#💬｜general-chat

1 messages · Page 81 of 1

rich kestrel
#

once u learn how to use kohya I can dm you the settings file I use

silent depot
#

hiii guy

#

i am no

small field
broken bloom
cinder ice
high swan
#

Hey everyone!

#

Good morning!

#

Anyone has experimented DeForum here?

crisp dagger
#

so i switched to NVDA GPU, fuck you AMD

zealous bluff
#

Hey yall, i was just wondering if there is a thing similar to stable diffusion that allows you to train a voice locally with audio files. Is there something like that?

regal raven
#

what is interesting is I can see the stable audio october event channel but not the stable audio channel

#

and I cant add an image here to show you

#

but it's supposed to be right above general chat correct?

#

when I click on the link above that @sudden ruin sent, itll show up for a sec then disappear agan

crisp dagger
zealous bluff
#

if that makes sense

crisp dagger
zealous bluff
#

fair, was just asking around if others maybe have

#

cause some german streamers have managed to get it to work with some colleagues

crisp dagger
solar condor
#

hey guys is it possible to train an AI on mac ˆˆ

#

??

broken bloom
#

I need image generation reviews for my fine tuned sdxl model with Loras
(running on nvidia A100) estimated time ~~ 15 seconds

https://ionium.web.app/sdxl

Can someone help, you just need to create images and give reviews everything is running on my own resource

teal pagoda
#

Do you know how some people can make videos without flickering and with a very good consistency?
Because the tutorials from youtube are not good
I mean there are some gifs/videos on civitai which are high quality, but I think that there's no way to use the reverse engineering to find out what settings were used for them.

wheat tulip
#

is there any way to put a text on an image?

sudden ruin
hushed ember
#

hello

#

is this the right place to ask questions about deforum ?

crisp dagger
#

Dudes. one can think that there woild be milon of clothong models on civitai

light crystal
fast sage
#

Could someone point me in the direction of how the ⁠bot-1 is doing videos?

rich kestrel
#

look at that creator's profile, he makes a crapload of clothing loras

rich kestrel
weak kettle
#

guys is there a possibility to add text that is not deformed on sd, not sdxl?

#

like for example "sprite" text

#

its always deformed on normal sd, prompts doesnt matter

slate imp
#

You mean besides controlnet?

weak kettle
#

it can be controlnet, is it possible in controlnet? To generate text

slate imp
#

SDXL doesn't do text good either, maybe you'll get lucky 1/10 times. You need to add it after the fact or with a controlnet unless you want to fight RNG

weak kettle
#

like character holding a sign with some text that im giving in prompt

#

always its giving me deformed text

#

how can i do this on controlnet? I only did poses in controlnet

slate imp
#

TLDR depth and softeedge on low weights

#

White text on black background

#

Either that or add it in post

#

Method 2 is generate image without text
photobash text with above image
then use IMG to IMG or Inpaint the area with the text, low denois or with a controlnet

#

Okay now I have a question:
What is the best extension to 1111 for viewing your output images after a generation. The standard one sucks dooty booty for a lot of images as a gallery

lavish current
#

I haven't touched SD in what feels like ages ever since DALL-E 3

#

I wish we could have something with the power of DALL-E 3 but the freedom and openness of SD

#

.

#

It's a great tool but belongs in the hands of someone better than Microsoft

upbeat tide
#

Hi guys,.. I have SD installed (Automatic1111 WebUser),... it's on my Brave browser, runs fine. However, earlier tonight, I installed Opera browser,. (NOT as default browser),.. but,.. BUT,. now my SD won't load in Brave,. I closed (shut) Opera and it still won't load,.. any ideas what I can do other than remove Opera? Ok, I got it working,.. I restarted PC,.. launched the .bat file again and a pop-up asked which browser I wanted to open the URL with, I chose my Brave browser and hey presto, it loads.

hasty hornet
#

so...you can just google how to change default file type for x in win x

upbeat tide
#

Oh, in Settings,.. default file types?

hasty hornet
#

It probably set itself up automatically...apps do that sometimes after installation

hasty hornet
upbeat tide
#

It didn't when I installed it (Opera),, but after a restart and lauched the .bat file, the pop-up came up to ask which browser I wanted to open the URL in

hasty hornet
#

oh I see

upbeat tide
#

W11

hasty hornet
#

no idea about win11, just search how to set up apps for different file types in win11

#

it shoudn't be too hard

upbeat tide
#

Yeah, ok,.. for future reference. Thank you. 🙂

#

Thing is,.. I doubt windows would know I had SD installed,. since it's not an actual .exe file

#

So it may not know a SD URL needed to be opened. Which in turn may mean there'd be no way to set a default file type..

hasty hornet
upbeat tide
#

Stable diffusion isn't a file type,.. it's a specialist browser installation 'thing',..

hasty hornet
#

technically it opens link...it should just open your browser

upbeat tide
#

I just looked,.. and I can open a link type,.. HTTP,..

hasty hornet
upbeat tide
#

So,. I guess that's how I'd do it in future.

noble lily
#

the videos the bot is making. What kind of GPU do you reckon you need to run something like that locally?

lost relic
#

Is there like a site to use while still be able to have workflows and get consistency cause I tried using comfyui and my laptop is just way too slow for it.

#

Really don't wanna be forced to buy a prebuilt pc or build my own just for stable (I don't touch pc games at all)

noble star
#

Just get a crappy pc with a good video card

#

And remote desktop into it

#

You can run sdxl on a 1080 ti for like 350 incl pc

hasty hornet
#

Thaaat's gonna be slooow...

stuck valley
#

hey dudes idk if this is the wrong chat but i just had to completely reinstall my stable diffusion anyone able to quickly give me a refresher how to install cpkts?

#

or like how to install different things? idk im trying really hard to remember how this thing was used

#

i just cannot for the life of me generate anything anime

paper seal
#

Im no expert so i might not be of much help but doesnt the model change art style the most?

stuck valley
#

so do i just go to hugging face for different models?

paper seal
#

I use civitai personally. Gives me a preview of the model results on the page and makes downloading simpler than Huggingface

stuck valley
#

but do i have the right idea?

paper seal
#

ABout the models or the cpkts thing?

stuck valley
#

ye

#

just download then drop in models folder

paper seal
#

Yep

stuck valley
#

thanks

#

im trying to generate some art for a dnd game im going to be a part of and for some reason i thought to steal my character design from some anime game

paper seal
#

another fun thing to do is IMG2IMG generation

#

I feed it mediocre art and it spits it out as professional art.

stuck valley
#

before i get on a tangent can i also drop in .json files into models or do i have to do something different?

paper seal
#

Hm... im actually not sure i havent done that yet.
Sorry to be unhelpful

stuck valley
#

nah no stress im not upset

#

i have no idea either :p

#

last thing then you got any reccomendations for anime esque models then?

paper seal
#

Yes

#

I use one called Manmaru mix but be advised it's more of a studio ghibil style than say a demon slayer or tokyo ghoul type of "modern anime" style

stuck valley
#

im going to keep it real with you i have no idea what any of those are

#

i am the last person who knows what anime is

paper seal
#

Oh lol

#

Studio ghibli has a unique style that looks almost like a mashup between Pixar animations and Anime

#

very famous in the 90s

stuck valley
#

👍

#

ive seen one anime and it was wierd and was not a fan but my friends like it

paper seal
#

It's a great genre but it's also a very unrestricted genre

#

There are some I consider peak fiction and others that are so weird and odd that I would not reccomend any human ingest it

#

When you simply dont have cancel culture, censorship, or really any restriction on what can be done, sometimes weirdos are allowed to make stories.

stuck valley
#

i saw berserk i liked the big sword guy but it was messed up and i saw some nasty things i didnt much like

paper seal
#

Berserk is a very mature and dark story. I would not reccomend it for a beginner, or anyone who is offended by 18+ sort of topics.

stuck valley
#

ah nah i liked teh idea of being a cool hero guy with a heavy sword im not trying to buy your pity but ive struggled a bit in my life and seeing a guy who keeps on kicking even when bad stuff happens is cool

#

but ive also had a couple of friends who are really into the stuff reccomending the most obscene things and i dont really want to be associated with it

paper seal
#

Well it's not pity, everyone has what they can and can't handle. I like dark stories but there are a specific couple tropes I can't withstand no matter how good the story is.

stuck valley
#

kudo to you if you like the stuff just aint for me

#

he wanted me to watch some sort of healer type dealio and i almost threw up

#

it was the most vile

paper seal
#

stories that put children at risk just to make you stressed, I can't stomach it. I dont care if it's a horror or action or whatever, its just sad.

#

Oh goodness, redo of a healer? Your friend hates you.

