🤝｜tech-support

[slow cedar]

I've literally never had this issue before

[red cobalt]

how can i make it faster btw

or is it system related

[slow cedar]

Get a better gpu

I'm not sure of there's a way to make ot faster, I just suffer

[red cobalt]

i have a 3070 bruh

i cant think of prompts

so ill just look at 34 for tags

[slow cedar]

Use wildcards

[red cobalt]

whats that how

[slow cedar]

Or yeah just look at danbooru tags

Wildcards

It just uhh simply put

Randomizes prompts for you

https://github.com/AUTOMATIC1111/stable-diffusion-webui-wildcards

You'll need to actually download the wildcards too but they're text files so they're small

[red cobalt]

u think i should try a 2048

x2048

it takes like a min for 1028x1028

[slow cedar]

No

Don't make it bigger

Just use hires fix to make it bigger or upscale it in img2img or smth

[red cobalt]

does wifi matter

[red cobalt]

oh

should i jsut make 512x and make it bigger

[slow cedar]

For what 😅

[red cobalt]

ai creation

generation

[slow cedar]

I just use uhhh 832x1216/1216x833

Depending on if I need a landscape or portrait

[red cobalt]

oh id prolly prefer height

ngl i could just try looking up stuff thats never posted on 34 and post them lmao

cuz why the fuck not

[slow cedar]

Then no, internet only matters in sd for installation/installing new extensions I think

[red cobalt]

@slow cedar how do i resize to 4k or 2048

[slow cedar]

Img2img works

Or while genning an image you can turn on hires fix and it will automatically upscale it as it gens?

Both take more time though

[red cobalt]

an u save prompts

[slow cedar]

Yes go to settings, scroll all the way down

You'll see a tab that says defaults

Just apply

It'll save everything currently on your ui

[ornate elk]

With 16gb of RAM you need to increase the Windows Pagefile.
That will fix the crashes and lags

[ornate elk]

Edit the webui-user.bat
Add --xformers --medvram-sdxl
To the webui-user.bat at the line commandline_args=

Which webui do you use?

[sharp osprey]

I will try this tomorrow ty

[ornate elk]

Ok, here is a link how to edit it:
https://www.tomshardware.com/news/how-to-manage-virtual-memory-pagefile-windows-10,36929.html

Make sure its only enabled for the C drive. And disabled for any other drive.
Set the custom size to 16000min and 24000max
Then reboot the PC and relaunch the webui-user.bat
Also make sure to have 15-20gb free space on C.

[sharp osprey]

TY!!!!

[empty cliff]

Can swarm run 1.5 SD?

And does it have dreambooth?

And does it work with 50 series out the box? Anyone know?

[lucid yoke]

yes

[lucid yoke]

no

[lucid yoke]

no

you need to get the comfyUI nightly build and use that as the backend but i dont have a guide for it rn

[empty cliff]

I got the portable to run

comfy that is

[lucid yoke]

oh that works?

nice

[empty cliff]

but it looks liek its RIP for A1111 and kohya

and dreambooth

[lucid yoke]

oh A1111 was already RIP

[empty cliff]

😦

[lucid yoke]

no updates for the past year

Koyha will get an update probably

[empty cliff]

with support for new cards?

[lucid yoke]

and dreambooth was always mid

probably

[empty cliff]

Looks like it will be wise not to sell my 3090.

This series cards have been out for 4 months already and the support looks abyssmal.

[lucid yoke]

buddy its mostly opensource so people do it for free

give it time as this is not like a simple update

[empty cliff]

is fooocus similar to swarm?

https://github.com/alibakhtiari2/fooocusrtx508090?tab=readme-ov-file

this looks promising

[lucid yoke]

hmm could work

but

i dont like fooocus

[empty cliff]

what is the problem with this card? CUDA is not compatible?

Dont 30 and 40 series have different architectures as well yet I dont recall having issues with those

[lucid yoke]

but ai got going after they were released

AND it does work for me ( using a older version of comfy but messed in some files) to disable torch audio etc and it booted just fine

[empty cliff]

Yea comfy does seem to work but others that depend on torch dont seem to

I guess I'll just have to be patient

BTW does anyone know if tiled upscaling is still the best way to upscale in 1.5?

I honestly havent used SD in like a year

And if it is, can someone help a brutha with the easiest way to tile upscale directly from a gen using Comfy

[glacial comet]

Haven't tried it, but the Efficiency Nodes has a script to du tiled upscaling. It seems easy: just a plug-in/plug-out option
https://github.com/jags111/efficiency-nodes-comfyui

From my side, I try to run WD14 captioning on kohya_ss, but it seems to have problems to locate ONNX model, even if I force to re-download it. Don't know if someone has a similar problem

[fair oxide]

Can ComfyUI tech problems be entertained here?

[upbeat moth]

Hi quick question does Stable diffusion still offer video generation API’s, image-to-video, and image generations? Also is there any restrictions or can you simply follow the API docs provided to get access to the API’s?

[vocal burrow]

Everything about video should be in here I believe : https://platform.stability.ai/docs/getting-started/stable-video / https://platform.stability.ai/docs/api-reference#tag/Image-to-Video

[vocal burrow]

doesnt hurt to ask.

[upbeat moth]

Yes I did read that, that’s why I am double checking because sometimes the docs are outdated and they can have changed the endpoints. I just want to make sure that the API endpoint still works exactly as mentioned in the documents and that they still are valid before I sit down and code it into my backend

[vocal burrow]

it s still clearly advertised so it should still be up.

[simple jetty]

I run the user one.

[ornate elk]

Can you show a full cmd log?

[red cobalt]

ive already had xformers what does --medvram-sdxl do

[ornate elk]

It makes sdxl models run faster on GPUs with 8gb vram

As it manages the vram better

[red cobalt]

how much would it normally use

[ornate elk]

Idk, but it can be multiple minutes difference

[red cobalt]

hm

should i run 2048x2048 or resize 1024 to 2048

[ornate elk]

As no model is trained on this resolution alwasy upscale

[red cobalt]

okay

im using some model called

[ornate elk]

Sdxl models are trained on 1024x1024

[red cobalt]

https://civitai.com/models/376130/nova-anime-xl

that started fast asf

but my pc sounds loud af now

with --medvram-sdxl

@ornate elk what do you use

[simple jetty]

I’ll send one when I’m home

[red cobalt]

how to fix 3 hnads

@ornate elk

[red cobalt]

tried to resize it

without prompts

do i just type 2 hands

is there a way to prevent weird shit like this

[lucid yoke]

Whats your negative prompt

[tribal anvil]

How do I use 2 loras at once

[lucid yoke]

What ui do you use

[tribal anvil]

Webui forge mb I forgot to say

[lucid yoke]

Ah, you just click the second lora

In the loras tab

[tribal anvil]

Click it?

