captioned, do you know how i go about training the style? im having trouble finding a guide for this

[ornate elk]

there are mostly no guides, only the quick start guide on onetrainers github wiki

[azure marsh]

oh i think i got it

last time it all went to not respoinding but i jsut checked the console and its doing something

[ornate elk]

you mean OneTrainer ?

[azure marsh]

yea i dont think its using my gpu though

there was a part at the start where it said cpu and i didnt change it to gpu ill try changing it after its done making the model

[lean ridge]

If i have a 3060 gpu of 12 gb. I Should use any command argument for faster generation?

i already have: --no-download-sd-model --xformers --no-half-vae --api

I've come back to automatic1111 but i remember my generations was faster

[ornate elk]

nope your fine with these arguments

make sure your nvidia driver is updated

[lean ridge]

oh ok ty!

[azure marsh]

i tried replacing cpu with gpu and i got this error. how do i know which one is mine?
also i figured out the training images thing i was messing with a dataset tool and i needed to create a configuration concept.

RuntimeError: Expected one of cpu, cuda, ipu, xpu, mkldnn, opengl, opencl, ideep, hip, ve, fpga, ort, xla, lazy, vulkan, mps, meta, hpu, mtia, privateuseone device type at start of device string: gpu

[oak mountain]

Guys anyways i can turn on the auto que option in comfui by default ?

[azure marsh]

i was told to change it to cuda and cpu but then i got
RuntimeError: CUDA unknown error - this may be due to an incorrectly set up environment, e.g. changing env variable CUDA_VISIBLE_DEVICES after program start. Setting the available devices to be zero.

[fickle panther]

Hey guys, sorry to bother everyone hope everyone has been having a good wednesday night. Unfortunately I've been having a problem using automatic1111, during high res steps, generation slows down to 2.6s /it down from average of 16+. I am running a 4090, so i feel like i should be getting more than that and I previously was getting more than that

my current args which have no changed since before the issue started

[random pagoda]

is there any way to set a lower cfg than 1 in auto1111

i want 0.5

[fickle panther]

"txt2img/CFG Scale/minimum": 1.0

change this to a negative value in the ui-config.json

[random pagoda]

ty

last time i tried this with the resolution it doesnt work, i will give it another try, hopefully it works with cfg 😄

[fickle panther]

good luck :)

[random pagoda]

it works , also the res works too now :D, thanks a lot

[fickle panther]

np dude anytime, enjoy

[surreal sentinel]

Do I need a separate install of python external to run SD on my PC, or is venv enough?

[calm cairn]

Here to ask about XL and XL Turbo models
When I run them, I get a cool picture, up until the very end at which point all the colors suddenly go bad.

Is this a mismatch of VAE with model? Wrong sampler? Wrong settings?

Thanks in advance, XL seems to be coming a long way fast and I'd love to be able to use it

[dim sequoia]

Hi, I'm trying to make requests to the Stable Diffusion 3 API, and I'm getting 400 content-type must be multipart/form-data, but what I've done is exactly as described in the documentation.

I've isolated the issue with requests:

from dotenv import load_dotenv
import requests, asyncio, os

load_dotenv()
STABILITYAI_API_KEY = os.getenv("STABILITYAI_API_KEY")
STABILITYAI_ENDPOINT = os.getenv("STABILITYAI_ENDPOINT")

payload = {
    "prompt": "A cat using a toaster.",
    "output_format": "png"
}
headers = {
    "authorization": f"Bearer {STABILITYAI_API_KEY}",
    "accept": "image/*",
}

response = requests.post(
    STABILITYAI_ENDPOINT,
    headers=headers,
    data=payload
)

try:
    response.raise_for_status()
    with open("image.png", "wb") as file:
        file.write(response.content)
except Exception as e:
    print(e)
    print(response.content)

