#💬｜general-chat

1 messages · Page 66 of 1

nocturne pewter
#

sorry is it possible to upload an image ?

wild steppe
nocturne pewter
#

ok thanks

wooden briar
#

Now that I've grasped the basics I want to look into upscaling - what's the best method for upscaling using A1111?

broken frigate
#

hi everyone, maybe someone can help me, since yesterday my stable diffusion web UI doesnt work, when i launch A1111 my computer freeze for many second and open the web UI, after that when i want to generate image, i have an error its a NansException, i dont know how to solve the prob. if anybody know u can send me a DM im open

THX guys

warm junco
oak tiger
#

bot stopped?

wild steppe
#

probably needed a breather, its working hard!

wise sparrow
#

is it normal that sometimes when im using automatic1111 i hit generate and my screen turn black then the GPU fans go full speed?

hasty hornet
wise sparrow
hasty hornet
#

idk, that doesn't happen to me...I had slight lags when I was going on max res (probably maxing vram), but never seen black screen

wise sparrow
#

hmm interesting thanks though

warm junco
wise sparrow
wooden briar
warm junco
wise sparrow
wise sparrow
wooden briar
wise sparrow
#

through some basic monitoring

#

the GPU clock was sitting at like 1650 at all times even though im not generating anything

wooden briar
#

Oh I've never had that happen. It just went black for a moment for me

#

Sorry

wise sparrow
#

no worries man, thank you

wooden briar
#

In A1111 I've used High Res Fix to upscale images from txt2img, but how do I upscale with img2img?

broken smelt
#

you can simply use the resize setting in the usual generation parameters or you could also use a script called ultimate SD upscale

warm junco
wooden briar
#

Is that better then the rescale checkbox?

warm junco
#

Yes

#

Its like hires fix

wooden briar
#

Awesome ty

warm junco
#

But uses less vram and can do higher output resolution.
But also takes a bit longer

wooden briar
#

Wonderful

wooden briar
warm junco
#

@wooden briar
The first is build in under scripts, the second has to be Installed as extension and then is also there as script

#

I would given you an example how sd upscale script works but im going to sleep now.

#

Ah found my comment example:
put your image in img2img, set the resolution to 512x512,
denois to 0.3
activate the sd upscale script at the bottom,
choose an upscaler like esrgan4x+ for example,
set upscale by 2
then press generate
if you get artifacts then you need to remove person specific tags from your prompt in img2img

#

Then you can play with the upscalers and denois value

fervent thunder
#

... did anyone else get kicked?

wooden briar
late blaze
stray tangle
barren sentinel
#

hoi folks how isi t going?

bleak matrix
#

As beautifully and wonderfully as one might expect. How about you, Mihoru?

blazing sorrel
#

@sudden marlin hey

sudden marlin
bleak matrix
#

How can I help you?

tepid dagger
#

Hi, anyone who could help me to enable controlnet - it was working for months. But after todays active1111 webui update + xformers 0.0.20 it just will not run. It appears like enabled but it is not loading models or doing anything, it behaves like disabled. # please help! 😗

bleak matrix
#

Hey there, David! I reccomend checking out #🤝｜tech-support for this, as someone might be able to best help you there

karmic brook
tepid dagger
bleak matrix
tepid dagger
#

I am starting to be desperate. 😄

worn scaffold
#

hello is anyone pro here at stable diffusion

rough barn
high lichen
#

Hi everyone, I’m new here and in need of some assistance. I downloaded the (ROOP v0 0.2) function it’s shows it’s installed and it’s not showing up in the text to image UI. I did an update and restarted the application still doesn’t show up. Is there a set I’m missing?

thorn pilot
high lichen
thorn pilot
#

you need to open a cmd window, navigate to the venv folder and run the activate.bat

high lichen
#

Not in power shell correct?

thorn pilot
#

cmd is better

#

i prefer it

high lichen
#

It ends with scripts?!?

#

Do still stick it in that folder the exe file is in there

thorn pilot
#

oh yeah scripts folder

high lichen
#

Did that and it’s still not showing up

#

Building wheel for insightface (pyproject.toml): started
Building wheel for insightface (pyproject.toml): finished with status 'error'
Failed to build insightface

stderr: error: subprocess-exited-with-error

Building wheel for insightface (pyproject.toml) did not run successfully.
exit code: 1

[214 lines of output]
WARNING: pandoc not enabled
running bdist_wheel
running build

high lichen
# thorn pilot oh yeah scripts folder

!!
check.warn(importable)
creating build\lib.win-amd64-cpython-310\insightface\thirdparty\face3d\mesh\cython
copying insightface\thirdparty\face3d\mesh\cython\mesh_core.cpp -> build\lib.win-amd64-cpython-310\insightface\thirdparty\face3d\mesh\cython
copying insightface\thirdparty\face3d\mesh\cython\mesh_core_cython.cpp -> build\lib.win-amd64-cpython-310\insightface\thirdparty\face3d\mesh\cython
creating build\lib.win-amd64-cpython-310\insightface\data\images
copying insightface\data\images\Tom_Hanks_54745.png -> build\lib.win-amd64-cpython-310\insightface\data\images
copying insightface\data\images\mask_black.jpg -> build\lib.win-amd64-cpython-310\insightface\data\images
copying insightface\data\images\mask_blue.jpg -> build\lib.win-amd64-cpython-310\insightface\data\images
copying insightface\data\images\mask_green.jpg -> build\lib.win-amd64-cpython-310\insightface\data\images
copying insightface\data\images\mask_white.jpg -> build\lib.win-amd64-cpython-310\insightface\data\images
copying insightface\data\images\t1.jpg -> build\lib.win-amd64-cpython-310\insightface\data\images
creating build\lib.win-amd64-cpython-310\insightface\data\objects
copying insightface\data\objects\meanshape_68.pkl -> build\lib.win-amd64-cpython-310\insightface\data\objects

#

creating build\lib.win-amd64-cpython-310\insightface\data\objects
copying insightface\data\objects\meanshape_68.pkl -> build\lib.win-amd64-cpython-310\insightface\data\objects
copying insightface\thirdparty\face3d\mesh\cython\mesh_core.h -> build\lib.win-amd64-cpython-310\insightface\thirdparty\face3d\mesh\cython
copying insightface\thirdparty\face3d\mesh\cython\mesh_core_cython.c -> build\lib.win-amd64-cpython-310\insightface\thirdparty\face3d\mesh\cython
copying insightface\thirdparty\face3d\mesh\cython\mesh_core_cython.pyx -> build\lib.win-amd64-cpython-310\insightface\thirdparty\face3d\mesh\cython
copying insightface\thirdparty\face3d\mesh\cython\setup.py -> build\lib.win-amd64-cpython-310\insightface\thirdparty\face3d\mesh\cython
running build_ext
building 'insightface.thirdparty.face3d.mesh.cython.mesh_core_cython' extension
error: Microsoft Visual C++ 14.0 or greater is required. Get it with "Microsoft C++ Build Tools": https://visualstudio.microsoft.com/visual-cpp-build-tools/
[end of output]

note: This error originates from a subprocess, and is likely not a problem with pip.
ERROR: Failed building wheel for insightface
ERROR: Could not build wheels for insightface, which is required to install pyproject.toml-based projects

Warning: Failed to install insightface==0.7.3, roop will not work.

stderr: Traceback (most recent call last):
File "C:\Users\CJ\Desktop\A1111\stable-diffusion-webui\extensions\sd-webui-roop\install.py", line 54, in <module>
raise e

#

File "C:\Users\CJ\Desktop\A1111\stable-diffusion-webui\extensions\sd-webui-roop\install.py", line 36, in <module>
launch.run_pip(
File "C:\Users\CJ\Desktop\A1111\stable-diffusion-webui\modules\launch_utils.py", line 124, in run_pip
return run(f'"{python}" -m pip {command} --prefer-binary{index_url_line}', desc=f"Installing {desc}", errdesc=f"Couldn't install {desc}", live=live)
File "C:\Users\CJ\Desktop\A1111\stable-diffusion-webui\modules\launch_utils.py", line 101, in run
raise RuntimeError("\n".join(error_bits))
RuntimeError: Couldn't install sd-webui-roop requirement: insightface==0.7.3.
Command: "C:\Users\CJ\Desktop\A1111\stable-diffusion-webui\venv\Scripts\python.exe" -m pip install insightface==0.7.3 --prefer-binary
Error code: 1
stdout: Collecting insightface==0.7.3