[lucid yoke]

?? What do you do now

[tribal anvil]

I just enter the prompt and click generate

Ye so idk what you mean when you say click

[lucid yoke]

There's a lora tab right

Wait

What lora are you using?

Because i think your confusing lora and checkpoint

[tribal anvil]

Like size?

I'm sorry I barely know anything about stable diffusion

Or the name

Oh no ik it's a lora

[lucid yoke]

Yeah whats the name of the lora your trying to use

[tribal anvil]

One sec

[lucid yoke]

Feeding the dog brb

[tribal anvil]

Ok

I tried a bunch of loras to see if if any of them work

I see the lora tab now I see what you meant by click

I'll try it out

[red cobalt]

none?

[lucid yoke]

Thats the problem, ill give you an example sec

[red cobalt]

how do i like

do i type

Weird shit

on neg prompt

[lucid yoke]

with a checkpoint you put in stuff you want to see right

but in the negative you type in the stuff you dont want

[red cobalt]

no like

wtf do u even call that

[lucid yoke]

so a good example could be worst quality, bad hands, extra digits, displeasing, worst aesthetic, old, early, recent, simple background, simple eyes, simple shading, simple drawing, simple face, simple coloring, mismatched pupils, artist name, text,

[lucid yoke]

bad hands, cojointed limbs, extra limbs,

[lucid yoke]

whats the resolution your trying to generate at and the model?

[red cobalt]

i was tryna make it 2048 2048 with img2img

since they told me not to generate as 2048

[lucid yoke]

ah

[red cobalt]

https://civitai.com/models/376130/nova-anime-xl

using this

[lucid yoke]

yeah you mean upscaling

[red cobalt]

and some random lora idr

do i even need the lora

to generate stuff

[lucid yoke]

depends on what the lora is for

want a specific style and a lora is for that? yes

[red cobalt]

idfk

i just got one

idr what even it was

[ornate elk]

Im on AMD and need different args

[lucid yoke]

https://www.youtube.com/watch?v=fjWcRU7Ue-0

[red cobalt]

is there one for creating loras

[lucid yoke]

its the best i can tell you with forge since personally im a swarm user

[red cobalt]

also

how to make it so it doesnt remove all my tags

auto saves?

prompts*

lmao im just trying to 2x a image as many times as i can rn

just to see how it changes and whats the end result

[ornate elk]

Don't use img2img to upscale them.
Better use the hires fix option in txt2img

[red cobalt]

bruh

[ornate elk]

Set it to 10 hires steps and denois 0.5 to test.
Upscaler to Resrgan4x for example

[red cobalt]

i can see why

it fucking crashed

[lucid yoke]

theres many opinions and eh Cs1o is the AMD expert

[red cobalt]

my chrome

[lucid yoke]

out of memory issues

[red cobalt]

hm

[lucid yoke]

though ill have you know, on nvidia i managed to get to a 9000x10000 pixel image

took ages though

[red cobalt]

OK

HOW TO SAVE PROMPTS

ITS FUCKING ANNOYING

TO WRITE EVERYTHING AGAIN

[lucid yoke]

hmm cant help you with that one in forge sorry

[red cobalt]

@ornate elk do u know

[red cobalt]

tried that

better ig

[red cobalt]

it still creates text even with that tho

[lucid yoke]

is that a screenshot? doesnt feel like 1024x1024

also its text for like on the image on its own

like

bottom right etc

[ornate elk]

They are saved in the image meta data.
You can drag an image into the PNG info tab to get it back.
If you want to save them use the styles option below the generate button.

[red cobalt]

yes its ss so it loads faster

this?

[ornate elk]

The button at the right of the empty dropdown

[red cobalt]

oh i see

edit styles

does simple background neg prompt remove background completely

last 2 images had like no background

[ornate elk]

In the negative you put stuff you dont want to see

So simple background in neg gives your more complex background

[red cobalt]

can u teach me how to train or make loras

[upbeat moth]

Do you need certain API key permissions to be able to use the image-to-video endpoint? I got an 400 error

https://tenor.com/view/sussa-gif-5432297967573209335

[ornate elk]

the api of stability ai is paid only yes

[ornate elk]

nope sry

[limber walrus]

Hello, did anyone ever had this error before ?

while generating

[ornate elk]

in img2img or txt2img?=

[limber walrus]

txt2img, but it usualy happen after HiRes fix

so it might be related to imgtoimg since hires uses this to upscale?

[upbeat moth]

Could you specify what you mean with its “pais” only ?

[red cobalt]

@ornate elk is it possible to talk with ai

[ornate elk]

paid

[ornate elk]

can you show your hires settings?

also whats your gpu and which image gen model ?

[limber walrus]

I don't think it's linked to my perf, because sometimes it works, like 2/3 generations it stops with that error but 1/3 renders properly

that's why i'm lost x)

[red cobalt]

SO BRIGHT 🧛

[red cobalt]

cab

canu delete that pic

it hurts my eyes

[ornate elk]

still whats your gpu? maybe your missing some performance launch args

[limber walrus]

Surely, my gpu is a RTX 2070

[ornate elk]

also 2 hires steps is to less

10 is the best value

[limber walrus]

Oh, didn't know

[ornate elk]

then you should have --xformers --medvram-sdxl in your webui-user.bat

[limber walrus]

I had xformers but not mdvram, might be this

thank you a lot

[upbeat moth]

Yes I was counting with that, I was wondering if additional requirements were necessary for me to be able to use the endpoint since I kept getting 400 error, but ended just up being my code and I got past it now and it works. Ty though

[ornate elk]

oh alright ^^ np

[hot briar]

@ornate elk I followed your guide to setup Comfyui with zluda, since I did this step "Rename the copies to cublas64_11.dll and cusparse64_11.dll and nvrtc64_112_0.dll. Copy these three files into ComfyUI\venv\Lib\site-packages\torch\lib and overwrite if asked." when i try to launch Start-ComfyUI.bat it closes by itself

[ornate elk]

ah okay, that means something isnt correct

did comfyui opened before copying the files? i guess no

means that something is wrong with the path entries in the environment path settings

[hot briar]

yes it opened but was giving error like cublas

when i clicked run

[ornate elk]

ah okay thats normal, and it shouldnt close after you copied over the dlls

can you show your environment path settings?