#

that is one of the only anime I rated 1 and wish I could rate -10 but the scale only went from 1-10

stuck valley
#

i dont like seeing evil people be evil im sort of an odd type of fella

paper seal
#

that show is the most degenerate, most disgusting, most vile thing I've ever seen on TV.

stuck valley
#

i like the over the top dramatic hero good guy type beat up the bad guy because i like it when the good guys win

#

that could just be my hick nature coming out tho

#

oh yea i did see one other one that i liked

#

it was like a boxing type one where the guy didnt have robot arms when everyone else had robot arms

paper seal
#

Megalo Box

#

One of my favorites

stuck valley
#

yea im a sucker for boxing and underdog fights

#

heck that and most mideval stuff

paper seal
#

I'm an avid writer for a trendy website called "Webtoon", I'm not sure if you've heard of it since you havent watched anime you likely haven't.

stuck valley
#

im faintly aware one of my coworkers like that thing but id prefer actual books

#

what ya make?

paper seal
#

Its basically the equivalent of Youtube but for comic creators instead of videos. A bunch of people make their stories avalible for free, in exchange if their popular enough they get an instant route to fame and money

#

I've made a couple quite popular comics. But just the storylines

#

Thats actually why I'm here, I'm using AI to give me examples of how to improve my own art so I can release a comic with 100% my own art for once

#

My self-rule is basically everything I publish must be done by my own hand, and AI is only to inspire and help teach me

#

I dont wanna cheat and just publish ai generated stuff

stuck valley
#

neatio i just have odd dreams i could make into stories that or i do a little dumb thing where i collect them

paper seal
#

I've tried the dream route before... Unfortunately my dreams are really dark and morbid. To the point where I cant really make a storyline out of it cause its just too sad

stuck valley
#

this is extremely personal so dont you feel any need to awnser but is there a story that you want to share with the world that youve never told someone NOT because its a super private secret but because its almost been forgotten

stuck valley
#

heck i could give you some ideas for you to try and make

#

itd be nice to share some of my dreams

paper seal
#

Hey, if you listen to some of mine, I've got no problem listening to some of yours.

stuck valley
#

neat

paper seal
#

My current main project is funnily enough inspired by a Micheal Jackson song... or at least that's where it started, and once the ball got rolling it evovlved crazily and wouldnt stop rolling.

#

I was tryna think of a story that was inspiring but also made the reader feel empathy and thats when I was listening to Billie Jean and it frickin clicked

#

I thought to myself "Wow a legendary guy who is just tryna convince the world this aint his son... Whatabout the reverse side of that? How would the son feel, truly believing a legend is his dad, maybe even wth prooof, but the guy just says no?"

stuck valley
#

sounds interesting enough

#

i aint have enough talent to do the art part but i can write a heck of a story

paper seal
#

I feel ya there, for sure. My arts getting there but ive got a lot of work to do

stuck valley
#

seriously friend if you want to ever talk about writing ideas id love to share some

paper seal
#

Absolutely.

stuck valley
paper seal
#

Oh my.
That sounds difficult....

stuck valley
#

imean it can be but its also got its upsides

paper seal
#

Though I will tell you what, it's somehow not impossible. I did know a fellow with aphasia who wrote and drew amazing things. I have no idea how, that sounds really tough to me.

stuck valley
paper seal
#

My writing and drawing proccess is 100% just fantasizing in my head with images and emotion. So i have no idea how the buddy did it.

stuck valley
#

if you dont focus on HOW it looks you can spend more time focusing on the why or cramming more details into it

#

you know "flow state"

paper seal
#

Wow

That actually is the exact advice I got for improving my imagination-drawing

stuck valley
#

i can sort of turn it off and on if i get into a rhythm most things happen by instinct so i may not be able to draw faces but the posing or bat crap creative stuff works wonders

#

i do believe we should take it to dms to not gunk up this chat if thats alright with you?

paper seal
#

Yes

stuck valley
#

but yea for the rest of you how do i use a json in my stable diffusion

wheat glacier
#

Ah fk did I miss the backdrop event

bold dirge
#

question, could I photoshop an image then use SD to correct the light and the color?

potent spire
#

There are also auto settings for matching colors etc

#

Oh, light

#

Depends what you mean by that

bold dirge
#

I use gpt to translate, im not good at english. my question is " Can I use Photoshop to composite an image and then use Stable Diffusion to make the composite image appear less abrupt? "

potent spire
bold dirge
#

I choose a bg photo and a portrait, I want to make these two blend more smoother

potent spire
#

I mean you could load a photoshoped image into A1111 and play around with a certain plugin for it for that but Photoshop is much better and efficient for that

bold dirge
potent spire
#

Yeah, Photoshop is best for that, there is even auto color correction amongst all for this

bold dirge
#

Oh i see! thanks for your help!!

#

I would try it, thanks!

potent spire
#

You can try it with SD but its really not near efficient as PS and i work a lot with PS

bold dirge
errant yacht
#

I saw an article talking about models for fine tuning and 3d model generation. Are there dates set for those releases?

sharp briar
#

Hey can I use the beats that I create with the stable audio and make a song and upload it on streaming platforms?

potent spire
red skiff
severe gate
#

Hopefully, I can post this resource as it's currently free and I've put about 3 months into it.

Made this 2D Character Animator course that shows how to clean up AI-generated characters and then Animate them in After Effects.

Have a video that covers it in more depth here:
https://youtu.be/g1vBjfHYrGY

Udemy lets me do an initial free discount for 5 days to promote it. Figure it might be of interest to you guys.

The direct link with the coupon already applied is here: https://www.udemy.com/course/2d-character-animator-photoshop-after-effects-midjourney/?couponCode=A42E2BAB1066AA982B8C

There are a couple of pre-set stock assets that I've made in there too.

dusky sapphire
#

first time working with sd, wanted to try face swap/morph
wanted to use reactor, i tried installing it twice, even in console it shows as reactor running but i dont see it when i run stable diffusion though, can someone help

errant yacht
rancid helm
#

Does paper space allows uncensored usage?

#

can we use custom lora or models in dezgo?

fathom kindle
#

Anyone seen the Deliberate v2? It seems to have vanished

regal raven
karmic brook
fervent thunder
#

my image been generating for 18 minutesagony

forest trout
#

Sounds like you are maxing out your GPU.

#

Too big of a starting image?

fervent thunder
#

no idea it was working before alright probably just bugged out

upbeat tide
#

AttributeError: 'ControlNet' object has no attribute 'label_emb' - I have updated all my installed extensions,.. problem persists. 😦 I need to outpaint but the controlnet refuses to play Cricket. 😦 Image is a tasteful NSFW image but her foot is cropped out of the image. It's 1024 x 1024 but I think 1280 might be just enough (ideally 1800 x 1800 but hey, that's asking a lot). Anything I can do?

golden finch
#

Can someone help me on Lora´s in SD? I cant see them all, I know there are tricker words, but if you cant remember that, what do you do?

solar condor
#

hey guys ! how can i use the models from A1111 with comfy ui

#

i dont want to duplicate SDXL twice since its huge haha

warm junco
jovial wraith
#

don't ask me how, but i managed to get AnimateDiff running on my 3gb card. my pc is not happy

solar condor
#

thanks for your answer man !

warm junco
solar condor
#

do you work in that domain ??

solar condor
#

ahhh i see the folder

#

and lots of yaml path

solar condor
#

ohh do i have to create a YAML file

warm junco
#

So that the file will be a .yaml

solar condor
#

and im using a mac m2 pro and damn its slow af when i use comfy ui haha

warm junco
solar condor
#

i added you btw !

high swan
#

is civit broken ?

#

atm

#

?

fading cloud
#

Couldn't download any attachments for awhile but it works now

wintry matrix
#

hiya

#

am here because im hoping amd gpus may have gotten a bit of love since my last attempt at getting stable diffusion to work

#

but i doubt it...

#

i remember looking into it and basically getting "yeah you can do it but you need to do some absurd roundabout command prompt magic and you'll never have a webui for it so good luck having any kind of accessibility"

verbal halo
#

Are there any fixes for the errror: "RuntimeError: mat1 and mat2 shapes cannot be multiplied (154x2048 and 768x320)" -- When trying to use Adetailer! on SDXL?

lost relic
wintry matrix
#

i want a local installation though, so i can use my own data sets

#

my friend sent me a stable diffusion data set for my favorite video game character and i wanna use it but i have an amd graphics card...

fickle spear
#

hey quick question, how do i change like format and style, cus im trying to generate stuff and it randomly adds formats when i dont want it to

verbal halo
#

Or well.. let me ask about the root cause of my issue. How can I get perfect hands on SDXL?

verbal halo
shut owlBOT
#

"TakeRep" command returned an error: strconv.ParseInt: parsing "--": invalid syntax

golden finch
#

webui wont run, it has been running all day, then this happend, it loads Python once and it starts but nothing more, I had it running one more time after this, but now it is the same. Does any body else experience the same?

