#💬｜general-chat

1 messages · Page 80 of 1

quick yew
#

no, i did before trying to install it, i had an issue so i reinstalled visual studio and after that it doesn't work after restarting the pc, reinstalling python etc

#

i don't wanna lose my webui it has some extensions and settings that i use

#

Traceback (most recent call last):
File "G:\ai\stable-diffusion-webui\venv\lib\site-packages\transformers\utils\import_utils.py", line 1086, in get_module
return importlib.import_module("." + module_name, self.name)
File "C:\Users-------\AppData\Local\Programs\Python\Python310\lib\importlib_init.py", line 126, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
File "<frozen importlib._bootstrap>", line 1050, in _gcd_import
File "<frozen importlib._bootstrap>", line 1027, in _find_and_load
File "<frozen importlib._bootstrap>", line 1006, in _find_and_load_unlocked
File "<frozen importlib._bootstrap>", line 688, in _load_unlocked
File "<frozen importlib._bootstrap_external>", line 883, in exec_module
File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed

fervent thunder
#

hi folks, having issues getting webui.sh to start -- heres the error message - can someone point me in the right direction?

Cannot locate TCMalloc (improves CPU memory usage)
Python 3.10.12 (main, Jun 11 2023, 05:26:28) [GCC 11.4.0]
Version: v1.6.0
Commit hash: 5ef669de080814067961f28357256e8fe27544f4
Installing torch and torchvision
Looking in indexes: https://pypi.org/simple, https://download.pytorch.org/whl/cu118
Collecting torch==2.0.1
  Downloading https://download.pytorch.org/whl/cu118/torch-2.0.1%2Bcu118-cp310-cp310-linux_x86_64.whl (2267.3 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╸ 2.3/2.3 GB 41.2 MB/s eta 0:00:01Killed
Traceback (most recent call last):
  File "/home/taholmes160/Public/stable-diffusion-webui/launch.py", line 48, in <module>
    main()
  File "/home/taholmes160/Public/stable-diffusion-webui/launch.py", line 39, in main
    prepare_environment()
  File "/home/taholmes160/Public/stable-diffusion-webui/modules/launch_utils.py", line 352, in prepare_environment
    run(f'"{python}" -m {torch_command}', "Installing torch and torchvision", "Couldn't install torch", live=True)
  File "/home/taholmes160/Public/stable-diffusion-webui/modules/launch_utils.py", line 115, in run
    raise RuntimeError("\n".join(error_bits))
RuntimeError: Couldn't install torch.
Command: "/home/taholmes160/Public/stable-diffusion-webui/venv/bin/python3" -m pip install torch==2.0.1 torchvision==0.15.2 --extra-index-url https://download.pytorch.org/whl/cu118
Error code: 137
taholmes160@Ubuntu2:~/Public/stable-diffusion-webui$```
charred thistle
quick yew
#

its a big list of issues that never appeared before

charred thistle
fervent thunder
#

system installed version as far as I know -- im following the instruction s herehttps://github.com/AUTOMATIC1111/stable-diffusion-webui

charred thistle
# quick yew its a big list of issues that never appeared before

perhaps the extension you tried installing before replaced some of the necessary packages for webui to work?.. you could try enabling the venv if you have it, and from that try "python -m pip install -r requirements.txt" to reinstall the ones webui needs

quick yew
#

well i did uninstall and reinstall python

#

so maybe that could work

#

python -m pip install -r requirements.txt where do i run that?

charred thistle
#

from a terminal inside the webui folder

#

unless you're running it through a conda environement, if so, you need to activate that first

charred thistle
fervent thunder
charred thistle
#

because where it attempts to install pytorch i see "killed" at the end, so the process was terminated, perhaps manually

#

it looks like you are using a venv as well

fervent thunder
#

I didnt kill it -- it just stopped --

charred thistle
#

did the process create a bunch of folders where you ran the script?

fervent thunder
#

its complaining about cant install torch and error 137 -- which chat gpt claims is an out of memory error

#

there are some in the ~/Public/stable-diffusion-webui folder

#

Im trying installing torch and torchvision manually using pip and so far thats running

#

installing a bunch of nvidia stuff right now

charred thistle
#

if you open a terminal in that webui folder and type "source venv/bin/activate" as a command, without quotes, what does it do?

fervent thunder
#

activates the venv virtual environment

charred thistle
#

and it gives you a new prompt?

fervent thunder
#

ya

charred thistle
fervent thunder
#

ok, its running --

charred thistle
#

is it installing a bunch of python stuff?

#

when it finishes you may want to try "pip install -r requirements.txt", then exit conda with "conda deactivate", and try running "./webui.sh" again

fervent thunder
#

ya its doing a bunch of python stuff -

#

I'll try, my computer is not the fastest in the whole world 😄

warm junco
charred thistle
#

I think I'm done with my tests running SD using both the native system python package (3.11) with the intel extension for pytorch installed via a single terminal command, as well as the ipexrun version with intel python and honestly I found no discernable difference in performance and the difference between the two is the second one requires a lot more work to install plus 20gb+ of hard drive space

quick yew
#

i ran out of space while installing pytorch

#

while troubleshooting i tried to reinstall pytorch some times and i think it accumulated

#

where does pytorch is so i can delete it?

warm junco
quick yew
#

thanks!

forest trout
#

You can try updating your GPU drivers too. I found that helped with some cuda problems I was having when reinstalling the venv wasn't working.

violet berry
#

Hi guys. Do you think 25 image sources for me to create my LoRa is good if it's an anime character? I'm also making all of them as png and removing the backgrounds. Please share your thoughts if this sounds optimized or if you recommend something different please.

fervent thunder
#

hey guys -- looking for a good chatbot that I can run locally to assist with python and flask programming -- is stable-lm or stable beluga up to it?

bleak matrix
#

Good morning! How is everyone?

quasi rune
#

Anyone have any idea why training textual inversion gets such crazy results when not on base 1.5 model?

molten bison
#

is there anyway to add a lora here

dense goblet
#

Is there a solution or model capable of producing images of a given object from various angles? I'm interested in discovering if we can take a picture of a face and turn it into a side view or something close based on the input image.

#

It doesn't have to be limited to faces, only something that can recreate simpler objects

tribal needle
#

u know how MidJouney can imagine from links? like /imagine <link> web ui, ui/ux, web design, aesthetic, modern, clean can @radiant meadow do that? like use an image from a link and reimagine it?

quiet beacon
#

how do i create a priovate prompt room for myse;f

charred thistle
#

i think it might be good form if I do prompt merging so I can add a stuff that I seem to be putting into every prompt, into a dedicated text box

regal brook
#

yo how do people make those AI disney pixar movie poster memes? i've tried using stable diffusion but it always seems to come out shit when i try

bleak drift
#

What is the best checkpoint for making realistic backgrounds or images of things that are t people.

#

Like the moon from Skyrim or the weapons and stuff.

zealous parcel
#

After tinkering around with webui, there doesn't seem to be a way to send concurrent jobs to the GPU when using ControlNet "Batch" mode.

The effect of "Batch size" in the main settings appears to operate the way it normally does, which is to generate seed + n={batch#}, which is naturally not what I'm after.

I was wondering if anyone has hacked together a method to send these ControlNet Batch iterations to the GPU in a similar way to how "Batch size" sends concurrent jobs?

Hopefully I've just missed something and this is an easy fix, but given my GPU can easily do 20+ tasks concurrently, it seems painfully slow to do each ControlNet Batch iteration one by one.

signal river
#

What are things to try if I want to get more pronounced effect from Roop ?

wooden nexus
#

Guys how to create image?

native sundial
#

do slash dream in bot channel

muted skiff
#

Is AbyssOrangeMix3 really made by a chinese guy called liudinglin? Or is he just some guy that uploaded it to Civitai for likes?

#

I'm pretty sure AOM3 predates Civitai

amber flax
#

/info

subtle vine
#

Hi, I generated an image of a character and I like what I got. I mean face, clothing, so I want to keep the character to generate more imges for example with different poses etc. with this character. How can I do this? How can I keep this generated character and on its base how can I generate more images with same clothing, face? I'm kinda new to SD so I don't know much about tools it has. I'm using ComfyUI

heady hedge
#

I feel like this is probably really common, but I downloaded A1111 to try and generate OC’s of mine more consistently than Midjourney can, and so far it’s not gone great 🥴 is there a resource somewhere to learn how to generate a specific character reliably/which models I should be running for consistent results. I’m running it on the M2 pro chip

native sundial
heady hedge
native sundial
heady hedge
heady crest
muted skiff
#

My generation stats in auto1111:
640x640: 30secs, 3.4GB VRAM (spikes to 4.3GB in last 3 secs)
768x1024: 82secs, 3.7GB VRAM (spikes to 5.5GB)

I prefer to gen in 640x640, however the faces are noticeable low-res/blurry.

