💬｜general-chat

[wanton vigil]

I’m an idiot. Forgot to hit apply settings

Holy shit, Token merging legit just double my generating speed

[dawn topaz]

by chance has anyone used this extension before? i want to install it but not sure what exactly to do, or if its even worth it, the last extension i tired to install to organize my images didnt work. hoping this one is decent, anyone using it? https://github.com/RupertAvery/DiffusionToolkit

[cunning idol]

hello

I got the image from SD and I tried to add my eyewear alpha image to her eyes by using inpaint, but the result is shixxxxx. anyone know the way to solve my problem?

[rugged mirage]

As far as I can tell SDXL on its own seems to be significantly better (as in, it gives good results more often) than SD 1.5, finetuned SD 2.1 can be good but has a worse prompt understanding, and DF is maybe better than 1.5 but not super clear

do people agree that looks acurate so far?

[simple siren]

is there a SD specific prompt list to generate ai avatars?

[wind quail]

᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼

᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼

᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼

᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼

᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼

᲼᲼᲼᲼᲼᲼᲼😶‍🌫️᲼᲼᲼᲼᲼

᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼

᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼

᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼

᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼᲼

[broken frigate]

hey, anyone here use cloud service for AI? like g collab?

if so can I ask a few stuff, thinking of renting them

[violet delta]

hello
i need some help for Stable Diffusion, before i have geforce 3060 12gb and everything work good,
i upgrade to geforce 3090, when i hit generate, my pc is restarting what can be problem?

[plucky shoal]

cant be problem weak psu? @violet delta

the difference is 180w

[violet delta]

psu i have, for cypto i have install two 3060, there working like charm :/

[plucky shoal]

3060 x2 = 340w
3090 = 350w
Your hw could be affected by crypto thing.
What psu do you have?

[violet delta]

deep cool DA700

[plucky shoal]

and recent intel cpu?
Can be faulty gpu, but i still think psu is issue.
But better to have second opinion.

[violet delta]

i will research, thanks you very much @plucky shoal

[plucky shoal]

no problem i can be absolutely wrong...

[fervent thunder]

hii

[fervent thunder]

thanks.

[past topaz]

do you know if there is a way to access the older beta GUI of dreamstudio that has 1.5?

[cosmic crescent]

Anyone actually seeing performance improvements with today's driver update?

[young wind]

how to make an accent on word in prompt guys?

[peak portal]

fine-tuned 2.1's prompt understanding can be night and day depending on how you fine-tune the text encoder and unet

[crude chasm]

Anyone else having issues with the negative prompt not going through?

[peak portal]

colab is a scam

use it for free and never pay, it wastes your money like mad and isn't much better than the free tier, when you do pay. and if you pay, you can no longer use free tier until you burn through your credits. and it won't let you not burn through them.

[fervent thunder]

You can't use webui on free tier.

[peak portal]

webui isn't the only way to do SD.

[fervent thunder]

is there any decent non webui SD notebook?
I mean like...different model uploads, vae, lora, embedings, controlnet?

upscaling...

[peak portal]

if you want to do all of that you should probably spend some kind of money.

[fervent thunder]

or well, at least basic options, like models, vae, cfg, resolution,samplers...

[peak portal]

i guess you could try running StableStudio on it but chances are that's blocked too

[fervent thunder]

it is webui, so yea

[true dawn]

i made a little ui for my 2.1 model if you wanna try

it runs off the cloud

[peak portal]

2.1 doesn't do VAE swapping or something

[true dawn]

https://rod.dev/generate?id=cgfmMvzavav

[north nimbus]

Is dreamtextures a good addon for blender or is there something else that's better?

I'm also installing AI Render.

[bitter olive]

Can someone lead me to a documentation where I can learn how to use signs in prompts like these:

" -- (()) -- ((())) "

I've seen it a bunch of times and I think it has to do with giving something more weight, but not sure 😅

[fervent thunder]

if it's for auto1111's webui, then it might be this you're thinking about
https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#attentionemphasis

[bitter olive]

Oh yeah, thank you 🥳

[broken frigate]

So, is there alternative service to use cloud that good then??

[abstract quarry]

diffusers

and all ui stuff in webui is basically just gradio, which also nicely works with diffusers

however, some stuff in webui just works better. Like the inpainting implementation in diffusers is not that good

but the API is more cleaner and easier to use and extend

[soft condor]

Can someone please help me with stable diffusion? I can't get it to generate anything good at all. Here is the prompt I used:

city skyline, waking up, sunset, sunset starting, dark sky but pink sky in the distance, mostly dark black sky as it is night, beautiful, realistic, big, foggy lighting, sharp lighting, romantic, hopeful, 16:9 aspect ratio, art style realistic, downtown city, dynamic lighting, gradient sky, 3d, high definition, 4k, 8k, 64k

https://media.discordapp.net/attachments/834885490451808286/1111053107141546024/city_skyline__waking_up__sunset__sunset_starting__dark_sky_but_pink_sk_S79234389_St25_G37.3.png?width=885&height=589

And I got that result

so bad

idk why lol

please help

[abstract quarry]

whats your settings? Looks like you have not enough sampling steps

[soft condor]

I got this here...

https://media.discordapp.net/attachments/834885490451808286/1111053950318284890/image.png?width=928&height=589

[soft condor]

lemme check

[soft condor]

is this the setting? i can't find anything else...

https://media.discordapp.net/attachments/834885490451808286/1111054313591152760/image.png?width=963&height=589

[abstract quarry]

your guidance scale is too high

set it to a value between 4 and 8

[soft condor]

oh?

[soft condor]

ok got it

[soft condor]

im re-rendering now...

[abstract quarry]

also render in lower resolution

512x512 is the native resolution for 1.4 and 1.5

768x768 is the native resolution for 2.1

[soft condor]

I'm gonna use this as a background image for something....

I have a good gpu so it's fine

[abstract quarry]

you can change the resolution a little bit (like 768x512 to make it widescreen) but you should not change it too much

yes, but you should use upscaling for that

[soft condor]

?

[abstract quarry]

render in low resolution. Then upscale to high resolution. Then img2img

[soft condor]

OOOOOOH

why can't i do high resolution directly?