But I need it to work in aiohttp, since I need to make the call asynchronously due to how long it can take:

from dotenv import load_dotenv
import aiohttp, asyncio, os

load_dotenv()
STABILITYAI_API_KEY = os.getenv("STABILITYAI_API_KEY")
STABILITYAI_ENDPOINT = os.getenv("STABILITYAI_ENDPOINT")

async def main():
    session = aiohttp.ClientSession()

    payload = {
        "prompt": "A cat using a toaster.",
        "output_format": "png"
    }
    headers = {
        "authorization": f"Bearer {STABILITYAI_API_KEY}",
        "accept": "image/*",
    }
    form = aiohttp.FormData()
    for key, value in payload.items():
        form.add_field(key, value)

    async with session.post(STABILITYAI_ENDPOINT, headers=headers, data=form) as response:
        text = None
        try:
            text = await response.text()
            response.raise_for_status()
            with open("image.png", "wb") as file:
                file.write(await response.read())
        except Exception as e:
            print(e)
            print(text)
        finally:
            await session.close()

if(__name__ == "__main__"):
    asyncio.run(main())

In either case I get the same json response:
b'{"errors":["content-type: must be multipart/form-data"],"id":"[...]","name":"bad_request"}'

[ornate elk]

Yes thats mostly a wrong vae then

[formal trout]

Hi, I'm trying to use SD Ultimate Upscaler but I keep getting seeems. In some image I haven't notice them (maybe they were there but now I'm paying more attention), I share some results and parameters

I noticed seems mostly when upscaling by 4, but with this images I see seems also upscaling x2

I've been using 0.18 denoising strenght

I'm trying seeem fixing options, now I kind of understand it. Band pass basically inpaints the seams, something I did manually for some images and it was a lot of work.

[scarlet igloo]

Can someone help me see where the problem lies

I have deployed a new webui, but its interface is incomplete or not running properly

[formal trout]

Have you restarted from the console and load it again?

[scarlet igloo]

Yes, I have tried. Yesterday, my Weiui was still running normally. After updating the latest WebUI, the Lobe Theme has been initializing the interface, so I redeployed a new WebUI and found that the interface is not working properly

[ornate elk]

Make sure the webui is whitelisted in any browser adblocker

[scarlet igloo]

I have ruled out the possibility of browser issues

The webui where I encountered the problem is on the cloud server

I have run and loaded my local webui normally

[ornate elk]

Oh ok

[calm cairn]

Turns out my XL VAE isn't loading! I just downloaded a different one and now it works fine

I appreciate your help pointing me to that!

[scarlet igloo]

I have solved my problem and restarted the entire cloud server. It is now functioning normally

[fair oxide]

hey idk if this is the channel to ask but can you help me with regional prompter??

my issue is that i cant use character loras without it drastically dropping the quality, making BOTH characters into whatever the character lora was (im only using on character lora, its oc shipping stuff dont judge 💀 ) or it blends both characters together into a malformed monster

[subtle cradle]

hello, i have a question, can we use a nvidia gpu and a amd gpu at the same time? I have a rx 7900 xtx and a rtx 4060 ti and i would like to use them at the same time in stable diffusion with the automatic1111 ui

[fair oxide]

ive looked into it before because i wanted to do the same and i dont think its an option based on what i read however if it is i too wish to know how so ping me if you figure anything out

[subtle cradle]

ok, im going to test later with the Ishqqtyger version of automatic1111 because it is optimized for amd

[fair oxide]

[ornate elk]

Hey, this should be possible.
You would need to follow my AMD Zluda install guide for the 7900xtx. Its in the pinned messages of this channel.

The guide is for lshqqtigers fork but with zluda Backend. Which is much faster than directml

I'm using a 7900xtx too and get 16it/s with zluda compared to 4-6it/s on directml

[ornate elk]

Yea this is the right channel.
I'm not a fan of regional prompter and still recommend the tiled diffusion extension, as it has a easier regional prompt feature

[fair oxide]

could you explain it to me

[ornate elk]

Do you have the extension installed?

[fair oxide]

its installing rn

tiled diffusion with tiled vae, right??

[fair oxide]

so i installed but but cant find it anywhere?? im checking my extensions and im not seeing it here either, am i blind

im blind right

its weird i really cant find it anywhere

not there, not in my files

but when i try to install it again it says its there??

is it put under a new name once installed??

[ornate elk]

Have you restarted the webui?

[fair oxide]

yes

three times

nothing has changed

[ornate elk]

Its called Multidiffusion

As it had a name change

[fair oxide]

yeah then its not in my a1111

at least not where its useful/where i can find it

[ornate elk]

Very strange

Is your webui updated?