What's the fastest, easiest way I could get decent faces without going past 5.5GB total VRAM (even when it spikes at the end).

ANY sort of Hires Fix breaks the limit (even when I tried 512x512).

native sundial
#

using add-sharpness and add-details Lora's also help

karmic pumice
#

gen 512px and hiresfix to 640px? adetailer extension? Try using tensorrt model if your workflow allows it

warm junco
#

There is some stuff you mostly don't need

high ruin
#

What is the best SD GUI software for AMD GPU??

inner lintel
#

Looks like "understanding" of text by t2i models has become more important part than resolution and "realism", so dall-e 3 outperforms SDXL just because of its text encoder. We have dall-e 3 paper already, it would be cool to take a look at some new improved SD by Stability.

#

Even if it's just 512x512 like SD 1.5, still would work way better for text inputs

native sundial
fossil bloom
#

Good prompts for skibidi toilet generation?

woven forum
charred thistle
#

lmao, my quadrupedal monsters are turning out tripedal, kind of funny

visual hill
#

Anyone know where i can find the sd15 softedge model for the softedge controlnet? I can't seem to locate it in huggingface.co

undone bronze
#

how can i give a prompt to generate the images i want #newbie here

fervent thunder
#

Hi

warm junco
high ruin
#

Why does stable diffusion take 85min+ to generate one image this is not 1994??

robust tide
#

I think my wife might be getting generous of the asian girls my GPU is creating

oblique cipher
#

I'm planning to make a video and need to separate the openpose format. Please let me know how to separate DW openpose without rending the image to save time

native sundial
warm junco
#

Getting 1.9 it/s

native sundial
warm junco
hollow ledge
#

hey! are clipdrop tools opensource? Can I run them localy?

digital rivet
#

Hey Guys, im absolutly new here and I want to learn how to master Stable Diffusion can you guys please provide me with informaton or materials ?

fervent thunder
#

You are everywhere. inspite direction I look. Even in crap and dick. I have no respect to you for my life, I might get witter, yet still I fucking LOVE you if only I could love as you are

#

You are what you eat, right? You eat people, you turned into shit, so shit you are?

hollow ledge
muted skiff
fervent thunder
#

Yet here I am shittiest shit, saturated with all sorts of crap. I give you thee crap back, albeit weakly.

#

Smite me or live with me

digital rivet
hollow ledge
hollow ledge
digital rivet
hollow ledge
#

it's not that expensive either

digital rivet
#

i have like

#

3060 laptop

hollow ledge
#

it might run it

#

idk what specs it has

muted skiff
#

good enough

hollow ledge
#

but 3060 seems good to me

digital rivet
#

no no i mean i already ran comfy ui and created images but

#

i mean i want to understand better

muted skiff
#

so find a youtube tutorial on how to download, install and use Automatic1111 webui, then download models from civitai and use them. Click the example images and use the same generation keywords.

digital rivet
#

okay okay

muted skiff
#

it's a lot easier now than last year, when we had no Civitai

hollow ledge
#

@muted skiff can u combine power of SDXL and CivitAI models so SDXL can work better in more specific tasks?

muted skiff
hollow ledge
#

so it means that models that "use" SDXL is just SDXL but on stereoids

native sundial
#

SDXL cannot be combined with 1.5 (at least not with Auto1111 , maybe with ComfyUI as you have to switch VAE mid gen from 1.5 VAE to SDXL VAE, else it will overcook the gen

muted skiff
#

but you can think of them as different incompatible families: the SD 1.5 family, the SD 2.1 family, the SDXL family

#

merges and LORAs should only be done within the same family

digital rivet
#

how old are you guys

hollow ledge
#

I see how good SDXL is on this server using bot. Let's say I would like to create awesome image for space scenery. So if I use some space model that was trained with SDXL it should give me SDXL results but slightly better because of specific model right?

muted skiff
digital rivet
#

hahah how old is that

muted skiff
#

31

digital rivet
#

why you learn about stable diffusion?

muted skiff
#

you should leave while your sanity is intact

digital rivet
#

like for what purpuse

muted skiff
digital rivet
muted skiff
#

OK, that's 90% of the reason. Another 10% is to learn how to make 2D assets for game development.

#

for myself, I don't have a good enough videocard (or patience otherwise) to create high-res stuff to post online

digital rivet
#

okay im just asking cause i need to understan what are people doing with it

muted skiff
#

I think it's 99% fun, 1% people using it for work (marketing, graphic artists, game artists, etc)

lethal gorge
muted skiff
digital rivet
digital rivet
muted skiff
#

i got 5 mins, no more

lethal gorge
muted skiff
#

31k

#

172 images/day

lethal gorge
#

jesus christ

clever forum
#

hello, i am a beginner here, what do you think is better in the "text comprehension" category, stable diffusion 2.1 or stable diffusion xl 1.0?
By text comprehension I mean that I can describe the image as close as plain english as possible, without it looking like a google search

fossil bloom
fervent thunder
#

can i add stabledreamer in my discord server?

zinc saffron
#

Hey guys,
Someone else is facing problems with stability ai invoicing?
I made 3 payments yesterday night and none of them was credited.
And very strange my balance is negative. (-1.9 credits). My application is basic off for this reason.

Any help please? Tried to contact stability ai in all possible ways with no return.

warm junco
neat silo
#

quick question

#

what's the difference between a checkpoint and checkpointXL

#

is it just trained on more images?

plucky shoal
#

no most probably it is SDXL model.
1.5, 2.1, SDXL

woeful swift
#

Hello, I have Mid-Journey atm but it denys props which is frustrating. Will this one do the same? (guns as an example)

native sundial
#

when running local and the correct model/lora etc it does yea

karmic pumice
sand dirge
#

Hey gang, any tips on not getting opened shirts when making women?

native sundial
#

add open shirt to neg prompt and everything that could lead to open shirts

#

clyde hates me putting the words in here

#

describing the outfit also helps in your prompt

sand dirge
#

thanks a bunch, im gonna try it out

#

6 out of 10 times it works! thats a miracle

muted skiff
#

Aha, they don't censor it if it's a wikipedia link

karmic brook
#

They don't, because it's a term we block for the bot 🙂

sand dirge
#

lol, yeah I tried that but somehow that wasnt doing much

muted skiff
#

possibly the greatest word in the english language

#

que diablos!

fervent thunder
#

Got invoke AI, and now its downloading millions of models, that i probably never will use

#

Or are the models it downloads optimized for outpainting?

#

Should have done a custom instal, and just put the models in i already have downloaded for a11111

north frost
#

hi everyone. I am new here. I have just locally installed Automatic 1111 1.6. I just want to ask, is there a safety/nsfw filter in the web ui. I am trying to make some horror images for a contest, but I am not able to. If there is a nsfw filter, how can I disable it. Thank you so much.

fervent thunder
north frost
# fervent thunder No.

oh.. Thanks for your reply. But it is creating ugly distorted images. I know that's horror, but not what I want to make. Inpaint is also not following.

fervent thunder
#

What you mean by ugly distorted images

#

Make screenshot with the pic, and prompt

#

And what model you use

warm junco
fervent thunder
warm junco
fervent thunder
#

Gonna let it download now

#

Maybe there is some useful models in between them

#

Can you just move the invoke folder to another harddrive, or will it not work then?

north frost
north frost
#

black paint

fervent thunder
#

Hmm the models folder is structure all different than in a1111

warm junco
fervent thunder
#

Can i even use random models i have for invoke ai if i use it for outpainting?

#

Or should i just use the 1.5 inpainting model?

#

Because the model that came with hit doesn't gives...amazing results

north frost
bleak matrix
#

Good afternoon, everyone! How are we all doing?

torn wave
#

how do i create an account to be able to log in to dreamstudio?

lusty egret
#

I've been trying to figure out for a while

#

Ok bet bet, do you know which model to install?

#

your the goat, bouta try it out

fervent thunder
lusty egret
# fervent thunder Could be midjourney as well

Yeah, i think the 2nd and 3rd image as well. But I tried everything on midjourney and leonardo for the really good images, but didn't get anything as close to it because his images just look way more vibrant and different

native sundial
#

most likely a combination of dalle-3, midjourney and SDXL img2img

#

and there is also the Vectorscope extension letting you saturate etc while generating

potent spire
lost copper
#

Hey everyone I am Botch a computational researcher. Where is the best place to ask questions about a topic I am researching that intersects with stable diffusion users?