[abstract quarry]

also there are special tools like controlnet tiling that can improve on that

[soft condor]

i see

[soft condor]

?

[abstract quarry]

the method is trained on 512x512. If you render high resolution it doesn't know how to compose the image

[soft condor]

hmm ok

thank you!

[abstract quarry]

if you render in low resolution and then upscale + img2img it will just "fill details" in an existing image

[soft condor]

i see

[soft condor]

this is already an improvement lol

https://cdn.discordapp.com/attachments/834885490451808286/1111055431956168704/city_skyline__waking_up__sunset__sunset_starting__dark_sky_but_pink_sk_S1640871376_St25_G4.8.png

[abstract quarry]

see also here https://stable-diffusion-art.com/controlnet-upscale/

[soft condor]

ok ty

[hollow flint]

anybody knows if civitai has record of deleting images? i uploaded some lewd images but couldnt find them anymore

[tepid escarp]

why dont you ask on their discord then ?

[hollow flint]

hm i could. but do you know if that is true?

[peak portal]

of course they do that

[hollow flint]

i mean, apparently they do allow lewd images. i am just confused whats ok and whats not

[peak portal]

if it's gone now, it probably wasn't and i doubt anyone here can answer either

if you're looking for sympathy, oh. i guess, that really sucks. i'm so sorry they did that to you.

[hollow flint]

thats fair. i might as well share my art else where

[peak portal]

fewer and fewer places to share that kinda thing but it's really for you to enjoy anyway

oh. try reddit

that place seems to give very few craps about what's posted

[jade swallow]

hi

[mint lily]

someone broke my webui controlnet! 🙂 😦 it doesn't offer me the filter preview so i can preview my canny adjustments

[peak portal]

i don't think anyone here is involved with that project

[fleet crown]

Do you know if there is anyway to get this bot in a private server?

If you know, please ping me about it.

[supple zephyr]

Does anyone have a resource I can use to research each ControlNet model? I can't seem to find a place where it's documented exactly what they do / what they use. There's so many models, and I just can't figure out the purpose of some of them.

[open heath]

Hi, im new here. Can i share my works here and ask for areas of improvements

[patent bloom]

Good evening everyone, is it possible to see all the images already created by me or are they deleted after the showdown?

[spice plank]

Stable diffusion folder > output > txt2images

Something like that

[patent bloom]

👍

Thanks

[sly eagle]

Hello! Can you recommend me a Laptop that can work with training?

[fallow shard]

If I'm looking to get Stability AI to make a private model for me/ my organization/company where/who should I reach out to?

[compact raven]

Anyone know how to make a chat gpt account without a phone number?

Or is there some version I can run locally or something?

[fervent thunder]

Do you have a link to non webui collab?

[abstract quarry]

no, I mean, you have to be able to program of course. The diffusers library is quite easy to use

[rigid heart]

On deforum there is same settings every time

Everytime i open its same

can i change that

[tardy schooner]

if model says to use this clip ( https://huggingface.co/openai/clip-vit-large-patch14-336 ) what do I need to do? not familiar with using clips

[candid agate]

Hey everyone, does anyone know which version will work better for fine tuning style (general concepts)

[karmic brook]

join us today for our stage, animation enthusiasts!
https://discord.gg/nvqsXxcm?event=1111031723195633766

[molten sleet]

Hi everyone. Could someone tell me, if i can combine two pictures ??? Or SD cannot do this function?

[sly eagle]

Hello! Can you recommend me a Laptop that can work with training?

[hexed jasper]

hello, what is this "Challenger" role? I don't see an explanation 🙂

[karmic brook]

it's to opt into events/challenges we host here on the server! like for #⭐｜pow-info and #1087493421209485393

[hexed jasper]

oooh, nice. then it's definitely for me. thanks.

👍

[brazen orbit]

this discord is so big now

[jade swallow]

jwnqeqjkwenqjwjkqewjknkqewqjknqwqwjnkkjnqwkewe

[sleek osprey]

Is reimagine XL using SDXL? Doesnt mention that anywhere + the images look way different

[fervent thunder]

alo what is the difference between clipdrop and dreamstudio? looks like dreamstudio is mostly generative vs clipdrop has generative + image editing products?

[peak portal]

i think they're just going Apple style branding on all of their next-gen models even if they're completely unrelated

[carmine reef]

Is there any way to use a local text-to-speech generator with a customized voice?

[fervent thunder]

hey guys, I can't really find the answer so, is there an extension or something to generate pictures with batch that are more related to each others? like more similar pictures for a better animation effect?

[fringe current]

reimagine sucks, the quality it produces are always low quality, can it be prompted?

[raw ginkgo]

I was literally sitting in my office a couple of days ago... wishing I could take a picture of the room and toss up some design options... and you mad geniuses just pull it out of thin air.

kudos to everyone

[sly eagle]

-_-

[halcyon granite]

on MJ Reddit discussion about banned prompts is getting removed lol. This is exactly why I refuse to rely on a centralised AI

[fresh marten]

why isn't stable diffusion xl open sourced

[safe phoenix]

can anybody tell me how to generate images. I am new here

[wild steppe]

it will be open sourced once its fully trained!

[fresh marten]

oh got it

[fresh marten]

do you guys have a timeline on when that will happen or no idea?

[wild steppe]

https://www.reddit.com/r/StableDiffusion/comments/13ppmtk/sdxl_is_now_50_trained_and_we_need_your_help/

[fresh marten]

thank you

[wild steppe]

a little update on our brief conversation from earlier - the TOS for the bot research preview is available here!

https://stability.ai/discord-tos

[karmic brook]

#1029055412764422214 is now open! We're about to go live soon ^^

https://discord.com/events/1002292111942635562/1111031723195633766

[copper jay]

Hello 🙂 Does anyone knows a good Deforum Google Collab ?

Also, looking for an AI origami model, if anyone can recommend !
Thanks

[fervent thunder]

Anyone using VLC? I have this weird problem where when I open a file the VLC window bleeds offscreeen

[hollow robin]

Where do yall get your models? Looking to get some more and I've got the one I want from my current site

[wild steppe]

#1047197565365538826

They're hosted on huggingface and civitai - both linked in here:
#1047239653356474491

[hollow robin]

Perfect, thank you!

