🤝｜tech-support

[main rune]

effing

spambots

[main rune]

Solved it, had to manually download everything but it was not hard with the logs at hand

[half wave]

Yes sir I am working on it for weeks and every day with a new strategy I think Tomorrow I will get it 🤣 but no

I have good pics from Astria and Open Art too a realistic Model and I need to get Finaly the safetensor base Model to continue my Project

[tribal glade]

I downloaded Stable Diffusion, what's wrong? I downloaded Git and Python.

[open fox]

Are you by hazard behind a large firewall? Sometimes user from
China without VPN Tools got trouble downloading models from the internet.

[maiden marsh]

Can someone help me with Stable Audio's input audio upload? I am trying to upload audios but they all just fail. One succeeded and the others are the exact same files as this one, but they fail. It just says "Contact support"

[placid zinc]

Turns out I was missing a c++ redis 2015-2019

[haughty minnow]

Every time i try and run this, i keep getting this error. I've been following the guide to stable diffusion by Corridor Digital and I've been doing things correctly but for some reason this just won't work. i am at the points where he says we need to run the json file. I'm just a beginner to this. Send help pls

who is the admin

[ornate elk]

its a scam, dont join

[haughty minnow]

figured

[ornate elk]

whats your gpu and what webui do you want to setup?

on the first link of the pinned messages, you'll find all setup guides for different webuis like Automatic1111, Comfyui, Forge, etc

[haughty minnow]

gpu is nividia geforce rtx 4060 and webui is automatic1111

[ornate elk]

alright, and the webui works? and you try to setup everydreamer2 it seems?

[haughty minnow]

yeah i can open stable diffusion when i open the webui-user.bat file. but i am not sure if i have set up everydreamer2

[ornate elk]

can you link the video/guide, so i can check what you try to do?

also dont install stuff in your Downloads folder

[haughty minnow]

ok. however, the video is on the corridor digital website and they have a subscription to watch some of their videos. this one included.

https://www.corridordigital.com/video/2681

[ornate elk]

oh ok

then you can try to setup EveryDream2 again in a different folder and with their manual guide from here:
https://github.com/victorchall/EveryDream2trainer/blob/main/doc/SETUP.md

i would go with python 3.10.11 64bit instead of 3.10.9
https://www.python.org/downloads/release/python-31011/

Also if you just want to train Loras, then you may want to checkout OneTrainer. Its much easier to setup and use:
Here is a install Guide for OneTrainer:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Lora-Trainer-Setup-Guides

[haughty minnow]

ok will try it out. thanks for the help

[ornate elk]

no problem!

[tranquil summit]

hi trying to get stable diffusion working with something to animate my existing images. don't really need anything fancy just something that will work. even camera shake and/or breathing would be good. i have animatediff installed but it's not working well just tearing the image apart in the animation. im open to anything to achieve the result im looking for (animating 24 frame or few second still images)

[hallow latch]

Hello .. I want to check if there is something with Pinokio .. I got a new GPU by GF and it was the "AMD Radeon RX 7800 XT" .. When I start Stable Auto1111 it runs up directly from 1-3 GB up to 14-15 GB so when I try to run a prompt .. I insta get

"
RuntimeError: Could not allocate tensor with 335544320 bytes. There is not enough GPU video memory available!
Time taken: 2.9 sec.

A: 12.37 GB, R: 15.44 GB, Sys: 15.4/15.4414 GB (100.0%)
"

I try to run with

" set COMMANDLINE_ARGS=--skip-torch-cuda-test --no-half --lowvram --xformers "

But still issue.. What can I do?

[ornate elk]

hey, you need to use Zluda for AMD GPUs

in the first link of the pinned messages you find all the AMD Guides.
Go for auto1111 or Forge with Zluda

[restive sparrow]

I use A1111, and resetly noticed when I use it my c drive(SSD) is temporarily siphoned maybe 20-30GB. Is it doing some sort of caching or something? I do use the "store models in RAM feature" but max im swapping between 2 6GB models (so that doest explain that). I have all my stuff on an HDD, only stuff on the SSD C drive is GIT and Python installs themselves.(so it must be some function of those2?) thanks for any insight,

[ornate elk]

how much RAM do you have?

when models get swapped the windows pagefile also sometimes gets actively used

that file is stored on C

[terse siren]

Ok so now afte rmonths of SD forge working fine, Im now getting this error when I click run

Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: f2.0.1v1.10.1-previous-659-gc055f2d4
Commit hash: c055f2d43b07cbfd87ac3da4899a6d7ee52ebab9
Traceback (most recent call last):
File "F:\SD Forge II\webui\launch.py", line 54, in <module>
main()
File "F:\SD Forge II\webui\launch.py", line 42, in main
prepare_environment()
File "F:\SD Forge II\webui\modules\launch_utils.py", line 434, in prepare_environment
raise RuntimeError(
RuntimeError: Your device does not support the current version of Torch/CUDA! Consider download another version:
https://github.com/lllyasviel/stable-diffusion-webui-forge/releases/tag/latest
Press any key to continue . . .

[ornate elk]

is your gpu driver updated?

[terse siren]

And when I try to run the web user bat I get this

Creating venv in directory F:\SD Forge II\webui\venv using python "C:\Users\Francesco\AppData\Local\Programs\Python\Python310\python.exe"
Unable to create venv in directory "F:\SD Forge II\webui\venv"

exit code: 1

stderr:
Error: [WinError 23] Data error (cyclic redundancy check): 'F:\SD Forge II\webui\venv'

Launch unsuccessful. Exiting.
Press any key to continue . . .

I read somewhere that WinError 23 means the drive is damaged but other than Forge, it works totally fine

[ornate elk]

upgrade your python to 3.10.11 64bit, then delete the venv folder.
Then run the update.bat in forge and then the run.bat

[terse siren]

Why that one in particular?

Why not any of the newer versions

[ornate elk]

because they are not as compatible as 3.10.11

[lucid yoke]

Lock the seed and Regen but without the lora

Or set the lora weight to 0 and generate it again

0 = does nothing

[ornate elk]

You can See in the meta data if a lora is used or not.
So when you generate the same image without the lora you can be sure it didnt changed the image

[lucid yoke]

whatr does the lora do even?

like if its a style one and its not changing anything its either your not using its trigger tag or its trained wrong

[restive sparrow]

I've got 32

I don't believe it maxes out, usually around 29gb with 2 models loaded

[terse siren]

I have a feeling my issue is that my external SSD is beginning to go

Im probably gonna have to get a new one

[zealous jolt]

so i ran update_comfyui_and_python_dependencies on my windows install and now i cant run comfy 😦

raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled

plz halp

[ornate elk]

Check it with CrystalDiskInfo

[ornate elk]

I still think the Pagefile is getting larger for the moment.

[lusty fractal]

Morning. so i have an issue with stable diffusion. it seems the only model i get to work is waiNsfwBranchRouwei. every other model gives me terrible results

[ornate elk]

Hey, do you have an example of the other models output and settings?

[lusty fractal]

Hi CS1o. well i just tried out dreamshaper xl which produces somewhat ok results, but i cant find a model that produces good results with regular loras.only the pony ones seem to work

this is what i get with revanimated

[ornate elk]

You need to check for which model the loras are made

For example 1.5 based loras won't work with sdxl
Pony loras work with pony based models and sometimes illustrious.
And xl loras work with Illustrious too

[fossil dew]

Where can I find VAEs to download?

I prefer the suggestions here rather than searching online myself; not sure what sites are best/most reliable ect ect

[lucid yoke]

civitAI has some but for what model type? XL has a lot of them baked in or youd use the basic xl vae useually

[fossil dew]

I can harly run XL

I'm trying to add VAEs and Models that work with 1.5 🤡

Hoping switching back will stop me from getting KUDA errors 🙄

Seems to crash a LOT more now that Im using XL

I fix it, but then after a little while I have to fix it again; exsausting

[fallow knoll]

Civit is where I got mine, but like they said a lot of the newer models have them baked in.

[ornate elk]

Civitai and huggingface for downloads.
for 1.5 there are a few good vae:
840000mse (for realism), kl-f8-anime2 (vibrant anime), blessed2 and clearvae_v23 (less vibrant anime)

[ornate elk]

what was your gpu again?

[fossil dew]

8 🙄

[fossil dew]

Thank you ❤️

maaaaaaan the quality of 1.5 compared to illustrious is ASS 🍑

[ornate elk]

and your using forge or auto1111 ?

[fossil dew]

auto1111

I could use forge, im just not sure how to install or download it or whatever- or where to find the link, you told me the channel its in before but I couldnt find it 🥲

[ornate elk]

its in the pinned messages of this channel

first link

or here:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides

[fossil dew]

appreciate you

ooo confusing ;')
I'll figure it out tho 🤭 ty again

[ornate elk]

np, i tried a 768x1024 image with an xl model on my old PC with a GTX1080 8gb vram and got around 1 min for an image without out of memory

[fossil dew]

interesting 🤔 I onder why mines giving me KUDA errors every 3 images 🤡

[ornate elk]

can you show me the settings you used?