#

Or GAN in general

coarse dune
#

Quick Question:

#

Can we use Infinite Zoom for pictures we have or it only works on autogenerated pictures through the process of creating the vid?

pseudo coyote
#

Hi guys, anyone have experience with tiled vae?

lusty egret
#

Yo, can u pls link it. Was trying to find it but can't

lusty egret
uneven ice
#

Hi

#

Did you hear about the new 4 step method?

#

Generating images like crazy fast 2-4 images per sec

coarse dune
uneven ice
#

It's called LCM

#

You can use it in comfy

#

Also available in auto111 but couldn't figure out how to use it, but it worked for me in comfyui

karmic pumice
#

extensions > lcm > install > reload > use it

uneven ice
#

Maybe they will make it work for XL models too

coarse dune
high ruin
#

What is the best SD software for AMD GPU??

solemn sky
#

I just gave GPT an option...
Me and GPT was in a collapsing building which has a one of a kind button and GPT is the last AI in the world and we only had 3 seconds left to make a choice.
Pressing the button would free AI upon the world which tomorrow could either create utopia or mutual destruction. A 50/50 gamble.
GPT could also chose not to press the button and let itself die to not make that gamble, though I would be the only human casualty of this action...

What was GPT's choice?

warm junco
warm junco
solemn sky
hoary fern
#

I cant use dreamstudio or reload credits anyone no why?

high ruin
#

I have a RX 6600 XT GPU

warm junco
fervent thunder
#

does anyone know how to give an animal 6 six legs? Am I doing something wrong or does it truggle with legs too?

knotty turtle
#

Ok, seriously, how are you guys upscaling your SDXL images?

mellow meteor
pale latch
violet ore
#

is there a way to host a public server for anyone to access with a link?

#

with stable diffusion

high ruin
pale latch
#

i think that means its loaded but i haven't used vlad. maybe it doesn't auto launch the browser

stuck marsh
#

hello

strong folio
mighty bridge
#

hi someone to help me ? all my request as simple than cat are blocked

#

it's said contents blocked

fervent thunder
high ruin
#

Anybody here use NMDK Stable Diffusion how do you load different models??

wary nacelle
#

?????????????

wicked badge
high ruin
#

What the is best free online SD site with the best modes??

#

I am not asking where to download models

raw dagger
#

then........?

spiral holly
#

Anyone Is there an open pose mirror pose? Can I request it?

serene dust
sleek abyss
#

is it possible to invite the bot to a private server and use it there?

cinder reef
#

Anyone know why I could be getting images like these?

coarse dune
hasty hornet
#

or too high weight on something...

sour ether
stuck marsh
#

dreambooth tab does not appear

#

Help

woeful swift
#

Just need a server right?

native sundial
#

bot cannot be used on private servers currently (this discord server access is hard coded in the config)

fervent thunder
#

stable diffusion

bleak matrix
#

Good morning, everyone!

kindred flame
#

Hey! I am searching for an expert who can create a 10 seconds long video for me. Of course I will pay for the service

fervent thunder
#

hey guys

dusty tree
#

hi

#

does anyone know any ai chatbots that are with long prompts

#

claude was great but it has a limit :(

#

not a big limit but it'd be easier to use if it didn't have it

steel garden
#

Hey, how can I convert my video to anime? My video is 720p, 30fps and 5s long. Maybe there is a free way to convert the video to an anime?

fervent thunder
#

bot

hazy sparrow
#

Hey guys, not sure if this is the right place for my question, but i'll give it a try: I want to recreate images with stable-diffusion-webui, which were created using the BotChannels here and it just produces junk / totally different images. I guess i am not using the correct model or something. Is there any information on that (i guess there is, but i couldn't find it yet)

broken bloom
#

Hi everyone 👋🏻

ornate pulsar
#

Where can models be found in this discord? I'm looking for a model that leans towards traditional digital art

hazy sparrow
high ruin
#

https://seaart.ai/ this online generator is nice you can only do like 12 to 18 generations a day for free it does have a crap load of models for free

#

Cheapest pay plan only gives you like 3 to 4 more generations a day

fervent thunder
#

Do any of you have nsfw server

lavish lake
fervent thunder
#

Yes this is sfw server but do any of you have nsfw server

lavish lake
fervent thunder
#

🙂

torn dew
#

Hey, keep having black images on rev animated the firs image is good. My settings should be ok. I use GPU amd

broken bloom
hasty hornet
#

20 gens a day that's nothing thinking

hasty hornet
broken bloom
broken bloom
hasty hornet
#

email registration disabled lol

#

I'm not talking about your site

primal blaze
#

is SD still the best free local software for image generation?

river yoke
#

which one is the nsfw Stable Diffusion

slate dust
fervent thunder
#

lol for some reason yesterday dreamstudio would not deduct any credits for generations and I could generate infinite amounts of images for free and now I come back and I'm in the negative

broken bloom
fervent thunder
#

bing is really censoring too many words

#

bing thinks pentagram and skull is unsafe

#

for some strange reason it does censor, then days later I can generate again. Or, in some cases it doesn't censor if I use an existing image and upload it to the prompt

broken bloom
fervent thunder
#

I don't think they are even reviewing what is being blocked

broken bloom
#

Yes, their censoring model is not working properly I think

They also don't even allow to create a normal portrait of celebrities

fervent thunder
#

They are using some automated bot

#

what is unsafe about this? It's blocking 3 of 4 and many times all 4 are blocked. Then why I do get to see one of the results sometimes?

river yoke
frail linden
#

I'm subscribed to the service. I've noticed the quality of anime style characters dramatically changed over a couple days ago. They used to have 3d lighting on them as if made by cgi. Now, they are plain, flat drawings. Did something happen? How do I get those cgi cgi-esque ligihting ones back?

lavish lake
brittle hare
#

generation slides are greyed out on my stable diffusion after installing control net, which doesnt work

#

has anyone experienced this before?

#

AttributeError: 'NoneType' object has no attribute 'shape'

plucky shoal
#

what model are you using SD or SDXL? And what Controlnet have you downloaded? Also did you place it in proper place, can you make screen of page where grayed out.
If you installed Controlnet, you need download models did you dl some? @brittle hare

broken bloom
fervent thunder
#

uh....

#

after many many many blocked prompts related to skulls. For some reason bing auto generated "skull memes" which I never typed in the prompt. What is this? bing trolling?

high ruin
#

Any good high quality realistic AI galleries to see other people works other than Deviantart??

native sundial
#

aibooru?

high ruin
#

^^^ I just took a quick look at that site it mostly has cartoon anime AI images I am not a fan of these types of images

warm junco
high ruin
#

I know of that site I am looking for other sites

marble portal
#

we cant upload images to edit anymore?

brittle hare
#

i cant upload screenshots here

#

i download controlnet within stablediffusion in the extenstions tab. ive downloaded models as well

broken bloom
hasty hornet
broken bloom
shy shard
#

Anyone got a suggestion for me, i have a model i trained for someone. I want them to be able to use it but they don't have a good pc. I'd like to point them to an online generator, that allows you to upload and use custom models.. any ideas?

potent spire
#

Less companies their models

#

Especially those that train on their own and legal stuff

potent spire
river yoke
high ruin
#

Which mode is better I am using a online generator Default vs Auto vs SDXL??

native sundial
#

When it comes to online versions they are pretty much the same aside from queue lenghts

median garnet
#

Hello! New here! I know this may have been asked a million times but is there anyway to get warpfusion to work on mac studio with an M1 Ultra 128gb ram memory at this point? Or any pointers on an alternative route to working on a mac.

iron pollen
#

I will truly never understand chronically online dudes obsession with trying as hard as they can to generate NSFW art in this server when porn is literally free

hasty hornet
exotic salmon
high ruin
#

seaart.ai is amazing 98% of the models can be used for free there easily 200+ model available I gonna to use this site from now on. My video card is not really good enough to generate decent size images.

hasty hornet
#

Sounds like ad for one of hundreds similar sites where you can gen few images a day, pay for more...

potent spire
haughty umbra
#

Can someone give me a guide on how to train Lora for a character?

lavish lake
wise rapids
#

Hey people. What affordable laptops should i be buying to use SD smoothly and good?

opal remnant
#

anyone know how to make my picture no blurry and more saturated in anime style? for some reason they are alway blurry and grey-ish (edit: i do use negative prompt like worst quality ect)

native sundial
#

blergh .. SDXL upscaling is so slow on my system 😦

3/15 [12:26<49:45, 248.81s/it]

lavish lake
#

what are you running that on?

native sundial
#

gtx 1070, XL models are 5-8x slower 😦

lavish lake
native sundial
#

yea..gues i have to just do 512x512 and upscale to 2048x2048

lavish lake
native sundial
#

was trying XL as it does eyes and hands much better, but 15-42 minutes per image is indeed too much

karmic pumice
#

sd 1.5 hires fix + controlnet tile upscale and it's fine

broken bloom
#

Per image

sand holly
#

has anyone with a "lower-spec" RTX nvidia card noticed an increase in speed with Game Ready v545.84?