[fair oxide]

im unsure, how do i check that

it says it exists, but like, where 😭

[ornate elk]

Its In tmp

You can delete it there

Then try reinstalling it.

[ornate elk]

Scroll to the bottom of the webui. There are the versions

[fair oxide]

i cant even find the extension in my extensions folder??

that means i cant delete it and fresh reinstall.

im cooked bro

[ornate elk]

No because its in the tmp folder

[fair oxide]

the what folder

[ornate elk]

Your on 1.7.0 but the latest version is 1.9.3

[fair oxide]

i found it in the tmp

[fair oxide]

so how do i update it

will updating it make the multidiffusion appear

also why do some extensions go to the tmp folder, whats the tmp folder, i think everything that gets yeeted there hasnt been showing up in my a1111

[ornate elk]

Tmp stands for temporary. Every extension gets downloaded in it before moved to the extensions folder

[subtle cradle]

Ok thanks im going to try it later

[fair oxide]

so should i move them from there to the extensions folder??

since its not moving them itself???

[ornate elk]

Something is wrong if it didn't moved them by itself but you can try.

[fair oxide]

alright lets see

[fair oxide]

before i do, should i move all the text files in the tmp folder as well

[fair oxide]

lmk if it works 💀

[ornate elk]

Text files?

[fair oxide]

text files

[subtle cradle]

Xd

[ornate elk]

No let everything in there thats not an extension folder

[fair oxide]

bet

alright i moved the folders only

starting up a1111 rn

lets hope

huh

[ornate elk]

Give me sec to deal with that kid

[fair oxide]

thatd be appreciated 💀

thank you 😭

anyways, its showing up in my extensions tab now

now i just cant find where its supposed to be in a1111

gonna try restarting the ui again

maybe that will help

it didnt, i still cant find it

ngl that attack felt targeted 💀

[ornate elk]

Then you need to reinstall the webui correctly like I said to you 21 days ago

[fair oxide]

most people spam every channel, bro came after you

[fair oxide]

💀

let me find that

[ornate elk]

The guide is in the pinned messages of this channel

[fair oxide]

[ornate elk]

Yea I guess too. Because its always the same guy with multiple accounts.

Banned him often enough

[fair oxide]

sorry you gotta deal with that man

im assuming ip banning isnt an option/isnt working either, huh

[ornate elk]

Nope that would need to do the discord stuff

We as server can only ban user IDs

[fair oxide]

😔

yeah that sucks

[ornate elk]

True but imagine how stupid someone must be to spam when a staff member is chatting in the exact same channel rn

[fair oxide]

yeah, thats why i assumed it was targeted

no one is that dumb, just that spiteful

[ornate elk]

He had fun for 10 seconds xD

[fair oxide]

bro must be really bored in his moms basement 💀

okay so i got it working

the issue was that i loaded it from this tab

i think the link is broken

so i installed it from url and it works like a charm

[ornate elk]

Ah nice

[fair oxide]

ready to help me figure out how to use it now

??

[ornate elk]

I guess it has to do with your old webui version xD

[fair oxide]

yeah i wanted to try before i updated

some friends of mine are heavy anti update for a1111

they got me scared to update LOL

[ornate elk]

Lol

[fair oxide]

anyways, ill reuse the prompt i was using on regional since it was working okay minus all the

weird stuff

whats the first step

[ornate elk]

Updates can break stuff yea but 1.9.3 and 1.8.0 had many improvements

[fair oxide]

oh??

[ornate elk]

Enabling tiled diffusion then enabling regional prompt control

[fair oxide]

they never tell me the cool stuff just 'update breaks stuff' LOL

[fair oxide]

using BREAK is still required right??

[ornate elk]

Nope

[fair oxide]

might need to change my prompt then

okay i enable the things you told me to

should i remove the BREAK words from my prompt

[ornate elk]

You don't need to but you need to split your prompt

[fair oxide]

alright, ive split into into my main idea, then characters

that should work, right??

[ornate elk]

First set the resolution to your likeness

Then in the regional prompt control settings press "create canvas" or whatever its called

[fair oxide]

[ornate elk]

Then you select the first dropdown and set it to Background

[fair oxide]

this??

like so??

nice

[ornate elk]

What's the resolution?