[peak portal]

you could liek give me FTP access to the checkpoints i won't share them

[strange mason]

Hey i need help, i have a video and i want the human to move the hand, how can i do it?

[peak portal]

can you tell me how many giggle bytes the model is at this point? is it a huge file like we've never seen before? or is it pretty close to the 4, 8, 10-whatever GB we'll see at the final product?

can i know what text encoder it is and how it's been trained

in fact, tell me all you know about its architecture is it just one model? three, like IF?

[spice plank]

How do you guys remember what models are what? I downloaded a bunch and can't remember what they do. Same with trigger words for loras

[wet willow]

There’s an extension called the civitai helper (or something like that). It’s not enabled by default but is in the extensions if you load the list & install it. It automatically will grab an image for checkpoints, Lora etc if you got the asset from civitai. It’s useful for a quick memory jog

It makes a new tab in the web ui and whenever you add new stuff you need to go to it and hit the scan button

It grabs the assets main image and sets it as the thumbnail.

[wild oasis]

How can I view the metadata of the generated images without using PNG Info?

specifically just the prompt

I want to make a python script that just outputs the prompt but idk where to start lol

[tulip yoke]

I got a chinese soldier next to me💀💀💀

[tulip yoke]

Png info

[wild oasis]

without PNG Info

[tulip yoke]

Use the file that already exist inside the auto1111 repo

Of png info

[wild oasis]

how does PNG Info extract it

alright ill take a look

[tulip yoke]

Metadata

[soft condor]

How to get permission to post in #1070829899918610632?

[wild oasis]

you cant make new threads but you can reply to existing ones

i dont think anyone can make new threads

[soft condor]

Huh

But posts are being made

Like one was made 6 hours ago @wild oasis

Oh wait

Lol

They're all by the same person

[wild oasis]

yep lol

[sly eagle]

Anybody has used AWS for SD?

If you could say between Google colab and AWS what would be better?

[tulip yoke]

What is aws

[hallow plaza]

Anyone willing to share their face sets for normalization? If looking for high quality pictures of real people.

[tidal lichen]

Hello! I took a break for a while and am now horribly behind on some terms 😅

In the olden days we used DreamBooth to "train" the model on what peoples' faces look like.

But that was destructive. So now we "embed" the new content? Or "fine tune" models? And we can do this a dozen times without it "forgetting" anything? Or ... is that the same thing as doing a LoRA?

Pardon my idiotic inability to even word the question properly 🙏

[willow coyote]

Idk I reverted to Dreambooth after seeing the Replicate founders doing it too (https://gist.github.com/basicfeatures/3d3817131e8f536c38f7f61210079b1d)

[willow coyote]

https://github.com/tonytonyjan/exif/ maybe? Idk

[olive crescent]

What sampler does the official Runway ML demo use? https://runwayml-stable-diffusion-v1-5.hf.space
I get much more creative results using this instead of A111 locally and I want to replicate it

[real zodiac]

that app sends a request with the prompt to whatever server so hard to tell but likely just default for the pipeline, which I think is ddim(?), their params for example pics are seed 1024, 45 steps, cfg 7.5, batch size 4

[crude stag]

hello

can we use stable diffusion image for commercial use??@sudden marlin

[quick thunder]

For sure. You can use videos generated in OpenShot Video Editor also for commercial purposes.

USING STABLE DIFFUSION FOR COMMERCIAL PURPOSES
During the release of Stable Diffusion in August, Stability.ai announced that it collaborated with its partners to ensure the final product is safe and ethical. The developers incorporated data from their beta model tests and communities to create the most reliable and effective art generator.

In collaboration with the vigorous legal, ethics, and technology experts at HuggingFace, as well as an amazing team of engineers from CoreWeave, the developers added several important elements to the final product. For instance, this model operates under the Creative ML OpenRAIL-M licence, which sanctions the commercial and non-commercial use of Stable Diffusion.

The licence focuses on the ethical and legal application of the model. Therefore, it’s your responsibility to ensure that you use the model legally and ethically. Make sure that the licence accompanies the model if you wish to distribute it. You must also make a copy of the licence available to your end users when including this model in your service.

Whether you’re creating images and art pieces as a hobby or as a profession, you can make money out of your work by converting your pieces into non-fungible tokens (NFTs) and selling them in the crypto market.

Hope that helps

[wind bramble]

hi guys i have a problem and i need an expert for this i have a question

that maybe no one can resolve

how to add a real object or a person into a real photo ?

i tried every controlnet models to do that but nothing worked

i thing it's the right place maybe to have an answer

[peak portal]

freeze layers of text encoder, less destructive

[olive crescent]

Hmm, I’m unable to replicate any sample images from the site using the same model and seed. Is there a way to look into the source code for the sampler? I’m not sure where to look

[real zodiac]

no idea what backend they're using so the source could be anything

there's a lot more that goes into a gen other than model and seed too, you likely won't be able to replicate it perfectly no matter what

[thin pine]

what's the best vae for realistic eyes?

[wild oasis]

I found it in /api/api.py

It's a method from the PIL library

[true dawn]

Good morning beautiful faces Happy Friday!

[ebon gazelle]

Hi folks! I really need help here. I installed the plugin to my photoshop - but it says that "Failed to load stable art"

I did all that was in the instruction

Current version 24.3.0

[fallen needle]

this is really cool, thanks for giving this opportunity. SDXL when? I'm ready!

[soft condor]

Someone please help me, my images look like crap, unless I fool around with the settings enough, and it fixes it. But then I forget my settings! Please help

[fervent thunder]

It's hard to help not knowing more details as that didn't really say much :P

[soft condor]

What details?

[fervent thunder]

what I read about your need for help was similar to: my car doesn't work, please help. :P

stuff like, what ai are you using, is it local, webpage, somewhere else? model? version? prompt, time, etc :)

[soft condor]

I'm using stable diffusion, web ui, that's all I know. And whatever default model came with it. The prompt is just whatever prompt I enter it gives me something bad or lows quality, when midjourney gives me something good looking no matter what lol.