[fossil dew]

Crashed after 1 image 🤡

[ornate elk]

wrong upscale settings

[fossil dew]

🤔

[ornate elk]

always set hires steps to 10

[fossil dew]

oki

[ornate elk]

and if you gen anime images then also enable tiled vae

[fossil dew]

yes- oh ok!

[ornate elk]

and dont enable tiled diffusion

[fossil dew]

Suggestion for upscale?

OHHHH I was using the wrong ok ok

wrong one*

[ornate elk]

yep

only use tiled vae, when you upscale and ignore tiled diffusion

[fossil dew]

ok I'll remember that

[ornate elk]

yandere neo is okay, or anime 50000 G

[fossil dew]

Using yandere mhm

[ornate elk]

some upscalers i use.
mostly fatal anime 50000 G or remacri, lollypop or ynxl200k

4x UltraSharp for realism

and 4x AnimeSharp is okay too for anime

nmkd siax 200k also good

Most can be found here:
https://openmodeldb.info/

[fossil dew]

thank you!

🤡

crashed after 1 image again

wheeeeeeeeeeeee ✨

tried running it a 2nd time and it crashed after 2 images

[terse siren]

Yeah...its saying caution with C5

I think its time for a new one

[zealous jolt]

so i havent updated comfy in like a year and i did it last night. now everything is different and confusing.

there used to be a menu to install/update nodes but i cant seem to find that anymore. is that feature removed, or somewhere i cant find it

[ornate elk]

can you try generate a 768x1024 without hires fix?

[ornate elk]

thats the comfyui manager, you may have to install it again

its a custom node

[zealous jolt]

ah got it. thanks

[zealous jolt]

why wont this let me choose a model? it was working until i tried to switch to another and now it just says undefinded and doesnt let me choose anything. there are models in the correct folder

nvm i restarted it and it works

[warped canyon]

whats the shortcut to interrupt a generation?

on a1111/forge

[hallow latch]

Is it normal than a simple 512x512 .. going to take like 67 seconds? with an AMD GPU like RX 7800 XT?

[fossil dew]

Ill do this right now

[ornate elk]

Nope

[fossil dew]

Crashed after 1 gen again 🤡

[ornate elk]

Can you show the settings?

[fossil dew]

In a moment, once I reopen it.

gimme like

5 mins 😂 it takes forever

Are those last 2 important?

half vae and uh vram thing

[ornate elk]

You can remove --no-half-vae if your GPU is rtx based

[fossil dew]

i believe it is RTX

ill have to ask my hubs when he gets home 🤡

how do you remove that? will it help with my issue?

Sorry its blurry- my screen does that and idk why 🤡

[lucid yoke]

you can check in the task manager, control + alt + delete. then the second tab (this icon) there should be GPU on the bottom

[ornate elk]

Just edit the webui-user.bat (right click) and then remove it and save.

[fossil dew]

settings look ok?

cool

[ornate elk]

Make sure you dont have anything open in the Background like games or Wallpaper engine or twitch streams.

[fossil dew]

yeah I shut EVERYTHING down when I use it mhm

[ornate elk]

Also make sure your GPU driver is updated

[fossil dew]

ill have hubs do that when he gets home, good idea

checked- its up to date

[ornate elk]

You can See the age of the driver in the same screen here

On the bottom right

[fossil dew]

oh interesting

oh so like when it was installed last

[ornate elk]

Yea when the last update was

[fossil dew]

yep- 5/6/25 uwu

[ornate elk]

Oh okay

[fossil dew]

thank you n_n This is so fustrating, you're a trooper dealing with my insistent questions 😂

🤡

[ornate elk]

Np xD

You should get more RAM in the future

Also try our forge webui. It could use less RAM

[fossil dew]

💢

[fossil dew]

yeah I think that will be what I have to do

[fossil dew]

Deff will when I get the $

I really dont get it

because it worked for HOURS like

3 days ago

[ornate elk]

You can try Firefox as browser. It should use less RAM than edge (and is more privacy friendly)

[ornate elk]

Also a PC restart (not shutdown) can help sometimes

[little bough]

also i then installed a new version of reforge and decided to just copy across files from the old install folder into the new one but it once again basically says they belong to another user?

[ornate elk]

Hey, the command should work and has to be done a few times for different folder paths it shows.
Easier is to add your user as permitted for the whole webui folder

Yea or in the security tab and there you can add your username and select include all subfolders

[little bough]

hmm didn't work, i must have done the wrong thing

its something in here?

[ornate elk]

Looks okay

[little bough]

oh i cropped out the part where it says 'owner' and change loll,looking up a guide now that goes over it but im on the right path, thanks ser

[ornate elk]

Np

[little bough]

btw is it considered safe to do this? i deleted the venv (and installed python etc) and let it reinstall with the new OS but aside from the ownership it should run as normal yeah?

yeah i followed this guide, handy to have for anyone else who might run into this, very simple https://www.howtogeek.com/301768/how-to-take-ownership-of-files-and-folders-in-windows/

[ornate elk]

Yep should work normal

[hallow latch]

Nope? Could be my commandline perhaps?

"set COMMANDLINE_ARGS=--use-zluda --update-check --skip-torch-cuda-test --no-half --precision full"

[fossil dew]

I use duck duck go

But I'll take you on your adbvide

Advice

In order to install forge, will I need to delete my stable diffusion file that has the batch file and everything else in it

[ornate elk]

Remove --skip-torch-cuda-test --no-half --precision full

These force the CPU mode

Add --skip-ort
Then delete the venv folder.

Then open up a cmd and run
Pip cache purge

Then relaunch the webui-user.bat

[ornate elk]

Nope you can have forge in a seperate folder

No need to delete auto1111

[fallow knoll]

you can even copy all of your checkpoints/loras into the new forge. id back everything up on an external of course.

[little bough]

@ornate elk would you happen to know what this redirect warning is about? it's appearing at the start of the launch stream (reforge) in my webui on both the old reforge and the new one? and then at the end of the stream after i've clicked on the screen to copy the ip address (and only after that) i get this error (interrupted wit signal..) and it basically closes after typing on both the old and new one? thanks

never had those errors before

maybe i should delete venv and re-install? could it be because i created the venv's before taking ownership?

[ornate elk]

Yep recreate the venv and try again.
Also add --cuda-stream to the webui-user.bat
That can help for the performance.

[hallow latch]

With these changes.. I get this error
"RuntimeError: mixed dtype (CPU): expect parameter to have scalar type of Float"

[vast crest]

Hi, I need help on Forge. When I want to create a Lora on sd 1111, I create a folder with the photos of the character I want to create. Then I upload the folder to stable diffusion 1111, by doing so I insert all the prompts in all the photos. Then I went to Google Collab to create the Lora. How can I do the same thing on Forge? the name of the tool is: Dataset Tag Editor

[ornate elk]

If your on AMD Adrenalin 25.5.1 you need to downgrade to 24.5.1 for zluda to work

Or you can try to get the latest Zluda version to work. It should ve compatible with 25.5.1 but idk if it really works.

[hybrid moat]

when training a lora on a specific style is it the same process as training on a character?

like what would u do differently when u want the art style

[hallow latch]

Hmm.. okay maybe I can try.. Not sure if I want to atm.. If I do add " --no-half --precision full
" it does work but maybe I can remove "--precision full"

[ornate elk]

Its using CPU only mode then.

[hallow latch]

Ohh

So is --no-half = cpu only?

[ornate elk]

No

--skip-torch-cuda-test means CPU only.
But as you added it one time.
It downloads torch for CPU mode.
So since that point its only CPU mode even if you remove --skip-torch-cuda-test.
Important is to delete the venv afterwards and cleaning the pip cache.
Then you have to either add --use-zluda or --use-directml to let it use the GPU

[hallow latch]

I am testing with set COMMANDLINE_ARGS=--use-zluda --update-check --skip-ort --no-half

But it takes still like 2-3 min 😄

[ornate elk]

Can you check in taskmanager if the CPU usage spikes when generating?

[hallow latch]

Yeah the CPU was at like 80%

Time taken: 4 min. 16.4 sec.

[ornate elk]

Ok

Most zluda users reverted to 25.4.1 or 25.3.1
But there is zluda version 3.9.5 which should be compatible with 25.5.1 but I can't test it rn

I'm on Adrenalin 25.3.1 without issues

[hallow latch]

i AM ON 25.5.1

Sorry caps

[ornate elk]

Yep then you have the two options or CPU mode

[hallow latch]

Uninstall 25.5.1 and install 25.3.1?

Or can I downgrade with the Adrellin?

[ornate elk]

Nope just uninstall and reinstall manualy

[restive sparrow]

never ever touched it, its at auto of 2048

[restive sparrow]

I just use models with included vaes and call it a day lol

[vast crest]

when you have time, remember me XD

[fossil dew]

@ornate elk Is this literally all I have to do?

under it it says "Old Forge"

Restarting my PC helped ! 😄
I'm still going to install forge

[last oak]

is there any free api for it

or i'll reverseengeenier it

[hybrid moat]

im having some issues installing kohya ss

theres issues with the timm compatability in the newest version

like the main branch on github

and i tried the last stable release, v25.0.3 but its also not working )):

says error during git operation

[haughty minnow]

how do i include the concepts in the training process

for one trainer

[hallow latch]

Morning :3, I installed 25.3.1 and remove --no-half but still getting "RuntimeError: mixed dtype (CPU): expect parameter to have scalar type of Float
Time taken: 5.9 sec."