#

Could it be some extensions I just installed?

echo frigate
#

Hi guys, it's been a few months since I last used stable diffusion. Would y'all say SDXL has progressed to the point of replacing previous versions? When SDXL models first came out I wasn't too big of a fan of the quality, curious if it's improved a lot since then

upbeat tide
#

SD needs a dedicated help channel,.. I've asked 3 important questions in here and only got help once. Edit Also, I see so many questions from other users, all going unanswered. SD, this is how you lose people using your product.

trim magnet
#

its a free product so you wont get costumer support here unless its from their paid services,only support u will get here is ppl who volunteer to help other ppl but they are not here 24/7

bleak matrix
#

Good morning, everyone! How are we all today?

tired pasture
#

From where can I get set of styles ?

warm junco
iron pollen
#

the number of detailed prompts i lose because i forget to copy it before the "GPU's are on break" message yeets it into the abyss, never to be seen again agony

wintry matrix
iron pollen
#

whyyyyy does it keep randomly doing videos

#

am I misunderstanding how "negative prompt" works? I'm entering "torso, full body, arms, legs, hands, belly" into the negative prompt and it's still generating full bodies

bleak matrix
warm junco
#

Its usable

#

What's your GPU?

wintry matrix
#

rx 6700 xt

warm junco
wintry matrix
#

woohoo! i looked it up on my phone earlier and saw it had lora support?

warm junco
wintry matrix
#

finally

#

i must go set this up

#

its the ishqqytiger one?

warm junco
lost relic
#

Is there like another site to use so I dont have to use my own gpu while still using workspaces? The generations just take too long.

lilac mortar
#

hey all, is anyone using SD with objects like soda cans? ive been able to get controlnet and use it on people, wondering if i can do this with objects

naive nexus
#

PLEASE , I NEED a HELP . Does Anyone knows some PROMPTS that I can to use correctly for to create some images of the female characters " WITH STRECHED ARMS TO THE FORWARD " more or less " as if wish to be embraced " on Stable Diffusion ?

autumn onyx
north pelican
#

Hello, why do I get a video and not the image?

neat cypress
brave swallow
#

So what is a Lora? Like a mini model? Is this like the difference between niji expressive cute scenic, or more for specific stuff that models have trouble with

open python
proper kite
#

Hi all. Sorry if this is not the right place to ask, but how do I go about training a model with my face? I’d like to generate images of myself but I am unsure where to start. Is there a way for me to train a model easily? Any recommendations on how to train one with my face?

hasty hornet
proper kite
leaden mural
#

Guys, anyone specializing in models modification? I need to ask.

#

Also, where is the artist channel?

abstract hull
#

What

hybrid kraken
#

Hello here

bleak matrix
#

Good morning, everyone! How are we all this morning?

fiery wasp
#

hello everyone, is there any better suggestion for command line parameters for M1 mac than those I currently use "--skip-torch-cuda-test --upcast-sampling --no-half-vae --use-cpu interrogate --medvram --opt-sub-quad-attention"

#

sorry, i am talking about automatic1111

bleak matrix
fiery wasp
vestal dew
#

We've come to this. With the latest update of A1111, SD no longer generates images. It runs for a few minutes, and then it throws a tensor error.

fiery wasp
#

i had that error a few days ago (i am a total beginner) but I fixed that without a problem. i can't remember if it was on windows or macos. which one do you use?

#

and what was the exact error message? developer in me complains about the incomplete question, but a fellow user in me understands your pain

vestal dew
#

windows

fiery wasp
#

nvidia with 6GB or less VRAM and SDXL or 2.x model, I presume?

vestal dew
#

"NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. 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"

fiery wasp
#

just add --no-half in command line

vestal dew
#

Why would I have to do that now, but didn't have to do it before?

fiery wasp
dire birch
#

hey, with my friend we've been building Social Network for AI-creators. You can create art, music and text together in public chats. And share it on your Public profile.

And we've just launched Early Access here https://aiphoria.world

If someone want to test and be one of the first, just DM me or apply on site! good luck

vestal dew
#

--no-half

fiery wasp
#

you can add it here in that file here
set COMMANDLINE_ARGS=
and run webui-user.bat

vestal dew
#

I think this: "call webui.bat --no-half"

#

"set COMMANDLINE_ARGS= - no-half"?

fiery wasp
# 
@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--no-half

call webui.bat
vestal dew
#

yes

#

two dashes

fiery wasp
#

export COMMANDLINE_ARGS="--skip-torch-cuda-test --upcast-sampling --no-half-vae --use-cpu interrogate --medvram --opt-sub-quad-attention"

vestal dew
#

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--no half

fiery wasp
#

and just run webui-user.bat

fiery wasp
#

do not use those

vestal dew
#

OK, now it won't run at all

fiery wasp
#

error?

vestal dew
#

Hold on, I will check. This is when I did:

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--no half

fiery wasp
#

do the test manually from cmd

#

liek this

#

webui.bat --no-half

#

then check the errors and fix one by one

#

when you are satisfied add them all to webui-user.bat

vestal dew
#

launch.py: error: ambiguous option: --no could match --no-half, --no-half-vae, --no-progressbar-hiding, --nowebui, --no-gradio-queue, --no-hashing, --no-download-sd-model

fiery wasp
#

you missed the - in the middle

vestal dew
#

Yes I did

fiery wasp
#

--no-half

#

btw, please go to tech-support, i will need to go soon, and someone there will help you even better than i can

vestal dew
#

sure

#

it's still not working...at least taking a long long time to generate

fiery wasp
vestal dew
#

Sure, but I've been using this for months

fiery wasp
#

if you have nvidia with at least 6GB like 1660 for example add this --xformers

vestal dew
#

these issues began after the last update

fiery wasp
#

it will make it work faster

vestal dew
#

I know.

#

Howeverm 15 minute generation time started after the last update

fiery wasp
#

but you will probably need this one --medvram too

vestal dew
#

Im now running one for 5 minutes, still at 0%\

#

I have 12 GB Ram

#

4770TI

#

4070TI

fiery wasp
#

than forget about --medvram

vestal dew
#

right

#

tyhis issue began after the last update

fiery wasp
#

and be sure to add --xformers

#

webui.bat --xformers --no-half

#

run this

#

is it better?

vestal dew
#

I did

#

it's maxing vream use,. 12 GB, but making no progress

warm junco
vestal dew
#

I have to noiw

warm junco
#

its not needed

vestal dew
#

without it, it crashes with a cuda error

fiery wasp
#

thanx @warm junco

warm junco
# vestal dew I have to noiw

your card only needs --xformers --no-half-vae
the reason you get an error is you probably try to load a lora as model and that wont work

fiery wasp
warm junco
fiery wasp
#

i am using --no-half-vae on ma mac m1, but i had to use --no-half on my sons 1660 super

warm junco
fiery wasp
#

it needs almost the same time as m1 to generate SD 1.5 image 🤷🏻‍♂️

warm junco
#

yesterday i helped someone with an M2 chip
he generates an SDXL image in 1:30 like my old GTX1080

#

thats okay imo

#

for an chip that doesnt have dedicated vram

fiery wasp
#

please take a look at my command line parms in support channel, and tell me if you have any advice

vestal dew
high swan
#

well this didnt go as expected

empty star
#

which upscaler is best for photos?

leaden mural
#

@austere marsh may need to ask a quick question.

#

What happened to the artist channel?

austere marsh
#

if you react to the community archive access in #👥｜roles you should be able to see the archived channel if there was a particular post you were looking for

grim grail
#

Hey guys. I'm wondering what the current state of AI is in regards to consistent character generation. Is it at the moment possible to use a dataset to define a character look, and have SD generate all kinds of scenarios and environments with the character remaining the same?

bleak matrix
grim grail
bleak matrix
#

np!

grim grail
#

Okay, so I've tried getting a good grasp of the different approaches after looking into the resources etc. I'm just gonna ask this:
I want to create a workflow where I can block out 3D scenes in Blender and output a render, then run it through SD to add the desired style, and replace the rough character with a trained consistent character. Is there certain methods that would be obvious to use of this process?

lean heart
#

I've been looking for near months now trying teh various incarnations / frontends for SD/SDXL and i'm not really sold on the ones i've found. I started using the standard method of an installer script to install a1111 gradio interface which served me well for a time. But then I started poking around the various other interfaces like invokeai, SDG, and most recently stablematrix which in-theory sounded great, and it has lots of ease-of-use features which are attractive / convenient.. but what with the inference module within the GUI, the discrete launcher concept, with multiple shared models/checkpoints it all gets fairly unwieldy IMO. I've now begun running into issues involving directories not being synced and LORAs / checkpoints not loading within the various frontends. It all just seems like a huge mess to me. and I'm wondering what is the BEST, most straightforward, least-sprawling local option to use? A large part of the problem is having to go into the various envs / packages in order to modify extensions, or forcibly remove things that have caused the app to break and/or not launch a111 in particular. a headache. and another thing is that i was having many issues saving generated media locally. Like I could not drag from the image browser/gallery to my hard drive, could not select multiple images at a time and save as... etc etc. But this cannot be the final solution it touts itself as .

#

thanks

warm junco
silent tartan
#

Good android games like mc and geometry dash pls

fickle siren
#

Can we disable the video and stick to static images on the bot image generation?

#

It is clearly having an impact on performance

#

sussed it thanks for the prompt help lol

lean heart
#

So, auto1111 is probably where I should have stayed, then. What is the best way to install that locally? I have been very lazy using installer scripts. And they're mostly failsafe it seems. But it might be good to have something that's not just a floating ENV in a directory, right? Should I Start all over , wipe my miniforge lib, etc? And what is the best place to install would people say? Probably cloning the git to the users folder?

lean heart
#

curious if defocus is the only option for animating? I imagine not..

modest vector
#

does anyone know if stability plans to release a diffusion model beyond SDXL 1.0 soon?

spice plank
#

Why does sd keep trying to download the default checkpoint when I already have a bunch of .safetensors in the folder?