thorn pilot
#

looks like you need to install visual studio build tools

#

it's having trouble compiling the c code for the library it's tryring to install

high lichen
#

its downloading as we speak

high lichen
#

Loading weights [c0d1994c73] from C:\Users\CJ\Desktop\A1111\stable-diffusion-webui\models\Stable-diffusion\realisticVisionV20_v20NoVAE.safetensors
Error executing callback ui_tabs_callback for C:\Users\CJ\Desktop\A1111\stable-diffusion-webui\extensions\sd-webui-depth-lib\scripts\main.py
Traceback (most recent call last):
File "C:\Users\CJ\Desktop\A1111\stable-diffusion-webui\modules\script_callbacks.py", line 151, in ui_tabs_callback
res += c.callback() or []
File "C:\Users\CJ\Desktop\A1111\stable-diffusion-webui\extensions\sd-webui-depth-lib\scripts\main.py", line 47, in on_ui_tabs
dataset = gr.Examples(examples=os.path.join(maps_path, t), inputs=[png_input_area],examples_per_page=24,label="Depth Maps", elem_id="examples")
File "C:\Users\CJ\Desktop\A1111\stable-diffusion-webui\venv\lib\site-packages\gradio\helpers.py", line 54, in create_examples
examples_obj = Examples(
File "C:\Users\CJ\Desktop\A1111\stable-diffusion-webui\venv\lib\site-packages\gradio\helpers.py", line 201, in init
self.processed_examples = [
File "C:\Users\CJ\Desktop\A1111\stable-diffusion-webui\venv\lib\site-packages\gradio\helpers.py", line 202, in <listcomp>
[
File "C:\Users\CJ\Desktop\A1111\stable-diffusion-webui\venv\lib\site-packages\gradio\helpers.py", line 203, in <listcomp>
component.postprocess(sample)
File "C:\Users\CJ\Desktop\A1111\stable-diffusion-webui\venv\lib\site-packages\gradio\components.py", line 1866, in postprocess
return client_utils.encode_url_or_file_to_base64(y)
File "C:\Users\CJ\Desktop\A1111\stable-diffusion-webui\venv\lib\site-packages\gradio_client\utils.py", line 349, in encode_url_or_file_to_base64
return encode_file_to_base64(path)
File "C:\Users\CJ\Desktop\A1111\stable-diffusion-webui\venv\lib\site-packages\gradio_client\utils.py", line 323, in encode_file_to_base64
with open(f, "rb") as file:
PermissionError: [Errno 13] Permission denied: 'extensions'

thorn pilot
#

is A1111 running?

#

I usually only get permission errors if there is a python instance running and using the file it's trying to update

unique hamlet
#

Other than YouTube is there a platform where SD content creators are growing big? Like live streaming SD sessions and stuff like this? For example Kick or Odyssey?

agile dragon
#

i got a permission error that looked like that once trying to install cudann simply because the cudann folder was inside another cudann folder, simply moving the folder back 1 fixed that error

high lichen
#

roop and all

thorn pilot
#

all good?

high lichen
warm junco
tender talon
#

hi guys
i am acutely getting a this type of error can you please help ?

TypeError: Script.run() missing 15 required positional arguments: 'frames', 'denoising_strength_change_amplitude', 'frames_per_wave', 'denoising_strength_change_offset', 'initial_image_number', 'prompts', 'save_prompts', 'save_video', 'output_dir', 'video_fps', 'video_quality', 'video_encoding', 'ffmpeg_path', 'segment_video', and 'video_segment_duration'

i am using it in img to img stable diffusion local

running on edge browser
rtx 2060

please help

warm junco
umbral spoke
#

Hi! Who I can contact for partnership proposal?

rough barn
modern basin
#

Hi. I want to create subjects looking like normal people in the average Facebook photograph. SD wants everyone to be professionally shot and made up whatever I seem to prompt. Any tips for overcoming this? Thanks

tender talon
hard musk
hard musk
sleek osprey
#

Its kinda funny how before the release of IF there was a decent amount of hype for it, but now no one uses it

sleek osprey
bleak matrix
#

Good morning, everyone!

#

How are we all this beautiful morning?

thorn pilot
bleak matrix
#

Also, I love your icon

thorn pilot
bleak matrix
thorn pilot
bleak matrix
#

I'll have to take a look at what you posted--haha, cuz I was going to say--did you post in #✨｜sdxl

warm junco
bleak matrix
#

Good morning, CS!

warm junco
#

Morning sunsmile !

bleak matrix
#

How are you today?

warm junco
#

Good, lunchtime is over now back at work stuff

thorn pilot
warm junco
#

And how are you ?

bleak matrix
bleak matrix
bleak matrix
#

I'll prolly work in Blender today

#

More simulations

warm junco
bleak matrix
#

I'm thinking about going for anti gravity.

#

For some. Might be interesting.

warm junco
bleak matrix
warm junco
#

I only changed the amount of Polygons for Vrchat characters with Blender xD

bleak matrix
#

I gotta set up the wind system today. I got the clouds, and was thinking of adding some burbs.

#

Hahaha, hey, that's something!

#

Good ole VRChat!

warm junco
thorn pilot
bleak matrix
tender talon
#

By the way can somebody write me down programs that required for stable diffusion?

Like python & ffmpeg ?
I think I should reinstall all required program to pix it

My this bug happened after a power cut because of storm

bleak matrix
#

I gotta add all the little details I've been working on. Might put out a tut today. We'll see

warm junco
tender talon
#

Ok thank you once again m8 😀

umbral spoke
rough barn
wooden root
#

Where do I need to go to get some help with setting up my Stable Diffusion? I got everything good its just the Lora files aren't being detected.

wild steppe
crimson belfry
#

I'm very excited to try this out

#

I have not been this excited since I learned how to code video games.

bright sand
#

Im new to lora training and I need someone to clarify for me a few things:
When some lora on civitai mentions it used for example 6000 steps, that means they used for example 50 images with a folder named 120_charactername?
And what exactly is Epoch in the kohya ss training? should I ever want to change it from 1?

fervent thunder
#

@steep star are you able to make the music ai samples longer than 15 seconds anywhere? (saw your message in sdxl section)

steep star
steep star
fervent thunder
steep star
steep star
#

it works

fervent thunder
steep star
#

follow this

fervent thunder
#

thanks

crimson belfry
#

what are the limits with this ai generator? How many prompts do i have ? Is there a limit?

warm junco
crimson belfry
#

That is amazing, Why does mid journey even exist. @warm junco Thank you for the feedback and I hope this continues. I will be donating asap.

warm junco
crimson belfry
#

im not sure if my laptop can handle it, From what i have been hearing we need a good set up to run it. Is this true? @warm junco

warm junco
primal current
#

Any way to improve the performance of the vram? I have been able to do it with the ram, watching tutorials, but I have not seen anything about vram. This is to improve AUTOMATIC1111. Is there a thread or tutorial that can help me?

crimson belfry
#

@warm junco I am running AMD Ryzen 3 3200U with Radeon Vega Mobile Gfx, 2600 Mhz, 2 Cores, 4 logical Processors I have 8gb of installed ram

#

ive never looked into running it local, I didnt even know we could do that.

warm junco
crimson belfry
#

Thank you for the help!

warm junco
wooden briar
warm junco
wooden briar
warm junco
wooden briar
#

Right, makes sense

#

Ty so much

warm junco
#

no problem 🙂

wooden briar
# warm junco no problem 🙂

Do you generate at 512, save seed and run again with high-res, or do you run all yours with high-res and hope for the best?

warm junco
#

takes more time but i let it run a few gens

wooden briar
#

Nice, ty

sacred canopy
#

like images I edited with photoshop and stuff

serene swan
#

Im reading a guide on making loras and it says to generate training images that are naked, but I've found once using the lora on said images, it bypasses my negative nsfw prompts and makes them naked anyway. So is it really necessary to generate naked training images?

fervent thunder
steep star
livid cloak
#

Is there a way to have the same options set everytime I open SD? They always go to some default. The Styles option only saves my prompts

bright sand
#

When training lora for a character that doesnt have much material in the internet, is it ok to add images with different style? or would it be better off without it?
Im talking about character lora training with about 10-20 images

#

Also im using a few small images from sprites in addition to very few decent resolution images what do u think

crimson belfry
#

Are we allowed to use our images for t-shirts?

waxen storm
#

opinions? Mikubill vs Automatic1111?

primal current
#

Do you know if there is any voice AI, that uses a trained model, but that speaks through written text?

waxen storm
primal current
harsh fox
#

When it says [Base Model SD 1.5] on civitai, Is that something I need to be concerned about or a setting I need to change?

livid cloak
#

Was wondering if that kind of info is in the metadata. Until now I've only seen the parameters and options used for gen the img

real kernel
floral rapids
#

Sa marche plus

twin pulsar
#

C’est kc

pale latch
#

I've been doing nothing BUT teleporting bread for 3 days ...

wild steppe
#

?

warm jasper
#

Is someone here using Stable Diffusion on an AMD RX 580 on Windows? I'll help to make a install with such environment, so I would like to be ready to something that may go wrong

pliant jungle
#

How to open negative promps

reef aspen
#

Did anyone get 0.9 working?

frail fable
#

Does anyone know if Stability AI plans to add ControlNet features to ClipDrop? (Or ControlNet-like features, if there are issues with the Apache license?)

fathom marten
#

Hey, are the bots 1 to 10 using 0.9 ? Seems a yes to me, but just wanna be sure ((:

#

Sry for the dumb question

thorn pilot
trail dew
#

how would i generate an image that has one part on one side and the next part on the other side?

thorn pilot
open night
#

so first time I "trained" i did not have xformers

#

test 2 copium will go better

#

for some reason I see no eta tho

ruby frigate
#

Does anyone know what would cause an image to have these sort of horizonal line patterns running across it?