[vast crest]

Hi, I need help on Forge. When I want to create a Lora on sd 1111, I create a folder with the photos of the character I want to create. Then I upload the folder to stable diffusion 1111, by doing so I insert all the prompts in all the photos. Then I went to Google Collab to create the Lora. How can I do the same thing on Forge? the name of the tool is: Dataset Tag Editor

and

i found dataset in Webui Forge,but after installing it successfully.it is not displayed in the program window 😦

[honest harness]

hey how do you add checkpoints to stable diffusion

[fallow knoll]

download the file (usually safetensors) and drag and drop or copy/paste it into your Stable Diffusion folder. The pathway is usually User > Webui/Forge/Whatever your using > Models > StableDiffusion. At least thats how mine is set up.

[honest harness]

thanks

[fierce moat]

lmao in this point we got too many variant of wan

there is vace and there is skyreels and there is different parameter

[lucid yoke]

Don't forget about causvid lora, its hella good

[fierce moat]

i want just want to do keyframe start and end for 1.3B model

any workflows can share

the vace workflow scare me off

[lucid yoke]

Oh hmm im not sure about 1.3b

[fierce moat]

the f is going

lol

[lucid yoke]

Ahh different versions

It makes sense actually

[fierce moat]

yea causvid is really hella good

tried it

improve speed like triple

[lucid yoke]

And fps

[fierce moat]

causvid is literally dream comes true

i donwant play with ltxv anymore

yea I think right now for start/end frame is to use VACE

its sad becoz skyreels v2 just add support to start/end frame (not vace)

[vast crest]

when i start webui forge,i always get all these errors.is this normal?

[granite bough]

Whenever I generate a prompt flux nothing happens no errors but no result

[fossil dew]

Okay- So @ornate elk When I do this; I just need to create a "forge" folder in my stable diffusion-webui folder?

Im finally getting around to doing it; so I just wanna make sure all my Is are dotted and Ts are crossed 🙂

[fallow knoll]

forge should be a separate set of folders from your old stable diffusion setup (i assume you were using A1111). It will have its own update, run, and environment.bat files as well as all of the associated folders and files that your old setup had. After that its a matter of running the update.bat to make sure its up to date, then run the run.bat file to make sure it loads properly. After that, copy all your saved checkpoints, VAEs, embeddings etc from the old folders to the corresponding folders in the forge path. Relaunch, make sure its all there, and create away!

[fossil dew]

ok so just create a new folder on my D drive?

Alot easier than I expected it to be, cool!

[fallow knoll]

Thats what i did. I have a folder that is AI stuff, but inside that i have separate folders for framepack, A1111 webui, and Forge webui. And inside each of those are the necessary folders for them to work

[fossil dew]

If I made an AI folder on my D drive, would I be able to take Stable Diffusion and move it there without it f*cking up?

[fallow knoll]

If you use the one click installer from the github repo that Illyasviel has it should have automatically set it up for you correctly, at least it did for me

[fossil dew]

I think Im going to do it manually tbh

Good to know theres another way to do it tho

[fallow knoll]

Yeah, as long as it isnt open and running when you do it. You may have to update your pathway for saved outputs if that was also inside your stable diffusion folder. Mine is separate and setup for an external harddrive, so i dont have to worry about that

[fossil dew]

Coolio ~

Thanks Retsubu 🙂

[fallow knoll]

Yep! I can help with simple stuff, once you get into code and debugging im not the guy lol

[fossil dew]

so after I unzip the file into the file I made `FORGE (Stable Diffusion) - When I do all the things (update.bat ect) and when I go to USE stable diffusion, I'm going to run it from this file from now on and not the one in my Stable Diffusion-webui folder?

So I'll run the FORGE Web UI?

silly question 🪿

[fallow knoll]

correct, you will launch with the run file that came with the Forge stuff

[fossil dew]

👍 cooooooool

[fallow knoll]

It will look familiar, but have a slightly better UI and some extensions baked in

[fossil dew]

okay one more question

Where would my models, vae, and loras go?

The screenshot is all the stuff that unzipped from the download

[fallow knoll]

inside the webui folder should be all of the same folders your used to

[fossil dew]

OH! Yep

there they are 🙂

[fallow knoll]

and just copy the file contents from the old folders into the new

and boom, ready to go

[fossil dew]

then when I run it when Im done, I use the same webui-user.bat file? (looks the same as auto1111)

just verifying

[fallow knoll]

yeah, its titled the same but the version it launches is noticeably different, at least it was to me

[fossil dew]

right on

interesting it looks the same as Automatic1111?

well- very similar

[fallow knoll]

yep. but its optimized, so it runs faster than A1111 did, and it should be less intensive on your pc

it was noticeably faster for me at least, but im using a RTX3080 GPU so that might be why

[zealous jolt]

can anyone explain what "speed" does in certain txt to img workflows?

hey mods

[fossil dew]

how will I know when run.bat is done?

it did alot of stuff, then opened the web-ui

hasnt done anything in like 2-3 mins

owo

[vocal burrow]

yup yup

[fossil dew]

assuming its done but I dont wanna close it early 😐 Auto1111 was slow

[zealous jolt]

i think opening the web-ui is the last step of the process. once that's done, are you able to start a generation?

[fossil dew]

i have a couple other thinsg I need to do first, but I think you're right

thank

[zealous jolt]

ok i helped ban a malicious actor and helped someone else. it'd be great if someone could take a gander at my request

oh wait hes not banned yet

[fossil dew]

ok well, here goes nothing!

🤞

ok so forge errored out

[fallow knoll]

yeah that is outside of my wheelhouse now lmao @ornate elk might know what went wrong

[fossil dew]

Yep! The go to guy for this kinda stuff, big brain.

[vocal burrow]

what's the whole error ?

[fossil dew]

I'll have to get that to you shortly

[fossil dew]

Sorry for double ping

[vocal burrow]

yes, but what's the log before that ?
how do you trigger the error ?

[fossil dew]

uh

Im trying to install forge

THERE WE GO

This was my first time running this:

[vocal burrow]

looks like it installed just fine

[fossil dew]

huh

[vocal burrow]

try rerunning it and make sure that you have whitelisted its url in your adblocker

[fossil dew]

its not doing anything?

uh

I use duck duck go

it worked with auto1111

ill try re-running

OH

I MIGHT NEED TO RESTART MY PC

RIGHT?

[vocal burrow]

unlikely but it's worth a try. Also duckduckgo is a website, not a browser to my knowledge.

[fossil dew]

its a browser 🙂

i think its both tbh

but anyways ima restart my pc then try running the webui-user.bat file again

ill brb if anything doesnt work

[vocal burrow]

oh yeah indeed looks like they have their chromium skin / fork.

I doubt that's what is causing the issue

[fossil dew]

what do you mean

oh

duck duck go

nvm

[vocal burrow]

you can always try to open the link with edge just to check.

[fossil dew]

how do i do that, set edge to default browserr?

[vocal burrow]

open edge and then copy paste the url (http://127.0.0.1:7860) into its top bar

[fossil dew]

ok ima see if it boops open in duck duck go first and if it doesn't I'll do that 🙂

it looks like it s doing a different process this time

YAY IT OPENED

COOOOOOOOL

I'll lyk if I have anymore questions 🙂

Are there any important settings here that I need to have turned on?

Ive never seen this before

[vocal burrow]

never OOM

[fossil dew]

ok

also can I still run AUTO1111 if I ran the right bat file?

[vocal burrow]

yes

[fossil dew]

i have codes saved in there ima need

COOL

thank!

I had uh

upscalers downloaded

what folder do I find them in so I can transfer them to forge?

ohp found it

ESRGAN

[honest harness]

Hey do yall have a text lora since i cant find one

[lucid yoke]

A lora to make text come out right you mean?

Theres some loras for certain words/shirts with words on them but for text in general? 1.5 & xl is outta luck

[honest harness]

i wanna get a licence plate with the correct text and i tried some but couldnt find any

[open fox]

Flux and SD 3.5 are quite good with text. If these don’t succeed I would recommend to use depth or canny controlnet to add the right text.
So generate the plate, remove the text on it. Add the right text in any font you like and use either image to image or controlnet to create a new image.

[mellow moat]

Stability ai doesnt even have autotopup for credit, what am i supposed to do if im building an application that use the api ? monitor everyday ?