Because for two characters its better to go into landscape mode

Instead of portrait

[fair oxide]

alright bet, lemme change that

is this better?

[ornate elk]

then you select type as background for Region 1.
Then push the sliders around until the red square fills out the whole image

In the prompt field for Region 1 you input your background prompt

[fair oxide]

??? like this

[ornate elk]

Then its relatively easy. For the region 2 and 3 you need to set the type to Foreground. Then make some squares where the characters should be. And then input each character in its own region

[fair oxide]

moving the red box is...weird

this sound more reliable

lets hope it works

knowing me ill mess it up

but thats why youre here 😭

so for region two and three, my characters

should i just copy and paste their prompts from my main prompt into the boxes

[ornate elk]

Yep

Be sure that only the characters are described and not a background with them

[fair oxide]

should i mess with feather at all or should i wait to see the results first

[ornate elk]

Feather defines how much they fade into each other

You have to test around with that, first leave it default

[fair oxide]

bet

testing

its gonna take a few minutes 💀

okay so while its not done yet its already looking better

but it gave bro a skirt

💀

if i overlap the fields im assuming itll allow characters to interact better?? like hugging or other stuff im not saying here

not telling yall my plans

[ornate elk]

Yes

[fair oxide]

https://tenor.com/view/grin-smiling-shouta-mha-boku-gif-12114496

omg the quality is already better

but its still mixing the prompts a bit??

like giving the guy the girls skirt and giving her his eye color

and his fucking mouth in general wtf

LOL

thats actually funny

hmmmm

its still pretty low quality but its not AS bad

what do you think i should do

[ornate elk]

What was the resolution?

[ornate elk]

Thats okay

Are you using sdxl or 1.5 ?

Models

[fair oxide]

1.5

[ornate elk]

Okay

And you used character loras?

[fair oxide]

yes

one

theres one character with no lora aside from an age slider to keep the bitch from looking 5 years old

and the other character has a character lora

[ornate elk]

Okay

[fair oxide]

im shipping an oc shhhh

[fair oxide]

okay so idk whats wrong with it now

the qual is still better than regional but not by much and now its added 4 people 💀

crying

[ornate elk]

Can't help much, check your prompts also don't forget to add negatives.
And make sure in your overall prompt at the top is no character or person mentioned

[fair oxide]

alright bet

[ornate elk]

I can show you an example later today

[gleaming steeple]

Thanks for your kindness

[mortal compass]

Morning! Noticed an error (yet again lol) showing up in the terminal. Seems like I cannot load my specific defaults for some reason?

[mortal compass]

Maybe they can help you at #📝｜prompting-help

[fair oxide]

I'll ask around there when I wake up

[mortal compass]

bedge

[mortal compass]

I'm not sure, but I think it sais that it is expecting something to be in the code, but cannot find it and therefore cannot load. Strange thing is, I haven't touched the ui_loadsave file, ever. Except in the SD settings when I set my defaults that is

github issues tells me my SD might be corrupt

[wicked leaf]

Trying to use Groq in my RAG webapp, but getting this error, can somebody please help me:

Error::
ValueError: Error raised by inference endpoint: HTTPConnectionPool(host='localhost', port=11434): Max retries exceeded with url: /api/embeddings (Caused by NewConnectionError('<urllib3.connection.HTTPConnection object at 0x7f24587c3040>: Failed to establish a new connection: [Errno 111] Connection refused'))

[ripe kraken]

it says on the site that for professional members there is a discord for professionals, is it here or is there still another discord channel?

[mortal compass]

be more specific

[mortal compass]

This ain't it

afaik

[ripe kraken]

maybe some directions how to find it? [e-mail] - doesn't reply

[mortal compass]

When did you contact them

[ripe kraken]

3 or 5 days ago

[mortal compass]

That's pretty shitty of them. Is there no link on the actual website?