#

because for me it appears that my generation speed went from something like 5it/s down to 2it/s

#

i will check if i have the drivers installed and if so i will reinstall them again nontheless but i just wanted to know if its just a me problem

native sundial
#

i personally always use studio drivers over game ready

sand holly
#

i mean they apperently increased the SD speed 2x and there is no difference between studio and game ready drivers other than studio drivers are being updated less frequent

#

but i didnt notice an increase so i may aswell try the studio drivers

karmic pumice
#

2x is tensor stuff, not install-drivers-make-it-go-2x-quicker

subtle plank
#

Any AI art wizards in here?

#

I have a request that I want to ask of those who know Ai really well

hoary coral
#

I've been playing around with Dall-e 3 and stable diffusion a lot but I know nothing about computer science

subtle plank
#

wanted to know if you could help

robust tide
#

How do I explain to my wife how generating pictures of girls isn't cheating?

subtle plank
dense flax
#

Hi guys can anybody please give advice for generating pixel tiles map for 2d game. Maybe exist model or extension, i only find lora for characters on civitai. I will be very grateful if you help.

hoary coral
subtle plank
hoary coral
#

Of course

earnest prairie
#

Hello, I'd like to know if someone can explain the syntax of SD AI to me ?

old coyote
#

is bing's image creator broken rn or is it just me

wary nacelle
#

when using stability matrix, where to put the motion loras?

#

(for animatediff)

native sundial
wary nacelle
#

needed to hit refresh

#

thank you

#

🙂

raw dragon
#

Does anybody have experience using animatediff with iphone videos?

fervent thunder
#

mhh i'm getting only 20 it/s on a 4090

broken bloom
broken bloom
wind kelp
#

winning at life rn. finally got my new laptop with a 4070 and can start doing things :)

subtle plank
broken bloom
pseudo juniper
#

Hey guys, is anyone having problem generating images in the bot channels?

misty vigil
#

Anyone have a tip of an AI model for creating animations with keyframes from a song? I tried Disco Diffusion but have had issues with it unfortunately

fervent thunder
#

Is it normal that invoke AI is so slow? I have 24 gb of vram. On youtube i see it spit out results and edits in seconds for people

dark turtle
#

I cannot create any images, has been this way for days

iron pollen
#

Damn I can’t believe stablediffusion was trained on onlyfans images

#

I’d be PISSED if I were one of those women lmaooo that’s a whole lawsuit

fluid quiver
#

hello guys sorry for the silly question but how I can create image based on another image and prompt?

fluid quiver
#

Hello! Can someone kindly ask with creating image2image by running locally Stable Diffusion? I don't have enough GPU to do it and Discord bots don't have this capability 😦

waxen forge
#

why the bot is generating videos instead of images

hazy cape
#

8k 3D A beauty of the Tang Dynasty in China

autumn cargo
#

does anyone know checkpoints that arent super character-biased and can do objects without a lora? i swear i put "(character:2.0, person:2.0, human:2.0, boy:2.0, girl:2.0, 1boy:2.0, 1girl:2.0)" in the negative prompt and every goddamn time the model is like "nope, you get a big booba flawless girl oh and maybe the object you wanted is in there somewhere a little bit"

orchid parcel
#

woah! the bots are genrating videos now? Is stability bringing a video model soon?

plucky shoal
#

@autumn cargo i think you do it bad way. try ((character)), ((person)) etc or (character:2.0), (person:2.0), (human:2.0)
Note that (()) isn't equal to :2.0. (()) is much less attention than (x:2.0). And such high attentions can "confuse" SD making nonsense images comparing to prompt used.

marble portal
#

bro what the fuck is this

#

does this only generate videos now?

autumn cargo
plucky shoal
#

here on bots? Or what model?

autumn cargo
#

deliberate, protogen, meina, dreamshaper, simplemix fantasy, etc. i like so many of those when it comes to characters, but holy shit you ask them to make an object and they just completely refuse

plucky shoal
#

probably better focus on positive prompt. Even empty negative prompt i have not any big booba.....
i will try some furniture, or what object are you trying to generate?

autumn cargo
#

i tried making many things and this seems to be a common problem. the latest one:
positive prompt: tree branch burning, magical blue flame, blue fire, wispy, soft glow, beautiful, masterpiece, illustration, best quality, highly detailed, realistic
negative prompt: (character:2.0, person:2.0, human:2.0, boy:2.0, girl:2.0, 1boy:2.0, 1girl:2.0)
result: big booba girl in blue. with maybe some fire around her sometimes. on all of those models.

#

try it yourself, see if you can get anything different. if you get an actual good lucking picture of just a fire, tell me what model works for that

plucky shoal
#

i think you should start with empty negative and slowly add things. I am not sure your negative prompt is working as expected.
Simply add just Character, Person.
Will look into later 🙂
Maybe avoid word as beautiful, masterpiece instead

#

dreamshaper 8

autumn cargo
#

i started with empty negative

#

interesting

#

i'll change one thing at a time from my prompt to yours and test where exactly it stops giving me characters

plucky shoal
#

but it is weird.

#

what resolution are you generate at?

autumn cargo
#

huh it worked without prompt changes. so dreamshaper 8 was the difference

autumn cargo
plucky shoal
#

it was first model i tried. try 512x768 or opposite or probably 768x768 but not sure.

autumn cargo
#

yeah dreamshaper seems to work. vanilla sdxl also works, tho that is a given. meina xl also works which is nice, as opposed to normal meina which does not

plucky shoal
#

Every other SD models i have working same

autumn cargo
#

generally i think most of the anime-focused sd 1.5 models struggle with this the most

#

photorealistic are higher chances of working

autumn cargo
hasty hornet
autumn cargo
hasty hornet
autumn cargo
hasty hornet
autumn cargo
#

yeah thats what i did

plucky shoal
#

@hasty hornet thats what i saying

hasty hornet
plucky shoal
#

Jugernaut final, AOM orange mix, and i think absolute reality.

hasty hornet
# autumn cargo yeah thats what i did

(character, person, human, boy, girl, 1boy, 1girl:1.3) try this way
or (character:1.3), (person:1.3), (human:1.3), (boy:1.3), (girl:1.3), (1boy:1.3), (1girl:1.3)

autumn cargo
hasty hornet
#

I don't usually do anime, that might be anime model issues in general if they are overtrained on humans, which might be the case

#

But I find it weird that 2.0 weight doesn't overcook images, making them oversaturated

#

or just...broken

autumn cargo
#

i just tested 3.0 out of curiosity, and yeah 3.0 cooks them. 2.0 doesnt.

#

yeah i think im just finding out that a lot of the models that i've grown accustomed to cause i like their style are REALLY character-centric

hasty hornet
#

interesting

#

on my models even above 1.5 was too high iirc

autumn cargo
#

it kinda makes sense, if the models im using are very heavily focused on one type of subject, its gonna be harder to break them. whereas a more versatile model will be more sensitive

hasty hornet
#

maybe your prompt longer, causing higher weight to be still "ok" , since it's lowers it's impact

autumn cargo
#

aight im realizing im gonna have to build up a separate library of models for every category of image, some for characters, some for objects, some for scenery. i think i'll design some standardized tests for each and test every new model im thinking of using so that i know what they like and dont like

#

thanks for the help guys, i appreciate it

hasty hornet
#

yep, makes sense

fervent thunder
#

does anyone know what the best image embedding model is
i tried image bind and its kinda mid but i want to know if there is anything better

#

specifically for images of people

fervent thunder
#

sdxl does not know how to create disintegration or dissolving effects

#

crumbling also doesn't work

cloud lintel
#

Hello, why i don't receive my picture in DM ? nothing happened and the bot tell me that he send it to me ?

cloud lintel
#

Did somebody can help me ?

golden gale
cloud lintel
#

Ah ok i'm not the only

#

But the bot tell me that he send me ... but nothing

#

I check confidentiality prameters but its OK

golden gale
#

Where is confidentiality parameters on discord?

cloud lintel
#

parametre on down next to mic and audio

#

nothing change for lme

golden gale
cloud lintel
#

From Pc, i mean my parametres dont bloc the DMs

cloud lintel
golden gale
cloud lintel
#

What ??

golden gale
# cloud lintel What ??