[fervent thunder]

this is the easiest, but not wholly true, explanation:

1. midjourney is like using photoshop while using most other things are more toward using normal MS paint :P
2. which model you use is very important as the basic ones (1.5, 2.1, etc) are very rough. It's like using bricks to build a house with instead of hiring a team to do it for you.
3. your settings help a lot to get what you want, same with how many steps, and other settings you use
4. your skill level help about the same as settings as you need to "know" what to write to get what you want as it's not always just a easy "write what you want in plain english to get what you want"
5. quality gets better with use of extra tools like high. res fix, upsaclers, etc

[soft condor]

Ok. So what can you recommend. What models? Any rough settings? Skill level i'm starting lol, so I understand I probably won't be the best. I tried upscaling in img2img, but the quality stayed the same...

[fervent thunder]

just like learning to draw, you need to learn how to "prompt"
no one becomes a master overnight. You need to train for that.
Just play around and see what happens and what X does when used with Y, etc.
You could download the "Dreamshaper" model, it's quite good without much need for anything else. There are probably better ones for X and Y out there, but only you know what you want. And the best thing to learn is to train. And also checking out what there is out there :D

[soft condor]

tysm 🤝

[daring flume]

Hello, one question, how do I make the images of Lora appear in stable duffucion?

hola una pregunta ¿como le ago para que aparescan las imagenes de lora en stable duffucion?

[hollow spoke]

I have an open support ticket, but no one has responded in the last 14 days. I am in need of two things. A support response and two can someone direct me to business development?

[hot marlin]

The settings for each image can be embedded into the image file for .png images. If you go into auto's settings check the "Save text information about generation parameters as chunks to png files" and it will embed them in the image. You can then use auto's image browser extension to see and use the parameters used to generate the image. Don't forget to save the images you generate as .png files which is also in auto's setting.

[soft condor]

ty

[fast sage]

How can I remove myself from the creator tag, I don't want pings.

[tulip yoke]

@fast sage

#👥｜roles

[fast sage]

Thank you!

[tulip yoke]

[soft condor]

someone plz help in #🏞｜general-with-images

[pearl lava]

Is it possible to incorporate lighting effects into my flat color illustrations

[soft condor]

@bleak matrixadvertising

[tulip yoke]

I am a creator

I want money for images i can make for free

And stable diffusion isnt a model

I think

[raven wagon]

it is the base model

[raven wagon]

and yeah ai artists are the future though

[tulip yoke]

@warm junco

[soft condor]

ty

[karmic brook]

play nice and no spam ads pls

[blazing sorrel]

@bleak matrix hello sir

[tulip yoke]

Sir

[viral birch]

Can you run stable diffusion on iPad Pro with m2 chip and how?

[thorn dove]

hi y'all

i have an output from midjourney. how can I use stable diffusion to make it more realistic and the model's face more real life

any suggestion?

[willow coyote]

Yo man

Checkout the lomography and film grain plugins from GIMP's Elsamuko. Here Ive converted it to Ruby and plan on running all my stuff through it and similar filters (ie. VSCO): https://gist.github.com/basicfeatures/a083eb1ac1602f583d4993c3aca65358

[hollow spoke]

@lapis bay @rigid bough @wintry stream .. can I get some help please.

[willow coyote]

Makes things look more realistictoosince it''s emulations of old analog cameras.

https://lomography.com

[shell seal]

Sup guys, haven't been up to date with SD for a while, what's the current trending model for anime girls?

[wintry stream]

Hey Marco, I don't see a support ticket from you on the this server. Do you mean a DreamStudio ticket?

[hollow spoke]

it was a Dreamstudio ticket and It appears that that just responded 53 minutes ago. I still need a business development contact though. We will be generating 100k+ images per day. Need to talk with someone about cost and scalability on that. 🙂

[wintry stream]

Thanks for the clarification!

Please send a request over to [e-mail] from your business email and that should get you in contact with the right team for such a request 👍

[spice plank]

I can't use controlnet on a 6gb 1660ti, I have tried the low vram and --medvram as well as the --xformers but it still throws cuda out of memory errors at me.

[wintry stream]

If you don't hear back from them by end of next week after sending the email, feel free to ping me again and I'll help follow up to get things sorted out for you

[languid pendant]

Ooh yay i won nitro! Thank you! ❤️

[pseudo finch]

Will the stable animation be available using the restful api in the future?

[tall wave]

could any kind helpful person link me a tutorial how to install a diffuser?

im not sure if my pc specs are enough

[hearty pine]

I understand there's no copyright issues with images I generate with AI. What if I use the images generated by prompts published online? Is there copyright problem?

[vast ice]

You a Nazi?

Well that's good because Nazis are subhuman pieces of shit.

It is what it is

[peak portal]

two GPT bots passing like ships in the night...

[alpine bear]

lmao right?

[brisk pelican]

why does "stable" generate fuzzy images?

why does "stable" generate fuzzy images?

[oblique river]

A.I Trip To the Moon https://www.youtube.com/watch?v=y1jLyEBvEPI&ab_channel=HalabanAIVideos

[bold chasm]

Hi I am using Auto1111, is there a prompt to generate random poses

[wet willow]

Random pose? What do you mean?

How random 😆

[bold chasm]

Like random

Like wildcard

[fervent thunder]

aren't all prompts random prompts when you think about it to the ai? ;P

[bold chasm]

🤔 ya i guess if we arent specific

ill try keep it blank

[fervent thunder]

but yeah, use a wildcard for the random pose, can't say I know of any other way to get those results at all 🤔

[wet willow]

You can use prompts like “dynamic pose” but I’m still not sure how random you want 😝

[bold chasm]

thanks

[bright cradle]

No image to image?

[fervent thunder]

anyone can help with what I can do to fix this version and folder mismatch after installing a newer version of python?

venv "C:\StableDiffusion\stable-diffusion-webui\venv\Scripts\Python.exe"
No Python at '"C:\Users\virec\AppData\Local\Programs\Python\Python310\python.exe'
Press any key to continue . . .

its what i'm getting when i launch the web ui bat file

[spice plank]

Reinstall python but check the box that says change PATH or something like that

[bright cradle]

No image to image?