Its an extremely simple functionality that would take any dev 1 day to set up, idk hows possible its not there while you guys are selling the api

[outer moth]

amd saw that there might be a chance for amd being useful in the ai department and wanted to change that

[serene notch]

hey, do you have any advices for posing characters? i'm using A1111 currently and struggling to make a character that justs looks straight in the canera with arms at rest

[vocal burrow]

i d look into using controlnet to set specific pose for characters

[serene notch]

do you have any info on that? searched that quickly, found this repo(https://github.com/fkunn1326/openpose-editor) but it looks like it's outdated as installing the extension in A1111 doesn't give me any new windows or anything

[vocal burrow]

there are many "pose editor" , for instance :

• https://github.com/ZhUyU1997/open-pose-editor
• https://github.com/nonnonstop/sd-webui-3d-open-pose-editor

Or you can use a reference image, extract the pose from it and apply it to your prompt

[serene notch]

ok so, either i'm stupid or my A1111 is broken (most probably the first option). i installed the first one you sent, it gives me a '3d openpose' window that shows up between 'train' and 'settings', i can edit the pose just fine, then once i'm done i go into 'send to controlnet', i see the image of my pose, and click 'send to txt2img', then nothing happens, it switches to the txt2img tab but there is nothing here, i tried searching but i cannot find an option either to activate anything

i read through the github too but didn't see any mention of how to actually use it

[vocal burrow]

screenshot your webuipage, so I can see how you ve setup everything

[serene notch]

so, here is what i have right now. when i wrote the message above i didn't have the controlnet panel, right now i'm trying to follow this tutorial: https://www.nextdiffusion.ai/tutorials/how-to-use-open-pose-controlnet-in-stable-diffusion , i installed the controlnet extension, downloaded the models and did as it says, but whatever pose i input in it doesn't seem to have an effect on the output, plus the 'send to txt2img' in the openpose editor doesn't work, so i have to manually download the pose image and put it into the controlnet

[vocal burrow]

what model are you using ?

[open fox]

What model are you using? A SD 1.5

[serene notch]

https://civitai.com/models/827184/wai-nsfw-illustrious-sdxl i'm using this model, and yeah figured out using this one the pose doesn;t work but with the default sb model it works

is the model just incompatible with poses or do i need to change something for it to work?

[vocal burrow]

so you re mixing controlnet 1.5 model and SDXL model

can t work

you need to get controlnet sdxl models or use sd1.5 models

cant mix and match

[serene notch]

aaah ok i see, yeah that make sense now, thanks

[proud ledge]

Hey, I want to create raw footages like the one in this video https://www.youtube.com/watch?v=QIQYzi7IPGk. I can act out myself but change the background and the face with of someone else's. what all softwares do you suggest me to use?

[terse siren]

An honest question, I know running SD locally is pretty intensive, but could it be theoretically possible that it could damage hardware

like an external SSD?

[terse siren]

Im just wondering if thats the reason why my old SSD took so much damage

[vocal burrow]

reading is """"basically free"""" most of the damage is done when writing.

[lucid yoke]

How old was your old sad

Ssd

[vocal burrow]

Also because of the nature of SSD, stored data are moved from cell to cell over time because they can t hold the charge for "too long". so in theory I d say the more data you have stored on your SSD the faster it will die. By how much ? Not sure...

SD itself is not doing too many writing on your disk, and the fact that you store a lot of heavy models on it should not impact it more than if you were storing a lot of linux ISOs.

[lucid yoke]

With normal use and a reputable brand it should last 5 years ish

My nvme was rated on 600gb a day write for a 5-6year lifespan

[vocal burrow]

It can probably last even longer. I havent changed my boot drive since i build my machine 6/7 years ago. CrystaldiskInfo/smart does not complain. I still have some spinning hard drivers from 10/15 years ago with no dead sectors. I m not using it that much nowadays, it s mosly my backup's backup drive for important stuff :). Just check from time to time to be sure. SSD / SD cards lock up in write when they re dying anyways. So you ll see it coming before you lose any data.

[honest harness]

yo yall have a lora that fixes fingers cuz holy shit is this bad at fingers

[brittle lantern]

Hey I need some help with forge on Mac only started today never happened before but Adetailer even after installing and uninstalling and reinstalling it always gives me a black screen on face detailer and it’s pretty much always now

Only once it has happened mid normal gen but didn’t happen after

I’ve restarted my Mac just in case

But same output

Any idea how to fix this

[fluid willow]

Hello! How can I upscale with Ultimate SD Upscale while avoiding each tiles to looks very different and avoid the tiling effect?

[strong path]

Hello i have there a problem with prompt s/r as you see in picture, but for positive prompt there is [scene], i try any way but found no solutions. What make i there mistake

[fossil dew]

Will PonyXL loras work with Illustrious models? 🤡

Im trying to find an XL Lora that makes Twitch Emotes but I can only find a PonyXL one and Im running an illustrious model.

[lucid yoke]

Theres a long and awnser but im gonna make it short lol, if its a illustrious v0.1 model theres a chance it works

The shorter awnser is to just try it

[fossil dew]

downloads it and tries it anyways

. . .
Seems to be working.

Coolio

Thank ~

[lucid yoke]

Quality is maybe a bit less but eh if it works is works

[fossil dew]

tbh quality is lookin' just as good, which is awesome. ty for answering n_n

[fossil dew]

Now it's genning the image, but it's not showing the complete image when its done 🙄

Goes though the whole process, but then the image goes blank when its done

but it doesn't say anything about an error in the python box thingy

well, as far as I can tell 🤡

[frosty drift]

anyone know kohya? that could help real quick

[terse siren]

I got it in 2021

[lucid yoke]

4 years, comes close to the 5 year average

[terse siren]

So the problem was that it's just old

makes sense, it was pretty dusty when I finally removed it

[lucid yoke]

Eh with ssd's its read/writes

Mostly writes

[lost solstice]

hello, i want to ask which one is better 4080s or 5070ti?

[dry elm]

Hey guys, I'm using style guide API for pattern generation, does anyone know how I can enable tiling? the patterns I generate rn can not repeat seamless. Much appreciated

[lucid yoke]

Scam

[dry elm]

yea ik

[daring reef]

When i enable this i am getting this error

[granite bough]

why do my flux models take 30 minutes to generate

[lucid yoke]

What gpu do you have?

[granite bough]

rx6950

[lucid yoke]

Ah amd, yeah makes sense then

[gilded umbra]

Hello!
Could someone please share a working download link for the SD 2.1 motion_adapter.safetensors file (v1.1, ~1.8GB)?
I’m looking for the official AnimateDiff v1.1 adapter (not SDXL or Lightning), SHA256:
5fd196988206cfac90bab1dc56cc59fdd12035492f74ff0662eb6eed5113c547
or
45b71fe98efe5f530b825dce6f5049d738e9c16869f10be4370ab81a9912d4a6

The official Hugging Face repo is down/private. If you have a Google Drive, Dropbox, or other trusted mirror, I’d really appreciate it!

(Please, no SDXL or Lightning adapters, only SD 2.1 v1.1!)

Thanks so much in advance! 🙏

[hard garnet]

hey there

why with the latest version of the requests package, it doesnt print IDs of 64 characters, and only of 32 characters
I think the code needs to be upgraded
I have the requests version 2.32.3

[grave hornet]

Does anyone have a good guide to help me set up stable
Diffusion with a and graphics card cause I don’t got the money to get an nvidia graphics card anytime soon

[ornate elk]

Wrong model size

[granite bough]

Wdym

[ornate elk]

How large is your flux model?

[granite bough]

Like 20gb

[ornate elk]

Yea and your GPU has 16gb vram

But the model needs to fit inside, to not use RAM instead

So use a smaller gguf variant or fp8 version

10-14gb would be good

[granite bough]

Where do I get those

[ornate elk]

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

[granite bough]

Ty

[grave hornet]

I meant amd*

[lucid yoke]

@pure tusk sorry to ping but schoolshooter threat/racism

1356568018032918538 userID

[ornate elk]

Hey, sure check the first link from the pinned messages for the AMD Guides.
Go with a Zluda one

[grave hornet]

Thank you I didn’t see that

[flat olive]

is there any way i can jump the automatic download of the model? during the installation?

[dapper wing]

please god can someone explain to me how to make stable diffusion install a newer version of torch

keeps collecting torch 2.1.2

torch 2.1.2 throws the error attribute "torch has no attribute 'get_default_device' "

this problem doesn't happen in torch 2.3.1 but god damn it stable diffusion refuses to install it

[dapper wing]

OK the solution was to execute stable diffusion by firing squad and fresh install. this is becoming a bit of a thing

[ornate elk]

Place a checkpoint in the models/stable-diffusion folder before launching the webui-user.bat

[lucid yoke]

Ahh a1111

Yeah outdated ui really

[cobalt mango]

https://platform.stability.ai/docs/legacy/grpc-api/features/text-to-image#TypeScript
Has there been any official communication from Stability regarding the deprecation or removal of legacy endpoints? We're relying on one of them for multi-sample generation, so it would be helpful to know if there's a timeline or roadmap for discontinuation.

[daring reef]

I am installing Vace and the vace example workflow, i am getting this error of a missing node "ModelPatchTorchSettings"
I open the manager to install missing nodes, but after updating and restarting, the same error appear and the same missing node appear

And there is not conflicts with the ModelPatchTorchSettings node

[daring reef]

Solved

[lavish jacinth]

Does anyone know if Forge works well with AMD GPU's im planning on getting a new pc and getting a 9070xt but i know ages ago AMD wasnt very good for stable diffusion but i figured enough time has passed that its probably good now. but i know it was a little more of a complicated install

[azure magnet]

Hi, i need help. How can i import a workflow (image) from civitai to comfyui?