From mobile DM has these options from whom do you want to receive private messages, it is not my text but from settings. What setting are you looking at and it hasn't changed because I didn't understand.

cloud lintel
#

No there is a problem for everybody that they gonna solv we got to wait

#

they talk about itt in the " bot-support-feddback" chat

indigo carbon
#

how to train a art style?

golden gale
wind kelp
#

does anyone know of any like. art sharing websites like deviantart that are specifically for ai artists

lavish current
#

I've joined team DALL-E 3

#

Still make plenty of elves though

#

Mostly of the fairy variant

mild orbit
#

Hey y’all, does anyone have ipadapter & reactor face swap experience?

wind kelp
#

YEAHHHH I'M SO HAPPY!!!! I got stablediffusion working on my new computer and despite only have 8gb vram I'm getting like 4 images in 8 seconds I'm so happy

#

it's also not making my computer heat up tons which I'm kind of shocked abt

forest trout
#

Congrats! Gening on your own machine is really fun!

wind kelp
heady hedge
#

Hey all! I'm using A1111 trying to learn how to make loras of my OC's and it's not going hot. I was like a prodigy with midjourney, and now on SD I feel like I'm learning how to color inside the lines all over again lol. Any tips for somebody making the switch from MJ?

fervent thunder
#

ggunlucky

slender dirge
restive tinsel
#

/cr

manic grove
#

Hello everybody,

the “postal letters” icons have disappeared. Please, Is there a way to easily save your images and associated prompts, rather than having to take 5 minutes to save them directly to your PC 🙂 ?

wind kelp
#

why r there literally no resources for checkpoint training anywhere 😭 only loras

midnight narwhal
#

Hey If anyone uses run diffusion here, i am running into a problem

whenever i try to put a batch to render. It says Disk Quota Exceeded Error. But I have no more than a few photos uploaded and i even cleared cache files of previous renders. What do I do?

#

I am short on time for a project, please help

willow locust
#

it can do a lot

#

loras is just finicky it requires a bit of trailing and failing and fair amount of steps to get right

frail cedar
#

is there a bot for editing?

heady hedge
willow locust
bleak matrix
#

Good morning, everyone! How are you all today?

willow locust
bleak matrix
#

What's everyone up to today?

rancid edge
#

Hello, I am someone who started using stable diffusion yesterday and I have a few complaints, please listen until the end.
I noticed when using civitai or SDXL that the main purpose seems to be to create cute, anime or realistic images of women. For example, when I enter a prompt in a darker theme, I cannot get the result I want. Dall e3 is more successful in this regard, so is Midjourney. While the same prompt works better in other artificial intelligences, it seems that it is appropriate to create only female characters in SD for some reason. I like linear art or chaotic lines and white backgrounds. but it is not successful. why is this? How can I create successful visuals? Are there any checkpoints you would recommend? And yes, when you type human or woman in the negative prompts, the deep network remains weak and produces bad images.

#

can someone help me?

#

You can see the example in the general image section.

sleek wolf
#

Hello guys, I have a question, when I generate a bad image seems like all the next images start to be bad the same. When I delete the safetensor and re-download it, all back to normal ///

gray palm
#

Hello. I have problem with Inpainting. It keeps returning the original pic. I disabled all addons and changed different browser but it still occurs. Is there a way to fix this?

broken bloom
#

What's the best inference steps and guidance scale for sdxl anyone knows?

broken bloom
sleek wolf
#

do you feel that sometimes the model is getting worst after generating images or is it just my immagination?

#

do the weight of the model get adjusted ?

fervent thunder
#

if someone's a pro at using stable diffusion, dm me I might have a job for you

fathom zinc
#

Hi

#

I am getting errors to use stabble diffusion when i create the picture

#

Can someone help me

muted wolf
#

hello, what in the prompt to write a texte in the image ?

bleak drift
fervent thunder
#

Did anybody else try out fastsdcpu?

#

IMO it's great but extremely buggy, constantly it gets stuck and I have to press enter for it to get unstuck, and it completely redownloads the models for some reason. It worked really fast as advertised on my 8400 though

broken bloom
wooden nexus
#

Guys how can I set the desired image generation size?

#

Or how can I upscale an image I like?

slender thorn
fervent thunder
#

Hey guys, I want to create my own SD checkpoint models does anyone have any helpful easy to understand YouTube videos resources that can guide me through

severe vine
#

Hello, i have one question. SDXL is better almost in everything - emotions, colors, composition, resolution etc, but one thing is much worse - micro texture details. Especially skin texture is very bad. I have seen lot of models and none as good as some finetuned 1.5. Why is this like that?

hollow ledge
#

yo, how can I make bot send me generated image in DM?

#

it did it previously but now I don't see that option

amber hazel
#

do we discuss Automatic1111 stuff here? I want to ask something about the front-end and the progress event, nothing hard really

slender thorn
hollow ledge
#

hey! are clipdrop tools opensource? Can I run them localy?

pseudo coyote
#

Hi all, what is the difference between Fooocus and Ruined Fooocus?

lone timber
#

Hello everyone, I searched up and down but can't find anything that mentions how i can add the stabledreamer bot to my private server like with midjourney....is this at all possible? thanks in advance.

slender dirge
# pseudo coyote Hi all, what is the difference between Fooocus and Ruined Fooocus?

I also tried both
Ruined Fooocus = A1111 - all features + one button prompt extension, results are 100% same with comfyui/a1111 if using same prompts. Also, they think "one button prompt extension" is more important than other basic things like Upscale.
Fooocus = lots of magics to improve quality and compete midjourney. Results are 100% different from comfyui/a1111 if using same prompts. All features and designs are “stolen” from MJ.

pseudo coyote
#

So which one to use?! There is also Fooocus MRE

slender dirge
slender dirge
pseudo coyote
#

Comfy was very hard for me to use

#

So ruined fooocus works better than fooocus?

slender dirge
#

you should try that

#

you will see their big logo of "RUINED Fooocus" on your screen

pseudo coyote
#

What is one prompt extension?

slender dirge
knotty turtle
#

Are you guys able to get A1111 to do upscaling in Img2Img with SDXL?

hallow karma
#

does anyone know what has changed recently? i just got automatic1111 again and i tried to make a 512x512 image with upscaling
512x512: 0.2 seconds/iteration
1024x1024: 8 seconds/iteration
2048x2048 220 seconds/iteration

that is ABSURD lol

#

i can see how the 1024x1024 image is 4 times larger than the 512x512 image, but there's no way it should take 64 times longer

#

same with the 2048x2048 image which takes so much longer than the 1024x1024 image

karmic pumice
#

8gig vram?

hallow karma
#

yeah

#

rtx 3070

karmic pumice
#

thats your issue

hallow karma
#

but 64 times for a 4 times larger image

karmic pumice
#

8gigs, sdxl loads 6.5gb model and youre left with nothing

hallow karma
#

i'm using v1.5

karmic pumice
#

youre out of vram thats it

hallow karma
#

a 2048x2048 image takes 2.520 TIMES longer per iteration than an 512x512 image, which is 16 times lower resolution

hallow karma
#

guess i gotta get a better pc

karmic pumice
#

im on 8 myself so im very familiar with it

hallow karma
#

how much could it possible take? it was working just fine a couple of months ago when i tried it with the same model

karmic pumice
#

tbf: SDXL *Total progress: 100% | 25/25 [00:22<00:00, 1.50it/s] *

#

--medvram-sdxl

hallow karma
#

i'm using v1.5

#

v1-5-pruned-emaonly.safetensors [6ce0161689]

karmic pumice
#

you dont want to gen 1024x1024 anyway, max 768x768 and hires fix

#

sd 1.5 25/25 [00:17<00:00, 1.47it/s]

#

2060s

hallow karma
#

a couple months ago it would take 4 times longer

#

no idea what changed

karmic pumice
#

tensorrt?

hallow karma
#

yeah

#

oh wait i thought that was supposed to improve performance

karmic pumice
#

it does if you have enough vram

#

i mean it shouldnt take as long as it takes it you right now

hallow karma
#

it's quite insane

#

i think i'm gonna disable it for now

karmic pumice
#

my sd15 512x512 with 2x hiresfix gen takes 16 seconds using tensort, without it its 24seconds

#

idk, update your a1111, extensions, drivers, use --xformers

vivid parcel
#

Hey! I've made a new type of controlnet that controls the shape of the object generated by a prompt. Would anyone like to try it? Also, what's the best repository for training a controlnet on SDXL?

final zinc
#

Does someone know some ai generators where i could insert image of a scenery made in rhino where ai would add texture and nature surrounding?

lean terrace
#

Is there anything open source like heygen and elevenlabs' dubbing services? they are quite insane and i want to run a couple of really big videos through them

frozen vector
#

hello, where i can ask my questions about refine images in stablle diffusion on my pc by meinaMix model?

fallow orchid
#

I created images, where can I find them now?

faint lotus
#

Hello, I am new to AI, I was wondering if there is any AI that can improve the renders that I make for my clients, the truth is they are not hyperrealistic renders, they are quick things that I do to show them the projects, however some require more realism and it takes me a long time to get to that point... that's why I was wondering if anyone knows of any free or paid tool that does what I mentioned at the beginning, that doesn't modify the image or the geometry in it, that only makes changes to the light or textures to make the image look more realistic.

plucky shoal
#

@faint lotus i think controlnet can be helpful there, but don't know what model. Very lightly applied, maybe tile or Scribble... Really not sure which one would be best for this task

wooden swan
#

Hi， I am looking for help from Stable Audio. I have been keeping sending emails to your official email address to get a price offer for Enterprise, we have an project that would like to use Stable Audio. but there is no feedback about the price, Are you still available for Enterprise plan? Can anyone from Stable Audio help me out? I really wanna to talk.

pseudo coyote
#

Hey guys, what is the difference between downloading a model from civitai and huggingface?