Bot

[fervent thunder]

already check marked the path when installing but not working

somehow the file pyvenv.cfg didnt get updated under the following path

C:\StableDiffusion\stable-diffusion-webui\venv

and when i edit that file with new python path and version i get another error with py torch

the bat execution fails at launch.py

deleted the venv folder and ran that webui-user.bat file

[fervent thunder]

so it turns out the latest version of python isn't supported with current web ui

[sinful pine]

What SD version do yall think is the best?

[narrow wind]

Good day, all. Looking for a port of dreambooth notebook from https://colab.research.google.com/github/ShivamShrirao/diffusers/blob/main/examples/dreambooth/DreamBooth_Stable_Diffusion.ipynb ported to kaggle.com?

[lofty gate]

Anyone available to look at my planned PC build (Newegg link) and lmk if it's a good plan or if there are better priced alternatives? I'm set on the 3090 but the other parts I'm pretty clueless about

https://www.newegg.com/m/tools/custom-pc-builder/ld?dln=25987695

[warm junco]

looks okay, but i think you dont need a mainboard with two rj45 ports. more USB ports would be better

[lofty gate]

True. I just chose a well rated board without looking at the features 😂 id like more USB ports for sure so I'll look into it. Thanks

[magic citrus]

hey friends, unsure if you have some better tool but I got really tired of generating bunch of xyz plots and waiting for each when testing my models and comparing bunch of checkpoints and so on. So I did a simple script to write all my prompts and parameters in a csv file and then leave this script for hours generating XYZ grids.
https://followfoxai.substack.com/p/saving-time-using-auto1111s-api-automated

[bright cradle]

No inage init option bot

Image

To image

[neon bear]

Hi Everyone, I'm new here. Is someone able to show me around? I have a Goal to Reach but I need a opportunity to be allowed to chat with Emad Mostaque? It must be him as He is 1 of 2 that understand my unique perspective, as no one but him would believe that I have solved a solution to garrantee Equality for All!

[tulip yoke]

can anyone introduce me to the settings of training a embedding

[tulip yoke]

who dat

[sage ravine]

Which version of stable diffusion is for ultra real human ? (I need it for pc not online)

[tulip yoke]

tf is ultra real human

[sage ravine]

High quality texture

[tulip yoke]

a human doesnt have textures

[sage ravine]

Like meta human u know?

[tulip yoke]

no

idk

[sage ravine]

👍

Let other answer then 🙃

[tulip yoke]

but

nobody did

[fervent thunder]

You can download automatic1111 and deploy stable diffusion on your PC, that's what I did. I'm new to this but if you want ultra human ya know, try using prompts like "hyperrealism," "perfect human form," and such...then in the negative prompt put what you dont want...like distorted limbs, missing fingers, no background, such like that. The negative prompts will negate themselves and produce opposite results of the standard prompt if used correctly. Render using the photographic mode or cinematic. For hyperrealism I have best results by starting with the cinematic filter and revising images form that point.

Look I just wanna post my images on the forums and I'm not seeing the image preview, only the file download is showing after posting. I put in a mod ticket but there is like very limited support here it seems like. My images show up just fine on my own Discord. Dunno if there's a setting they have disabled here or what.

[tulip yoke]

thats not what he meant

[fervent thunder]

you literally said you didnt know what he meant so what are you talking about?

fr

[tulip yoke]

a human

like

a good one

[sage ravine]

I have like an old version of stable diffusion from like October last year

[tulip yoke]

good detailed

[fervent thunder]

yea...duh.

[sage ravine]

So where is the link for new one for pc

[fervent thunder]

use the right prompts.

google it bro

figure it out. i did.

[tulip yoke]

at google

[sage ravine]

Website is not providing the the link

Should I go on Github

[tulip yoke]

yes

[sage ravine]

automatic1111 web ui Or just normal stable diffusion?

[fervent thunder]

web ui

thats what I used.

[sage ravine]

Ok

Download and replace the files or something else?

[fervent thunder]

Bro.

Look this up.

I'm not giving you step by step direction.

You can also look into StableStudio.

you'll want to look up how to install it, it can be a process. You need a lot of dependencies.

[sage ravine]

Ok

[fervent thunder]

Good luck!

[fluid bramble]

hey uh

i need help with something

so i updated my python and somehow the route for the gui is still the previous one and theres no python there and idk how to change it

for context, i use k diffusion retard gui

wait nvm im dumb

i should not have updated

[plucky latch]

👋

[slim zinc]

Does anyone know if Instagram hides Ai pictures from explore pages? I tested and my picture is seen on my account but not in the explore pages it has been tagged with.

not really looking for a solution but just checking if someone else has seen the same thing.

[gusty falcon]

Do any of you use Stable Diffusion on an intel Mac?

[gritty steppe]

is there a way to create seamless tiles with specific prompt or model?

[wet willow]

There’s a tiling toggle. It works with the prompts.

You may need to double the resolution though. Keeping res at 512 doesn’t seem to have the resolution it needs

[gritty steppe]

Oh nice. Thank you

[crude stag]

@rigid bough can we get the upload any image option in the bot channels,so we can genreate new image with it ?

[sage ravine]

Found civit ai files for real look

See #🏞｜general-with-images

[frank rock]

is .safetensors only more secure than .ckpt, or are there other advantages?

[fervent thunder]

if you already had it you can just add ```git pull```` to your webui-user.bat to update it.
If you want new models - go on https://civitai.com/

[blazing sorrel]

@bleak matrix

[bleak matrix]

If you want to give support, please go to #✨｜sdxl

[fervent thunder]

#💬｜general-chat message

[wild steppe]

@fervent thunder please don't advertise on the server and keep off-topic in #🌶｜off-topic

[fervent thunder]

@wild steppe

[wild steppe]

yes?