[hot briar]

that one?

automatic1111 was working with same things

[ornate elk]

is it in the system environment path or user environment?

also can you show the folder content of C:\zluda?

[ornate elk]

looks okay

can you open up a cmd inside the comfyui folder and there launch the bat in?

[hot briar]

like that?

[ornate elk]

your in cmd but in powershell

you need to launch it in cmd only

[hot briar]

like this now

[ornate elk]

the files are still missing

the dlls

can you copy them over again

[hot briar]

i did but same error

[hot briar]

it's in

[ornate elk]

hmm strange

can you delete the venv folder again then relaunch

after that, close the comfyui in browser and cmd and copy the files again

[hot briar]

okay

[hot briar]

same, copied files and again it is closing itself

[ornate elk]

hm okay can you show me the error again?

[ornate elk]

and your not running it as admin hopefully

[hot briar]

yeah should i?

[ornate elk]

no

never run any webui as admin

that can cause issues

[hot briar]

oh got it

[ornate elk]

is the folder name ZLUDA and not zluda?

[ornate elk]

oh your missing the original files in that folder

it should look like this:

[hot briar]

yes..

[hot briar]

ah i directly changed name of them

[ornate elk]

what was your gpu ?

[hot briar]

rx 7800 xt

C:\ZLUDA\zluda

[ornate elk]

thats not good

in your environment path its different

C:\ZLUDA

[hot briar]

C:\ZLUDA

okay then it's good

should i do something

like delete venv

[ornate elk]

yes delete the venv and then relaunch and show the the cmd log

before pasting the dlls

[hot briar]

okay

@ornate elk

[ornate elk]

looks good

now replace the 3 dll files

let me know if it asks to overwrite. if it doesnt ask something is wrong

[hot briar]

copied this three

[hot briar]

it asked

C:\Users\niceb\a\SD-Zluda\ComfyUI\venv\Lib\site-packages\torch\lib pasting here

[ornate elk]

can you redownload the zluda files, maybe you have the wrong version.
Here is the direct link:
https://github.com/lshqqytiger/ZLUDA/releases/download/rel.dba64c0966df2c71e82255e942c96e2e1cea3a2d/ZLUDA-windows-rocm6-amd64.zip

delete everything in C:\ZLUDA

and then put the files in there, make the 3 copys and rename and replace in comfyui again

[hot briar]

C:\ZLUDA

[median olive]

I have expertise in dev of AI generator
So if you are looking for a AI developer, feel free to reach out

[hot briar]

same @ornate elk

[ornate elk]

delet the zluda folder and the zip file from there

[ornate elk]

thats very strange and im running out of ideas

[hot briar]

yeah it is, it worked with 1111 very well

[ornate elk]

yea but comfyui is harder to setup

some strange stuff on the system can easily break it

mostly its related to path settings

[hot briar]

hmm

i dont know if it works with doing everything from 0

will try later

[ornate elk]

okay let me know

also run
pip cache purge
in a cmd before

[keen sequoia]

@ornate elk Hey, am the guy from yesterday, with the openpose nonfunctioning 😅

You was saying that you have controlnet union model working, and maybe we can test and compere them...

If you still dont mind giving it a try ofc... and if that doesnt help, then am gonna gave up on that AI thing for now

[ornate elk]

hey yea we can test xD

you used auto1111 no forge right?

[pliant plover]

Can anyone tell me what's the best choice these days to make a lora? Any extension? I never made one. I'd like to create OCs from scratch, starting with a single image and then building on that.

[keen sequoia]

auto1111 yeah

[ornate elk]

okay, and you got the controlnet union from here?
https://civitai.com/models/136070?modelVersionId=655836

or here:
https://huggingface.co/xinsir/controlnet-union-sdxl-1.0/tree/main

both are ok

[keen sequoia]

Nope i wasnt, but downloading now

oh wait

[keen sequoia]

I do have this one from here tho, the diffusion_pytorch_model_promax.safetensors

[keen sequoia]

Tried that one, still not working

[ornate elk]

try these settings:
here is the pose

worked for me with the model
Nova Anime XL

[keen sequoia]

omg its working no way

alright so wtf i was doing wrong then, is it the width/Height setting... or?

My bones img are looking different btw

[ornate elk]

hmm idk what yours looked like

[keen sequoia]

heres my bones img that i got

and yours

looks different maybe they the problem 🤔

[ornate elk]

tested it , works too

[keen sequoia]

not working for me

[keen sequoia]

wtf

[ornate elk]

i only switched the controlnet image to yours

[keen sequoia]

okay wtf is this then xDDD

[ornate elk]

very strange

try input my image as source and set the preprocessor to openpose_full (tested openpose face and it didnt worked)

[keen sequoia]

yeah

same

only works with preprocessor none

[ornate elk]

what?

can you screenshot ?

[keen sequoia]

can i just show you my screen? xD

[ornate elk]

sure

[red cobalt]

how to add pose

[ornate elk]

you need the sd-webui-controlnet extension and the models for it

[tame sun]

I'm getting 1min gens at 832x1216 res on an AMD 9070 (non-xt), is that about right?

@ornate elk

[azure magnet]

Someone can help me with this?

[azure magnet]

Is this the last version of ComfyUI version: 0.3.26?

[empty cliff]

use the portable version G 😮

the installable version always gives me problems

[red cobalt]

can u add objects too to the pose

to make it better

[red cobalt]

guys why does 4096x4096 take so fucking long

[onyx solar]

Well, to give you part of the reason...
1024x1024 is a mere 1,048,576 pixels
4096x4096 is a huge 16,777,216 pixels

That's 16x.

[red cobalt]

it takes longer than 16 1024 s

[onyx solar]

Because you likely begin swapping VRAM.

It's not just compute.

At some point you have to offload from VRAM into system RAM back and forth, and that's a slow operation.

[red cobalt]

it took like an hour rn

and its at 97%

100% on cmd thinh

97 on chrome

[onyx solar]

And?

Sounds like you should lower your expectations.