[open fox]

If the workflow is embeded simple drag the image into your comfyui browser window

[azure magnet]

already tried but doesn't work idk why?

[open fox]

Then the image does not have the workflow embedded or you have downloaded a preview

[azure magnet]

I have the generation data copied, i made the image myself. Do you know if i could paste it somewhere?

[open fox]

Well the whole workflow is embedded to the meta data of the images.

[azure magnet]

Download and dropped but doesn't work

[open fox]

is it sfw?

then you can send the link

[azure magnet]

Is sfw but could i send you privately?

[open fox]

yes

[fair oxide]

Finally got this working using ComfyUI... Do i need models? I'm brand new to all this so not sure what i need to begin making images

[open fox]

For the start open a simple workflow from the "browse templates" menu. Start with the workflow templates they ask to download the models too

[fair oxide]

Perfect thank you!

[open fox]

Depending on the GPU you have you could look for the flux.schnell workflow

[fair oxide]

Nvidia Tesla P100

I'm using a linux server (headless) is there a command to just download all of these?

[open fox]

Well i downloaded most with wget from the cmd and huggingface or civitai as source

[fair oxide]

where do you put the downloaded files?

im so confuserd

[open fox]

Well depends on the files. Text encoders etc. went into the clip folder. Models from SD1.5 SDXL into checkpoints, flux and hidream into unet, loras into loras,....

Yes it is confusing.

[fair oxide]

Steep learning curve haha. So for example where does v1-5-pruned-emaonly-fp16.safetensors go?

which folder(s) ?

ah nevermind, models 😄

[open fox]

To start i would suggest you look at:
https://docs.comfy.org/tutorials/flux/flux-1-text-to-image
In the Menu you will find tutorials and within the text you find such file structures to help orientation

[rustic roost]

Was working jsut fine yesterday, now it no longer recognises the model?

[ornate elk]

GGUF models doesn't work with Auto1111

[rustic roost]

Very peculiar considering it worked with A1111 when I installed it yesterday... Very well, will keep in mind!

[ornate elk]

But you can use them in Forge webui

[ivory plover]

I'm trying image to video workflow in Comfyui, but most of the nodes are missing, even I can't find the missing nodes in the manager..

[nocturne juniper]

guys how do i train a specific style

like lora

would appreciate if yall send me like a tutorial or sum

[ornate elk]

You can use tools like OneTrainer or Kohya_ss to train loras

[daring reef]

I am getting this error using this to upscale videos

[ruby yoke]

you installed comfy to your desktop. that's the wrong place and a lot of things won't run.

[daring reef]

So i just move the folder?

[ruby yoke]

nope. you have to delete comfy, and i would advise also uninstalling python, then go get the comfyUI windows installer, run it, and click manual install, then find the C: drive for location, and install it on root - right on C: - and make sure you ahve room. then you can put yoru models on some other drive if you want, and just list them in the extra model path's file

[azure magnet]

Hi, how to fix it?

what could be?

[flat olive]

which arguments line are the best for a 1660super 6gb? like to optimize the time it takes to generate, i'm using Forge as cs1o recommended me a long time ago, this are the one's i'm using rn

[pale garden]

Are there any SD Zluda AMD users here? I feel like generations are abnormally slow. A flux 1000x1200 2 image batch literally estimates 1 hour 30 minute completion time
Illustrious is maybe like 2 minutes for 2 images

[daring reef]

But i have the portable version

[ornate elk]

Yea what's your gpu?

Your flux model is maybe to large for your vram

[flat olive]

is forge compatible with animatediff?

bc i tried to install it, and then hit apply and restart but i'm not seeing it anywhere

[ruby yoke]

just becasuse comfy itself will run happily anywhere does not mean comfy manager will or that the custom nodes for it will

[pale garden]

7800xt which is 16gb vram

[ornate elk]

And which size has your flux model?

[dapper wing]

hahahaha yep really need to make the switch

[pale garden]

flux_dev.safetensors

[ornate elk]

Size like GB, MB, bytes,

[delicate urchin]

910 second generation time for a 20 setep flux dev fill workflow on a 1MP image in ComfyUI with a 4090 card. Why is it now so slow and how do I fix it?

ah my gpu is at 0% usage

[delicate urchin]

Comfyui says 100% gpu usage and taskmanager says 0% usage

as shown in cmd pytorch is also running on the 4090 card and cuda is available

[daring reef]

I am getting this error

[lusty fractal]

Hi. Anybody on who knows swarm ui?

[lucid yoke]

Yes but instead of asking that why not ask the problem your having?

[lusty fractal]

fair point^^ well i didnt install it over github because i didnt figure out how. also i dont know how to find out if xformers is active. the video generation is very slow. 480p 4 sek takes about 1hour30min

[lucid yoke]

What, what do you mean you didn't install over github?

Its just a .bat file to install

Also video generation can take a while, what gpu do you have?

[lusty fractal]

im pretty new to ai so i didnt really know what i was doing. i mean installing wan 2.1 was quite difficult for me. i tried adding xformers to the launch file and it wrote a lot of error messages saying didnt install over git hub

[lucid yoke]

Hmm well what gpu do you have in your machine?

Nvidia? Amd?

Intel?

[lusty fractal]

amd ryzen 7 7840hs

[lucid yoke]

Ahhh

Thats the cpu?

[lusty fractal]

i am getting a new tower pc which is more powerful but i thought xformers could help nonetheless

[lucid yoke]

I meant graphics card, or do you only have intergrated graphics?

You can check in task manager (control alt delete) and then the second tab

[lusty fractal]

nvidea geforce rtx 4070

[lucid yoke]

Yes thats what we needed, hmm 12gb vram should be able to run the 1.3b version no problem

Not sure about the 14b

Swarm will auto apply the settings for your gpu

But how exactly did you install swarm?

[lusty fractal]

actually i cant really remember...i tried lots of stuff in the end i followed a youtube video...

[lucid yoke]

https://github.com/mcmonkeyprojects/SwarmUI?tab=readme-ov-file#installing-on-windows

[lusty fractal]

https://www.youtube.com/watch?v=4mVCJSqazh0

thanks for the assist. but do you think the videos should be generated faster?

[lucid yoke]

Hmm if you use causvid lora maybe

12gb vram might not be enough for the 14b version

[lusty fractal]

i see. what video generator are you using?

[lucid yoke]

swarmUI with 16gb vram

2-4min

[lusty fractal]

damn thats fast.

but does swarm ai come with an integrated video generator? i installed wan

[lucid yoke]

You might be better off joining the swarmUI discord (link on the github above)

[lucid yoke]

No, wan is the model to generate video

[lusty fractal]

ok

well im also using stable difusion automatic 1111. do you know a way to create vids in that?

i tried deforum and animate diff but with unsatisfying results

[lucid yoke]

Animatediff but comparible to wan its absolutely awful

[lusty fractal]

i see so ill stick with swarm for vid creating

[lucid yoke]

Yeah, swarm can handle all models not just video

A1111 only stable diffusion models

[lusty fractal]

i see. well im going to join their discord server. thanks for the help.

[ruby yoke]

they also have a channel on this one

[fair oxide]

does anyone have any advice on making accurate faces? I've been trying and I'm using a few LoRA's based on the subject matter but they are all distorted some how and not accurate they look bizare. I'm at a road block.

[open fox]

It depends on the model. I guess you are using SDXL and then you might trying use either a refinement step (second pass run) or a facedetailer / repair procedure.

[fair oxide]

I may sound silly here but how do i know if im using SDXL?

[open fox]

Well different generations of AI Models have been around for a few years now. Some models architecture as generations tend to specific problems. If you would asked for aweful hands i would have guessed you use SD 1.5.
New generation Models like Flux oder SD 3.5 are quite good in creating faces.
So i took a wild guess that you might have trouble with a SDXL Model (would also include some evolved models like Pony or Illus). To check it would be helpful to know the model (size) and the front end you are using

[fair oxide]

I'm using ComfyUI

Sorry I'm new to all this, I'm using a small model of just the subjects name, high quality photo realistic etc.

[open fox]

And the name of the checkpoint is? Not the lora model.

[fair oxide]

ultrarealisticbystable.v10

[open fox]

Ok i would say this i even stable diffusion 1.5 Architecture. So the bad faces and some bad anatomy (limbs, fingers,...) is pretty normal for these models. If you have installed the "impact pack" within ComfyUI you have a node called face detailer.

[fair oxide]

I have just installed face detailer but im struggling to find where you load it from haha sorry this is such a steep learning curve

[open fox]

doubleclick on an empty space of the node editor and search for face detailer or use the node library

[fair oxide]

thank you very much

i cant work out how to connect everything haha colors are missing on the things im trying to connect to

[fair oxide]

how do i get these files?

Prompt outputs failed validation:
InstantIDModelLoader:

• Required input is missing: instantid_file
  ControlNetLoader:
• Required input is missing: control_net_name

[pale garden]

18.6gb model

[open fox]

Sorry thought you found help on the comfy discord channel...
Well i am not even sure if you want to use these nodes. As i am not sure if it does what you want to achieve.
The models shown are used for "face swap". Your Question earlier was how to fix bad faces. Sure you could always swap another in but for repairing there are other ways.