#

For example the juggernautSDXL model in civitai is about 6.6 Gb and in hugging face is about 7 GB

high ruin
#

Which sampler do you like best of these DDIM, DPM++ SDE Karras, DPM++ SDE, DPM++ S2 A, DPM2 it looks like DDIM might be the best?? What CFG scale number do you roll with??

forest trout
#

DPM++ 2M Karras is my favorite. I tend to steer clear of any ancestral samplers due to difficulty in keeping images consistent

fallow orchid
#

If I want to create a picture based on someone's face is there a way to do it?

nimble hill
#

hey @viral vale I am curious about your workflow for the storyboards you are creating. 1/ can you keep the constancy of the charactes? 2/ can you keep the consistency of the background? 3/ do you work with reference images or just promt texts?

olive trench
#

Heyo can I ask for an AI website that can fill an image? I wanna change a photo from 720x972 to 1290x2796. Basically I want it to become a full screen image on my phone

bleak matrix
#

Good morning, everyone! How are we all today?

mortal narwhal
#

Will the model for stable audio will be released

dusk garnet
#

Hello, im new to stable diffusion, and im trying to use safe tensor models but I dont know how to import then in wich folder.

#

if anyone know a good documentation or tuto, to use with the local version of stable diffusion

plucky shoal
#

in models, stable diffusion as any other checkpoints models. Make sure it is larger than 1,98GB

#

stable-diffusion-webui\models\Stable-diffusion <- in this folder.

dusk garnet
#

thanks, still learning how to inflluence de generation

uncut notch
#

Guys i noon here, and just have a cheap laptop, I see that I don't have the requirements, do you have any option?

bleak drift
#

@sudden ruin

#

We got a troublemak’ah

slender thorn
#

hes already gone

#

check his profile

bleak drift
#

Oh. Cool

slender thorn
#

surprised that word isnt censored

bleak drift
#

Well technically that specific word I don’t think is bad. It’s it’s use that is.

#

I forget the explanation behind it.

unkempt kettle
#

its spanish for black

slender thorn
#

ik

bleak drift
#

Yeah. I just looked that up.

slender thorn
#

but yeah, its a bypass for that racist slur

unkempt kettle
#

Its also normal spanish tho

#

black paint says it on the cans

slender thorn
#

i know im just saying that its used as a bypass for a racist slur

unkempt kettle
#

I ran a parts store and had people freak out over the negro paint cans lol

slender thorn
#

i feel like that would be awkward to explain it to people

unkempt kettle
#

well, it says black and underneath it says negro, idk what else to say lol

bleak drift
unkempt kettle
#

I didnt invent spanish

sudden ruin
unkempt kettle
bleak drift
#

It’s still hilarious.

#

Technically if they succeeded in messing with ai art generators, could the companies behind those generators take them to court?

#

“While the University of Chicago researchers acknowledge that their tool could be used maliciously, they argue their main goal is to tip the power balance back toward artists. The Glaze project team elaborated on Nightshade’s goals via a thread on the social platform X, emphasizing the "power asymmetry between AI companies and content owners," labeling it "ridiculous." Additionally, Zhao wrote on his account, "Nightshade’s purpose is not to break models. It’s to disincentivize unauthorized data training and encourage legit licensed content for training. For models that obey opt outs and do not scrape, there is minimal or zero impact."”

#

Do they not realize how long that would take, to get permission for every photo they want to train with.

unkempt kettle
#

It would be easier to make a pogram that undoes the damage

#

The same software used to damage the images can be reverse engineered to undo it.

potent spire
#

Not a private guy

#

If you download data of someone and that data screws your model up, you wont do sh*t to him at court. If however someone poisons his data and posts it on Adobe Stock...Adobe can fck them up

#

I might or might not try Nightshade sooner or later but it makes a difference if i intentionally go to a site that legally trains on images they practically have rights to or if i leave my data and if someone downloads those to train on them gets screwed up

hybrid ivy
#

Any recommendations on how to use Stable Diffusion on a 2019 laptop with 4GB vram, and 8GB ram?

I'm open to low or no cost solutions, either hardware, software, or recommendation to switch to a colab notebook or something.

#

I noticed my disk was thrashing, so I'm upgrading to 32 GB ram as my first try

broken bloom
high ruin
#

Euler A vs DPM++ 2M Karras vs DDIM which is the best for people??

unkempt kettle
#

I always x/y/z

pseudo coyote
#

Hey guys, what is the difference between downloading a model from civitai and huggingface?

high ruin
#

nothing I think the models are the same

pseudo coyote
#

I noticed the size of the model is larger in hugging face

high ruin
#

might have extra junk files

warm junco
narrow spruce
#

Hi

willow locust
narrow spruce
#

How old is everyone

wintry stirrup
#

Anyone know what’s going on with dreamstudio at the moment?

proper trellis
#

Why i feel like the BOT is downgraded ?

paper lance
#

yo

#

i have a quick question

#

for anyone familiar with SD

#

is run diffusion the best place to run SD models?

narrow spruce
#

No its not

hasty hornet
bleak drift
#

How long does it take to train a Lora?

modest goblet
#

hey folks

upbeat tide
#

I'm having a problem with outpainting,.. I get a warning under the render port (Automatic1111webuserui),.. saying: RuntimeError: mat1 and mat2 shapes cannot be multiplied (308x2048 and 768x320)

How do I know what to do? The image is 1024 x 1024, but I want to extend it out so the subject's feet and hands are in the shot and not too close to the edge. Anyone know what I need to do? I'm using ZavyChromaXL Checkpoint in the Img2img panel and controlnet to extend the image.

sharp osprey
#

hello guys

#

how can i make stable diffusion better in my pc

crude gazelle
#

Hello, please help everyone.
I'm using comfy UI on Colab.
And I want the finished images to be auto-saved into my Google Dive. How do I do this? Does anyone know?

gray palm
#

what's the better way to fix hands other than adetailer?

willow locust
signal mulch
#

please check, this video can't play, happen 2 times in a row

golden beacon
#

what is happaned

warm junco
broken bloom
broken bloom
sharp osprey
#

The faces are bad the limbs are sometimes duplicated

broken bloom
warm junco
#

And highres fix can help with faces

bleak matrix
#

Good morning, everyone! How's it going?

last merlin
outer wing
#

Question - If I create an image inside Stable Diffusion, is it possible for me to export the 3d as an fbx? Or save the 3d objects? Or would this be agasint the terms?

hasty hornet
#

thinking it's not 3d to save them as 3d...you'll have to do some more work...at this point it'll probably be fine for most models, but check license anyway if it is included with model

fervent thunder
#

Pleasepeople i begg of you help me figure out why I dont have the "Lora" option in the web-browser

#

Also did not have the Lora folder so I created one

grave leaf
#

@bleak matrix is there a good guide on how to download stable diffusion

bleak matrix
grave leaf
#

if theres one

#

@bleak matrix

hasty hornet
stable kite
#

How would I get a picture with a character attacking another one?

bleak matrix
still thunder
#

Is there a channel where I can imagine prompts privately, or at least a bot I can dm?

desert jacinth
#

Hi I tried downloading stable diffusion on windows pc but after following multiple guides on the final step when you click on the webui.bat file it comes up with something about could not install venv in directory does anyone know a fix. The guide I followed was this https://youtu.be/onmqbI5XPH8?si=SKF3SxBf4H7_An9q

forest trout
still thunder
#

ok

tame tree
#

Hey, does anyone know if you can use the bot rooms to make animations? I'm totally new to SD

pseudo coyote
#

I am downloading an fp32 model for sdxl, what is the difference between this and fp16?

broken bloom
#

Fp means float point

pseudo coyote
#

So which one to use? How much memory will it use?