[red cobalt]

is it usual to take an hour

im asking

[onyx solar]

That depends on a lot of variables.

[red cobalt]

I THINK IT JUST FUCKING CRASHED
AFTER AN HOUR

yeah rip

[tame sun]

https://tenor.com/view/cat-cat-turning-head-confused-cat-rizalalthur-orange-cat-behavior-gif-10491959385063137392

[pseudo forge]

hello

Guys, i have an AMD 6750 XT 12GB vram. After some troubleshooting i managed to get SD working. But now I have been having problems with inpainting. When i do inpainting it just looks noisy and oversaturated in the inpainted area, plus nothing changes even with high denosing strength . Do you think this problem is because of my AMD GPU?

[tame sun]

Do you have an example you can post here?

[pseudo forge]

yeah, here i inpainted the eyes, you can see how weird it looks

These are my settings. I have no idea what im doing tbh. Just learned the basics from YT tutorials.

[pseudo forge]

Ok, something is definitely wrong with my img2img tab. These two images are the result of me trying to make a normal building look futuristic and cyberpunk. The one on the right looks alright, not great but the generation at least was working properly. The image one left is me trying to to the exact same thing but now it looks horrible. I have no idea what is causing this. It was working perfectly yesterday and today its all screwed up. Text2img still works fine though.

[fair oxide]

i am giving a try to comfyui, should i directly run main.py or should i create a venv for that, i cant see the piece of code that suggestes it will automatically create a venv

[ornate elk]

In Auto1111 or Forge?

[ornate elk]

Hey this issue is because you installed the webui with the directml Backend and not zluda.
Best is to follow the AMD Zluda Guides from the first link of the pinned messages.

Its much faster and inpaint will work

[ornate elk]

What's your GPU again?

[fair oxide]

4050

[ornate elk]

Then I wouldn't use comfyui. As it doesnt give you any benefits over Forge In terms of performance.
But if you want to try comfyui go for the zip it provides.
It doesnt create a venv but stores its stuff in the site packages folder

[wraith orbit]

Anyone here on forge and using the SD Hub extension?

[tame sun]

Forge

[ornate elk]

Do you have --cuda-stream --attention-quad in your webui-user.bat?

[tame sun]

I do not.

[ornate elk]

Then add these after --use-zluda

And test again

[tame sun]

Sure thing, I'll let you know how it goes in a bit.

[ornate elk]

Also make sure that your not running Wallpaper engine or games in the background

[hot briar]

@ornate elk automatic1111 idk if its new, automatic1111 gives same error but it's opening after error

[tame sun]

Seems to be a little faster as it was clocking in little over a minute prior.

[ornate elk]

Try use Euler a or DPM++ 2m for illustrious

Both are a bit faster than SDE

[ornate elk]

Okay so everything works ?

[tame sun]

Euler A and DPM++ 2M

[ornate elk]

Okay

[tame sun]

Also, for whatever reason, it's not saving defaults when I try to save settings for resolution and sampling method/schedule type.

It saves prompts but not those.

Minor annoyance.

On an AMD card?

[ornate elk]

You can't save those. But all Settings are saved in the image metadata.
Drop a image into the prompt field and press the blue button to auto apply them

[tame sun]

Oh yeah.

[ornate elk]

Nope they wont work for AMD
Xformers is only for nvidia.
SDP is not supported by zluda rn

[tame sun]

I imagine it'll get a lot better as they flesh out ZLUDA and the new series of AMD cards.

Supposedly the 9070s are great for LLMs but I haven't tried it much myself.

[ornate elk]

You need to wait until Someone releases optimised gfx files for hip and also the new hip SDK which supports rx9070 by default

[tame sun]

Yeah, I feel it's held together by rubber-bands at the moment with compatibility.

How long would it take for a 7800 to run image gen via stable diffusion at similar parameters?

[ornate elk]

Idk

[tame sun]

Fair enough. Thank you for the help.

[ornate elk]

NP, also get the tiled VAE extension for faster upscaling.
And the booru tag autocompletion extension for better prompting with illustrous

[ornate elk]

On my 7900xtx it takes around 15 seconds

I think the rx 9070 will perform nearly the same when the new hip SDK with native support for it drops

[tame sun]

That's reassuring.

[pseudo forge]

Thanks, I will try installing this Zluda to see if it works. Reporting back when finished.

I was about to buy an Nvidia card just to avoid these issues, but maybe it is not necessary after all

[ornate elk]

Let me know if you need help.
What's your GPU BTW?

[hot briar]

don't know yet i deleted comfyui

[gusty rivet]

I'm a Stable Audio 2.0 Pro subscriber. I'd like to use the Python requests module to make API calls against the Stable Audio 2.0 models. When I use the access token from the developer section of my pro account, it appears my alloted number of 3 minute audio files (250) is not represented in the credits portion of the billing. Is there a way to use the Pro account from a Python script? A link to working code would be great.

[pseudo forge]

Yo, it worked. Not only inpaint works correctly but generates images much much faster. Thanks a lot man! I have a 6750 XT with 12GB of vram. Also the version of a1111 i was using before was constantly telling me there was not enough ram and with this new one everything is flawless. Thanks again.

I am an old school illustrator and these stuff is so new and confusing to me. But am on the way to improve.

[ornate elk]

Perfect, no problem 🙂

[hot briar]

@ornate elk everything looks fine, idk what changed. Thanks for all, something worked

[hot briar]

fooocus looks wrong..

nop

[ornate elk]

When do you get that error?

[languid steppe]

Hi, brand new to using SD and my first text to image attempts are giving this error "NansException: A tensor with 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. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check."

I checked the upcast cross attention layer to float32 option already and nothing changed, I don't know where to put these commandline arguments

[ornate elk]

Hey, what's your gpu?
And how large is the checkpoint you try to use?

Also set the float setting back

[languid steppe]

Oh I didn't have a checkpoint selected maybe that was the issue

I have an nvidia gtx 1650

i jumped into this without learning anything beforehand so all the settings are greek to me tbh

[ornate elk]

Ah okay that GPU needs
--xformers --medvram --no-half
In the webui-user.bat at the line commandline_args=

You can edit the webui-user.bat by right click edit with notepad

[languid steppe]

thank you, okay attempting again!

hmmm gave me the same error

is there something I should have downloaded besides the SD program itself?