#

How to fix? Im on linux if that matters

fallow veldt
bleak ice
#

How to change model ?

topaz parcel
#

Can I ask a 'noob' question. Using Auto1111 web UI, Is there a way to go back to previous generations and prompt settings from a previous images generated in a single session, or is it once you make changes, and generate a new image, anything gone before is gone? (other than the save of the images)?

quasi saffron
modest meadow
#

/promt

#

/prompt

long aspen
#

hi, guys can anyone recommend any creators or courses where i can begin learning about SD? Im new so would be helpful to hear from experianced people)

bleak matrix
#

Good morning, everyone!

#

How are we all today?

wind glacier
#

i just found this can someone explain me what this is i comparison with midjourney ?

#

and is this free t use ?

#

can you also do all aspect ratios ?

#

and based on which models is this trained ?

#

is there any website ?

#

so many questions im sorry im hope someon ca answer them

wind glacier
#

Anyone ???

wind glacier
#

???

#

A great model like this and no one that can answer my questions?

mild ferry
#

anyone got a jupiter lab i can run for vast ai?

topaz parcel
wind glacier
#

I asked multiple but the first thing what Is this server is this the same as midjourney?

#

What's the difference ?

#

@topaz parcel

topaz parcel
wind glacier
#

Is this opensource ? Thanks !

#

Are there any colabs where I can train my own models with this engine ?

#

@topaz parcel

topaz parcel
# wind glacier Is this opensource ? Thanks !

Stable Diffusion is licensed under the Apache License 2.0, which is an Open Source license. Although Stable Diffusion is open-source and has a permissive licence, there are some restrictions.

topaz parcel
wind glacier
#

@topaz parcel thanks but can you use this model on your own server ?

#

And are there any documentation about this

#

And can I finetunen with my own hardware ?

#

@topaz parcel

#

And if I can finetune it are there any Google colabs that I can use to try and research it ?

#

@mild ferry <@&1112843154362740796>

#

<@&1025179534330433656>

#

@still glacier @elfin dome

#

@bleak matrix

#

@fervent thunder

still glacier
#

Pinging everyone in here is probably not the best way to get people in a mood to help you.

#

Also, I don't use google colabs

pale latch
#

@wind glacier that's really lame. i would offer to help you out, but you seem to think everyone is at your beck and call. So if i start helping you seem like one of those people who would latch on but then also argue because you know better

#

colab bans people trying to abuse their services anyways. it's meant for students of machine learning to learn the concepts and understand things. It's made for experimentation. dropping a premade script someone else produced into it and expecting it to push button go, is against the spirit of colaboratory

waxen heath
#

So Easy Diffusion is frozen - no new features will be coming unless someone other than CMDR forks it and takes up the mantle. I will miss it greatly. I know it will continue to work, but I'll miss the excitement, remembering back when updates were almost weekly and full of new goodies. CMDR recommends Fooocus or Invoke. I've got them both, but theither seems to fill the void. Any suggestions?

pale latch
#

i just stick to automatic lately. invoke is nice for sure but they don't make their ui too extensible and they seem to be late with updates. I can't use animatediff with invoke still

clever grotto
#

hey guys, I'm trying SDXL for the first time after using SD1.4. It's taking several minutes to do an image with my 3070 with 8GB RAM. SD1.4 was a few seconds for me, is that normal switching to SDXL? I'm using automatic1111.

wind glacier
#

@pale latch I'm sorry battery died in sorry I'm just very impressed and I need to know everything about it I really want to fine-tune my own models an I hope this is the place ?

#

I don't abuse the colab it's just to learn then I will run on my own hardware

#

It really own but my AWS servers

#

It = not

#

@zinc turtle

bleak matrix
wind glacier
#

@bleak matrix thanks sunny

bleak matrix
#

np!

fervent thunder
#

Hi, is it possible to download all controlnet models for SD1.5 and SDXL from one simple source? I used google but not found all-in-val source.

regal meteor
#

Just noticed that I only have 2GB VRAM memory.

#

Is there any way to get around this restriction, maybe exchanging it with slowdowns instead?

#

(I would prefer minutes of waiting actually to not getting pictures at all)

north valve
#

what python version does stable diffusion use?

#

3.10.6?

#

because it's not working somehow

regal meteor
north valve
# regal meteor What is the error message?

D:\stable-diffusion\stable-diffusion-webui>git pull
Already up to date.
venv "D:\stable-diffusion\stable-diffusion-webui\venv\Scripts\Python.exe"
No Python at '"C:\Users\Metability\AppData\Local\Programs\Python\Python312\python.exe'
Press any key to continue . . .

#

I downloaded 3.10.6 and i checked the folder and it's saying 310 instead of 312

#

do I change the name?

#

do i download 3.12.0?

#

@regal meteor i fixed it

#

delete the "venv" folder and run "webui-user.bat" again to create a new "venu" folder, then the problem is solved. It took some time, but it worked

#

here

amber wave
#

All the license stuff is confusing to me.
Is there someone here willing to answer a few questions for me?

broken gull
#

How accessible is stablechat to write blender addons for example?

fervent thunder
#

could i pay someone here to make me a perfect prompt for a realistic girl?

hazy karma
#

Where can I find a guide on how to use the sdxl code download?

regal meteor
#

With google colab (not webui), is there a way to draw a mask inside google colab, or must I use separate program and upload it manually?

stark fog
#

Hi guys, question,

is it possible to have a style and a character(lora) in one checkpoint?

pulsar harbor
#

there's an error when running webui-user.bat

warm junco
upbeat tide
#

For WebUserUI A1111: Has anyone found (or got that you'd recommend, obv') a SDXL (not 1.5) Checkpoint which is really good for producing period piece imagery,.. classic style,.. medieval, renaissance, French revolution, Battle of Waterloo, Napoleon, Marie Antoinette, WWI, WWII, etc.

I know about the Renaissance Checkpoint but that's not for SDXL AFAIK.

I currently have: JuggernautXL, CopaxTimelessSDXL, Cyberrealistic, Dreamshaper, NightvisionXL, ZavyChromaXL, SVDN6RealXL, PirsusEpicRealism, Reliberate, Retromix, SDXLV1VAEFix, RoXL, RaynaeVisionXL. Photon.

bleak matrix
#

Good morning, everyone! How are we today?

rustic osprey
upbeat tide
bleak matrix
#

Dandy like candy, then!

#

I'm eating my bowl of cereal, ready to plot my way into whatever it is I'm doing today.

upbeat tide
upbeat tide
bleak matrix
upbeat tide
bleak matrix
upbeat tide
#

lol 😄

bleak matrix
#

You know, one of those cats that likes water.

upbeat tide
bleak matrix
#

Unfortunately, I'm afraid this is also the case.

#

Mahahahaha!

#

The land of opposites, doncha know?

upbeat tide
shrewd marlin
#

Hi uhh

bleak matrix
bleak matrix
shrewd marlin
#

How do i make consitent character ?

upbeat tide
upbeat tide
shrewd marlin
#

Oh Yes thank you, ive searched video like that, but it is not show up in my page 😅

#

Thank youu dude

upbeat tide
#

No worries. 👍

bleak matrix
bleak matrix
rustic osprey
upbeat tide
#

No. If I did, then I assume your CGPT would work. Sorry, you might want to specify that when you post a link to it. 🙂

rustic osprey
upbeat tide
#

AFAIK, the company operating ChatGPT is losing SERIOUS money every day to run it.

rustic osprey
upbeat tide
#

For WebUserUI A1111: Has anyone found (or any that you'd recommend) a SDXL (not 1.5) Checkpoint which is really good for producing period piece imagery,.. classic style,.. medieval, renaissance, French revolution, Battle of Waterloo, Napoleon, Marie Antoinette, WWI, WWII, etc.

untold trail
#

Which bot is the best one to create a sorrowful image having proper teary eyes

#

It never creates tears

dense haven
#

how to get sd without my pc explode

rich kestrel
#

git a gud gpu I mean

gilded blade
#

sdxl blurry

torpid solar
#

What should i do with all generated images. Publish on website, blog, children's book, book. Any suggestions?

torpid solar
silent hamlet
upbeat tide
# torpid solar nightvision

I'm using Nightvision and I just cannot get Paul-Emile Becat's pencil and watercolour style. *Note,. singular. Style.

#

I also can't find a Lora for his style either.

torpid solar
#

Yes it can be hard to tweak exactly as you want. But i find it to perform quite good with my historical images.