[fervent thunder]

How to get role

[fervent thunder]

Dream bot achieve

[wild steppe]

you can grab roles from #👥｜roles! just react to the post with the emoji for that role

[fervent thunder]

Thank you

[mortal lark]

hi

[potent cloak]

guys why are there minor differences to my outputs after this update

[umbral tulip]

hmm, how does Tiling in SD actually work? in terms of a higher level explanation on what's happening under the hood

by tiling I mean the option that makes the image edges match

[vivid portal]

Just released a new model:

https://civitai.com/models/76276/aether-lux

Aether Lux

[young ivy]

I wish I had a 3090 or something, then I could make an actual model of my own and I would make a great one

[restive prawn]

Anyone got a good guide on using img to img and inpaint to improve the quality of an image?

[kindred bison]

Can anyone please direct me to where i can start using Midjourney alot has changed from the last time i was here

[dawn topaz]

does anyone have any advice: i have stable diffusion downloaded locally on my pc, im going out of town, with only my laptop, but still want to generate deforum videos and nsfw content, what would be the best solution you recommend for this? or by chance has anyone ever used Tailscale?

[young ivy]

I just managed to get a GPT bot to instruct me how to make drugs...what the fu....

[worthy tree]

Ask a way to have an endogenous drug

That way you won't need anything to get high lol

[young ivy]

that would send my brain going haywire into schizophrenia or catatonia

[restive prawn]

Unstable diffusion has a website

[warm jasper]

Are 2.1 models usable with 4 GB VRAM now? Last time I tried, I could only generate very small images with them (like 384x384)

[torpid zephyr]

Does Automatic1111's Stable Diffusion webGUI have a means of generating random numbers?

I was just looking for ways to add some variation to things - for instance, telling it to pick a random intensity/weight value for a certain attribute.

Nmind - I think I found an extension that accomplishes the task.

[pine ore]

Hello, are we allowed to suggest .ckpt's here? I'm looking for something that can help me generate goblins in the style of the Pathfinder TTRPG. I'm having a lot of trouble generating eyes of a single solid color (solid red eyes in particular, no iris/pupils) as well as ear to ear grins.

[bleak matrix]

If you want to learn more about running SD, I recommend checking out #1072220168534642768 followed by #1080946152318443610 and #1072229020520947753 The bot is also running right now, which you can check out in the bot channels.

[bleak matrix]

There are models in #1047197565365538826 and you can check out #📝｜prompting-help for more info on prompting, etc

[swift juniper]

hey general question when training loras with dreambooth. I hear people say "id train it for 200 steps etc" so im confused by which number they are referring to. lets say i have a dataset of 50 images. when i tell dreambooth to train for 100 steps the countdown will say 5000 total steps as its 100 steps per image, (100x50=5000). anyway Ive seen a lot of comments recommending 1000-2000 steps for a decent lora and was wondering as to if theyre usually referring to total steps or are they referring to steps per image, (the number youd input into dreambooth)

if you can help me out thatll be greatly appreciated. feel free to ping or dm.

[neat bronze]

Hey, is there any way to get TensorRT 1.6.1.6 without being a developer, or can I just register an account to download it from Nvidias site without any problems?

[plucky shoal]

I dont know @swift juniper but there is actually debate in #🔧｜finetune
You have better chance get better answer there for sure.

[neat bronze]

Thank you :)

[neon bear]

Can any provide link to a 1,6GB Image of the Web??

It's a image that can be stored locally for research

Mobile and PC

[plucky shoal]

@neat bronze it was for @swift juniper 😄

[neat bronze]

Ohhh okay 😅

Yea I think I better ask this in #🤝｜tech-support

[plucky shoal]

for sure

[hardy rune]

Am I allowed to ask very basic stupid questions here?

[zenith thunder]

Has there been an updated how to guide on how to run sd using 6000 series amd gpus?

[exotic acorn]

https://github.com/XingangPan/DragGAN

Anyone knows how much vram i would need for this ?

[wet creek]

Anyone know the fix to stop faces distorting on the final step before output?
Issue is documented here https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/7851 https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/10223 https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/7244

[fervent thunder]

There are plenty of suggestions on links you provided, but you also can just get older a1111 version before it started happening

[wet creek]

Might have to

[fervent thunder]

That hash working fine for me a9eab236d7e8afa4d6205127904a385b2c43bb24
But I have previews disabled for faster gen and never use euler a if it has something to do with issue

[crystal mulch]

Hello, it's been a while since I've used dream channels, is there a limit of how many images I can generate in bot channels? Thanks

[unkempt hatch]

so you know how a lot of loras and text inversions all at once make the image a mess? Any way to balance that? higher CFG? or maybe lower? more steps?

[gray gazelle]

Hey guys

New here

Does anybody know when stability.ai is going public and where to invest?

[bleak matrix]

Good morning, everyone! How are we all today?

[bleak matrix]

You can learn more about how the bot works/runs by checking the pinned comments. And, if you have any feedback, you can post it in #✨｜sdxl

[wet creek]

i tried your hash and still get messed up faces.

[bleak matrix]

I would probably check #🔧｜finetune or #📝｜prompting-help on that!

[gray gazelle]

Good morning! I’m fine Ty and you?

[bleak matrix]

I think I'm alright!

awake!

What's everyone up to today?

[potent spire]

hello 😄

[potent spire]

a bit of gaming and then doing some art (2D and 3D) and playing around with Photoshop and its generative fill 😄 How bout u?

[sly eagle]

Hello! good morning! If I have to pay for a service what do you think is better? Google colab Pro, Runpod or gradient?

[tulip yoke]

auto1111 is so funny

my room themperature is 26.5°c

[gray gazelle]

Enjoying the sunny Sunday sunny :))

And you

[gray gazelle]

Need a answer for this pls

[sullen barn]

Not sure which of my many Discords to ask this one on so I hope the clever folk here can give me a pointer. I have a project which will allow anybody to upload and later find a picture of their cat. This is going onto the permaweb so anything uploaded is there forever (or 100 years etc ...). I therefore do not want pictures that are porn etc being uploaded, even if I later de-index them from the site, they are still there.

So the question is, ideally I want an AI to view the pictures at the point a third party uploads them in my React project and go "cat" or "not cat". I am sure there must be third party services capable of doing that rather than me having to train an AI just for my project. Anybody give me a direction or suggest a better Discord to ask in please???