[ornate elk]

You need a checkpoint

In the models/stable-diffusion folder

It should be 2gb - 6gb

Not smaller than 2gb

[languid steppe]

there's a file called v1-5-pruned-emaonly.safetensors

in that folder, 4 gb

[ornate elk]

Thats the base model. Try select that in the webui

[languid steppe]

I did yeah

it gave me that error again qq

I've tried freeing up memory by closing other stuff but no dice

[ornate elk]

Can you delete the file and try downloading an other model (checkpoint)

For example this:
https://civitai.com/models/4384/dreamshaper

[languid steppe]

downloading this but it's taking a few

[languid steppe]

got the same error, this is under it to the right if it means anything "A: 2.83 GB, R: 2.91 GB, Sys: 4.0/4 GB (100.0%)"

[steep palm]

C:\AUTO1111\sd.webui\system\python\lib\site-packages\timm\models\layers_init_.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers
warnings.warn(f"Importing from {name} is deprecated, please import via timm.layers", FutureWarning)
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.

Model runs well but what is happening 😭

[mystic basin]

can anyone explain me why i get these result when trying to generate an image?

[wicked nest]

I don't have the answer but depending on checkpoints (in my experience) used some do this. not sure if I'm missing something or what... but others work fine. If you find or get the answer could you let me know.

[mystic basin]

sure i can, if i find hte answer

i'm not missing anything though, maybe it's the vae causing the issue, i can't find the vae that the creator of the model used for all his image

so i used another one instead

[mystic basin]

or it could be caused by this

but i'm not sure, i copied the entire workflow of the person who made the model, so i don't know why this would happen if it was something else than the vae because the vae is the only thing i changed if we don't count the amount of steps and image size

[knotty turret]

anyone know how to fix it?

what is that?

idk man never did this before prefer wait for CS1o

nvm fix it

[mystic basin]

nvm it's the clip set last layer, it was set to -1, i removed it and it work but the issue is that being able to use that thing without having weird issue could have been nice to get certain specific image

some my issue is kind of fixed

[clear pawn]

Does anyone here have experience with Lora Training in Flux Gym? Ive been working at it and keep running into an issue were it says my training complete after like a minute. No sample images or anything.

[red cobalt]

guys

im worried

txt2img is weird

[lucid yoke]

Those are wild images but yeah thats anime finetune stable-diffusion

Though id delete em since its quite nsfw

[red cobalt]

wouldnt rlly call it that nsfw but okay

[lucid yoke]

I don't mind but the panties shot gets cleaned by staff regularly

[red cobalt]

would it turn people into anime if i wrote like jennifer lawrence

[lucid yoke]

It could depends if the model was trained for it

Illustrious? Probably not

[red cobalt]

so far i typed a whole lot of stuff and it just did them

[lucid yoke]

Sdxl anime? With a lora sure

[red cobalt]

novaanimexl or smth rn

im using

didnt rly try out other stuff bc it takes a good 5 mins or 10 to download

[red cobalt]

do u know a good model i can use

checkpoint?

[lucid yoke]

Ah thats a good one

I prefer wai nsfw for illustrious since it knows a lot of characters

[red cobalt]

yea its not so bad except uh

chins look kinda weird ig

the shape i mean

@lucid yoke do u know a good lora i can try

like outfit or smth

[lucid yoke]

You need adetailer

[red cobalt]

i have it.

[lucid yoke]

Hmm then probably work on your prompt a lil but theres hunderds

Thousands even

[red cobalt]

loras?

i tried one but it didnt work

[lucid yoke]

Yeah

[red cobalt]

oh also

[red cobalt]

how do u do like 2 characters in one uh

like

different prompts

[lucid yoke]

Hmm either lets say kafka and ganyu, hugging

And luck

[red cobalt]

oh

bruh

so u cant specify different ones

like

black hair and white hair for one

can i send u the one i tried

and didnt work?

@lucid yoke https://civitai.com/models/99631/lora-yuzuriha-inori-cos-guilty-crown-clothing-with-multires-noise-version

this one didnt work

when i tried it

[lucid yoke]

It should work

Did you use it's trigger word?
outfit-inori

[red cobalt]

yes

[lucid yoke]

Did you enable the lora so its in your prompt box?

[red cobalt]

either i did or it didnt show up on loras

does it matter if i rename the file i downloaded it again

to try again

[lucid yoke]

Did you put it in the loras folder

[red cobalt]

yes

i did it again but i cant see it there

[lucid yoke]

Hmm refresh the ui

[red cobalt]

how

i literally closed and re launched

[lucid yoke]

Oh that should work

[red cobalt]

but it doesnt.

[lucid yoke]

Hmm I'm unfamiliar with the ui your using imo but those the the logical steps

[red cobalt]

automatic1111

[lucid yoke]

Unless you used a lora for a different version of model

[red cobalt]

hm idk

eh fuck loras

i hate loras now

[hot briar]

a couple minutes after click to generate

[red cobalt]

i never had an error while generating

what do u use

[hot briar]

fooocus

[red cobalt]

idk that

i use automatic1111 thing

idk what to call it

dont blame me

webui?

[ornate elk]

Can you share the cmd log?

[ornate elk]

Its for 1.5 based models

It won't work for larger models like sdxl/illust

[ornate elk]

Did you disabled Fooocus v2 style in the advanced tab before pressing Generate?

[hot briar]

just entered prompt and clicked generate

[ornate elk]

Yea that won't work

Also restart fooocus before trying that.

[hot briar]

thanks didn't see that

[red cobalt]

what are the system reqs for automatic1111

someones asking

@ornate elk ?

[ornate elk]

A Nvidia or AMD GPU with at least 4gb vram and 16gb of RAM or more

CPU doesnt matter

[red cobalt]

k

@simple bough

this channel to ask for help

[simple bough]

[red cobalt]

we haev 7 servers now :3

[red cobalt]

@pure tusk

@ornate elk i have a question r u here

[ornate elk]

Yes

[red cobalt]

im like experimenting and i found out that

i cant change what a pool is filled with

like

if i try to make it blood

its just adding blood

do u know how to fix

[ornate elk]

Hmm which model do you use?

[red cobalt]

novaanimexl

[ornate elk]

Okay, and what's the prompt?