#

I want to imagine how it looked like in old times. photostyle

upbeat tide
#

I find it incredibly hard to GET a BASIC style and tweak it toward what I want,. there's no guarantee Nightvision has the data it needs to reproduce his style (not content, that's something different).

torpid solar
#

i see. think i answered wrong. you meant historical artist styles 🙂

upbeat tide
#

Yes,.. though some of the figures and interior design also (to a degree)

#

Most Checkpoints are portrait oriented, so tend not to grasp I want a small sketch of 3 characters, created in his (or other artists') style/s.

torpid solar
#

agree. its hard to not get a portrait style of picture. but its possible with a lot of work of course.

upbeat tide
#

This is why I think SD needs to do away with singular downloads and have 1 massssssive database of ALL the datasets so that we can use it like Midjourney

torpid solar
#

sounds good

upbeat tide
#

You can never be too sure what data went into a checkpoint,.. and,. let's face the obvious,.. 1 file,. is,... 1 file,.. that leaves out a TRULY EPIC amount of data that ISN'T in the 1 file.

#

It's like trying to build a house with ONLY 20 matchsticks,.. no glue, no more match-sticks, no nails, no metal, no glass,... etc etc etc.

#

If I asked Midjourney to render me a beautiful woman in Picasso's or Frank Frazetta's art styles,.. it's almost guaranteed I'd get an accurate result for each.

#

For a lesser-known artist,.. in SD,.. no guarantee whatsoever.

torpid solar
#

i see. still i am amazed by the output i get but haven't tried Midjourney

upbeat tide
#

Midjourney, will, I admit, likely blow your mind.

#

It's a million times better than SD for its data.

hasty hornet
upbeat tide
#

And time consuming too.

torpid solar
#

what plan is best= 10 /30/90 a month?

hasty hornet
#

MJ is good at generating...
SD is good at tons of other different functions, and community content

upbeat tide
#

90 I presume is the best for getting the most out of Midge, but damn that's expensive.

#

I was on the 60,. but they've increased the prices. 😦

torpid solar
#

thankss

upbeat tide
#

£60GBP per month... I don't know if you meant $ or £.

torpid solar
#

ok thanks

#

anyway. have you tried dynavision or dreamshaperXL?

upbeat tide
#

Dreamshaper,.. ha,. almost all I get from that is cartoon-like and very rubbery ugly looking figures.

#

Will look at Dynavision (is it SDXL or SD 1.5?).

#

Ahh,. it's not SDXL. Bummer.

torpid solar
#

Dynavision SDXL 1.0

upbeat tide
#

Nightvision is persistently giving me its own 'oil-painting' renders,. despite my best efforts to make it into a competent and expertly created mature (as in skill, not content) pencil sketch

upbeat tide
torpid solar
#

yes

upbeat tide
#

Author says it's: " stylized rendered anime-ish".. I definitely don't want ANYTHING anime in my work.

torpid solar
#

dynavisionXL try : eastern bloc hydraulic gopnik, zulu background, rococo watercolor painting, highly detailed,dreamy, a busy street in Paris, a cafe, people in high hats

upbeat tide
#

That's a big download to just test lol

torpid solar
#

nightvisionXL. or : RAW photo,full body shot of a drawing of a man walking to his cabin door to open it and a bird on his shoulder, storybook illustration, a storybook illustration, serial art, digital Aquarell painting

#

yes it is 🙂

upbeat tide
#

I can't see a Lora I downloaded,.. it's a pencil sketch (Anders Zorn),... in my list of Lora,.. any particular reason why that might be?

#

*Yes, it's in the correct folder.

torpid solar
#

not compatible with sdxl?

upbeat tide
#

The civitai page says it is

torpid solar
#

i get that some time with loras. works with one model but not the next

upbeat tide
#

Hm,.. it wasn't in the folder,.. but I had it

#

Redownloading it.

torpid solar
#

can i ask? what do you do with all your renders?

upbeat tide
#

I only 'accept' about 1%,... the ones that are really good. Only about 50% of thoese get printed and only about 50% of those get put in my soon-to-exist portfolio.

torpid solar
#

nice. so you print ( tshirt?) the ones you like

upbeat tide
#

Prints, on paper.

#

square on A3 'square' format to go into the books (2 books).

#

Most are midjourney renders, some newer stuff is SD

torpid solar
#

nice that you make something with it 🙂

#

mine are on the hdd only. (some in insta)

upbeat tide
#

Well, thanks, yeah. It's a damned sight cheaper than photography which is what I had hoped to do but can't

torpid solar
#

why not do both?

upbeat tide
#

Ask yourself this,.. If he can't afford 1,. how is he gonna afford to do both?

#

😉

torpid solar
#

a artist did a exhibition with photos taken with mobile phone

upbeat tide
#

Photographic portfolio,.. done properly,.. several thousand pounds,... MJ or SD,.. a few thousand.

torpid solar
#

true

#

got to go. nice to chat! bye

upbeat tide
#

Cool. 🙂 Cya. 🙂

torpid solar
#

👍

stiff topaz
#

what model is the bot currently using? and is it the same model for video as well?

#

(assume a different model, or something else it does by generating multiple images on the same model?)

left star
#

is there something like reposer for facial expressions? or what would be the best way to handle that?

knotty viper
#

Hey everybody, would anyone be able to help me out? Ive been trying to make an AI pic of "a kobold wielding a scorpion on a stick battling a beholder in a cave" in that popular 1970s dark fantasy style, and I can't get it to work. If you have to the time / want to help can ya DM me!

knotty viper
#

And also, are there any tips or specific AI models that you can create a detailed description of multiple characters, or like a recurring character that can then just be kind of name-dropped?

regal meteor
left star
#

hm i thought it was an extension but looks like it might actually just be a comfyui workflow, my bad

regal meteor
#

@left star Might be off-topic, and I never used it either, but you might like to use GFPGAN

#

I happen to need similar feature, but for anime faces, do anyone know about it, or does GFPGAN also work for that too?

left star
#

that looks interesting i'll check that out, thanks. https://github.com/nerdyrodent/AVeryComfyNerd If you look at the reposer images there you can see the workflow that shows how it takes the face from the first image, the physical pose from the 2nd image, and the outfit from the third image. I'd like to figure out a way to incorporate a facial expression into that too but I'm still kind of new to all this

regal meteor
#

Oh, I thought you want to fix faces

left star
#

not exactly but it's related and could be useful

fervent thunder
#

hey guys, I used to use stable diffusion 1.5 locally in my computer, then took a long break. what is happening right now? What's SDXL? and where can I find more news and how to install?

young ivy
#

I greet everyone

broken cave
#

where did deepfloyd go?

median escarp
#

Uhh...I don't know where I should bring this up so please don't ban me, it'll be like pulling off a bandaid here goes: SD doesn't like gay kissing

#

it just makes the people each kiss the opposite gender

#

like, you want 2 people but you get 4

#

any ideas on how to uhhh...compensate for that prompt wise?

median escarp
#

uhh..you also can't put men in lingerie...

hasty hornet
#

there's probably models or loras for that

bleak garnet
#

Hi folks I don't know if I am in the right place so sorry if I am not.

Can someone help me identify the model/Lora for this style ? Thank you so much .

sick onyx
#

o/ how to train SDXL with 24GB ram?

bleak matrix
#

good morning! How are you all today?

pure fox
#

can i add @radiant meadow to my server?

warm junco
gaunt spruce
#

Anyone know any ai that can listen to a song and tell me what piano notes are being played? I've tried a couple but it's only the chords not the fingering for the notes

serene depot
#

Hi, can someone please explain to me the relation between hires fix and base model? Say I want to render at 512x512 and hires upscale by 2. That works good, but if i increase hires upscale to 4, things go weird. From my messing around i found out that for upscale 4 you need 256x256, so what is the point? I'm kinda new to this.

candid rain
serene depot
#

ahh, ok, it looks good now, thanks

solar canopy
#

░░░░░▐▀█▀▌░░░░▀█▄░░░
░░░░░▐█▄█▌░░░░░░▀█▄░░
░░░░░░▀▄▀░░░▄▄▄▄▄▀▀░░
░░░░▄▄▄██▀▀▀▀░░░░░░░
░░░█▀▄▄▄█░▀▀░░
░░░▌░▄▄▄▐▌▀▀▀░░ This is Bob
▄░▐░░░▄▄░█░▀▀ ░░ Copy And Paste
▀█▌░░░▄░▀█▀░▀ ░░ in Every server
░░░░░░░▄▄▐▌▄▄░░░ So He Can Take
░░░░░░░▀███▀█░▄░░ Over Discord
░░░░░░▐▌▀▄▀▄▀▐▄░░ (don't spam him)
░░░░░░▐▀░░░░░░▐▌░░
░░░░░░█░░░░░░░░█░░░░░░░
░░░░░░█░░░░░░░░█░░░░░░░
░░░░░░█░░░░░░░░█░░░░░░░
░░░░▄██▄░░░░░▄██▄░░░░░░

tough saddle
#

yo

tropic belfry
#

Hi

#

I don't know if it's the correct channel to ask this question

#

I'm a total newbie to stable diffusion

#

I'm searching for an api solution that takes as input an existing image and a prompt and change detail of the image, like change the hair color, or the background

#

Tried the sandbox but it return only GRPC request failed message

#

Is SD the correct solution for this job?

mystic sedge
#

is there somethings new to know about SD or it's still the same since fews month ? there were any update or something ?

rich kestrel
#

1.6 is tha bomb diggity. Never looked back.