[ripe kraken]

nope, tnx for helping 🙂

[wet flame]

how do i solve this issue?

i installed cuda with pytorch but it still says that

[mortal compass]

what versions

[wet flame]

well this is my new gpu

[mortal compass]

You need to always pay attention to which versions the README tells you to download. You have Cuda 12.4, and I'm pretty sure you want an older version of that

[ornate elk]

Update your GPU driver. If you switched from an other GPU creator like AMD make sure the old drivers are completely removed

[wet flame]

yeah done that

its latest gpu driver just installed

was expensive gpu

[mortal compass]

even though you get a gpu for a million dollar, it won't preform magic and make things work

by itself

[wet flame]

yes i know that its applicable for you too

[ornate elk]

Is your webui the same as before?

Did you used AMD before?

Then you would need to do a fresh install following my nvidia guide

[wet flame]

No this is new install

On my desktop

It works good on my laptop with same setup only difference is med ram on my laptop

[cosmic furnace]

I wonder how much fast amd now on zluda ? Well the almost top & top compared to nvidia almost top & top or somewhere near it. Its still faster on linux so far y ? Well compared to zluda

[mortal compass]

Holy fuck man it feels like I'm just troubleshooting my SD all the time rather than actually creating anything 😩 Wanted to give wildcards another go, downloaded the ones I wanted, restarted and all that jazz. Now I'm getting insanely long error messages in my terminal

I have no idea what I'm looking at lol, but I think this might be of interest
WARNING:dynamicprompts.wildcards.collection.structured:Wildcard file C:\Stable_Diffusion\Forge\webui\extensions\sd-dynamic-prompts\wildcards\artistStylesFor_v10.json has unsupported value for key modelId

[wicked leaf]

Can anyone give a look at my code?? It's very small and simple yet I cannot understand the error's cause.... please help

[mortal compass]

I'm not a coder, sry man

[wicked leaf]

No problem buddy

[fair oxide]

lol sorry to hear that, that could be frustrating, are you using some uncommon apps or something?

[mortal compass]

yeah, it's very tiresome. At this point I'm afraid to get too fed up with it and just say fuck it and just move forward in my spare-time elsewhere. But SD is so fun and the community is really nice, I wanna stay 😭 lol
Yeah I think so. CS1 look at my extensions the other day and told me to not have 2 civitai ones for example. The all-in-one extension also interfered with another one so had to turn that off too

How many extensions do you guys have? I have no clue what's too much xd

[fair oxide]

not a good practice to use tons of extension, i personally keep very essential ones, a lot of the extensions out there are buggy so you gotta watch for that, and a lot of them are not maintained / patched.

[mortal compass]

That also grinds my gears, there so many cool and handy extensions! But I can only have a small amount enabled at the same time, otherwise program becomes doo doo 😂

My self control ain't that good lol

[fair oxide]

well there are some extensions i can't do without, like controlnet, adetailer, tag completer, dynamic prompt, and few others, there are bunch of others that are not really essential but kind of give you convenience and if they are buggy and getting in the way of your creative workflow then they are not worth wasting time on

[rocky owl]

Hey is there some experienced user of SD when it comes to own models/loras with real products that will be trained to replace an object using inpaint / replace etc. Would appreciate any help/tip/links to blogs etc.

[wet flame]

Anyways fixed them myself

[violet swift]

any solution for this?

[ornate elk]

make sure you have python 3.10.11 installed and also disable the windows path lenght limit at the end

[violet swift]

i go to check

[ornate elk]

also check for any red nodes

[violet swift]

ok i have Python version: 3.11.8

i go to check

[ornate elk]

so here is the easiest example, i generated an anime image in txt2image.
then i switch to the img2img tab, load the image in, then use the same prompt or describe the image as detailed as possible, then set the denois at 0.5 or higher. The more denois, the more it will change from the input. i used 0.75 as an example.
Then set a proper resolution similar to the input resolution.
then switch to an realistic model and generate the image:

[violet swift]

I have 3.10 installed on my PC, but within ComfyUI 3.11 appears, how do I make it detect 3.10?