#

Only happened with one of my images I made with the prompt, but would like to know why

#

Sorry, actually I'll move to the images chat

meager marsh
#

Can I still run SD in Google CoLab if I have a paid account?

thorn pilot
open night
#

I'm still training the model, the results are ok. You can see some coherence to the style I think

#

no clue how to tell the ai what is wrong so imma just keep training and see if it improves

#

i'm glad it's training even if I have no idea what it's doing

#

it didn't work before because having preview images enabled apparently is bugged

fervent thunder
#

Anyone knows a website there's only AI art?

frank spruce
#

can always do a search on deviantart or something tho, not a very good site tho

pearl fog
#

I know one but it's furry lol

harsh fox
#

Are these models limited to like one sex? I am using darkSushi25D and changing (1 girl) to (1 male) still produces a female.

broken smelt
#

to answer that question you have to ask yourself, what are the odds that the creator of the model trained it on both males and females

#

when it comes to anime

#

im assuming pretty low

#

take a gander at civit.ai and notice how many males you see there 😛

harsh fox
#

Yea, I see very few lol

nocturne pewter
#

Have a question, does midjourney and stable diffusion work the same way? Coz I'm seeing some genius using the prompt as if it was midjourney

hushed quarry
nocturne pewter
#

Like the v-5 👁️👄👁️

desert crest
lusty dirge
#

Hello everyone, is this the correct chat to ask a quick question?

indigo berry
#

guys, which checkpoint is the best in generating medical images?

inland flax
desert crest
inland flax
desert crest
inland flax
primal quest
#

sorry for stupid question: but can i install sdxl 0.9 localy on my pc like automatic1111 ?

radiant nacelle
#

Hello

dusk agate
#

How can I make pictures

radiant nacelle
humble spear
bright sand
humble spear
#

Yes as I understand it more variations are good ..
Re : pose , background and other details so the model can learn ..
How to work around the " main concept " producing differences in style and orientation ..

fervent thunder
#

Hey guys. I am new to this server, as my badge suggests. Anyway, how can we upscale a result image?

bleak matrix
#

Good morning, everyone. How are we all today? sunsmile

bleak matrix
#

It has links to all kinds of info

fervent thunder
bleak matrix
#

It is fun, to be fair! Hahaha! 🙂

#

With the bot, though, you press reSize

bleak matrix
fervent thunder
stable pawn
#

Not long ago I saw on YT a demo of a web site that generated images from text, one function was that it could point out unnecessary words that would not affect the image or affect it in a bad way, like contradictions, is there any extension in SD/A1111 that can do something simelar?

bleak matrix
slender field
reef aspen
#

Anyone get 0.9 API yet?

zealous geode
#

Anyone know when 0.9 will be available through the clipdrop API?

fervent thunder
#

Hi guys

#

might get a new GPU

#

in the $300 range

#

which one should I get?

radiant nacelle
#

Is there a place for sharing open source model creations?

pearl fog
#

Like the model itself?

radiant nacelle
#

No, the arts made through models

#

There is a channel for sdxl creations an example

#

I am using automatic1111 on my pc so i am using open source models too making art with them

shrewd whale
#

hey i have a question, if any one can help me figure it out, it will be highly appreciated. During in painting, i have seen people promot on focus, sharpness, and also enhance the image without changing anything to the image at all. On NC... Can anyone help me figure those promts or details out please??

#

prompt

rustic canopy
#

I got a Rx 6700xt with 12 gb vram but everytime i render an Image on Stable Diffusion with hires.fix with everything above 1x upscale it crashes halfway through

harsh fox
#

Any tips on how to change the colour of armour or hair? Should I write each detail individually or something like, "long brown combed back hair" ?

sudden violet
#

Heyo. What is the definition of NSFW for the purposes of this forum? Everybody has a different definition.

sudden ruin
sudden violet
#

Cool. I like the outlook. So the stray nipple in a fine art nude isn't gonna get me flagged from the sounds of it, AND I don't have to deal with pervert weirdos? I'm gonna like this place. Sounds much better than the group I just got banned from.

frank spruce
sudden violet
#

Naw, man. They were... Gross, but surprisingly tolerant of my opinions, even though my opinion is that they all belong in jail. No, Mage.

peak portal
#

@bleak matrix in #sdxl there's a pretty shitty conversation ongoing for a while

#

well they stopped - but an unsavoury pattern of dogwhistles

frank spruce
sudden violet
#

I didn't say I desired tolerance, but I like people to practice what they preach. And MAPs belong in Jail, and chibi hentai weirdos with them.

frank spruce
#

okay well, idk about jail, but i dont feel like listening to you anymore then, bye

sudden violet
#

I've got a flag of the archangel Michael trampling the Gadsden snake over my bed.

fast lichen
#

I just installed A1111 WebUI on a headless Debian system. The github page for the WebUI mentioned running webui.sh to install dependencies and to run the UI. Is it normal practice to continue to run this script to restart the UI after turning off the system?

peak portal
#

yup

spice plank
#

Whys this discord have so many channels?

peak portal
#

to make it look busy

fast lichen
#

thanks @peak portal for the input! I think I'll setup a demon to autolaunch the script at bootup, and then move onto learning how to add checkpoints and other fun addons.

peak portal
hasty hornet
raven gate
#

Anyone know where to download new controlnet 1.1 preprocessors and respective models?

thorn pilot
primal current
#

I was wondering, is there an AI that changes the language of the voice? 🤔

low compass
#

Is it faster to upscale a 512 picture 4x or a 1024 picture 2x or same?

remote urchin
#

Guys, what's the easiest way to add two bright glowing eyes to a faceless face? This has been killing me. I've used inpainting, controlnet inpainting, masked area only, whole picture, it takes me like 40 attepmpts to just get two white orbs on a stupid face. I'm stupid, can somebody suggest me the best way to do something like that please?

pearl fog
#

Photoshop

remote urchin
# pearl fog Photoshop

if I just put two white dots on the face in photoshop, then put it back to img2img, will it make it look nice?

#

or nice-r

pearl fog
#

Yeah that might work.

remote urchin
pearl fog
#

Easiest way in Photoshop though. Make a new layer with white dots

duplicate layer
go to the layer on the bottom and gaussian blur it until it gets a glowing effect

#

Works better if the layer below it is darker

remote urchin
pearl fog
#

Yeah photopea will work. It will work in any program that has layers

remote urchin
broken smelt
#

you cant just say im not doing x while doing x lmaoo

foggy spear
#

ok, i apologize

#

there deleted

#

the short version:
I'm looking for beta testers to join the Crystal-Team

fickle juniper
#

Is prompting with image supported?

jolly goblet
#

I've got a question. Is it a good idea to use a 16gb ram stick with a 8 gb ram stick that I already have? Or is it better to just upgrade my ram from 8gb to 16gb? Both will be ddr4.

crimson belfry
#

I love stable diffusion XL

cunning marsh
warm junco
narrow pagoda
#

With Steam now prohibiting the sale of games featuring AI-generated artwork unless creators can verify that the underlying model was trained on copyright-compliant data, I've started looking to see if there's any work toward checkpoints that would meet this criteria and come up empty. It's clear that 2.1 was a step closer, but I'm curious if anyone knows of any SD plans toward meeting this regulatory hurdle (like Firefly does) that isn't going away any time soon?

rare viper
#

How can I do I painting?

rigid steeple
#

have any plugin for automatically predict prompt?

jolly goblet
#

Ok, thanks for the answer CS1o

broken smelt
narrow pagoda
#

so if you've got something with clear giveaways like alien hands they're going to flag it

broken vale
#

Where would I go if I need help downloading

supple oasis
#

Hey guys, how do you organise all the prompts you generate?

plucky shoal
#

you mean storing them (prompts)? @supple oasis

supple oasis
plucky shoal
#

i am using A1111 so i organize only pictures, and prompt and all settings are part of image metadata. Which can be shown in pnginfo tab and sending to txt2img

bleak matrix
#

Good morning, everyone! How are we all this lovely day? 🌅

bleak matrix
tired vigil
#

Hi All, is hugging face the best way to learn about SD and implement the model locally?

bleak matrix
#

You can learn more by following the links I just posted above!

tired vigil
#

ok thanks!

#

haha the resources are mainly hugging face so assume the answer is yes

surreal lance
#

Rocm 5.6 just released but I didn't see that is compatile with windows. Any news?

bleak matrix
#

There's a lot of tutorials on YT if you are looking for something more visual.

#

If you are looking for something specific, though, there are channels here dedicated to certain topics

#

There's also plenty of community written guides

tired vigil
#

ok thank you

bleak matrix
#

Np!