[red cobalt]

like

i think

a pool filled with blood

and i add water to negative prompts

and it still would put water

@ornate elk what does cfg scale do

can that be why

[simple bough]

Interesting

[red cobalt]

wait

im going to try to swap negative and normal prompts

[simple bough]

Run webui-user.bat from Windows Explorer as normal, non-administrator, user.

also i dont underst and this step

[red cobalt]

cmd

copy paste the clone github thing

[simple bough]

webui-user.bat

this?

[red cobalt]

oh

u did that

wait

dont open the bottom one

open the one above that

[simple bough]

what u mean

opeN WHAT

[red cobalt]

ss

it

send

[simple bough]

send what bro

[red cobalt]

what r u trying to do rn

did u do the uh

git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git

did u do this

or

[simple bough]

Install Python 3.10.6 (Newer version of Python does not support torch), checking "Add Python to PATH".
Install git.
Download the stable-diffusion-webui repository, for example by running git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git.
Run webui-user.bat from Windows Explorer as normal, non-administrator, user.

[red cobalt]

u got git and python?

[simple bough]

Yes

[red cobalt]

u downloaded that and got git and python so

[simple bough]

i did the first 3 steps

[red cobalt]

run webui-user.bat

the one above the bottom one

[simple bough]

dosent work

WaiT

wym above the bottom

[red cobalt]

theres 2 rightr

[simple bough]

Where can i find that

[red cobalt]

can u screen share..

[simple bough]

No im a secret agent and i cant

just what do i open to get htat

files?

[red cobalt]

u downloadedhttps://github.com/AUTOMATIC1111/stable-diffusion-webui.git.

open it

its there

[simple bough]

idk i just did the instructions

[red cobalt]

..

okay delete https://github.com/AUTOMATIC1111/stable-diffusion-webui.git. this

if u installed it

and do the automatic installation

[simple bough]

Wait i tHink

This?

[red cobalt]

.bat

[simple bough]

yes i did

now

what

its installing smt

[red cobalt]

yes

wait.

[red cobalt]

its gonna look like shit at first

u need to download a model from civitai.com

[simple bough]

ah

[red cobalt]

https://civitai.com/models/376130/nova-anime-xl

i use this

[red cobalt]

@ornate elk wtf is going on

thats 3 people now

its spreading

@pure tusk

mind doing something?

@simple bough

howsit goin

[simple bough]

NICE

[red cobalt]

did u get model yet

[simple bough]

95%

[red cobalt]

hm

maybe u can help me with smth

it might be a bit weird but check dms

[simple bough]

yes

[simple bough]

Sm1 help

[red cobalt]

@ornate elk do u know

[lucid yoke]

did you close the cmd window?

[simple bough]

n-n-n-no

[lucid yoke]

hmmm

can you show me the cmd log ?

[simple bough]

oh now i closed it tho

[red cobalt]

lol

[simple bough]

OH MB

[lucid yoke]

what GPU do you have?

[simple bough]

3050

IT worked before

but i got a new model

[lucid yoke]

yeah hassasku right?

did you run SDXL before?>

[simple bough]

no

[lucid yoke]

hmm personaly im not too familliar with A1111 enough to say on how to get sdxl running

maybe Cs1o can once he comes back online but please dont ping him

[brazen sun]

hello there, I dont know if this would be the right channel to check this but I been trying amuse for amd and it goes much faster but it kinda sucks because of the restrictions on choice of model and no loras. I tried automatic1111 with the amd instalation but it takes too long, specially compared, any ideas?

or recomendations?

[daring reef]

Is it possible to use two gpus to speed up image generation in stable diffusion? A 3060 and a 5070, I usually use them with lossless scaling to play games using two gpus and I was wondering if it was possible with stable diffusion.

[ornate elk]

Nope you can only use both with their own instance to generate images simultaneously

[red cobalt]

cant he geenerate 2 halves and combine them?

[ornate elk]

Hey, you should checkout the Zluda guides for AMD.
They are on the first link of the pinned messages in this channel

[ornate elk]

Nope

[red cobalt]

isnt it possible

if u rly think about it

[ornate elk]

You should use Forge Webui with that GPU

[ornate elk]

Maybe with Comfyui and a big workflow

[red cobalt]

cant u do it with a1111 if u set the same seed btw

discord link fail

[brazen sun]

this one?

[red cobalt]

what

[ornate elk]

Yep this one

[red cobalt]

link didnt uh

link failed to link

thats why

i said that

[ornate elk]

Nvm xD I meant your other question

[red cobalt]

oh

well

does it hurt the gpu btw

[brazen sun]

tried it fully and it goes slow as hell for some reason, like not the first image only but anything

[red cobalt]

or other parts

ik it heats it a bit

like 55*

[ornate elk]

Then something isnt setup correctly

What's your GPU?

[brazen sun]

RX 6600

not the best one out there but it was running amuse well enough

[ornate elk]

Should work good with zluda too

Let me know if you try again and get any issues

[brazen sun]

right now I cleaned all up so I'll try again from the start in case I did something wrong

ill tell if still goes slow

[ornate elk]

its like playing games

[red cobalt]

no my pc doesnt freeze while playing games

near the end atleast

[ornate elk]

that means you should enable tiled vae (in auto1111) or never oom (in forge)

because the last step is the vae, the vram usage is heavy

[brazen sun]

redoing it I realized I made a mistake, it goes a lot faster now but it does get stuck a bit at 95% which I assume is due tipical loading things, although maybe am getting a bit greedy but is there any way to speed it up? if not I am more than happy with this speed but would like to know

[ornate elk]

you need to install the tiled vae extension if you want to use hires fix for upscale

it speeds it up at the end

also add --medvram-sdxl to the webui-user.bat

[brazen sun]

its on the extension tab right?

[ornate elk]

yep, click on available, then on load from

to get the list

[brazen sun]

once installed i have to restart and how do I use it? and sorry about asking too much but what do medram and sdxl do?

[ornate elk]

yep restart and then just enable the Tiled VAE checkbox. thats all

it will do its stuff when needed

and dont forget this:
also add --medvram-sdxl to the webui-user.bat

its needed for 8gb vram gpus

[cobalt coral]

is there even a github stable diffusion gradio project with tiling or a method to search, anyone know?

[ornate elk]

Stable-diffusion-webui can do tiling

Forge Webui too

Reforge too

All based on gradio

[hot briar]

Device cuda:0 does not support the torch.fft functions used in the FreeU node, switching to CPU. @ornate elk you know what it is?