[ornate elk]

maybe ocmfyui comes already with 3.11, then it would be fine

maybe you need to reinstall the animatediff extension

as it cant find a specific file there it seems

[violet swift]

i go to check

[shell crystal]

Currently using runpod with stable diffusion, but does anyone know how to install a model from civit ai using wget? I try to do it, but I keep getting the message "Username/Password Authentication Failed" and I don't know how to get around it

[mortal compass]

@ornate elk any idea what this could be? 🤔

[ornate elk]

would need the full cmd log

[mortal compass]

I warn you, its long xd

[ornate elk]

you have to much extensions xD

[mortal compass]

Your hurting their feelings :((

I'll reproduce the error, after that I can turn everything off and turn it on one by one I suppose

[ornate elk]

in your screenshot i see dynamic prompt warning and a lot of stuff loaded from civitai browser

[mortal compass]

its trying to load wildcards, which I got from the site. I'm not sure why it would want to load it from civitai browser though.. I already downloaded it? xd

Btw I duno how to export from the terminal, I'll see it txt works

Alright I'm gonna trust you with the log file, don't judge me ok

A VETERAN

How did u make it through

[ornate elk]

venv\Scripts\activate pip install albumentations==1.4.3 pip install pydantic==1.10.15

[mortal compass]

It's all rutine by now 😖

true

pitchforks

[ornate elk]

i got the offer, but didnt wanted to

[mortal compass]

D:

[ornate elk]

okay its only a dynamic prompt error because of unsupported stuff in the json files of these wildcards

[mortal compass]

Why would they not put json files in the wildcards? Half of dynamic prompts is about wildcards, no?

ops

Ment to say

[ornate elk]

for me wildcards are .txt files

but i dont use dynamic prompts

[mortal compass]

why would they put unsupported things in it, seems counter productive

[ornate elk]

idk ^^

[mortal compass]

jeeez

Guess I'll just open an issue on their github and pray to AI

[violet swift]

nothing aaaaaa

[violet swift]

help pls

[ornate elk]

you could try if it would work on auto1111

because forge has some issues with certain extensions

[mortal compass]

Hm yeah, I've actually been thinking about switching back anyway. I'll try tomorrow

[ornate elk]

which modtionmodule? sdxlbeta?

can you show me the full cmd log when you get that error?

thats not the full cmd log

whats your webui and torch version?

okay, and whats your gpu?

also xformers version?

also a full cmd log would be good also dont use any other extensions when testing animatediff

[mortal compass]

poor extensions

[ornate elk]

when using animatediff in auto1111 you have to enable this setting:

if not you can only use up to 75 tokens

i heard it has some more featurs in comfyui, and can be used better in workflows there

but i dont think the quality is affectet, because it still relies on the model and motionmodule + resolution

[vast apex]

yo can sb help me :,)

"An item with the same key has already been added. Key" comes up when i try launching comfyui

[pulsar vessel]

I have an AMD driver so I understand there will be some limitations, but after I got everything running, I’m noticing that image generation is taking a while. Is it possible to get the system to use my GPU over my CPU? Or have I hit the bottle neck since I don’t have nvidia?

[mint mauve]

Hello, no matter what I have tried I can't get open webui to interact with a1111. I keep on getting something went wrong string indices must be integers not str. Has anyone seen this before? My openwebui instance runs on Ubuntu with python 3.11 and my a1111 I stance runs on win11 with python 3.10. I can access a1111 from ooenwebui host just fine. Also, running --api --share --listen flags on a1111 side

Somewhere I read something that could be related however I'm unsure.. regarding mistmatch of expected format from a1111 api's output (JSON)?

[fair oxide]

guys how many minutes or hours it takes to convert 1 minute 1920x1080p video @30 fps into anime?

[ornate elk]

What's your GPU?
Stable diffusion can work with AMD, you need to follow my AMD install guide. Its in the pinned messages of this channel.
Zluda is the one for the most newer GPUs. And the directml one is for older stuff.

[vapid sluice]

Paste by mask from Masquerade is really a useless node, why can’t i pass the combined image directly into blend image node ?

The only way is to pass the image through a ksampler by img to img but this decreases the quality exponentially

[cursive cobalt]

Right! hello!

So, I'm trying to add some styles to my SD. So far, so good...
I've learned that the styles are stored in the Styles.csv document. 👍 .

What i do not understand, is when trying to get a new style from Civitai, the file i download is marked as a LoRA file, and is downloaded in the SAFETENSOR extenstion.

Am i missing something here?

I'm using the Automatic1111 UI

[uncut jewel]

Heyo everyone! Can someone help me regarding that issue ?