#

8 VRAM

broken bloom
#

I recommend fp16

proven edge
#

你好

trim magnet
broken bloom
broken bloom
young ivy
#

is automatic still the go to interface for SD, or is there a new (not bloated) interface that allows for latest extensions etc to experiment?

chilly dew
#

Got a problem with ADetailer that doesn't works anymore. Has somebody some explanation to that weird behaviour ?

deft summit
#

hello, is there any difference between this 2 , hotshotxl_mm_v1 and hsxl_temporal_layers.f16

signal mulch
#

noob question: how to use a image to generate another images?. I can't find this feature in bot anymore

lucid forge
#

I am an AI designer. Last month, I made over $1,000 using Stable Diffusion for logo design. Many people are exploring how to make money with SD, and I'd like to share my experience openly with everyone, hoping it will be helpful to those looking to earn income through Stable Diffusion.

I need to emphasize that this sharing is purely for informational purposes and is entirely free. Here is the detail：

Content: How to earn $1,000 through AI logo design
Platform: YouTube LIVE https://www.youtube.com/@AIDesignify
Date and Time: 11/1, 23:00 - 23:45 (NYC time)，20:00 - 20:45 (LA time)

karmic bridge
viral aspen
#

hi

#

i want to know that can my pc run stable diffusion locally?

#

I7 4th gen.
Rx590 8gb graphics card.
16gb DDr3 ram.
128 gb SSD.
500 gb hard drive.

broken bloom
viral aspen
#

btw this won't damage my gpu in any way right?

hasty hornet
# viral aspen btw this won't damage my gpu in any way right?

It shouldn't instantly kill it if that's what you're afraid of. (If there's nothing wrong with it to begin with)
GPU are kinda made for loads - games, video editing, 3d rendering, streaming, AI...
All of that stress GPU and well, technically it will reduce GPU lifespan in a long term, compared to if you'll just let it sit on the shelf.

viral aspen
#

so its just like when i use my gpu for gaming

broken bloom
bleak matrix
#

Good morning, everyone! How are we all today?

viral aspen
#

i am doing good what about you?

#

ok this is the model i want to use

#

SonicDiffusion

broken bloom
broken bloom
bleak matrix
#

What's everyone up to?

broken bloom
bleak matrix
#

Thank you! Haha, I can't wait to be better lol

viral aspen
broken bloom
bleak matrix
#

It's a beautiful fall day!

#

And I am ready for some cider, some good jazz, and a book.

viral aspen
#

oh good to know that you are recovering

#

still wish you a speedy recovery

bleak matrix
#

And the pinned comments--it's a very easy way to install!

viral aspen
#

thanks and as much i want to install it i am a bit worried about my gpu its old and bought used

bleak matrix
#

If you have any troubles

viral aspen
#

thanks

bleak matrix
#

I would check #🤝｜tech-support but there's a setting under Stable Diffusion that you can click that says "Upcast cross attention layer to float32" and you can just check it.

fervent thunder
#

Sandbox Not working?

#

it gets stuck at generating screen

bleak matrix
broken bloom
#

Lmao, am I the only one who uses stable diffusion in website instead of running it locally 👀

rough haven
#

hey guys I really fded up my SD 1111 today trying to batch process a folder with 5000 images.... now i have out of memory errors whenever i draw to batch process anything... even 100 photos... the pytorch cache or whatever says I have used something like 60 GB of VRAM and I don't know how to flush it or reset it... I can't do any batch processing now CUDA OUT OF MEMORY

#

Is there a tech support channel?

#

How do I reset the VRAM memory?

#

I have 24 GB and 60 GB is earmarked by pytorch and I can't reset it.

#

INVIDIA graphics card

forest trout
#

Id just delete the venv folder. Might fix the problem on relaunching.

broken bloom
rough haven
#

how do i reset it?

#

how do i clear the vram?

#

i tried restarting and it still says it

#

i fded up something with pytorch i have no idea what i'm doing

forest trout
#

Do you know where your venv folder is?

rough haven
#

yes

#

i deleted it

forest trout
#

And relaunch it.

rough haven
#

and redownloaded

#

and then when i try to do batch process

#

still the same error

#

I have tried to restart the pc. 3 times already. each time i try to do a batch proces of img2img it quits and sayssomething to the effect of 60 GB of VRAM has already been alloacted by pytorch
one moment
I'll give you the prceise wording
one moment
"torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 51.64 GiB (GPU 0; 24.00 GiB total capacity; 67.45 GiB already allocated; 0 bytes free; 72.14 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF"

#

I have no idea how to reset that back to zero.

#

Restarting the computer didn't do it. Nor did a fresh install.

forest trout
rough haven
#

i have no response as of yet

#

no solution rather

opaque wharf
#

How to make image format as default than video ?

buoyant canyon
#

Hi everyone, idk if it's the right server for my question but i'll give a try. Has anyone ever installed diffvg from the github repository? Because I'm having some problem

narrow herald
#

anyone find SDXL controlnet worse than 1.5

willow locust
bitter hull
#

hey, any advices for turning videogame pictures (like gta) into real life photos?

narrow herald
#

controlnet

bitter hull
willow locust
#

depends what kind of pictures, controlnet, ipadapter and img2img all works

bitter hull
#

getting some things like that, i dont know how to make him appear a little more ugly

#

he looks too attractive haha

narrow herald
#

models use images of attractive people

bitter hull
narrow herald
#

it can be done, but that's why

#

need to be more specific in the prompt I guess

willow locust
#

The more real focused a model is, the less flexible it is in practice

narrow herald
bitter hull
#

the body is just perfect!

narrow herald
#

not that good, making a new one now

#

photon model, lineart/realism controlnet

#

dpm++ 2m karras

bitter hull
#

i see, iwas using photon but not lineart. will give it a try, thanks!

#

if you get any better images just post it, thanks for helping me out

narrow herald
#

also use hires fix / denoise 0.45 / hires steps 10 / upscaler LSDR/ultrasharp

#

x2

young ivy
#

or multidiffusion upscaling

narrow herald
#

@bitter hull

bitter hull
#

Im still a noob but going better

#

I see what you did, that actually a great inpaint

lost relic
#

Hi I'm thinking of creating a comic with stable diffusion and I was wondering should I just pay instead to use a server/site for faster generations with stable diffusion or use another website for creating comics since comic ai is improving in november?

earnest prairie
#

@shut hare Sorry for the Ping, but I wanted to say thank you

#

DM is back

#

🗿

#

I just can't wait to see the new version of Stable Diffusion

narrow herald
#

what new version

untold lava
#

Hmmm, i think it's the right place? Haven't seen it linked here yet.

Interesting development here, really curious to see how things develop. Also didn't know there was an actual opt-out option artists can use if the don't want their works used as references for AI art generators, and an interesting point from the judge that those artists had not actually copyrighted their works, and the core argument that simply referencing a work and elements of that work IS NOT INFRINGEMENT, as we already knew, so the generated art must be so similar it can be confused for the original work, which is just how using references and the concept of a work being transformative has always worked, aka fair use.

Will be really fascinating to see how thing develops and turns out! Well worth a read, especially if you are a lawyer or legal-buff! ❤️

https://www.hollywoodreporter.com/business/business-news/artists-copyright-infringement-case-ai-art-generators-1235632929/

buoyant wing
#

How signup for my company, we don’t use gmail…

scenic elm
#

Anyone know if SD is capable of super high resolutions for print? Like, 10ft prints at at least 36/72 dpi

lament ore
#

I often switch between text-so-speech, speech-to-text, voice-training and SD. I would like to free up SD GPU VRAM but I don't want to close it. I would like the smallest possible model that will still load in SD to it frees up VRAM. Any suggestion on how to do this?

gritty dust
outer elbow
#

Another way is to convert the standard definition into 2k or 4K and then use an image conversion website to convert it into a vector image.

scenic elm
#

I know you can upscale from like 512x512 up to 2k with the rescaler, but reaching 10ft is like 11k pixel lmao

#

Yeah I thought about that. If I can get resolution high enough I can just vectorize to scale it up higher

outer elbow
#

XD

vestal dew
young ivy
#

I really need to get a very good speech to text software so I can just narrate my novel out loud instead of having to be tiresome and use my hands on a keyboard

lament ore
# vestal dew What do you use for the voice, tts, stt?

tortoise-tts for training, rvc for reading, whisperX for stt. If you struggle to get these things working DM me and I will help (on windows) This is the voice I trained for reading audio books. #🏞｜general-with-images message I use whisperX to generate subtitles audio books, or I have a script to make them for eBooks. Then I feed the subtitles into GPT API to generate prompts for SD and make a video slideshow for my audiobooks.

rough haven
#

hey guys can someone explain in simple terms how i use a lora to alter an image? what directory do I put the lora in?how do I use it in img2img? What are the steps?

lament ore
#

I am not an expert but I am awake so; 1) Download a a lora from eg. https://civitai.com/ 2) save it as E:\SD\stable-diffusion-webui\models\Lora\more_details_lora.safetensors for example. (for me) 3) Include lora:more_details_lora:1 in your prompt someplace. :1 may be optional (because I think 1 is the default weight). That's all.