[chrome rampart]

hello all. anybody pratical experience of running WAN 2.1 within comfy on 3060 card and let say processing time?

[brazen sun]

added it just, sorry about asking again, wht do those do? if there is a resource you can point me to I can also see it there

[brazen sun]

is this normal?

[radiant osprey]

Hi all, this morning I registered for stable diffusion but haven’t received any verification email yet.
I tried several times to get the email sent again but still haven’t received yet. I also contacted stable diffusion but they suggest to place my request in Discord community.
Anyone can help? thanks

[lucid yoke]

if its not in your spam inbox theres little we could do there. if its for #🗣｜artisan-support-feedback id recommend checking there

its mostly a community

[radiant osprey]

nothing in spam... I'll try artisan support. thanks

[mystic basin]

Does anyone know why clip set last layer to -1 cause this issue when I generate my image? I use two KSampler in my comfy workflow, I don’t understand why this happens. (You can copy the entire workflow by copying the image if anyone can help me)

[ornate elk]

It has a function that won't work with zluda

[ornate elk]

It manages the vram better for sdxl based models.
Which means faster generation speed

[ornate elk]

Only enable tiled VAE in txt2img is enough. No adjustment needed

[brazen sun]

ahh okay, thank you

[maiden frost]

Can somebody tell me what I'm doing wrong here? It very slightly added in some pixels to the model's but no other perceivable change.

[daring reef]

is there any other way to download the models? i can't download it using civitai

it does not loading

[red cobalt]

@ornate elk can a1111 create jennifer lawrence

[red cobalt]

do u have a vpn on

[daring reef]

no, do i need one?

[red cobalt]

idk

[daring reef]

wow, now is working

with a vpn

[ornate elk]

Yes with a lora

[cobalt coral]

is there a tutorial to use tiling on all of them?

[late tree]

It's been a while since I played around with SD, I had to reinstall SD Next because I somehow managed to prevent it from updating. It seems to be trying to use my CPU instead of my GPU and I forget how to change that. The target line in the bat file is set to:
"\SD_Next\webui.bat" --use-zluda --debug --autolaunch"
Which last time I tried to use it, worked with my GPU

[lucid yoke]

Spam

@late tree don't join these

[late tree]

Wasn't going to, considering this is the tech support channel for SD. Lol

Thanks though. ^.^

[sturdy cedar]

/settings

[faint fossil]

how can i fix this?

did i got scammed?

sorry but who are you?

[faint fossil]

are they trying to scam?

captcove

because he sent me url and i joined the server but i realized that i might get scammed

[lucid yoke]

Dont join

Leave that server its spam

[faint fossil]

yeah yeah thanks

but i clicked the url

[lucid yoke]

just joining isnt harmful but dont give permissions/give out deets

[ornate elk]

What's your GPU?

[faint fossil]

rtx 3050

[ornate elk]

Then make sure your GPU driver is updated.
Then delete the venv folder and relaunch the webui-user.bat

Also make sure to have --xformers --medvram in the webui-user.bat

[faint fossil]

okay let me try

@ornate elk

now i cant relaunch webui-user.bat

should i edit the path of python?

[ornate elk]

Reinstall Python and check "add python to path"

[faint fossil]

i have 2 different path of python

[ornate elk]

Open up a CMD and run
Where Python

[faint fossil]

310 is the one with pips

[ornate elk]

Uninstall python 3.13, its not compatible with ai stuff

[faint fossil]

oh alright

is it the reason of error?

[ornate elk]

No the reason is that you have to either reinstall Python 3.10.11 and check add python to PATH in the first screen.
Or you run the Python installer again and click Modify, Next, "add python to system variables"

Both should work

[faint fossil]

okay let me try second one

but i think i install 3.10.6 first

not 3.10.11

[ornate elk]

Should work too

Ist just a bit older

[faint fossil]

alright

is it because i deleted venv folder?

[ornate elk]

Move the sd.webui folder to the C drive.
Not in desktop, downloads or documents

Then run the update bat first

Then relaunch

And dont run it as administrator

[faint fossil]

it says "already up to date"

and i moved it to c drive

now should i check webui-user.bat again?

same thing

i dont run it as administrator too

[ornate elk]

Okay then you have to reinstall the python 3.10 too

[faint fossil]

alright

should i delete venv folder after i reinstall it again?

[ornate elk]

Make sure to install python with administration rights

[faint fossil]

alright

its done

same thing

but i have to check paths again

nothing wrong about paths

python path leads to pip files

[ornate elk]

Can you open up a cmd and run
Where python

That should show the path

Then copy that and edit the webui-user.bat file and paste it in there at
PYTHON=

[faint fossil]

should i change the current path with the path that cmd shows me?

[ornate elk]

Screenshot pls

[faint fossil]

alr

[ornate elk]

Ah okay its using the wrong python

From Microsoft store

And the python 3.10 is not in system path lol

[faint fossil]

what should i do

[ornate elk]

Download and install Python 3.10.11 64bit
And show me the first screen when Installing

[faint fossil]

alright

should i delete my python?

3.10.6

[ornate elk]

Yep uninstall

[faint fossil]

do i have to open installer as administrator?

should i click add python.exe to path?

[ornate elk]

Exactly and then Click install now

[faint fossil]

alr

its done

[ornate elk]

Alright open up a cmd again and run
Where python

[faint fossil]

now it says both

[ornate elk]

Looks good

Now relaunch the webui-user.bat

[faint fossil]

let me try

oh

now it works ig

installing something about 2.5 gb

[ornate elk]

Also have you added --xformers --medvram to the webui-user.bat?

[faint fossil]

how can i add?

just copy-past?

[ornate elk]

After the install is done, close the cmd.
Then right click and edit the webui-user.bat.
At the line commandline_args= add --xformers --medvram

Then save and relaunch

[faint fossil]

understood

@ornate elk oh, it worked

thanks for your help

[ornate elk]

Perfect, no problem 🙂

[strange horizon]

Hi! Most of the time my WAN video looks good, but sometimes the motion of the characters are hectic and accelerated somehow, what could cause it? I'm generating with 40steps/2CFG/uni_pc/simple, teacache, SLG 9,10, Shift 9, 121 length

[upbeat moth]

So I am trying my image-to-video endpoint on postman but the issue is the video url I am getting is over 1 800 000 characters, like it contains the entire video encoded as a one big data url and obviously postman is going to crash because it can’t handle all of that. I have rewritten my code to upload the storage on AWS and supabase and also added this to my environment variables “MAX_UPLOAD_SIZE= 1024 * 1024 * 50” to be able to handle more but so far, nothing is working. Even when I go out of postman to try it on browser directly it crashes too. Could someone please help me with this

[hot thorn]

Can someone explain to me how do I properly merge 2 checkpoints?
I've tried normal merger and supermerger but everytime i try to test them it just generates static.
both checkpoints are made by the same guy, one of them is based on pony, the other on illustrious.
My goal of combining them is to get compatibility of pony loras and get the style and detail from the illustrious one, any help?