#

ReActor also launched not long ago which is a godsend

mystic sedge
#

I mean sd is still the same since i use it, there had been never news things since ?

sullen walrus
#

Does anyone know some ai that find images on the Internet with a similar Strukturen like Mine?

left star
hasty hornet
hazy karma
#

just got stable diffusion to run on my macbook

vivid parcel
#

Hey guys! If you have any time could you please check out this user study I made for CVPR? I'm trying to get some human-evaluation results for our new dataset, and we're doing a last-minute user study by the submission deadline Friday lol https://forms.gle/wKPFtTm8178adL1r6

It's a new massive image generation dataset specifically made for generating transparent images - we're hoping to release it to the public soon

Right now I'm hoping to get at least 20 people to do this study...it's not going to be very large so every person that fills it out helps tremendously

If anyone wants access to the dataset early please let me know!

small seal
#

hi everyone

#

i just got into stable diffusion webui

fluid summit
somber mortar
#

Anybody here using Fooocus?

hallow plaza
#

i there an easy to use ai or model to touch up images? i went on vacation and I have several images where i'm not looking at the camera that i would like to fix.

eager lantern
#

anyone know how to do clothing change or undress on a image

high sun
#

hey guys, someone have a link for the dreambooth discord server please ?

pure sorrel
#

Hi. Longtime NMKD user. I'm wondering what are people's favorite GUIs?
I tried Auto1111, and it was ok . . . but I'm looking for something even deeper and more comprehensive.

gilded cradle
#

yo guys

#

i m new to using S.D

#

need some help with

#

how to use extensions

main pulsar
#

Join over 10,000+ readers to become smarter about AI with new tips, tricks, and hacks - all in a free bite-sized 5-min newsletter. Read by subscribers from Amazon, Anthropic, Cohere, and Google.

Subscribe and invite friends to access leaderboards for top AI courses, research reports, events, stocks, investors, grants, repos, AI tools (including this one), and AI tech stack companies.

https://www.theaicrunch.com/

smoky kernel
#

hello guys a few months ago i was using SD webui automatic 111 on google colab...now i tried again but it's not working i tried running the last notebook from "the last ben" but some models do not work , for instance sdxl and protogen and dreamshaper xl....they all give garbage output...only dreashaper works...what could be the problem? anyone using automatic 111 on Colab here?

somber mortar
#

Anybody here using Fooocus?

craggy tartan
#

Besides using Linux, is there any way to use AMD GPU on A1111?

#

I'm currently running off my CPU

pale latch
#

do it. use arch. go full linux. make windows the dual boot and linux the main. dooo it

#

/whispers do it

#

kde plasma, the windows wobble when you drag them. wibble wobble windows! i'm not even joking bruh

pale latch
#

@craggy tartan ||do it||

sly ether
#

Just tried SDXL 1.0 1024x1024p and it took like an hour

#

I am trying to understand if this is my fate running on an AMD Ryzen 5 3400G with Radeon Vega with 32 GB RAM

#

Or if it is due to me having the wrong torch or xformers installed, or something else

#

Anybody who knows if this is to be expected, or am I doing something wrong?

weak kettle
#

guys how can I generate image using 2 different lora's (celebrities)?

#

the trigger word for both is sks person

#

cant make it work

glass lynx
#

how do i report dm weirdos

#

or like advertisers

craggy tartan
glass lynx
#

might take some time to learn how to use it though

#

using sdxl as a base and img2img sd 1.5 takes about 1 min per gen

craggy tartan
#

Does inpainting require specific LORAs, or you can use any LORA during inpainting?

pale latch
#

might as well be cpu tier

#

get on rocm you'll be like "woah!"

craggy tartan
pale latch
#

click it its relevant

craggy tartan
#

i did

fervent thunder
#

Would this be the right channel to ask a question related to creating custom models?

#

If not, where would be best?

hollow dust
#

Latelyy I've been getting the word "arafed" when I do Interrorgator. Can't seem to find any solid definition via google etc. Anyone know what this word means in relation to prompts?

warm junco
somber mortar
#

Anybody here using Fooocus?

feral eagle
#

Boys went from directml in window on my 6700xt, then Rocm in Linux and that was a big jump, but prefer windows then I got a 4070 today. Good god so much easier and so fast.

#

And all my generative ai stuff I’ve been learning for work…. So much easier. AMD needs to step it up

somber mortar
#

what model are you guys using on invoke AI?

hasty hornet
#

Not using invoke or whatever foocus is

somber mortar
#

So what are people using the most

#

cause Im getting bad results with Invoke

hasty hornet
#

a1111, comfy

somber mortar
#

is that the best?

hasty hornet
#

results shouldn't be affected too much by ui anyway

hasty hornet
ivory ridge
#

hi where can I ask a question about comfy ui

valid echo
#

is it worth to switch to linux for the performance increase?

slender dirge
slender dirge
# hasty hornet results shouldn't be affected too much by ui anyway

it should. invoke use different grammar and results will be out of order when copying civitai prompts/data. also i find images from invoke in most cases less saturated than a1111 even after I double checked vae things. besides, fooocus use inner GPT to handle inputs and no matter what i type, the results are always not bad

hasty hornet
#

idk about getting exactly same results, but while I was trying out invoke it wasn't that bad, I noticed weird thing tho...
Maybe it was old version or maybe I had something bugged out, but it used DDIM nomatter what I chose

#

Which is fine for testing things out, but as for getting actual good image...nah

slender dirge
primal cradle
#

I found out that the term "execution" is on the list of prohibited words for use with the Stability AI API. Could anyone please provide me with the complete list of banned words to ensure I adhere to the guidelines when making requests?

lucid lynx
bleak matrix
#

Good morning, everyone!

somber mortar
#

like hours

#

do you have any idea why?

fervent thunder
#

Will Stable finetune be a FOSS application?

keen gate
#

how do I check if my stable diffusion is auto updating?

ocean mango
warm junco
#

or if you edit the webui-user.bat and find git pull in it

keen gate
#

ok thanks

green heart
#

Greetings everyone, Pardon me for my intentions.
I am a software engineer in pursue of producing a AI based application
And I am deeply in need of a part time designer ( 2d )
If anyone is looking for a part time job, My DM is always open.

#

( You do not have to be a professional designer nor programmer, since all aspects of QT designing envrironments are user-friendly and made easy. )

vast fog
#

https://youtu.be/0pGR6JJk2n8 Guys I asked AI to analyse Interstellar. Some really dififcult science concepts broken down so did a great job. Let me know what you think

primal current
#

With what tools can I create a virtual assistant that answers my questions? or who has experience doing it? It's for a job.

somber mortar
#

Where can I download models?

slender dirge
somber mortar
slender dirge
#

are you using 20xx or 30xx or 40xx cards

vast fog
#

https://youtu.be/0pGR6JJk2n8 Guys I asked AI to analyse Interstellar. Some really dififcult science concepts broken down so did a great job. Let me know what you think

somber mortar
#

Im using 2070 super

slender dirge
topaz sundial
#

Can someone direct me to a place where I can ask questions? I have just installed the software and the batch file says press any key to continue and then just disapears

vague pond
forest trout
#

There's a 1.6 model? News to me...

warm junco
lavish lake
somber mortar
warm junco
#

but on civitai youll find more stuff and better visualised/sorted

somber mortar
#

Am I missing something or is the model a1111 and comui a completly different program or models you need to add to your existing program?

warm junco
#

The models itself are .safetensor files

#

These need to be downloaded to be used in a webui

rough sail
#

I found a nice suggestion from someone on Reddit on how the next SD could beat DALL-E 3:
Just in time, please reserve at least a million for this:

How a new SD could beat DALLE-3 at the base level without any extension

A full training on the biggest LAION instead of a smaller LAION subset -> (1M$)

A new CLIP -> (*LLaVA or even better)

RLHF training using vision feedback loop concept -> (idea2img https://arxiv.org/pdf/2310.08541.pdf
Sounds like a nice idea for @wise stratus to consider when SD3 is being made IMHO.

somber mortar
#

Do you know what's the best model for Invoke AI?