#

You can always checked the pinned comments for more. If you run into issues with installation, you can check out #🤝｜tech-support

remote urchin
#

Okay guys, need help I’m stupid. So I have two generations of one picture. I really like the legs of how one turned out, but I like the rest of the body on the other one. What’s the best way to mush them together and have them be combined? I’m trying to use control net in painting and using one image as the base and the other one as in paint reference and it doesn’t seem to do anything. What am I doing wrong?

thorn pilot
#

If they are close enough, I use krita to add the two images as layers and then erase the part on the top layer to expose the part from the second image. Then can use img2img to clean up

bleak matrix
remote urchin
bleak matrix
#

There's a lot of tutorials online, but really, it's super simple

#

Not necessarily

#

You can make it matcg

#

Or

remote urchin
bleak matrix
#

You can take the leg, and put it in Krita, etc, then do it in img2img again

remote urchin
#

I’ll try to do that thanks. Which one is easier to use gimp or Krita though?

bleak matrix
#

You can always change the hue/saturation, and transform the shape if you need to

#

Tough call. It's a personal preference. Either will work

#

I'd use Krita

thorn pilot
#

I've stuck with Krita for a while, seems to have everything I need so far for quick touch ups

remote urchin
hasty hornet
#

Can I give downvote instead? thinking

young ivy
#

dont suppose anyone know what model is the best for training realism LORAs on? For max compatibility

hasty hornet
#

all this people trying to capitalise on someone else opensource work...

young ivy
#

what are you talking about?

hasty hornet
#

about qr code generation services.

young ivy
#

I see

bleak matrix
stable pawn
#

I have used SD1.5 as standard model, but I now see that is not as good as other models and can sometime be just plain weird, so what model would you recommend if I still want a bit of that look SD1.5 had but also want to be able to do furries, anime and other fantasy that not tend to go over to the sexual?

bleak matrix
#

You can watch a basic video on how to use it and I am sure you'll be able to

#

I would recommend checking SDXL, to be perfectly honest ❤️

#

You can check out the bot for now to see a preview

stable pawn
#

@bleak matrix I use A1111 now locally for I sometime go out to the coast where I have no connection, so for now I look for a model that are current. In the future when I get a better data plan for my mobile online sites may be more of intrest, if they do not get too expensive.

But I agree with you that it is things folk should keep an eye on and test.

grim frigate
#

Any tips on how can I avoid generating random structures on the background?
Whether its a indoor view of a room or a landscape. Some structures doesn't just make sense at all. Cityscapes are just pure hell.
I tried using deep negative, unspeakable horrors and other textual inversions but no luck.
ControlNet with MLSD model no luck too.

fervent thunder
#

we live in AI era

plucky shoal
#

Plain background? Solid color background? Single color background? Pure white background?
I know somebody asked for this not so long ago. But forgot it. @grim frigate

serene swan
#

keep running out of mem. Trying to make a lora, I already got my training images and tagged, are there any others options for training a simple character lora?

grim frigate
# plucky shoal Plain background? Solid color background? Single color background? Pure white ba...

Ok let me put in other terms. If there is a fix for ugly hands, ugly face, eyes, etc
Is there a fix for backgrounds that don't make sense at all?

Example of this is an indoor view of a room. If you put chair in the prompt sometimes the chair will clip to another object you put like a table.
Or if you put 'window' it will just generate something that is a bit NOT like a window.

Another good example is a cityscape. Sometimes the buildings just appear wobbly or sometimes the buildings just appear to cluttered or sometimes the windows are even realistic at all especially if viewed from a far away distance.

fervent thunder
#

Huuuuh

plucky shoal
#

i know probably what you mean, for example astronaut and planet. if astronaut break planet edge, than planet looks like elipse in best case, but usualy much worse.
Try search for some embeddings, i personaly realy dont know.
Or you can try probably Regional prompter. Where befor addcom you add your background, which shouldnt be affected with what is in foreground. pfff... now i dont know what i am writing 😄
@grim frigate

#

check general with images, realy random images made with regional prompter, if it is o.k. with your standards or not. @grim frigate

grim frigate
plucky shoal
#

today i dl some anime models and surprised how lines are perfectly shaped...

#

try then regional prompter, it is worth. Bit of learning but worth.

rare viper
#

Where i can find a PDF file of the terms and conditions of the generated images?

fading wraith
#

what's the best website / paper / tutorial if I want to understand how stable-diffusion works, and start programming with it a little bit?

#

(already have knowledge in ML)

glad meteor
#

Out of curiousity, anyone know of an AI powered text to speech platform that I can run on my pc? 15.ai has been dead for months

cosmic crescent
harsh fox
#

Any prompts for making the character look at the camera? Or controlling the direction of the eyes in general?

cosmic crescent
serene swan
glad meteor
#

I'd prefer running things locally so I don't become dependent on some website

cosmic crescent
glad meteor
#

ios and google aren't very dynamic, you just don't really notice it because the majority of stuff they say is preprogrammed and heavily vetted

serene swan
#

it'd be cool if I could record my voice for lines in my visual novel and use ai to make it sound sexy

plucky shoal
#

@harsh fox i think looking at viewer. Not on camera, because it add camera to the picture.

glad meteor
#

I notice google's voice making mistakes quite easily once I go off script

#

also it just sounds unnatural

cosmic crescent
serene swan
glad meteor
#

@cosmic crescent welp, that was easy, I now have a running docker image with web ui lol

cosmic crescent
#

nice

#

always nice when you get it done before losing interest

glad meteor
#

haha yeah

#

it's generating speech at a pretty nice rate too

#

especially since I'm using CPU

raw marten
#

im new here, how to i change the image ratio. --ar 16:9 isnt working?

plucky shoal
#

using bots here?
There should be button

karmic brook
raw marten
plucky shoal
#

There is aspect, where you entering your dream there are 3 other things. One is aspect, probably this is it?

harsh fox
#

Is there a way to add things to an image that do not seem to be known by the AI model? For example a certain type of item from a video game or a hair style or something along those lines? Or I guess the question should be, can I teach this thing new tricks? lol

outer crescent
#

using lora's assumes you run SD locally, but you can look up training textual inversions and loras on youtube

austere current
#

hello - i am new to this community and looking for direction on how to connect with and find the best stability.ai developers interested in collaboration on generative ai and workflow apps for film and TV. Please advise best channels or people to reach out to

wild steppe
gaunt pulsar
#

how am I supposed to follow Stability and Emad's tweets now that Twitter is locked for users without an account?

hasty hornet
#

You can follow there? thinking

#

I'd say use reddit instead cause feck Elon, but uuuh....there is another story...

final valve
#

hiiiii, my name is Florencia, I'm wanting to launch a beta version of a cooking app and I want to know what kind of prompts can work to generate recipe images. If you can give me a hand I would be very grateful

wild steppe
knotty turtle
#

Bit of a long shot... anyone familiar with the One Button Prompt extension?
I'm trying to figure out how to:

  • Interrupt a generation/s
  • Generate more than 50 images at a time
final valve
late laurel
#

guys

#

what is the easiest way to face swap in SD

fading geode
coral meteor
#

Anyone know some good discords or slacks for general gen ai dev chat? (DM me if can’t post here). Thanks!

wispy light
#

Hey guys 👋 we are looking to host our SD model in the cloud to be able to scale easily, do you have any recommendations of service that do this well ?

wispy light
fervent thunder
#

Why is modifying clothing so hard even with Controlnet?

#

I'm using the anythingv4.5 model just for testing

wide solar
#

hey guys just wanted to tell you to have a great day and a good time. Hope yall are doing well🫂

still glacier
still glacier
warm junco
bleak matrix
#

Good morning, everyone!

#

How are we all today? sunsmile

thorn pilot
#

Morning Sunny! 🙂 Feeling pretty pretty pretty good. You?

fervent thunder
bleak matrix
#

What's everyone up to today?

thorn pilot
#

I played diablo 4 most of the day

#

now just relaxing and playing around with SDXL

bleak matrix
#

Also, that sounds like a lovely way to end the day

thorn pilot
#

how I spend most of my nights for a while now 😄

bleak matrix
#

Any particular goals?

#

I think I'll prolly do some cloth simulations today. Or maybe I'll take a break (yeah right), and play some games. I'm going through the Bunker on hard mode.

#

It's so fun to just do everything in the dark.

thorn pilot
bleak matrix
solemn marsh
#

we need a option on the bot that allows us to remove the background and get the img dmed to us

bleak matrix
wild steppe
#

Happy Canada Day to all Canadians here! 🇨🇦❤️

clever pebble
#

ow to run stable diffusion

sudden ruin
clever pebble
heavy drum
#

In the bot channels when we slash-dream, why does the "seed" value appear to always have the top bit of a 32 unsigned int always set? Every case seems to be between 2 billion and 4 billion and change.

It doesn't matter, but seems odd to not half of the available seed.