[upper linden]

I recently installed comfyui and i get errors trying to generate videos from image, not sure if i can ask the question here

[lucid yoke]

What gpu do you have?

Also what workflow do you have? Though the problem is in the last line

[upper linden]

i have a 7800xt

[lucid yoke]

Hmmm and what video model?

Ltx?

[upper linden]

how do i copy workflow to show here

[lucid yoke]

Hmm well i think i do know the problem

But what model are you trying to run? Wan? Hunyuan? Ltx?

Just to be sure

[upper linden]

https://civitai.com/models/1329096/wan-video?modelVersionId=1501125 this is the model im trying to run

[lucid yoke]

Wanvideo? Text isnt very visible but I'm gonna be honest chief I'm not sure you can run it with your card

[upper linden]

damn

[lucid yoke]

Maybe a zluda installation could help but cs1o might know more

[lucid yoke]

But with my 5080 running wan consumes 15.5gb vram

Amd needs to use directML or zluda to emulate cuda iirc and that uses some more

[upper linden]

i am running direct ml, itshow i launchh it

F:\ComfyUI_windows_portable_nvidia\ComfyUI_windows_portable\ComfyUI>python main.py --directml
Checkpoint files will always be loaded safely.
Using directml with device:
Total VRAM 1024 MB, total RAM 31899 MB
pytorch version: 2.4.1+cpu
Set vram state to: NORMAL_VRAM
Device: privateuseone
Using sub quadratic optimization for attention, if you have memory or speed issues try using: --use-split-cross-attention
ComfyUI version: 0.3.27
ComfyUI frontend version: 1.14.6
[Prompt Server] web root: C:\Users\crazy\AppData\Local\Programs\Python\Python310\lib\site-packages\comfyui_frontend_package\static

Import times for custom nodes:
0.0 seconds: F:\ComfyUI_windows_portable_nvidia\ComfyUI_windows_portable\ComfyUI\custom_nodes\websocket_image_save.py

Starting server

To see the GUI go to: http://127.0.0.1:8188

[lucid yoke]

Vram= 1gb

You launced the cpu version?

[upper linden]

i was trying to launch the gpu

idk how to though followed a guide and said to type python main.py --directml

is there a way to edit to the gpu and more vram? i have 16gb

[lucid yoke]

Hmm well your best bet is to wait for a amd user or join the comfyUI discord

Iirc its on the github

[upper linden]

okay thanks

[hot thorn]

does anyone know if its even possible to successfully merge pony and illustrious checkpoints? everytime i try its just static

[upper linden]

scam

dont trust him

[upbeat moth]

Is it really?

I was about to send my issue there smfh

[upper linden]

yes

why send a link to a completely diffrent discord where this channel is litterally for help

might aswell close the channel lmao

but its open for a reason

[upbeat moth]

People need to get out and get a life tbh , I’m already annoyed rn like what’s the point? @gritty talon , can the admins do something about that please

[upbeat moth]

Idk 🤷🏽‍♀️ all I know is that I need help with my video URL and this is the tech support channel

[upper linden]

exactly

so ask here

[upbeat moth]

I didddd

No answer yet

[upper linden]

and someone will help eventually

[upbeat moth]

I’m waiting patiently

[upper linden]

me too

[upbeat moth]

twins

[upper linden]

[ornate elk]

you need to install the AMD compatible Version of Comfyui then:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides

[hot thorn]

@ornate elk is it possible to successfully merge Pony and illustrious checkpoints? all results i got were just noise

[ornate elk]

i dont think so

but most pony loras work on illustrious too

[hot thorn]

ye, i switched to illustrious and about 70% of them dont work

• the illustrious model has what i desperately wanted, good quality eyes

since the pony model I use struggles with eyes a lot

[ornate elk]

adetailer can fix the eyes too

[hot thorn]

ye, thing is, i just leave SD to run for the night, and make 100 pics, not gonna edit every single pic

[ornate elk]

adetailer is fully automatic, or do you use comfyui?

you can also regen a image with adetailer enabled at the next day, for the few you like

[hot thorn]

i can enable adetailer for my generations? so its enabled as I generate images at night? how so

[ornate elk]

which webui do you use?

[hot thorn]

ReForge

[ornate elk]

okay, do you have adetailer installed as extension ?

[hot thorn]

no, but i will now

does it extend generation time by a lot?

[ornate elk]

when you installed it you can just enable it. it will then facefix every image you make:

[ornate elk]

i would say by 10 seconds maybe

can test real quick

[hot thorn]

im usually doing 1080x1080 with 1,5 upscale, is it really 10 sec with that big of an image?

i got it installed what now?

[ornate elk]

yes because it only needs to detect the faces and upscale them, it doesnt matter how large the initial image is

[hot thorn]

or do i need to restart the entire SD?

[ornate elk]

after installing extensions yes

also get yourself the booru tag autocomplete extension if you work with illust models

helps a lot

[hot thorn]

i got it already

[hot thorn]

figured, i opened a new tab and while its in the installed list, i didnt get any new tabs or options

[ornate elk]

so tested it, adetailer adds 8 seconds on top

[hot thorn]

i assume that is it?

[ornate elk]

yep it downloads the models

[hot thorn]

k what now

[ornate elk]

you can define if it should only fix the face or only the eyes, lips, etc

[hot thorn]

so in 1/2/3/4 i choose what to fix im guessing, what about the other options on the bottom?

[ornate elk]

you can leave everything on default.
if you want specific extras like a forced eye color, glasses or an face expression then add that to the prompt field of it

if its empty it will use the same stuff as the prompt

[hot thorn]

what do all these do? I can guess the face and hand, but others?

[ornate elk]

the rest are settings you only need when you want to adjust stuff

person will replace the whole body, its not that great

idk what world does xD