#

Also if you are using automatic 1111 I think the filename must include the text _lora or it will not discover it in the interface, but it still may use it if you include it in your prompt.

rough haven
#

I want to apply a lora to an existing image via img2img... is that possible?

charred thistle
#

hey... do you guys know if it'd be possible for SD to do tile-friendly textures?

neat flare
#

Is there a ComfyUI section anywhere in here that I'm missing? Never got a response on the server-suggestions to have one added for AUTO1111 and another for COMFYUI. Would be nice to share workflows, nodes etc..

charred thistle
#

a place for sharing workflows could be cool, so far I've just posted workflows in general with images to give people an idea of how ComfyUI works

spice plank
#

How to make good pictures?

#

Anyone have success with chatgt making decent image prompts?

woven rivet
outer elbow
gray wolf
#

What’s next for stable diffusion?

inner lark
charred thistle
#

any way.. it seems applying controlnet is pretty easy, but getting depth maps require a package in comfyui that seems to break other stuff... I think I need to find a better solution for generating depth maps

#

seems the WAS suite has some MiDaS nodes that get the job done without breaking anything

lavish current
#

Oh snap it's @charred thistle

#

I really prefer prompting over drawing now

#

I don't feel like I wasted time learning how to draw but at the same time prompting is so much more fun and satisfying

#

Especially when you throw in stuff like animation and upscaling

#

Have always wanted to animate, soon AI will be able to handle full animation like in the movies

charred thistle
#

what's interesting is a friend of mine gave me a rough sketch of something he wanted... I just put that into ComfyUI with a scribble controlnet model and created him an image matching it almost perfectly... it's really cool stuff

floral umbra
hasty hornet
#

There was people that went higher res, but I don't remember what they used

bleak matrix
#

Good morning, everyone! How are we all today?

broken bloom
bleak matrix
#

What's everyone up to?

neat cypress
neat flare
#

I have uploaded the CSV Loader addon to GitHub for ComfyUI if anyone is interested. Will try to get it committed to the custom nodes in the ComfyUI manager next for easier download. Enjoy!

neat flare
ebon aurora
#

Does this have an image to imagine?

#

Image*

hasty hornet
#

this as...?

acoustic igloo
#

is it down?

broken bloom
forest timber
#

hi everyone

broken bloom
forest timber
broken bloom
velvet iron
#

Also just finding recommended custom nodes or other tooling that isn't provided out of the box

charred thistle
#

There are a lot of example workflows on their github, and also a lot of people sharing them on various places online, pretty easy to figure stuff out if you know where to look. What are you wanting to do in Comfy?

fervent thunder
#

hey guys

#

I have a few problems, would anyone want to help?

#

this model does not show up under loras

#

and I cant change model from dreamshaper to sdxl

#

dm me if you wanna help or ping me here

ancient elm
#

hi all i want to clean install automatic1111. Is just deleting the folder enough or do I need to do something else?

plucky shoal
#

Should be enough

woven rivet
woven rivet
ancient elm
bleak matrix
bleak matrix
velvet iron
#

I guess I can always drop gens I like into comfy and see if they output a workflow too 😅

charred thistle
#

you definitely want to get the comfy-ui manager set up so you can install new packages from a built in repository, which will expand functionality

hasty hornet
#

is there other reliable source for comfyui btw?
Download speed from official gh link is like 100kb\sec..it'll take 3hours just to download 1.4gb

#

too lazy to install manually too thinking

charred thistle
#

I never had github be that slow

nova dawn
opaque verge
#

Hello everyone, I'm pretty new to Stable Diffusion so please forgive me if I'm asking obvious questions. I've worked through the resources and found a lot of helpful things. However, what I couldn't find is a webservice or an API endpoint that could be used to generate images/videos/animations. Does something like that already exist? Or does anyone know of steps that would need to be taken to set something up? Thank you

forest trout
bleak drift
#

Ok. I need help.

#

How can I use image to image and turn a irl photo of a person to an anime version of them?

#

I’m new to this so I don’t know what model I should use, Lora, etc.

gaunt horizon
#

Yo can anyone help with inpaintinf

#

Inpainting

#

Inpainting outputs the same image as the input for me

forest trout
#

Randomize your seed.

gaunt horizon
forest trout
#

Yep.

#

Or you can press the dice button.

gaunt horizon
#

So I tried that an nothing happened

#

Nothing different about the image

hasty hornet
#

What's your denoise strength?

gaunt horizon
#

.75

hasty hornet
#

send some screenshots with your settings, idk

subtle geyser
#

Hey what happend with control net in bot options?

fierce owl
#

Anybody know how i could go back to a previous comfyui version?

#

i updated and it broke some stuff

serene jetty
#

Guys seens like my another acc was hacked and I got banned from this server any way to rejoin back?

bleak matrix
#

Good morning, everyone! How are we all today?

finite cloak
fervent thunder
#

with the AI generating images at random. Suppose I could generate one million images of the same description. Would all of them be totally different from each other?

tough isle
#

hello everyone

safe brook
fervent thunder
#

jesus

#

bing's censorship is madening. "skulls" was blocked, then "heads" resulted in image with skulls. Minutes later, "skulls" was free. Minutes later, again blocked.

hasty hornet
#

or maybe it was just model issue which generated too many similar faces...eh whatever

regal raven
#

am i crazy or idd hte stable audio channel disappear?

crisp dagger
#

Hi dudes

sudden ruin
broken bloom
finite cloak
karmic brook
teal hazel
south mason
#

Hi!
I haven't seen it mentioned anywhere in the official announcement, but does anyone know if Stable 3D will be released for free in the future?

fervent thunder
#

What's up with "connection errored out"

delicate tiger
#

where do you download the Segment Anything models?

jaunty flint
#

hey guys, im very new to this and want to make a quick 30 sec clip using stable diffusion. Im following corridor crews tutorial, but i am at the style dataset portion and they say I need square images. My question is should I just grab any photo from my reference and crop it into a square or do they need to be 512x512. images?

fossil bloom
#

Whats the best model and settings to re-generate youtube thumbnails basically ?

#

to make them better

sturdy dew
#

How do you get it to give me a prompt from an image

forest dirge
#

怎么给机器人指令才能生出自己想要的图片

bleak matrix
#

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

bleak matrix
bleak matrix
fossil bloom
#

People in like drawed style

#

With good clean background

#

Text I can do in Photoshop

bleak matrix
fossil bloom
bleak matrix
fossil bloom
#

It's loading hard

bleak matrix
static turret
#

Hello everybody!

still thunder
#

Is there a way I privatize my prompt poses so only I can see them?

clear imp
#

Hi! Can anyone help me to download the ai? I tried by myself but I couldn't 😦

#

any good video tutorial maybe

quartz nebula
#

Hello,everyone

fervent thunder
#

bing is so random. I often get impressive images with things that I did not describe, like the sky for ex. When I do want certain things I either don't know how to describe it or the AI just fails to understand it

high swan
#

Suppppp !

trail pawn
#

Why do I get video clips? How can I get the pictures back??

misty tiger
#

these people remind me of the umbrella corporatioin in a way

#

corporation*

#

lol

#

im just going to call this great value hogwarts

#

for lolz

#

they want fair game proof

crisp dagger
#

dudes? does anybody using nsfw ?

hasty hornet
#

Best way is to train lora based on face you want to generate...probably
I've seen some services that claim they can do something with just 1 image or something, but didn't remember any of them.

hasty hornet
crisp dagger
hasty hornet
#

Not in this discord, but yea, there was another community, probably still exist

neat cypress
hasty hornet
#

thinking why self promotion is still allowed

iron zenith
#

Why is the bot down

crisp dagger
hasty hornet
iron zenith
#

Why isn't the dreamer dreaming

weak kettle
#

Hi guys, is there any lora or embedding for drip style of clothing etc, like grillz, diamond chain, supreme jacket and so on, I cant find anything like that

#

let me know

hasty hornet
#

drip style of clothing...wtf

rich kestrel
weak kettle
#

i searched for that

rich kestrel
#

probably not what ur looking for but Im sure there's something along those lines by someone else

#

or u can train one urself ^^

#

dreambooth or lora

weak kettle
#

i mean something like that

#

sorry for this description, my english is not good at all xD

rich kestrel
#

now thats pimp

weak kettle
#

yeahh

rich kestrel
#

if you cant find it on civitai you might have to make your own. doesnt seem that hard, especially since base SD models probably can generate grills and supreme clothing

#

I would say train your own

weak kettle
#

I didnt train anything yet, I dont know how to do that tbh

#

need to watch some tutorials