Just curious

lofty sentinel
#

how to invite this bot to my server?

fervent thunder
#

links allowed here? i wanna ask something

blazing sorrel
#

@bleak matrix sir

bleak matrix
quiet dagger
#

Does StabilityAI have mastodon account? Twitter seems to be shutting down

pearl fog
#

You need an account to view Twitter now

plucky solar
#

I have a question about a prompt that is being rejected but I can’t understand what content is forbidden. It’s basically a picture of jesus as king. I did use the word “submission” but I tried again with a new algorithm, removed what I thought might be misconstrued, and still and not passing muster. Thoughts? Not trying to rock the boat just make a cool picture of heaven.

quiet dagger
# pearl fog You need an account to view Twitter now

*paid subscription account to view timeline. With recent changes daily limit hits within 5 read threads. Free accounts without blue checkmarks are able to view only 300 tweets daily (replies in large threads fill the quota too)

stable pawn
#

One thing that I think is fun in SD and A1111 is the "Interrogate CLIP" that try to describe the content of the image, but would it be possible to do the same but where it just try to describe the graphical style to make it simpler to create a "style".

plucky solar
#

Is there a way to recover a seed?

stable pawn
#

Images you render is stored in a Output folder, you can go there, find the image and then drop it in PNG info

#

C:\StableDiffusion\stable-diffusion-webui\outputs\img2img-images

#

or "txt2img-images", that is how I recover seeds and get ashamed over things I made 😄

plucky solar
#

Being on a mac, sadly, that is not what's happening. Is there a tutorial on using img2img? I've generated an image I would like to continue using and am too much a noob to know how to pull it back into my prompts.

stable pawn
pearl fog
#

You can talk to other people about stable diffusion and test the new model in the bot channels

stable pawn
#

I can not answer on that, but I have got great help here from other users when it come to settings and models.

pine fulcrum
#

how do people get these amazing 3d images with backgrounds and full body characters, it could just be the modle isnt ment for that but like, this is really annoying

orchid wasp
#

hi why when i try generate img2img they generate me a ugly face:(

warm junco
empty dove
#

anyone know where i can get access to and start using SDXL? Thank you in advance

broken smelt
pine fulcrum
#

Hi

stable pawn
#

I not tested the bot or SDXL yet, I do not know why, but I seen some result.

#

Darn. As I have explained, I had a stroke not long ago and my logic and memory sometime drop so I feel a bit stupid: What do I need to do to test the bot or find some fast tutorial/instruction.

#

I feel okay and can walk and talk and so, no need to be extra nice, I can take some "What a noob" jokes 😄

pearl fog
#

It will let you add negative values and choose a style too

stable pawn
#

Thank you. I was reading up, then YT got my attention, I get distracted so easy 😄

stable ether
#

Excuse me people, I installed a local stable a while ago, I was looking for a way to generate a high qual texture of an old game texture but tryint to keep it similar to before, is there a name or term to learn that?

pine fulcrum
#

Anyone know a good model for generating dnd character art?

stable pawn
#

My last prompt indicate that I need to sleep; "whatever trending on artstation but in 8k and masterpiece".

pearl fog
frank spruce
#

working with two tabs in automatic1111, i get a bug where it thinks its using one model but it actually loaded another, and its impossible to load the one it thinks its on without switching to a third model to load and waiting. just thought of a workaround, i could switch back to the original model in the second tab, but its still a bug to fix

livid river
#

Anyone who knows how to train human LoRAs, hmu, have a big business inquiry

night hornet
#

What's the style of art used for this Discord channel's banner?(Lady giving ice cream to the dog) is it from a particular artist or style?

#

I want to get a similar art style in my results

frank spruce
dusky stream
#

hi

night hornet
#

Do anyone have any links or sites that go into the details about prompting?
I saw someone using ((parenthesis)) and "BREAK" but couldn't find any guides that explain these things

The resources channel doesn't have what im looking for

plucky shoal
#

(()) more value on word inside. More parenthesis, more attention. BREAK should add another 75 tokens available.
Or is in prompts from regional prompter i mean BREAK.

night hornet
pine fulcrum
frank spruce
#

ngl if i go over 75 tokens i just get rid of stuff i didnt need in my prompt, simpler prompts is better

fervent thunder
low heart
#

hi

#

how can i delete completely stable diffusion ?

warm junco
magic solar
#

im thinking of setting up SD on my PC but can you generate images higher than 512 locally like 1440p? most ai image generators i have seen all have the same 512x512 limit.

warm junco
magic solar
warm junco
magic solar
warm junco
magic solar
#

alright thanks man

warm junco
magic solar
warm junco
magic solar
warm junco
#

No problem 🙂

magic solar
warm junco
past pumice
#

are there any models using sdxl 0.9 in civit?

umbral heron
#

It wasn't publicly released so if there are any they violate the NDA (?)

bleak matrix
#

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

shell quail
sour needle
#

Can anybody tell me a website for free generation?!?waow

fervent thunder
#

Can anyone create a lora out of images of a friend? She did consent

#

The 'v's are killing me

knotty turtle
#

G'mornin' friends. Question about SDXL (as a non-techie)...
Once SDXL releases and we can use it locally, will it require a new installation of SD? Or will it be as simple as downloading a checkpoint and using it in A1111 like any other checkpoint?

frank spruce
#

who knows, im assuming they will do everything they can to control how we use it

#

and my GPU wont be stronk enough T,T

fervent thunder
#

Guys

#

how to transform pt into bin file?

#

or safetensors into bin

tribal pagoda
#

is there a way yet to edit faces using controlnet? been googling like crazy but cant find anything

unkempt glacier
#

Would you guys choose a 7900xtx or a 3090?

mental lintel
#

How do i put an image and stable modify the image guys

bleak matrix
dapper ferry
#

i would choose a 3090 i presume it has tensor cores that give it more versitility.. i could be wrong

sudden ruin
#

Also asking in one channel should be enough

pale latch
#

3090 for gaming too

pearl fog
#

I've heard that a 24GB AMD card is equivalent to a 6GB Nividia card on windows, at least when it comes to SD/ML stuff

unkempt glacier
#

Alright

pale latch
#

i was an amd guy for a long long time. ML made me go with nvidia. it's the obvious choice when you consider ML purposes

unkempt glacier
#

I’m concerned more because I’ll also be doing video editing, which the 7900xtx seems to benchmark higher in

pearl fog
#

It depends on what's more important to you and what you'll be doing on it more

unkempt glacier
#

I guess I’ll just get a 3090 for now and later on a decent amd gpu

pale latch
#

AMD used to be a good budget choice for gaming but now their cards are "competitive" with amd. They price them as high as they can . Back when i bought my rx 480 with 8gb, it was cheaper and better than the gtx equivalent. They're not that AMD anymore

pearl fog
#

I wonder what will happen when Intel finally gets around to making GPUs

pale latch
#

they are already. its okay

#

first gen is nothing to rave about but it' nice to see a new competitor

pearl fog
#

I heard they're planning higher end ones in the next few years

pale latch
#

and in a few years, nvidias will still be better then too. intel has to make serious moves and breakthroughs to significantly compete here. It's a good card with performance vs pricepoint being competitive, but it's nothing extraordinary.

#

nvidia is just sooo far ahead of everybody else in machine learning. they made smart moves years ago by investing r&d

#

jim kellers got his tenstorrent company that i think will make some industry breakthroughs that may level the playing field a bit.

mighty gazelle
#

anyone using comfy ui here?

pearl fog
#

Comfy is actually in this chat

mighty gazelle
#

hello @neon oriole hi sorry for ping i er. need small help

neon oriole
#

yes?

mighty gazelle
#

its not a problem with the software but I was wondering if I could save the metadata in the output images, currently I can't upload to civitai due to missing metadata

neon oriole
#

comfyui does save metadata to the images so you can drag them on the ui to get the workflow

mighty gazelle
peak portal
#

@shadow trellis have you seen ADVChina, or SerpentZA?

shadow trellis
peak portal
#

ohhh i was just saying because they're white husbands too

#

they have some funny stories about it

#

i don't pay attention to their fearmongering or propaganda

#

they really used to love China and then they left and now it's like they're really upset

strange cove
#

hey someone can tell me how to use Stable Diffusion? I use to use it want it was free and I'd like to buy it but dk how

light solar
#

hey let me know if this is the place, ive been trying to run 'git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git.' an i keep getting 'fatal: could not create work tree dir 'stable-diffusion-webui.git.': Invalid argument'.

I was suggested 'git clone https://github.com/deforum-art/sd-webui-deforum extensions/deforum
Cloning into 'extensions/deforum'...' and that is currently working, but surrely that still requires the stable diffusion!? I have GIT and python, can anyone tell me what is going wrong, or should i try the 'installation on Windows 10/11 with NVidia-GPUs using release package' route instead ?

warm junco
vocal kettle
broken smelt
#

Until a model trained on freely available images comes out of course

sleek raven
#

can we dreammbooth 1.0 that's going to be released?

pearl fog
#

I wonder what it would look like if someone trained off of public domain images like old paintings, Wikipedia stuff, etc.

vocal kettle
vocal kettle
#

Btw, what status will be of the art, that will be 100% completelly painted over AI art by hand? I mean, you will copy by hand existing art, but you are copying art that does not belong to anyone, a public domain art.

pearl fog
#

I mean, public domain means that no one owns it so anyone can do anything with it. I already know there are redbubble accounts that rip AI art from other people and resell it on merch.

vocal kettle
# pearl fog I mean, public domain means that no one owns it so anyone can do anything with i...

yeah, but if art will be 100% drawn by hand, and only using ai art as guidline it's perfectly fine for being an art that you can own and claim. I think. if you will drawn over existing copyrighted art that does not allows derivatives than you will be in trouble. Like some artists that painted over photos of real people without right to do so. but what about AI art? Who can claim it and accouse you of creating derivatives of it using it as guidline and base? I could be wrong.

pearl fog
#

I guess it depends on whether that specific AI image is public domain. It's actually possible to copyright an AI image if there's additional human effort put into it.