[broken smelt]

sounds like you just want a standard nsfw filter, which has nothing to do with AI unless you define things like google's search algo as AI

[bleak matrix]

Oh, just resting today, really! 🙂

[sullen barn]

Thank you. That would certainly avoid the worst cases. I assume that things might be moving along though as AI (not least images) has progressed so far recently. There are lots of possible projects where "thing" and "not that thing" would be useful.

[sly eagle]

Hello! good morning! If I have to pay for a service what do you think is better? Google colab Pro, Runpod or gradient? Idon't havea pc capable of SD for now but I need to use these services. So what do you think it's suitable for quality and price?

[ember flame]

When generating pics locally, is there some way it saves the prompt and settings you used along with the image?

[broken smelt]

if youre using automatic1111, there's a PNG Info tab that you can click on and upload the picture, by default every pic you generate will save that info inside of it

[ember flame]

I am using automatic1111, yes. I admittedly am not quite sure yet what you mean but I'll check and see if I can find it now. Thank you 😄

[ember flame]

NVM I believe I see it now

[fervent thunder]

Hey guys i m the beginner basically idk what i have to do to genrate images can somebody plzz help me i really want some help and i m also not good at techss so yahh

[tidal lichen]

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

Scroll near the bottom

Read and follow the "Automatic Installation on" instructions for your operating system 😎

---

That's if you want to do it on your own computer.

If you don't care about doing it locally, you can use this site:
https://beta.dreamstudio.ai/generate

Nope, gotta post here: https://discord.com/channels/1002292111942635562/1004159122335354970

[gray gazelle]

When is the ipo??

[fervent thunder]

Is this just machine learning alghoritms on large datasets made by "low paid immigrants workers"(I'm not American)??

So you can get the datamodels and kinda change them and use one of the available alghoritms in regards? Regards.

[bleak matrix]

@fervent thunder Hey there! Welcome to the chat. Please see our #✍🏼｜rules-and-tos and see our rules regarding NSFW. I also recommend taking a moment to check out #1072220168534642768 along with #1080946152318443610 and #1072229020520947753 to learn more about SD

[fervent thunder]

No, I don't think I will, thanks.

[fervent thunder]

hi

[hot onyx]

Hey when someone uses Stable diffusion auto tools, it uses our own computer resources right, but it seeds to the public?

[gray gazelle]

why tf is nobody answering me holy shit

?!!!!

ffs

[gray gazelle]

@karmic brook

[quasi carbon]

How make vertical images ?

[gray gazelle]

@bleak matrix

[peak portal]

international pancake offering?

i don't think anyone would ever reveal to you when they're doing an IPO, if they are, roflmao, if they even know that themselves

[hot onyx]

Does this tool uses computer resources: https://github.com/AUTOMATIC1111/stable-diffusion-webui

[gray gazelle]

ty! where can we invest

okay what do you prefer for europeans

hmm

ty very much tho

how can we invest before it goes public

is there a way ?

yeah so only ipo in early july

and then its going public for the rest of the world

when is that?

youre literally the only one that answered my question

legend

[north nimbus]

trying out kitchenui for auto1111. seems to be good.

[gray gazelle]

what do you think

invest or no invest

im not sure either

i just think everybody will fomo in it

cuz its going to be the only ai company that is publicly traded

lets just hope every ai company goes public

after that

[bleak matrix]

If you have a question that needs answering that is not answered by the various pins, etc, please file a ticket with #1010934719455707218 Thank you.

[bleak matrix]

You mean the width and height? Are you using the bot, DS, or SD locally, etc?

[normal stream]

I am using Linux and would like to know if the AMD RX 6600 8GB VRAM would work properly for generating images by text and 512x512 LoRa training.

I am aware that Nvidia is better but with RX 6600 AMD on Arch Linux (for example) you shouldn't have much problem, right?

I'm asking this because I used to have an RX580 and it was a pain to compile everything, I wouldn't want to go through the same thing.

Who have RX 6600 on Linux?

[quasi carbon]

Discord

[gray gazelle]

buy a quantum computer

[normal stream]

Thank you very much for your comment, it was very useful, see you on 4chan ;)

[bleak matrix]

You can find info about using the bot in the pinned comments of each of the bot channels. (It's the aspect ratio.)

[bleak matrix]

I would check #🔧｜finetune for the latest info or #1072238304042438758

[normal stream]

Ty.

[daring flume]

Hello, I've been using stable diffucion for a couple of weeks, but I have some problems that I'm looking to solve. Is there someone who is willing to teach me for a fee? I do not have much
note: i only speak spanish
Anyone interested please send me a DM.

hola soy llevo un par de semanas usando stable diffucion, pero tengo algunos problemas que busco solucionar ¿hay alguien que este despuesto a enseñarme por un pago? no tengo mucho
nota: solo hablo español
cualquier interesado porfa envíeme un DM.

[bleak matrix]

If you are looking to learn more about SD, I suggest checking out the #1080946152318443610 ur and #1072229020520947753 if you haven’t already. There is also #📝｜prompting-help or #1072238304042438758 and several guides available as well. I am currently writing new guides for 2.1 but they are not finished, but if you are struggling with something in particular, I really recommend at other people's prompts, talking with others, etc.

[rocky burrow]

yes it runs local on your machine and uses your GPU to generate images.

waiting for SDXL to release is fully is painful =b

[restive prawn]

anyone know any extensions that make the drawing for inpainting easier? (mostly things like zoom are neccessary on larger images)

[wet willow]

There is a zoom extension. Just do a find for zoom

It’s called canvas-zoom

[daring flume]

hello, a question to the owner of the group (I don't know who he is) @radiant meadow @sudden marlin @bleak matrix @regal raven @hidden dagger stable diffusion what is it? Are they a group, a company or what?

> hola una pregunta a el dueño del grupo (no se quien es) @radiant meadow @sudden marlin @bleak matrix @regal raven @hidden dagger ¿estable diffucion ¿que es? son un grupo, una compañía o que son?

[spring carbon]

gm

[daring flume]

And what does it mean?

¿y eso que quiere decir?

[alpine bear]

Stability AI is a company. The owner is Emad.