[serene notch]

hey, so i have a 2070super with 8Gb of vram. i'm trying to generate sdxl images using A1111 and it works great, only issue is it takes from 5 to 10 minutes per image at a resolution of 768x1024. What can i do to speed up the process apart from buying a new gpu, wich i canot really do rn?

[fair oxide]

noo not at all, no need for any sorry you've been so helpful. i gave up earlier as this stuff is just so complicated to me.

I was told to copy someone elses project. i cant do it without those two. I'm not sure what it does haha

[fair oxide]

What?

I'm using a much older Nvidia Tesla P100 with 12GB VRAM and it does images even at that res in far less time. sounds like something is very wrong there. a 2070 super should destroy my card.

was told to replicate this

Just did a image, 1920x1080, Requested to load SDXL
loaded completely 10311.825019073487 4897.0483474731445 True
100%|███████████████████████████████████████████████████████████████████████████████████| 20/20 [01:41<00:00, 5.08s/it]

Your 2070 super should be much faster. I would say its maybe VRAM but mine is using 56% of my VRAM so you should be well within spec.

[serene notch]

damn, i have no problem in games though, and when generating images it's usage goes up to 99% so idk what's the problem exactly

i'm currently generating an image using this model: https://civitai.com/models/827184/wai-nsfw-illustrious-sdxl , i'll send the exact settings and time it took once it's done

i also have a ryzen 5 7600x and 32Gb of ddr5 6000 memory so i doubt it's a hardware issue

so, without any loras this generated in 5:45 minutes, sent the image output if you wanna try and import the settings (sorr yhe's naked, forgot to prompt him clothes lol)

here is also what my gpu usage looks like while it is running

[fair oxide]

yeah, should be fine. im fairly new so perhaps im missing something but seems like a long time to me

processor at 5Ghz turbo, i like it 😄

[serene notch]

yeah ryzen always running at full throttle lol

[fair oxide]

My main machine is a Ryzen they are good processors. Your machine is better than mine so i have no idea why its taking so long.

[serene notch]

yeah i'm not sure either, probably just because of different models if there isn't any issue i guess

[fair oxide]

yeah

[lucid yoke]

did you install with zluda?

[serene notch]

i don't think so? just cloned the git repo and ran the update and start.bat

https://github.com/AUTOMATIC1111/stable-diffusion-webui/releases/tag/v1.0.0-pre i downloaded this version exactly, it's the one they tell you to use for nvidia gpus

[lucid yoke]

ah you have 2 gpus

mb

saw gpu1 and saw amd

[mystic basin]

maybe deleting venv will fix the issue

[pale garden]

run wmic path win32_VideoController get name in cmd

[hallow latch]

Need some help

This is how it looks at like 90%

then it's

[lucid yoke]

whats the model?

also normal quality:2 isnt helping imo

[hallow latch]

The model is toonyou_beta6 something. I try to use Euler a but wont work.. If I use like DPM++ 3M it works :/

[ornate elk]

Yea thats to large. Use a smaller gguf model. Like Q4 or an fp8

[fair oxide]

is there anyway to increase the number of cores/threads used when CPU work is being carried out. its only using one of mine which leads to 4% CPU.

[ornate elk]

You need to have --xformers --medvram-sdxl in your webui-user.bat for sdxl to work faster

[ornate elk]

Wrong vae mostly

[serene notch]

omg thank you so much, i went from generating an image in 5 minutes to generating the same image upscaled 2x in 1:30 with these enabled

[ornate elk]

Perfect! No problem 😁

[neon orchid]

hey guys a question

Can you install stable diffusion with 2GB of VRAM?

[neon orchid]

My friend stayed here, the program is not progressing.

[lucid yoke]

You probably can run SD 1.5 at a slower speed

[hallow latch]

It goes from good to shit 😛

There is nothing else than may cause it?

[ornate elk]

No its the vae

Your trying to use a pony VAE which is based on SDXL for an 1.5 model (toonyou)

[hallow latch]

I downloaded both SD1.5 and Pony and used automatic but kinda went weird when I tried to use Euler A

[lucid yoke]

Then why dont you set it to manual?

[hallow latch]

This is with an VAE based on Pony, Euler A and still

[lucid yoke]

Pony is made for 1024x1024

SD1.5 is made for 512x512

[hallow latch]

I like using 768 with Pony.

The only thing that changed on these 2 images was I changed from 24 steps to 25..

The vae is set to None as it says on the checkpoint that vae is built in.

Ofc the seed is different due to I am using -1 but it should not affect it that hard right?

[lucid yoke]

A LOT more changed since the pose etc is different

Hmmm

Seed doesn't matter that much but something else changed too

But idk the ui your using

[hallow latch]

I am using auto1111 stable diffusion

The pose is prob different because I don't tell that much to the ai. It only gets "score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, solo girl, bikini, striped bikini, brown hair" so it will make it very different

[wheat flower]

I have no Idea what I'm doing wrong. I put them in embedding it doesn't appear, I put it in ESGRAN it doesn't appear. What in the world am I supposed to be doing to make these Embeddings appear?

[flat olive]

how i can make img2vid? without using comfyui?

[lucid yoke]

scam

[lucid yoke]

if you joined that server leave it as its a scam

[ornate elk]

Remacro goes into esrgan.
The others in embeddings.
Then relaunch the webui and select a compatible model for the embeddings to show up

They won't show up if they are not compatible

[wheat flower]

Got remacri to work. the embeddings not so much

[ornate elk]

What's the checkpoint you tried to use them.with?

Because the embeddings you downloaded are for 1.5 based checkpoints

[warped canyon]

is there a way to make it so that i can name the folder that batch process creates?

[hallow latch]

This is good command? ( set COMMANDLINE_ARGS=--use-directml --opt-sub-quad-attention --no-half --disable-nan-check --autolaunch --medvram )

I used these settings and it took me Time taken: 1 min. 42.9 sec. with an Radeon RX 7800 XT. Can I get it to do better?

[lucid yoke]

By using zluda yeah if your card supports it

Instead of direct ml

[hallow latch]

--use-zluda?

[lucid yoke]

Check pinned messages install guide from cs1o

If your card can be run with zluda or not

[hallow latch]

If I turn on the --use-zluda .. it uses the CPU .. but my card is supposed to handle zluda

In the command prompt it says: ZLUDA support: experimental

[lucid yoke]

Maybe your better off waiting for cs1o to come online, he's the amd expert

Did you follow the steps in full?

[hallow latch]

I followed his steps previously with zluda etc and did not get it to work to 100% so I skipt it for a few days. Watched a video and got it to work with the GPU

But using zluda will only force it it to use CPU atm

[mystic basin]

zluda should work with your RX 7800 XT

you can probably add CUDA_VISIBLE_DEVICES=1 in the launch file (or 0 instead of 1, it depend if you have an igpu with your gpu or only a gpu)

you should be able to do :

wmic path win32_VideoController get name

in cmd to get the name of your gpu / igpu

Name
AMD Radeon(TM) Graphics
AMD Radeon(TM) RX 6850M XT

for me the result is this which mean my RX 6850M XT = 1 and that my igpu = 0 since my igpu is the first result and my gpu the second

[mystic basin]

unless it's set HIP_VISIBLE_DEVICES=1 with zluda (i'm not sure, i switched to comfy and it's not the same way to change it so it uses my gpu)

(sorry if my explanation are bad) but i had to do add something like that in the launch file so it knows which one should be used (igpu or gpu) when i switched from direct ml to zluda when i was using automatic1111

it should be the same thing for forge and idk about the other (except for comfy)

[tame sun]

Do we got optimized gfx/ROCM for AMD's 9070s yet?

https://tenor.com/view/pokemon-3000years-gif-12110509

[pale garden]

I tried to run a smaller flux model (11.6gb) (GGUF file), and i am getting some kind of clip error. (File "C:\Sd-Zluda\stable-diffusion-webui-amdgpu-forge\backend\loader.py", line 62, in load_huggingface_component
assert isinstance(state_dict, dict) and len(state_dict) > 16, 'You do not have CLIP state dict!'
AssertionError: You do not have CLIP state dict!
You do not have CLIP state dict!)

[mystic basin]

you have the clip file?

https://huggingface.co/comfyanonymous/flux_text_encoders/tree/main

you can find it here if you do not have it

can anyone help me with this? i keep getting this error for no reason, it wasn't happening the last time i was using comfy, i don't remember updating it since then (comfy with zluda)

it was working perfectly fine the last time i used it (i don't think i updated comfy since then) (yeah this is a repost but this is an issue i haven't managed to fix)

the issue is not caused by changing the cuda visible device since even if i change it or keep it like it was before i changed it, i still get the same issue

even after completely redownloading comfy completely

[ornate elk]

I'm on vacation, so I can't help that much for the next 2 weeks.

[lucid yoke]

ah fair

[pale garden]

Shouldnt you be setting CUDA to :0 if you are using an AMD gpu? RocM does the cuda process automatically on zluda

Also another thing to note is that the recent amd gpu updates are incompatible with zluda.