Example: someone sets up the scene/composition with a sketch or 3D model and then uses a controlnet to keep it consistent in img2img. They then manually edit on top of that to finish the picture. That image could be copyrighted.

#

So if you're generating your own images, then you can be sure that it's public domain. If you're using random AI images, you might be infringing on copyright. Also, the rules vary by country. I heard that unedited AI images can be copyrighted in the UK, for example.

#

But copyright is a grey zone. Is something transformative enough to be unique or is it infringement?

#

It can be hard to say. Certain images started out okay, only to later be considered infringement

vocal kettle
#

There is so much uncertainty right now. I guess we will have to wait till all the lawsuits got resolved and companies like Steam will make consistent policy regarding AI works.

light solar
vocal kettle
#

Right now I'm thinking about 100% drawing over AI art, so there will be no pixels of original work left. I hope that will make everyone satisfied.

#

IMO that will be a a completely new work that you can own the rights to and use in your projects.

light solar
#

my hope is to use my 3D animation or my artwork and use control net and deforum to either add to that or allow me to create very simple 3d scenes that i can then flesh out with deforum and control net

#

wondering if it will save time over creating every single element in blender/compositing in after effects etc

hasty hornet
hasty hornet
stark flint
#

"the application is not responding" when I try a "/dream" someone can explain?

crimson belfry
#

Im seeing a ton of hate towards people like us that make ai art.

#

its sad really because most of the hate i have been getting is from regular artists

pearl fog
#

They're worried about their jobs.

crimson belfry
#

i think so as well

#

flo mine is buggy also, my prompts stopped going through

stark flint
crimson belfry
#

I agree!

stark flint
#

i can't dream, i'm addicte, help

vocal kettle
hexed jasper
#

most of the hate i have been getting is from regular artists
even regular ppl dislike AI art 😢

crimson belfry
#

I was in the middle of making a pretty funny caricature

#

We are at the start of a new era and people do not accept change at first

#

They do not understand the power that we have now

#

while we open our minds all the way up they will catch on

foggy star
#

This is the same as automation in factories taking away jobs from unskilled laborers.

crimson belfry
#

creative expression is going to thrive, You can say that this is compared to the industrial boom in the usa vide

vocal kettle
# pearl fog They're worried about their jobs.

Artists have a point. But then there just a lot of cases where if job will not be done by AI than it will not be done by anyone. Small indie projects without budget struggles to find artists willing to take a risks of wasting time for nothing (if project dies before being finished or won't meet expectations).

crimson belfry
#

Artists will always have an advantage over someone who does not know what they are looking at

pearl fog
#

Yeah it's easy to fix stuff like broken hands if you know what you're doing

foggy star
#

It allows for more people to throw their hat in the ring and make larger projects that don't require as much time or as many people as previously

hasty hornet
pearl fog
#

Even artists could use it for longer form content like comics or animations. I've seen so many webcomics stop because they don't have time for it

crimson belfry
#

Stable diffusion will thrive because its so open

#

there is nothing to hide and i like that

#

as artists usuing ai we want to tell people that we work for that they can be confident in the product

#

i do not see a private ai doing this well except maybe adobe

#

we all know where adobe is going......money,money and more money

pearl fog
#

Stuff like controlnet and inpainting gives you greater control than midjourney where you have to take whatever you can get

crimson belfry
#

midjourney will last as long as people do not have knowledge about ai, I do not see why we would pay so much. Its like paying taxes on your taxes.

#

Ai is as good as the people allow it to be

pearl fog
#

I'm hoping future versions of SD are easier to install and take lower requirements. That's a big hurdle I'm seeing now

crimson belfry
#

Yeah for me i am using a decent laptop but i dont think i can handle it. i agree with you.

hasty hornet
#

requirements are going to be higher...

pearl fog
#

Yeah I can see that too as it becomes more capable and people call for higher resolutions.

#

Maybe Nividia or someone will start adding more VRAM :U

frank spruce
crimson belfry
#

Is everything halted for everyone?

#

I am getting errors still

#

nvm i think its back

light solar
#

i mean i work in film and art, and esspecially in film cus most people are older no one gets AI at all so they are less hating it more thinking u are doing magic

#

and it doesnt really ever superceed classic workflows and art making, it can only help

pearl fog
#

Look at the online art communities and AI is either NFT2.0 or the apocalypse

light solar
#

AI cant get anything as specific as u can imagine but it can help push what u imagine to have more polish quickly

#

yeah i know

#

its funny, people are scared of anything new though arnt they

#

and for compositors and other specialised VFX people it is pretty bad lol

pearl fog
#

People freaked out about recorded music, they freaked out about photography

light solar
#

exactly

#

jobs always go

#

creativity always continues

#

people seem to think AI is a replacement for that

#

but just look online, at the amount of generic AI art

#

just goes to show u cant replace actual creativity

#

but NFT's really arnt there yet either

#

when they work as a way for artists to keep profits from resale of physical work by having a digital sort of certificate as well as 'the jpeg' they will be something really good

#

thats one of the main problems in art world today so, if it can help, great.

pearl fog
#

But artists rejected them even though they're just a 2020's version of an adoptable

light solar
#

im not

#

but also not publicly shouting about them as its got such a bad rep as sort of a ponzy scheme from sleezy artists

#

just trying to find the tools that make the workflow easier to get the ideas out

#

used midjourney extensively on the last HBO project i was on

#

but not as a final product

#

just a helping hand

#

it did confuse the fuck out of them though, as they always need to 'clear' reference imagery for copywrite etc and i was like, well i made this in midjourney then painted on it in photoshop and my account says it belongs to me. So had to sort of make them some kind of certificate to show them this and also ask midjourney to make a way to make things u generate private

#

i guess loads of concept artists asked that though cus after that they came out with a new tier of subscription lol

pearl fog
#

It is very good for quick concepts

light solar
#

amazing for what i was doing

#

when u cant get an actress to pose for the picture needed, u can just invent it with a bit of training from google images

#

was doing, prop paintings

#

no way i could of done it without AI

#

had 5 weeks to do 6 2m oil paintings

#

an they didnt know what they wanted

#

my fear is it sets a bad precedent

#

by me doing that

#

it normalises such insane workloads

#

like how VFX houses outbid each other and then overwork and underpay

pearl fog
#

In an ideal world, the workload would stay the same and the AI would help lighten the load, but this ain't an ideal world.

light solar
#

its up to artists to make sure, but as a younger guy in that world i have to worry about impression i leave

stable pawn
#

I did it again, I wanted to depict an artist working on digital art and I wrote, "editing in gimp", that is the program I often use, but my model just put the human in a vert tight leather outfit that is very NSFW.

pearl fog
#

Yeeeeeah that has always been an unfortunate program name.

light solar
#

btw do u need to run the cmd in the stable diffusion folder every time u open the browser program to update it?

#

or can u just paste the URL in for your diffusion without worrying

#

cus ive installed deforum and control net now, and im wondering if i need to do that

#

im still not getting the models i installed when i click the drop down i get them in preprocessor drop down but not models

light solar
#

restarting did work im just being a ludite

light solar
#

one question though, do u need internet connection to use it at all times? i restarted my router so wondering if that will effect when i generate video. Or can u run SD without internet connection?

pearl fog
#

SD doesn't need internet

old ruin
#

hello

#

we can do a 16:9 here ?

pearl fog
#

Yeah, there are multiple options for the aspect ratio

old ruin
#

what is the propt for do 16:9

light solar
pearl fog
light solar
#

i wish there was a way to tag/mark a character like wonder dynamics are doing so you could target a vid more specifically with deforum and control net

willow coyote
#

i always post-process my stuff using my favorite filters ported to ruby, makes things look even more realistic

past vortex
#

huh, Dreamshaper was taken down off Civitai

#

wonder what happened

frosty jackal
#

Guys, is the platform for windows and Mac OS only or can I use it on android phones and iPad OS as well?

pearl fog
#

Not on Android (at least not locally), but there is an ipad app

pearl fog
old ruin
#

bro how i can have same result at this image here ?

#

i send them in general with image

toxic oyster
#

Sup guys.

#

I’m new here

#

And I keep dancing around how to learn SD. Any good torrents/tutorial videos I can jump into?