#

I am using that atm

somber mortar
#

actually what's the best model to make a logo

weak kettle
#

can anyone explain to me how to train it via this guide? It's so messy but results from this guy are great

#

lmk

fervent thunder
clear oyster
#

Hello I need help with using SD on CPU

#

I'm getting "Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check"

ocean aspen
#

i setup automatic1111 and it worked okay, but when i installed control net, and installed the models for it, and placed them in the models folder in its exetension folder, and opened the web ui, i can see it, it shows me the preprocessor and stuff, but when i click generate its like its not using controlnet at all, and when i open settings and go to control net, not all the settings are there, 2 are missing which are the config file for adapter and control net models location.

broken bloom
#

Hmm anyone interested in AI image generation models

Looking for collaborators 🗝️

potent frigate
#

Did you add --skip-torch-cuda-test to the commandline arguments? My full command to turn on automatic1111 with cpu uses --use-cpu all --no-half --skip-torch-cuda-test

native heart
odd harness
#

WHEN CONTROLNET FEATURE WILL BE AVAILABLE

prime crag
lavish lake
rapid peak
#

One message removed from a suspended account.

delicate yoke
molten tangle
#

Is there any realtime sketch2img app somewhere ? Where you have a context prompt and then draw a sketch for controlnet ? (a kind of img2img)

sour mica
#

is clipdrop paid now?

fervent thunder
#

anyone know how the taylor swift x oscar the grouch NSFW image was made like what website cuz it was really well done but every website ive tried is bad

sudden pawn
#

soyjak.party

#

nah just kidding i don't even know what you're talking about

clear oyster
clear oyster
#

an do you add wuotes

#

quotes*

potent frigate
#

type into your command line "/Path/To/Your/webui.sh" --use-cpu all --no-half --skip-torch-cuda-test

#

And that should work. Maybe.

hollow ridge
#

hi all I'm just wondering if these tools like stable diffusion, midjourney, etc generate high enough quality / high resolution images that I would be able to take to a print shop and print in a large size for wall art, like 30x40 inches or greater?

clear oyster
late garnet
#

Do we need to install the CUDA toolkit to benefit from performance or stability/compatibility (using comfyui or a1111 on win11 or WSL2-Ubuntu)? And are there any benefits of manually replacing the 7 cudnn files, under the site-package/torch/lib path, with updated releases?

slate dust
lavish lake
shy cedar
#

2060 super is still a decent gpu right?

eternal pendant
#

The video features is a little disappointing I ask it for a car with flicking headlights and all I get is a simple static image that pans left to right

spice plank
#

I have a few hundred images I need to run through png to text2image with a hirespass on. Is there any easy way to automate that with auto1111?

warm junco
ionic nexus
#

Аааааа

plain crane
#

I trained my lora for 2400 steps using 34 photos and i used it. man, my pictures looked worse with this lora than without. What should i do, train more?

tawdry sun
#

Hello everyone，If I want to generate some 3D game scene concept, what should I do?

hasty hornet
#

thinking pick model and prompt what you want on scene

bleak matrix
#

Good morning, good morning! How are we all today?

calm geyser
#

como cria imagem ?

dark cipher
brazen tulip
#

Hello everyone! Where can I find an overview about the frameworks and tools, like Stable Diffusion A1111 / Stable Studio / ComfyUI / ... maybe a table or so?

wild bane
#

I once read about an option where you can generate an image where a prompt tag is only applied to the top half of the image for example. Rings a bell for anyone? I'm trying to find some info about it
Edit: I found it, it's called "Regional Prompter"

coral mist
#

cc

#

who speak french here please ?

#

salut

#

comment on fait pour utiliser stable difusion stp ?

glacial mason
#

hi can i specify seed in the bot?

#

there is no bot parameter

#

-seed in the prompt?

#

also no outpainting here yet?

dawn anvil
inner lintel
#

How difficult would it be to prepare a large dataset like LAION but with LLAVA-generated captions? 🤔

#

Tried the online demo for LLAVA, it's awesome and seems to outperform BLIP

lucid lynx
#

thanks!

pliant ice
#

Is there an extension in which I can make cumulative changes based on the resulting image?

Like, let's say I got a simple full body image of a man and I want to make his hands bigger. I send the pic to img2img, inpaint his hands, then write in the prompt something "bigger/larger hands"?

abstract pivot
#

new hires method by kohya

velvet jackal
novel stirrup
#

HEEEEEELP please ;(;(

#

No module 'xformers'. Proceeding without it.
Warning: caught exception 'Torch not compiled with CUDA enabled', memory monitor disabled
2023-11-16 11:45:05,216 - ControlNet - INFO - ControlNet v1.1.417
ControlNet preprocessor location: /Users/stable-diffusion-webui/extensions/sd-webui-controlnet/annotator/downloads
2023-11-16 11:45:05,276 - ControlNet - INFO - ControlNet v1.1.417
Loading weights [6ce0161689] from /Users/stable-diffusion-webui/models/Stable-diffusion/v1-5-pruned-emaonly.safetensors
Deforum ControlNet support: enabled
Creating model from config: /Users/stable-diffusion-webui/configs/v1-inference.yaml
Running on local URL:

#

To create a public link, set share=True in launch().
Startup time: 5.0s (prepare environment: 0.2s, import torch: 1.7s, import gradio: 0.5s, setup paths: 0.5s, initialize shared: 0.1s, other imports: 0.6s, load scripts: 0.5s, create ui: 0.5s, gradio launch: 0.2s).
Loading VAE weights specified in settings: /Users/stable-diffusion-webui/models/VAE/vae-ft-ema-560000-ema-pruned.ckpt
Applying attention optimization: sub-quadratic... done.
Model loaded in 8.0s (load weights from disk: 0.5s, create model: 0.7s, apply weights to model: 5.3s, apply half(): 0.4s, load VAE: 0.8s, calculate empty prompt: 0.3s).
Restoring base VAE
Applying attention optimization: sub-quadratic... done.
VAE weights loaded.
Restarting UI...
Closing server running on port: 7860
2023-11-16 11:45:31,729 - ControlNet - INFO - ControlNet v1.1.417
/Users/stable-diffusion-webui/extensions/deforum-for-automatic1111-webui/scripts/deforum_helpers/ui_right.py:38: GradioDeprecationWarning: The style method is deprecated. Please set these arguments in the constructor instead.
with gr.Row(elem_id='deforum_progress_row').style(equal_height=False, variant='compact'):
Running on local URL:

#

To create a public link, set share=True in launch().
Startup time: 0.7s (load scripts: 0.3s, create ui: 0.4s).
0%| | 0/20 [00:00<?, ?it/s]2023-11-16 11:45:57.324 Python[10468:375542] Error = Error Domain=com.apple.appleneuralengine Code=6 "createProgramInstanceForModel:modelToken:qos:isPreCompiled:enablePowerSaving:skipPreparePhase:statsMask:memoryPoolID:enableLateLatch:modelIdentityStr:owningPid:cacheUrlIdentifier:aotCacheUrlIdentifier:error:: Program load failure (0x5)" UserInfo={NSLocalizedDescription=createProgramInstanceForModel:modelToken:qos:isPreCompiled:enablePowerSaving:skipPreparePhase:statsMask:memoryPoolID:enableLateLatch:modelIdentityStr:owningPid:cacheUrlIdentifier:aotCacheUrlIdentifier:error:: Program load failure (0x5)}
2023-11-16 11:45:57.325 Python[10468:375542] ANE load failed!
2023-11-16 11:45:57.325 Python[10468:375542] Issues were found in compilation, falling back to GPU

#

Trouble with SD from the start. Initially, faced NAN errors, and now nothing is happening in the program. Need assistance agony

warm junco
coral mist
bleak matrix
#

Good morning, everyone!

#

How are we all today?

lavish lake
clear osprey
#

hi, what's the last version of SD and SDXL and where do i download them?
i have 1.5 but nothing more

lavish lake
white otter
#

Anyone know of a discord that talks specifically about LoRA training? They're all so scattered

loud spoke
#

why i cannot use more that 50 steps in my API

#

??

#

text to image sxdl v1.0?

white otter
#

I'm finding a lot of success with training sdxl but I'm doing a project with sd1.5 and they are turning out horribly

lavish lake
raven wagon
#

hey guys quick question
for SDXL training
what model do I use? what's the best one? because there are different ones out there

could you send me hugging face link please?

ornate flame
#

Any news about SD3?

halcyon granite
#

can anyone recommend me some free windows software that will allow me to realtime(no delay) preview still jog as a video? I used to have a software like this but can't remember the name

sour dome
#

i installed a model and try to make a picture but it seems like no matter what model i use stable diffusion uses its own model and the output is bad, can somene help me to solve this thank you.

frozen rover
#

Hey guys, how do I add an additional conditioning to a stable diffusion model? Has anyone does it before whose script I can refer to?

pseudo coyote
#

hey all, I am downloading control net 1.1 and came up with two pages to download models from, one has 14 models and the other has models for sdxl. i am using sd 1.6. which page do I use?

lavish lake
pseudo coyote
#

stable-diffusion-webui-1.6.0 this is what it says

lavish lake
#

oh the webUI okok, thought you were referring to the models

#

the webUI is one thing and the SD models another

pseudo coyote
#

yeah the model are 1.5 and sdxl

lavish lake
#

for 1.5 there are quite a few, depending on what you are doing (there's ones specific for anime for instance)

undone mango
#

This is going to be a game changer

fallen mason
#

where does the images i make save to?

#

it says it autosaves them but idk where to look

weak kettle
#

guys, if i want to train model using kohya dreambooth, I should save it as diffusers, safetensors, or ckpt?

#

i want to later convert to lora

clear oyster
#

when I uninstall SD do the images stay?

weak kettle
#

NotImplementedError: Cannot copy out of meta tensor; no data! anyone?