[pale garden]

I have yet to use any model that needed clips. Also flux 16gb apparently did not need me to download clips, it worked it was just super slow, Also that link says comfyUI, i am running forge

[mystic basin]

yeah but the clip is basically a file needed for certain flux model (it's the same for any ui)

some already has it some does not

idk why

[mystic basin]

never worked, setting it as one used to work

[mystic basin]

i didn't know that

that must be why

ty

[pale garden]

Does it need a specific clip? I downloaded flux-schnell-dev-merge-q8-0.gguf but i cant even find it anymore on civitai

[ornate elk]

You need clip l and the t5 clip model for Flux to work

They go into the clip folder

Only flux and newer models need extra clip files

[sudden pike]

ya know how i can update Zluda?

have seen that it "works" with 25.5.1 now

but idk how to get the version

i just cloned it into the folder but doesnt work 😦

[wheat flower]

Hassaku

[ruby yoke]

https://huggingface.co/mhnakif/flux-schnell-dev-merge-gguf/blob/main/fluxunchained-schnell-dev-merge-q8-0.gguf you'll find it on huggingface

[ornate elk]

Check the size. If its a 6gb model it won't work with easynegative for 1.5

[past hemlock]

Hey everyone, i downloaded and installed Framepack using the one click package on the github page,

but any time i try to turn a image into a ai video, the program starts sampling it fine, but once it finishes sampling, the Gradio tab becomes completely unresponsive and filled with errors, and the program seems to constantly transfer the model data to itself for a couple times before stopping..

I have a NVIDIA Rtx 4080 graphics card im running it off of, with 32 gb of ram.

if anyone has like a list of things i need to do or make sure i have installed to make it work, i'd SUPER SUPER appreciate it. im at a loss.

i know that the ffprobe not found thing meant that i needed to install ffmpeg, so i downloaded it and installed it but i cant figure out how to get framepack to recognise it.

[fair oxide]

is there anyway to make this use all CPU cores?

[vocal burrow]

add it to your system PATH

[vocal burrow]

SD runs on gpu (unless specified otherwise) so there is no point trying to allocate more cpu cores to it.

[slow hill]

Not sure where to post this but:
⚠️ user linda06_zee seems to be spamming users of this discord server with dms.

Also the "report to staff" feature is still broken

[fair oxide]

I'm trying to train a LoRA using this guide: https://thundaxsoftware.blogspot.com/2024/08/train-your-own-lora-with-comfyui-step.html but im getting this error upon actually training. any clue why? The directory is where it should be and i have png files and generated txt files

[atomic eagle]

Hey all, I came here hoping to find some news relating to the persistent API error I'm receiving "Image generation failed: Stability API error: 404 - : An upstream service tried to fulfill this request, but a required resource was not found." This seems to be elated to the API status as reported on the status page. Thank you so much for making that info available, btw. Is there any way to anticipate when this might be cleared up? I am an app developer driving the image generation elements of my project with Core and Ultra. Both very impressive models aside from the present connectivity issue. Hopefully we can look forward to services resuming shortly - any forecast?

*** Edit - I was just able to generate an image. It appears that the API is back up. Is this an issue that is expected to be kept under control?

[pale garden]

So this is flux schnell with the clip added? Why would the civitai person not have it already included

[ruby yoke]

Its the gguf version

[pale garden]

?? i already had the gguf version installed and i got a clip error

[fossil dew]

Is it easy to install the animation thing to Forge

I forget what it's called.

[valid pewter]

hi guys im using sd webui replacer, and every time i generate images, the models SAM and grounding dino are loaded each time. is there anyway to prevent this for faster speeds?

[iron lantern]

hi there folks ... needed some advice on getting a new gpu i have a z390 pro 4 mother board ... and a 970gtx it takes like 30mins to make an image ... i rang the pc store ... and told them i have a i59600k .... they told me i can get a 3060 ... but it will run at like 75 percent .... then he said if you can find a 2nd hand 2080ti or something like with warrenty ... your better off with that because your cpu wont bottle neck the 2080ti ... what you guys think i am wanting to make images in under 30seconds or so .... not 30 mins XD thanks ❤️

[flat olive]

hey, i been trying to fix the issue with sd-forge animatediff but i can't, someone's got an idea about how to make it work?

[flat olive]

fuck that was scam?

[lucid yoke]

Any link out of this discord is most likely a scam

Like "helpdesk here" or that kinda stuff

[lucid yoke]

But animate diff is kinda old, any specific reason you wanna use that?

[flat olive]

not really, i'd like to know how i can make img2vid locally with my gpu

but im not finding any way

i've seen some videos about svd but idk if i can run it with my gpu or even how to install it or make it work

[fallow knoll]

framepack is the only low vram budget way to make img2vid that i know of, it only requires 6gb of vram, but at that amount its slow.

[lucid yoke]

Ltx also worked but the quality is rather bad unless its a scenery shot

[flat olive]

i will try framepack

[flat olive]

what do you mean by scenery shot?

[lucid yoke]

Landscape, nature

[flat olive]

like no complex movement?

[lucid yoke]

Nothing human that's for sure

[plucky basalt]

Do you need stuff like comfy ui to install stable diff?

[past hemlock]

@vocal burrow Hey just wanted to say that what you said to do worked, after i looked up how to do that, it seems to generate videos from images fine now!

[shut apex]

@half island

This is the problem

[half island]

What's your specs?

Also ,why so old torch? We're on torch 2.7 on the latest today.

[open fox]

You are missing python pip which handles the installation of modules

or you have forgotten to active your virtual environment

[half island]

It's active.

[shut apex]

However I downloaded python. That's what I don't understand. Did I get the wrong version?

[half island]

@shut apex What's your specs? And i recommend forge instead. It's auto, but not broken/outdated.

Ah, both of us oversighted xD

[open fox]