#

I keep seeing some wonderful work online of wonderful anime characters generated. I want to try to generate some of my own stuff, but I don’t quite understand how it all works

marsh compass
#

question, how big should stable diffusion datasets be for good performance?

high lichen
#

Has anyone been able to get perfectly spelled text or words to display consistently? Lora training? or best prompts?

sonic tundra
potent spire
#

Basically everyone can take your assets and use them as they want

#

One of 3 major reasons serious companies dont use SD

#

Or any of the AI art generators

#

Adobe Firefly and Nvidia might change it

late laurel
#

any stable diffusoion prompt guides?

shadow trellis
#

baltimore, Maryland,what happened just?

fervent thunder
#

is it possible to dream from an image with the discord bot ?

desert aspen
#

QUESTION: Is it ever possible to replicate "text + picture" like Midjourney in Stable Deffusion?
For example, I take a picture in Midjorney and add text to it. Midjorney somehow understands what is drawn in the picture and mixes it with the text of the prompt. If you do the same in img2img in Stable diffusion, you get a completely different result. Even if you use control net - reference. This is not the same. Is it possible to repeat this in Stable diffusion? Or is this a unique feature of Midjourney?

cyan comet
#

is there a "tech" help here

thorn pilot
cyan comet
#

oop

thorn pilot
slim gorge
#

just need feedback and to see if it works across other peoples diffusions

fervent thunder
#

a

real plover
#

hey

eternal otter
#

Hi everyone ! Do you know how I can have a more detailled image ? I did this but I don't really like
the character smoothing #1101178553900478464 message

past vortex
#

Hires fix

#

Despite all the YouTubers banging the image to image drum, hrf inserts more detail

serene swan
#

any news on civitai?

#

crap crap crap

rigid oracle
#

@serene swan regarding?

serene swan
#

its been down and up, barely loads anything

inner tapir
#

Is there a limit to how many times we can use the SDXL dream thing?

peak portal
past vortex
#

it is, but in my experience it works better than manually img2imging

#

depending on the model

peak portal
#

it's just literally the same thing

jaunty hazel
#

So, I use SD locally buuuut, I want to better stylize images like photos and I struggle with that. What can I do to use SD for the whole avatar creator thing?

tulip yoke
#

here comes JOHN CENA

#

DUDUDUDFUUU

frank spruce
#

but i agree increasing resolution is vital for fixing certain things

past vortex
#

Yeah that's what I meant, actually scaling up the resolution

#

A lot of people do straight img2img and go to the same res in tutorials

#

Then upscale after with extras

#

That's a stupid way to do it because if you do the img2img and upscale together, the quality is way higher

frank spruce
#

you can do img2img at any resolution with controlnet and ultimate sd upscale script

crimson belfry
#

How is everyone today!

tulip yoke
radiant nacelle
#

Bro who the fuck is steve jobs?

pearl fog
#

A spooky skeleton

sweet night
#

Hello!

#

Any channel for ClipDrop help, suggestions, support?

frank spruce
#

so is CFG what makes your image explode halfway through generation in the previews

toxic oyster
#

Sup guys

willow needle
#

One thing I can never get right is the damn hands and fingers. I try negative prompting them out but they're like herpes, they keep coming back

pearl fog
#

The only reliable way to fix them is with an openpose hands controlnet.

tepid spindle
#

i am the first person to say "serial killer" in this server

junior jetty
#

Hello

#

Where can I ask a question about LoRAs?

fervent thunder
#

anyone know of a job board for people interested in working in AI? I have architecture and product management experience

teal oxide
#

can gtx 1650 4gb use stable diffusion?

pearl fog
#

Yeah, I believe so. It will be slow, though.

valid charm
#

I am currently dead because of @radiant meadow

pearl fog
#

Is it going full terminator?

heady flax
#

How do yall get such consistent animation results, im still learning controlnet but i haven't been able to render anything as good as some of these

plain copper
heady flax
plain copper
tulip yoke
maiden drift
#

hey y'all i'm new to stable, how do i make good prompts? i dont know how to use any of the --xxx tools, only the words, and bc of that it doesnt always come out the way i was thinking it would

marsh compass
#

how do you start creating your own dataset

#

is there like a software or something?

small vigil
#

this is a dumb question. i added SD discord to my discord. I'm in the app now typing this. But it won't show up on my server list on the left. I have 15 other servers, not sure why this one won't pop in there

hidden sail
#

hello, hope y'all are doing good. i have a question. is there any way i can make these new trendy ART AI QR codes with the server without needing to download whole SD model on my machine?

sudden ruin
fervent thunder
#

If I move my stable diffusion folder into another location, will SD still work?

warm junco
fervent thunder
#

Wdym

fervent thunder
#

Can i git clone from one folder in my computer to another?

warm junco
#

Nope git clone is from the internet to your computer

fervent thunder
#

Hm

#

I know that the last version of the webui broke the extensions I use

#

So that's why i'm a little hesitant

warm junco
#

Which extension?

fervent thunder
#

But doesn't matter, i'll just try to clone the previous version

#

What do i do?

warm junco
#

You dont know which extension is broken ? But you think an update broke it ?

fervent thunder
warm junco
#

The first thing would be to Update the extension

fervent thunder
warm junco
fervent thunder
#

But still

#

I don't feel like updating yet

warm junco
#

Yea sure you dont need to.

#

But if you move the complete sd webui folder to somewhere else, you break the file path for git.
Meaning that you cant update the webui anymore with git pull

#

To fix that you would need to run the git clone command inside the webui folder again but that would also update it to the newest version

fervent thunder
#

Ok

#

How do i run that command?

#

Just git clone or do i put something else in there?

warm junco
# fervent thunder How do i run that command?

After you moved your stable-diffusion-webui folder go into that folder. Then click in the Explorer bar and type cmd then press enter.
Then you copy and paste this command:
git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git
Press enter and after its done you can close the cmd and launch the webui-user.bat

fervent thunder
#

Or the settings I had

warm junco
#

Nope it shouldnt delete Anything

fervent thunder
#

Hm ok

warm junco
#

May i ask why you want to move it ?

fervent thunder
warm junco
#

Or by creating a symlink with the linkshell extension programm

fervent thunder
#

The what

warm junco
# fervent thunder The what

You can edit the webui-user.bat and there at the line Commandline_ARGS=
add --ckpt-dir "D:\YourPathToTheNewModelsFolder"

quasi coral
#

Hi guys, I'm trying to use the A1111 api and I can't figure it out how to interact with the scripts, for instance I'm trying to access the "depth image IO" script and pass the parameter required like "append" and "the depth image", is there any docs or any tips, thanks in advance, any help is much appreciated

marsh compass
tribal pagoda
#

can u install automatic1111 sd on external drive

#

i have like 6gb left on main drive and the process takes up all my space to install it on conda

#

cant figure out how to have tmp folder on another drive

wind hill
#

how can I generate image to image with promps?

sudden ruin
wind hill
#

yes, i mean in the UI, ¿how can I do that?

bleak matrix
#

How's everyone this morning? sunsmile

sudden ruin
weak mist
#

So, I saw audiopalm paper. I wonder how fast something similar will be implemented in open source space. Than we will be able to enjoy language models with built-in ASR and TTS

tired dawn
#

bot down ?

#

@wild steppe @sudden marlin @wintry stream ?

warm junco
wild steppe
tired dawn
#

okey, thanks !

marsh compass
amber wolf
#

if im in the wrong spot, please let me know where i should be asking. I'm trying to find a docker image for lora training, and one for dreambooth. I'm struggling to do either on my ubuntu system. My 12year old is interested in doing a AI related project for a science fair, we have image classification working but she wants to do the reverse, and generate images. SD is working for that, other than a lack of a specific model she is after

rancid tinsel
#

How I can see nswf content?

amber wolf
north spade
#

Please help me every time i try to generate an image in auto 1111 i get an error, so much that ive lost count, this makes me want to smash my setup and microwave a baby, I can’t stand this anymore

#

Should i redownload 😭

harsh fox
#

You could start with what error you are getting... @north spade

tidal lion
#

hi guys , i am learning about prompts in civitai . in the prompt there are keywords about lora like lora:add_detail:0.9 . Does anyone know how to find and download lora with keywords in such prompt?

harsh fox
tidal lion
warm junco
north spade
#

Ok thanks

#

I’ll post when i get an error later

harsh fox
tidal lion
#

woa love you so much 😍

#

you are wonderful

north spade
#

Wait now there aren’t any errors showing in the command but the generated image won’t appear

harsh fox
#

Open your output folder and see if it is saved there.

spark kite
north spade
#

Thanks

#

Found it

#

Thanks a lot

#

One last question, does arg: precision full no half make generation slower ?