#

anyone can help, knowing kohya?

primal holly
#

Hey

mild dock
#

Hey guys, currently I'm using stability.ai rest API, so what I want to achieve is, I put image, and regenerate that image with new aspect ratio, is it possible? anyone can help?

undone mango
mellow lynx
#

hello

low nimbus
#

Hey fam! I'm new in this. Dumb question, how safe is to install SDxl locally in my pc?

bleak matrix
#

Good morning, everyone! How are we all today?

bleak matrix
low nimbus
#

I mean, there is no risk about any malware or malicious code that could be hidden in this installations?

bleak matrix
visual pier
#

Heya ^-^
Im new to this discord. I have a local Web UI Stable Diffusion and since I started there is hardly a day where I not engage with it ^-^

bleak matrix
visual pier
#

I like that it tickles both my tech brain and my art brain. And that even though I have a terribly old graphics card I can still use SD ^-^
And obviously I love the images that it generates

#

Oh and since then I helped 4 people install SD on their own devices! Since once its set up its surprisingly easy to use I feel at ease helping my friends set it up

bleak matrix
#

I also love the part of helping people---that's what's really fun about SD as well--having a community of like minded people, as well as being able to help family and friends enjoy this new technology.

#

What kind of art do you like to do?

clear osprey
#

does stable animation allow to get gifs from absolutely nothing except prompts and a base image? basically img2gif

visual pier
clear osprey
visual pier
clear osprey
#

bruh animate diff is so slow though
it takes several minutes to generate some 3-4s 24fps animations in low res
even 4K upscales are faster

bleak matrix
visual pier
clear osprey
bleak matrix
clear osprey
#

no need to send so much stuff xD
i only asked to see if stable animation was as good as animate diff

royal ruin
clear osprey
visual pier
clear osprey
#

anyone knows where i can put a bunch of my best generations?

visual pier
bleak matrix
bleak matrix
#

You can use #✨｜sdxl for showing off your work, of course!

visual pier
#

Idk if its SDXL tho :O
I am still figuring out where the difference even is

bleak matrix
#

Between it and...?

visual pier
#

SD

#

I also remember the indigo furry mix model stating that its not SDXL yet :D

clear osprey
bleak matrix
bleak matrix
clear osprey
clear osprey
bleak matrix
visual pier
clear osprey
#

yeah
BUT sdxl can generate in large resolution without hires fix

clear osprey
visual pier
visual pier
royal ruin
#

i know i can drag the image in, but then i need to go back and format my prompts, set a bunch of settings that aren't in the image data, etc.

visual pier
royal ruin
#

so i want to be able to save point-in-time versions of my workflow to iterate on

visual pier
royal ruin
clear osprey
visual pier
royal ruin
visual pier
visual pier
clear osprey
visual pier
royal ruin
#

the other thing i was running into with A1111 is that i was having problems sometimes replicating an image even with exact settings. i did a little research and it seemed like it might be a bug that can creep in when you're switching models and reuse a previously cached version

visual pier
royal ruin
visual pier
royal ruin
#

oh the seed was definitely correct, i was double-checking that.

#

i saw a github issue that talked about the problem, but didn't dig into it too deeply yet

#

i was getting frustrated and decided maybe it was a good time to try out SDXL since i'd only used SD1.5 so far

visual pier
royal ruin
#

which led me to Comfy because SDXL on A1111 was juuuuust a smidge going over my VRAM (3070 8GB) and i'd seen that Comfy was more efficient (proven since, i can fo SDXL gens on Comfy as quick as 1.5 on A1111)

bleak matrix
royal ruin
bleak matrix
royal ruin
hasty hornet
bleak matrix
#

There's literally so many effecient workflows that people have made that it's super super easy to just...find what you need.

royal ruin
bleak matrix
hasty hornet
#

I wish we had like a library of most basic workflows we can adjust...tutorials are good , but kinda takes too much time when all you need is just to take one little thing

bleak matrix
#

For sure~! I would check out the prompting guide that I posted earlier if you want to learn more about very specific things.

#

I'll be updating it when I have time, but there's enough info for you (and others) to get started!

royal ruin
bleak matrix
#

As well, and there is a ton of information here as well to get you started--there are tons of resources available to you to learn!

royal ruin
#

oof i am just skating below my VRAM with basic gens on SDXL. i think my xmas present to myself is gonna have to be a new card

visual pier
#

I uploaded some of my best ofs on the character channel ^-^

visual pier
#

Since I have an old graphics card I figured out how to generate high res images while using as little ressources as possible ^-^

royal ruin
visual pier
# royal ruin this just doing upscaling in post or do you have a secret technique?

Kind of. I first generate a 512 x 512 or 800 x 450 image that I like the composition and promt stuff of with as little steps as necessary (around 25). Then I pin the seed and generate it with hires fix. Upscale it by 2 and around 25 steps are enough for most images. If theres lots of background details I crank it up to around 35 hires steps

thorny frost
#

Hello. Completely new to AI generation! I've installed A1111 and this is running SD1.5... but presumably there's a way to install SDXL? Are there any guides for this?

hasty hornet
visual pier
hasty hornet
#

I have no idea how most things work in comfy, and not sure if I bother, I just want to use something that already works without spending hours going through tutorials on something I'll need just once and then just save workflow and gen gen gen

royal ruin
visual pier
royal ruin
#

has made it a lot easier to understand going from A1111 to Comfy (and SDXL)

#

im only trying out the move because i've got to the point in my current project where i felt like i was missing a deeper understanding of stable diffusion's behavior and wanted a better view into the guts

thorny frost
# royal ruin im only trying out the move because i've got to the point in my current project ...

Apologies for jumping in here... but is Comfy better than A1111? I'm a complete newbie with this stuff, so it feels as if I'm just stumbling around in the dark. I've installed A1111 but haven't installed any models yet, so my results are pretty rubbish at the moment. I want to generate scenery/backgrounds/environments so that I can comp my 3D stuff in... but all models seem to be aimed at character generation, which I don't want/need. Can I just install a suitable/relevant model and then use a negative prompt to ensure that I don't get any characters... or is that not how this works?

visual pier
royal ruin
# thorny frost Apologies for jumping in here... but is Comfy better than A1111? I'm a complete ...

i wouldn't say that one is better across the board, and also i've only been using stable diffusion for a few weeks now so i'm far from an expert so take what i say here with a grain of salt..

A1111 was great to get into everything and it's super powerful and easy to work with. what drew me to check out Comfy (and again, i've only been using it for less than 24 hours) is that the way Comfy lays things out you can see (and fiddle with) all of the elements in the pipeline.

#

also fully half of why i wanted to try Comfy is that on A1111 with SDXL my 8GB VRAM card takes 3-5 minutes to gen but Comfy is more efficient and it takes ~20-30 seconds, which is similar to SD1.5 on A1111

thorny frost
royal ruin
smoky kernel
#

hello guys sorry to interrupt, can anyone explain again the use of the little blue arrow (is the one next to the bin?)...thanks

royal ruin
smoky kernel
#

saved

#

i always have a myriad of images i want to go back and rework but i completely forgotten the parameters...

royal ruin
smoky kernel
#

interesting,,,,i wonder where can i find more info about this setting that adds the parameters inside the png ....

royal ruin
#

i pretty much learned everything i know about A1111 from reddit

clear osprey
visual pier
thorny frost
smoky kernel
#

i just realized for the first time there's a "png info" tab!!! but unfortunately my png s have no paramaters saved for some reasons even though in the settings is said that the parameters are saved in png chunk...also in the settings there's a
Create a text file next to every image with generation parameters
i will try that from now on

royal ruin
#

if you don't have them super personalized, might want to try resetting your settings to defaults

visual pier
smoky kernel
#

i confirm that automatically saving a txt file with the parameters works...

thorny frost
visual pier
jade shale
#

What's the best way to generate a character into a background image using ControlNet Pose? I know you can just inpaint the area where the pose image is, but I'd like something that doesn't requiroe guess work as to where the character will be and not be cut off or change the backgroun too much

bitter laurel
jade shale
bitter laurel
jade shale
bitter laurel
# jade shale I don't want to change the background at all though

hmm. that's a good question. I saw Krita just got a plugin - you'd just overlay it and have the image generate. I haven't played with that too much yet, but I have a drawing tablet, so I am considering putting in the work to see what happens.

otherwise you'd make the character in openpose, then cut it out, and have img2img clean it up a little bit with the lighting, etc.

#

but you could also try img2img and mix in the controlnet into it, with a low setting. I feel like there should be some way to make it work.

ocean mango
low nimbus
low nimbus
foggy estuary
#

Does anyone know of a good model (Lora, checkpoint, anything) or good prompts to use to make pictures look more realistic? As in like look like a photo taken on a smartphone or cheaper camera? I want things to feel more authentic in both composition and quality

visual pier
wispy jasper
#

Am I the only one who thinks SDXL kinda sucks?

visual pier
compact summit
#

hello

visual pier
compact summit
visual pier
#

Yes :D

compact summit
#

nice to now

forest trout
bitter laurel
winter iris
#

Is there a job board somewhere? I don't see any channels to post SD related jobs

visual pier
forest trout
#

Sounds like commissions to me.

left star
#

is there any workflow using a1111 or comfyui that can reproduce similar effects as runwayml gen-2 (image+text)-to-video? if so, can anyone point me in the right direction?

pale latch
#

what is sd1.6? anything about it other than it just existing on the api?

pale latch
#

and you'll need lotsa vram since it's all the frames generated in one big batch