You could try "py -m ensurepip --upgrade" to install pip

[half island]

````py/python -m ensurepip --upgrade` ```

Add those ` symbols to make a fancy bracket

[shut apex]

I can't write in the same window. I have to open a new one with cmd?

[half island]

Yep. Is this a fresh install? And is it a portable release? Or manual?

[shut apex]

the files ? I downloaded them 2 days ago

[half island]

Automatic's last push for new stuff is almost a year outdated lol

https://github.com/lllyasviel/stable-diffusion-webui-forge

Here. Get forge instead. AIO package for easy install for newcomers

[shut apex]

Ok I will try this

[shut apex]

@half island so it won't download it to me. It tells me that the file is not available on the site

[half island]

Screenie please

[shut apex]

Yes I'm french

[vocal burrow]

That s not exactly what it tells you tho. where are you downloading it ? do you have the rights for that folder ?

Might be worth to check your antivirus too. It might be intercepting the files and doing weird stuff.

[half island]

Full window screenshot. Got no idea what that is from.

Or is that the browser?

@shut apexHere. Packed for shipping over discord. Just extract 001 file with 7zip and rest follows.

[shut apex]

@half island

[half island]

Specs?

Or specifically, which gpu?

[shut apex]

finally the download worked, but by clicking on the "run" file, it gives me this

[shut apex]

What is "specs" ?

[half island]

Specifications of your pc. Graphic card, cpu, etc

You can check specs in task manager. Cpu by clicking on it, gpu at the bottom of the performance device list.

[shut apex]

GPU : AMD Radeon RX 6800
CPU : AMD Ryzen 5 5600 6-Core @half island

[ruby yoke]

read through the pinned posts in this channel - you're looking for @ornate elk 's AMD gpu guides

[shut apex]

Didn't work

[vocal burrow]

don't install it somewhere with spaces in its path

(and overall don't install anything on your desktop)

@shut apex

[ruby yoke]

install it to c:\A1111

[shut apex]

@vocal burrow @ruby yoke same shit

Is it because there are other files in A1111?

[vocal burrow]

delete the venv folder and try again

[shut apex]

Different result, but always a problem with Torch

[ruby yoke]

how about - get rid of auto1111 and use https://github.com/mcmonkeyprojects/SwarmUI - it generates, it runs comfyUI, so better than a1111 by a very long distance, and the developer @jagged plover is here in this discord if you need him

[half island]

Because you don't have cuda cores. And i recommend not using automatic one, it's ass, and quite outdated.

And i will guess you've not installed ROCm?

Here, try this one. Forked and made for amd with zluda in mind. A middle layer that from what i've heard will give amd gpu's quite a boost in performance.

https://github.com/likelovewant/stable-diffusion-webui-forge-on-amd

Or try comfyui if you wanna try node based workflows. Also doesn't go through the entire process from start to finish like forge/auto does. It only re-does whenever you make a change.

https://github.com/comfyanonymous/ComfyUI

It has several mentions down on the git's page with various ways to use amd gpu's

[fossil dew]

Is this a memmory issue (again) ?

[fossil dew]

I closed it and tried to open it again, but its doing the same thing it did before.

It will gen another image, but when its complete, it dissapears and shows the previous image

no errors in python

Where's CS1o when you need em 🤭 hehe

[fallow knoll]

said he wouldnt be back til after the holiday i think

[fossil dew]

😮

WHELL- Hopefully it figures itself out then I guess 🤭

Guess Ima try restarting the PC 🤷‍♂️

brb

[fallow knoll]

here

[fossil dew]

insane I hope he has a great time tho

[fossil dew]

If anyone could help me with this issue, i'd be greatly appreciated ❤️

[vocal burrow]

what's the full log ? copy paste everything in a .txt and drop it in here

[vocal burrow]

you updated driver (automatically maybe ?) since seting up sd zluda ?

[fossil dew]

Ill try that 🙂 Thank you

[vocal burrow]

nooooo

no no no no

don't

that's why I'm asking

If you update your driver you basically have to reset your zluda install to some extent. (don't know the details off the top of my head)

So in case you did that I would have suggested you to reinstall. If you didn't, then that's not the problem.

(but if it got installed automatically....)

[fossil dew]

ohhh

i haven't done the thing

I haven't updated any drivers that I know of

uh

there is a new one that needs installed tho

[shut apex]

What is "log" ?

[vocal burrow]

what s displayed in the console, so basically what you screenshot.

but the whole thing, including what s hidden above

[quasi jolt]

its not allowing me in?

can you please check whats happening?

why does your tree have a . in the middle?

[ornate elk]

Its a scam. Dont join their server

Dont give away personal information

[quasi jolt]

thanks

i figured as soon as he liked his own shit and reported him right away

[ornate elk]

Do you still need help?

[quasi jolt]

i also fixed the issue, i thought i need to run using webui user batch to run diffusion

but its just the run bax

batch lol

silly me

[ornate elk]

Its run bat only if you downloaded it as zip file

[quasi jolt]

yeah i got the zip file, from the 60sec youtube guide lol

thanks for your time

[ornate elk]

Ah np 🙂

[sudden pike]

Can you help with Kohya_ss stuff too?

i wanna train Lora's now

[north zodiac]

Hi all,
Forgive me if this is the wrong place to post this. I performed a clean install of my ForgeUI due to an issue with an AMD driver update breaking my original install. I've done the git clone stage and when I launch webui.bat for the first time, this appears on screen.

I tried performing a clean install of Python & ROCm files, but it always comes up with this. Anyone know what I need to do to fix it? Let me know if any further info is required 🙂

[upper linden]

good afternoon, im getting a new error after trying to make a image

[sudden pike]

iirc newest AMD version does not work

i downgrade to 25.4.1 everytime

[north zodiac]

Thing is, it did work fine until about a week ago. I updated the drivers, but I got a separate issue. I tried to perform a clean install but instead got the problem mentioned above :/

Would downgrading to 25.4.1 fix my problem?

[sudden pike]

iii dont know

u can try

[north zodiac]

Is there a way to downgrade through the AMD Software interface? Or do I have to download a specific file?

[sudden pike]

Uhh go on the AMD website there you can download the driver file and start the setup

It's around 800-900mb big

[north zodiac]

Thanks mate, I'll give that a try 🙂

[shut apex]

I have this when I click on "run"

[shut apex]

This ?

[sudden pike]

did it work?

@pure tusk Sorry for the ping. But there is an Token Grabber again...

Very fast reaction damn, respect

[north zodiac]

I've reinstalled 25.4.1 and tried opening the webui.bat file again, but had the same error. I've not yet had the opportunity to try restarting my PC yet though

Will restart shortly though

[sudden pike]

was worth a try 😅
im also facing some problems with Kohya_ss atm bcs i wanna train Lora

[north zodiac]

I was going to attempt to try that as a challenge for myself until AMD decided to shat up my plans lol

[sudden pike]

xD

[north zodiac]

Just tried restarting and trying again. Didn't work :/ I did a bit of research online and it looks to be something to do with Python somehow

[sudden pike]

@pure tusk again

[vocal burrow]

no

[vocal burrow]

This. But what's above it too, the whole install log

[shut apex]

Ah okay

Now it's this :/

I've deleted the Forge folders that I've downloaded, and now I have this...

[ornate elk]

Dont install the webui into the Documents folder

@sudden pike @north zodiac @upper linden
Sry guys, dont have time rn to help, check my guides and that you have AMD Adrenalin 25.4.1 or 25.3.1

[ornate elk]

Follow the AMD Zluda guides on the first link of the pinned messages.

[north zodiac]

No worries mate. I'm currently reinstalling the webui in my user directory. Didn't realise that made a difference. Will let you know if it works though

[shut apex]

this one ? https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides

[ornate elk]

Yep

Go with Auto1111 Zluda or Forge with Zluda

[shut apex]

I can't access it

[ornate elk]

The website?

If github is blocked in your country, try change the DNS Server or use a VPN

[north zodiac]

I know you said you've not got time to help, so please reply if/when you have time.

I've tried to install it in my user directory, but I am still receiving the same issue as my initial message :/

[half island]

Don't even know what you downloaded. Forge fork with amd, or comfyui

[shut apex]

Same issue

[shut apex]

The first

Forked

Eh... the support guy ask me to send him my wallet adress. Should I be worried?

[sudden pike]

yes?

For support they dont need any personal information

related to Stable Diffusion

[shut apex]

That's what it seemed to me, I see no connection

[sudden pike]

Stable Diffusion is open source iirc

so its FREE

[shut apex]

I left the server. Don't want to get scammed

[sudden pike]

was it Stable Diffusion related?

[shut apex]

I don't know

the link was send by @azure spindle

[sudden pike]

@pure tusk

Sorry again. idk why it's SO common here

nancy deleted a message. should be in the logs if you guys have logs 😅

[pure tusk]

banned

[sudden pike]

❤️

why is it SO common here?

[shut apex]

Don't worry. I immediately had doubts when they asked me for my wallet address. All I sent was one of the cmd windows screens

[sudden pike]

Easy thing. If people want you on a Discord server or click on links to get support

ignore them

espacially here happens alot that people want to invite on servers etc

[shut apex]

Ok I note

Thanks

[sudden pike]

no problem

[upper linden]

its okay, i have reinstalled, and deleted my venv just incase and it still gives me the same error, whenever you have the time, i would like assistence

[shut apex]

In the meantime, that doesn't solve my problem I'll see tomorrow

[sudden pike]

yeah he has holidays

[half island]

yeah, that's called phishing. Stay FAR away from those.

If you hold mouse over link, you can see the true address. It's like this. civitai.com

click it, it's not civitai, it's huggingface. That's how they fake it. Usually with the scam you faced and fell for, they do 9%3discord.com/"inviteurl" in the actual link. Discord in their vast knowledge just keeps adding ammo for scammers lol

[half island]

Did you properly read the description on the git page?

[jaunty nymph]

Hi. Is the install guide pinned here up to date? 11 months seems like a lot

[lucid yoke]

Installation guides dont really change over time unless its a really big change, what UI are you talking about?

[undone bramble]

im wondering if i should get the 5070ti or the 3090, any suggestions?

[vocal burrow]

mostly depends of what else you re gonna do with it.

5070ti will get longer support overall than 3090 when it comes to games. but AI wise they both have 16gb and that s what really matter.

And 5070ti should be faster too accoding to some benchmarks from : https://vladmandic.github.io/sd-extension-system-info/pages/benchmark.html

oh wait the 3090 is using wsl2....

let me get better numbers

that s probably closer to the reality

not quite the same exact setup but yeah.... overall they seem similar ish performance wise for AI workflow

[potent sun]

does anyone know hot to setup image inputs in text generation webUI?

tehre should be an option for multimodal but I cant see it

[shut apex]

Then I realized that my graphics card was not up to date. So I updated it, and now I have this

And yes I read correctly. You need Python: I have it, and it does Git: I have it too

[vocal burrow]

there s still something broken above ..... still can t help if I don t have the full log

got to read errors from the top to bottom.

[shut apex]

The problem is that I don't understand what you mean by "log" (I'm bad at computers). Isn't that the entire cmd?

[vocal burrow]

if you scroll in the window, isnt there anything written above ?

cause activate.bat not being recognized as an executable is bad.

[shut apex]

No exactly. Everything marked in the window is there

[vocal burrow]

ok

[shut apex]

The problem remains the same: there is something preventing me from installing torch

[vocal burrow]

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

[vocal burrow]

does not matter for now.

if you can t even find the keys for your car it does not matter if the engine needs oil. you find the keys first.

activate.bat not being able to run fine, means broken venv, means broken install

[vocal burrow]

ok. Does this file actually exist ?

[shut apex]

Yes. This is the one I used for the files

Ah wait

[shut apex]

There is no "activate.bat"

[vocal burrow]

ok, so most likely then it has been removed by your antivirus for some reasons.

check that, add an exception for it, restore the file and it should be fine.

[shut apex]

Weird... there's nothing in the Avast chest virus :/

[vocal burrow]

avast should not have any issues with it regardless so probably not the anti virus

Then the initial setup must have gone wrong somehow.
So delete the venv folder again, rerun webui-user.bat.
Copy paste the whole log (white lines in the black window thingy) in a .txt file and send said file in here.

[shut apex]

What log ? I don't understand that part

[vocal burrow]

This !

and everything above it. the whole text, not the last few lines

once you delete venv and rerun webui-user.bat it will try to rebuild the whole venv folder and I m guessing that stuff should go bad and throw some useful error messages then.