When i do the Pip uniostall Pillow wether on pythonembebed or root folder, i get that error aswell..

[ornate elk]

Styles in the CSV are just words that go in the prompt field. Its like a shortcut to input multiple words that alter the style like
Anime style, retro, 2d, masterpiece..

Loras are small files that got trained on specific artstyles, characters or other things. These go on top of a model to help the model generate a specific style/character.

For example a style in styles CSV can't help generate Homer Simpson if the model doesn't "know" him.
A trained Lora of Homer Simpson can help the model generate him.

Loras go into models/lora

[cursive cobalt]

So when people have several resources listed in a post in Civitai, but no apparent lora:pack:1 in the promt window, what, and how they did done it?

[ornate elk]

What do you mean by several resources?

[cursive cobalt]

Sec. Gotta find a sfw example xD

[ornate elk]

If people have used a Lora but the lora:name:1 isnt in the prompt, then they have used an other webui like comfyui mostly

[cursive cobalt]

I guess Automatic1111 can't do that then?

[ornate elk]

Automatic1111 needs it

[cursive cobalt]

needs it`?

[ornate elk]

The lora:name:1 In the prompt

[cursive cobalt]

Oh, so it allows more than 1 lora to be used, at the cost of promt character limit?

[ornate elk]

There is no prompt limit and yes you can use multiple Lora's at once

In Auto1111 and comfyui

[cursive cobalt]

Wait, so the 0/150 is a lie?

[ornate elk]

At the beginning of stable-diffusion 75 tokens was the limit

Auto1111 disabled this limit

[cursive cobalt]

Oh, so since Automatic has a weird update cycle, its a relic from a bygone version?

[ornate elk]

The counter is still there, so you can check how long your prompt is.
Its good to know because the longer the prompt the less attention every word gets. Also loras increase the token count too

[languid tinsel]

Is there an extension or something similar I can add to my SD to have an ability to do live painting like this?

[cursive cobalt]

But even then, can't you just (, ((, (((, whichever promt is the most important?

[ornate elk]

That can only be done by Krita-ai diffusion which uses comfyui as Backend

[ornate elk]

Exactly or using word:1.2 its better than making a lot of ()

[languid tinsel]

thanks

[cursive cobalt]

I'm guessing the :0000, is the strength value of a promt/lora?

So

Lora:pack1:2Lora:pack2:0.7 would put pack1 at 200%, and pack2 at 70%?

Coding makes my brain hurt 😐

[languid tinsel]

bruh this is not coding 😂

[cursive cobalt]

To me it is xD

Is all lines and words

[languid tinsel]

but yes, that's 200% and 70%

[cursive cobalt]

ADD sometimes makes stuff hard 😄

[ornate elk]

Yep, but make sure to separate loras with ,

[cursive cobalt]

ofc. One last question, at least of this hour. The <lora> stuff. The main lora selected, i assume is just there, and any <lora> you add in the promt, is 2nd, 3rd and 4th etc?

[ornate elk]

I don't get it sry

The attention of words in a prompt goes from left to right

[cursive cobalt]

That Lora is always selected even with multiple lora's are added in the promt window, ye?

[ornate elk]

Yes, wouldn't use that dropdown for loras

Its always active

[cursive cobalt]

Which r u supposed to use then for auto1111?

[ornate elk]

The lora tab

In txt2img

[cursive cobalt]

Does that add the promt to the promt window?

[ornate elk]

It adds the specific lora:name:1 to the prompt

By Clicking on the lora

[cursive cobalt]

ooooh

Installed this stuff like 3 hrs ago xD

[ornate elk]

Np ^^

[cursive cobalt]

My poor gpu is struggling

[languid tinsel]

pro tip is to have a photoshop/gimp running on 2nd moitor and just do manual adjustments durign img2img process, it saves a lot of headaches

[cursive cobalt]

Cute. I can barely have discord open, thats how many resources its using atm 😄 Lagging the entire system

already using 10.8 gb / 11.3 gb (1080 Ti from 17')

[languid tinsel]

you only use photoshop between generations

so it doesn't matter

[cursive cobalt]

Is fine though. Not really intending on using img2img anyways

[ornate elk]

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