#

There’s not much info about it in Google

pine fulcrum
#

When I use stable diffusion to generate art of characters I can’t have them holding stuff, the image just breaks down, anyone know how to fix that?

pearl fog
#

You probably have to generate the character and then generate the object separately

#

Inpainting

#

Depends on what you're trying to do, though

pine fulcrum
#

Any good guides on how the process works?

ember ivy
#

Hello, can anyone send me ulzzang-6500-v1.1? It is not available for download anymore and I was looking forward to trying it out.

frank spruce
ember ivy
verbal harness
#

day one - being homo
day two - being gay
day three - being transgender
day four - being straight
day five - becoming emo
day six - wearing girl clothes
day seven - wearing makeup
day eight - change of name
day nine - starting HRT known as hormone replacement therapy
day ten - lost interested in transitioning.

silent parrot
#

hello guys. im NOOB here. is there any website that i cant learn how to write promte for SDXL ? or if its here can u help me to find the channel pls?

light solar
#

does anyone find running on D drive slows things down rather than C, im currently trying to install xformers but, lmk any other things i can do do aide the process. thankyou

thorn pilot
light solar
light solar
ember ivy
silent parrot
#

guys pls help me im super noob, how can i download stable diffusion ? i mean i want to use it on my PC not discord. i cant find any link .

light solar
#

im really hoping xformers can be a big help, its saying 1.5x faster from what i read

light solar
#

is this an issue also? '[MSC v.1932 64 bit (AMD64)]' my GPU is nvidia? but my SD runs saying this, is that just part of it, or do i need to change ?

silent parrot
light solar
#

im pretty new here too, im afraid, and i ran what that same guy said would work with deforum and controlnet

#

in his other video

#

im just trying to get xformers running with it!

pearl fog
silent parrot
pearl fog
#

SDXL is a separate new model that hasn't been publicly released yet

silent parrot
light solar
#

i think im on 1.5 for control net

#

but i have a bunch of models

silent parrot
#

can i use v2 models from this site if i use SD 1.5 locally ?

pearl fog
#

The models are standalone, so you only use one at a time

light solar
#

@pearl fog do you know how i can get xformers going with my deforum and controlnet by any chance, ive not found much online to help, i dont really want to delete and reinstall everything as some suggest

pearl fog
#

You can use whatever you download on your computer. Think of them like game downloads where you can switch between them

silent parrot
pearl fog
light solar
pearl fog
#

/shrug

light solar
#

will try there now, thanks

waxen storm
#

any word on the new .9 model release for local?

frank spruce
#

sdxl is a myth, they actually just put a bunch of midjourney pics as a joke to trick us all

iron pendant
sudden cedar
#

I don’t know how to use stable diffusion, it hurt my brain

frank spruce
#

Is there a reasonable way to create LoRAs for normal users? I watched two installation videos and ended up with multiple bugs and errors

#

i dont want to handle the kohya bullsh**

#

reminds me of when i tried to use linux once

#

the user process should be downloading an exe, running the exe, and having it run instantly and perfectly without any command line inputs

#

whole thing reeks of "this is really a linux product but with some workarounds it's technically possible even windows users could use it kind of"

#

kohya_ss

weak mist
frank spruce
#

and then the thing dont work anyway like, maybe ill try this again in 5 years

stark mason
#

curious

sudden cedar
weak mist
pine fulcrum
#

does the order you place your prompt in matter?

frank spruce
pine fulcrum
frank spruce
#

its a file you can link in your prompt to make something the model would normally struggle with

pine fulcrum
#

im asuming its an acrynym

frank spruce
#

theoretically you should be able to select the files you want to train, drop them onto an EXE, and it outputs the lora safetensors file, no we have to go full linux on you

stark mason
weak mist
tawdry bramble
# frank spruce the user process should be downloading an exe, running the exe, and having it ru...

this is a very entitled take, you realize the stable diffusion people are just AI software engineers (mostly), they just make the AI, everything else is 3rd party and not in their job description, stable diffusion is made to run in basic command terminals it doesn't even have a user interface, all the GUI sites are 3rd party, automatic1111 is third party, by people who did it for free, for everyone to use and are not affiliated with stable diffusion (with some exceptions)

packaging, the AI + GUI + python + frameworks and other dependencies into a single executable is work that is in nobodies job description or responsibility, luckily for you though someone did it, for free, and it wasn't their job to do it https://stable-diffusion-ui.github.io

I realize now you are talking about Lora's, and to that my argument applies even more-so

weak mist
broken smelt
blazing pasture
#

Does anyone know what model stable diffusion xl's "cinematic" is? I'd like to get it for my desktop.

sleek jasper
#

Hi trying to create a textual conversion should i add my character emotions (smiling, angry, looking at camera) in text files

drowsy robin
#

Damn, I am so excited for SDXL...

hardy swift
# frank spruce linuxy, but okay. but then you have to change security settings, install a bunch...

the "linuxy" steps are an unintended barrier of entry to new people probably. i think they dont want to make a .exe yet for this technology yet because its new.

its a gold mine for security breaches since most people that want to generate art on their machine have no idea whats actually happening in the background, hence, it can become a security issue.

i can imagine someone making a .exe file that claims to train LoRA's but its actually just a trojan

willow needle
#

I've been using an insanely long negative prompt for a long while, but now when adding in multiple LORAs(Character, style, etc) I get no result with the "Try adding --no-half-vae commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.". I'm assuming long negative prompt should just be snipped to the basics instead of adding more commandline arguments?

warm junco
#

Also yea dont use insane long negative prompts

#

You could use the negative embeddings instead

willow needle
#

Fair enough. I'm still a bit new to this so glad to know I need the argument. I'm unfamiliar with neg embeds... Those are like LORAs trained to avoid specific/usual problems?

warm junco
willow needle
#

so "set COMMANDLINE_ARGS=--xformers --no-half-vae"
Is what my commandline should look like? Without quotes ofc

I'll def grab a few of them to help with certain character models, seems like they were trained on pastels and oils and blur easily, not to mention fingers. Just grabbed a VAE for eyes, it's working nice.

#

Last question for the day if I may. Steps for your pictures are pointless to go above certain amounts when your character/art has already come together? Like doing 150 steps on every image is pointless or maybe counterproductive?

#

Seems like higher values have better results overall but I'm not experienced enough to know for sure

warm junco
willow needle
#

I see. Thanks a bunch for all the help man.

warm junco
#

No problem 🙂

warm junco
#

Usually done with the Highres fix

willow needle
#

I use 4x_fatal_anime_500000_G lately, it works great. My issues arise when using character LORAs more than anything else. Maybe they're not trained on a huge dataset, or maybe I'm just adding extra crap

warm junco
#

Some are overtrained

willow needle
#

Aye, if i understand it well enough, taking a LORA at 0.8 would give it serious impact on the picture but another LORA at 1.0 would override it with its nuances?

#

I find most characters work well 0.6-0.8, the ones that the uploaders recommend at 1.0 usually give me conniptions

warm junco
#

Yea it also depends on how much loras your using. Mostly 2-3 should be the max

willow needle
#

Understood, figured as much when they came out like Picasso on drugs

warm junco
#

Also always seperate them by ,

willow needle
#

I find some LORAs come out quite splotchy, sometimes blurred and unfocused. Then all of a sudden I create a super sharp detailed image using the same one that's been giving me problems... Have you encountered someone messing up their pictures the same? Sometimes I even find myself using the exact same creation data with the same seed and upscaling just amplifies the problems

#

I figured it must be mismatching of model, prompt, and LORA, but if there's one thing I know it's I don't know a lot

#

Oh, wow, --xformers makes a huge change

#

I'm seriously glad you told me that lol

warm junco
#

or some models too

willow needle
#

I see, thank you. So I probably just need to keep experimenting and noting what works with what

#

Oh yeah, CodeFormer or GFPGAN face restore?

warm junco
warm junco
willow needle
#

Oh, face restore is for lower res pictures? good to know

#

... That's probably part of what's screwing me over lol

#

Dude xxformers argument is amazing, I was waiting 15minutes for a single upscaled image

warm junco
# willow needle Oh, face restore is for lower res pictures? good to know

no face restore is an option that is good to get better faces on non upscaled images.
But highres fix can do it mostly better while increasing the image resolution.
Both together will break the face/eyes so only use highres fix.
Face restores is still a good option if you use it in the extras tab

willow needle
#

I see, thank you. I shouldn't have been adding the two together then, certainly part of my mistakes

warm junco
willow needle
#

Did not know about controlnet, looking into it now, holy crap is the SD guide a big one lol

warm junco
willow needle
#

Thank you, but for tonight I think I'll mark and save it, probably want some sleep before I go at this one. I'll read up all I can before heading to those rooms ideally

warm junco
#

it can be a bit overwhelming

willow needle
#

Thank you for all the help. I'll try to do the same if someone asks the same questions

warm junco
#

nice and no problem!

willow needle
#

Here I keep saying last question, but one more... If there's a particular LORA(POV perspective), that has a tendency to change characters hair, hair color, and other things... Is that a result of too strong of training and nothing I can do other than try and counter its tendencies with negative/positive prompting?

warm junco
marsh compass
#

how do you start creating a SD dataset with LoRA?

willow needle
#

I see, I hadn't even thought of upping the strength of the countering clauses, lol

marsh compass
#

Sorry, you have been blocked
You are unable to access aituts.com

#

big bruh

willow needle
#

wat

marsh compass
#

it just says im blocked

#

:/

willow needle
#

use a proxy

bleak matrix
#

Good morning, everyone! How are you all this beautiful Wednesday?

compact nimbus
#

hellooo

bleak matrix
#

👋

silent parrot
#

hello guys do you know if its possible to use Image in Bot 1- 10 for using imgtoimg ?

modern ferry
#

can i make a lora/embedding of a specific object but it has to be same. For example, I want to create a chair object and I want to assign it to a specific code or prompt. Is it possible? to be more clear, let's say I wrote "ikea2" on prompt line and will i get the same image of the chair that I have already created(lora/embedding) on the output? but it has to be EXACT image.

sudden ruin
#

Maybe include what you need help with