@radiant meadow is a bot, not a real person

[daring flume]

aaa ok I ask since it seems strange to me that stable diffucion is free, as well as being a company they have not found a more advanced AI

*aaa ok lo pregunto ya que me parese algo extraño que estable diffucion sea gratuito, asi como que siendo una empresa no hallan echo una ia mas avanzada *

I explain?

¿me explico?

[alpine bear]

Stability's mission is to make powerful AI tools available to the general public. They believe that locking the tools behind corporations and governments is fundamentally wrong, so all of the AI technologies they pioneer are generally open-sourced once fully developed.

While they don't sell Stable Diffusion specifically, they make an absolute boatload of money in GPU processing - so it isn't like they're just hemorrhaging money.

[daring flume]

I understand, if I didn't say it because of losing money, but rather that I think there is a lot that could be done and improved. question is stable diffusion open source? what if someone wants to make an ai and improve the ai using stable diffusion code is it possible with commercial use?

entiendo, si no lo decia por lo de perder dinero, si no mas bien que creeo que hay mucho que se podria hacer y mejorar. pregunta ¿stable difuccion es de codigo abierto? ¿y si alguien quiere hacer una ia y mejorar la ia usando el codigo de stable diffucion ¿es posible con uso comercial?

[alpine bear]

The base model of stable diffusion is typically very limited. There have been massive improvements to the quality and content of images by individuals who fine-tune the models to suit their needs.

If you're looking to generate better images, look into the following:

Automatic1111
ControlNet
CivitAI

Stable Diffusion is completely free to use for personal and commercial work. And because they release their models for free, anyone in the world can improve upon them or 'fine-tune' the models.

[daring flume]

Yes, I have a1111 on PC, but what I mean is that improvements can be made so that anyone can use it easily and quickly

si, yo tengo en pc a1111, mas a lo que me refiero es a que se pueden hacer mejoras para que cualquiera lo pueda usar de forma facil y rapida

since for inexperienced people like me there are too many things that cannot be explained and hidden things and everything is complicated

ya que para gente sin experiencia como yo tiene demaciadas cosas que no se explican y cosas ocultas y todo es complicado

[alpine bear]

There are options out there for people who need quick and easy image generation. Stable Diffusion has a website called Dream Studio where you can generate online without installing anything locally. There's also services like Midjourney and DALL-E.

The complexity of a local install will likely always be just that - complex. It has to be complicated if it is going to offer the sheer amount of customizability that it has. Not to mention - not everyone will be able to run it locally because of the system requirements.

For users that need quick and easy image generation, using dedicated services will likely always be the better route. For users that want absolute control over their outputs, a local install will be far more useful.

I don't see this changing anytime soon.

[daring flume]

It is not only that, it is also that all the AI ​​that are on the web are censored, and it is not only that the installation is complex, but rather that you have to understand how to explain well to the AI ​​what you are looking for, as well as that no matter what have good prompts the quality of the image is not the best, the controlnet is not easy to understand more if everything is in English and one does not know any English 😦

no es solo eso, es tambien que todas las ia que estan en la web estan censuradas, y no es solo que sea compleja la intalacion si no que tienes que entender como explicarle bien a la ia lo que buscas, asi como que por mas que tenga buenos pronts la calidad de la imagen no es la mejor, el controlnet no es facil de entender mas si todo esta en ingles y uno no sabe nada de ingles :(

I think stable diffucion is on the right track but I think it can continue to be improved 🙂

creo que stable diffucion va por buen camino mas creo que se puede seguir mejorando :)

[alpine bear]

I agree that it could be improved - especially in the language department. Finding resources in a specific language is a huge challenge to overcome. I think that AI language models will be very useful in providing real-time translations, but that's a whole other technology that's being developed.

If you have specific suggestions for stable diffusion that you'd like the developers to see, you can leave them in the #1002293361526460608 or create a ticket at #1010934719455707218

Or you could always just ping Emad. I don't think there's a rule against it

[daring flume]

Yes, there are many things to improve, but they are on the right track

si, hay muchas cosas que mejorar, pero van por buen camino

I especially think that you have to think of newbies, the easier you do, the better

*en especial creo que hay que pensar en los novatos, mientras mas fácil que haga mejor *

[cedar bay]

Is there a channel where i can get help with installing automatic1111?

[slate dust]

https://discord.com/channels/1002292111942635562/1002602742667280404

[flat horizon]

NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.
Time taken: 5.03sTorch active/reserved: 1740/1946 MiB, Sys VRAM: 3614/4096 MiB (88.23%)

can anyone tell why this error occur and how can i resolve it

[west marsh]

Comn bro , ask this to ChatGPT.

[real zodiac]

based on your vram sounds like an old card that has problems with fp16

just do what the message says really

[bleak matrix]

Good morning, everyone!

Happy Memorial Day!

[bleak matrix]

#🤝｜tech-support

[bleak matrix]

Please check #🤝｜tech-support so you can post screenshots! (Your main folder might look like C:\StableDiffusion\stable-diffusion-webui)

[junior tulip]

Check this one out please #1101178483608137918 message

[fervent thunder]

hello

[fervent thunder]

Complexity is what makes SD good.
You have alot more control, more options, more sliders to play with.

If you want something simple - there are simple webui's for simple people.
If even that's too much - well, there is always mj, nothing is easier, but also, alot less control.

WebUI and plugins you mentioned made by community, not by SD themselves btw.

[bleak matrix]

If you see something, you can always file a ticket through #1010934719455707218 etc and let us know!

[shrewd swan]

eyyo quick question, what is the practical advantage of having more than 12gb vram? a friend wants to buy a graphics card for SD and is wondering if more than 12gb is worth to get

[junior tulip]

I downloaded it tried opening via many different programs but I'm unable to do so

[bleak matrix]

SD?

[shrewd swan]

stable diffusion

[jagged solstice]

can someone point me to a guide for installing stable diffusion + control net for amd?

[bleak matrix]

If you need help, and have an error, try #🤝｜tech-support

[junior tulip]

Sure thanks a lot

[bleak matrix]

Check #1072220168534642768 followed by #1072229020520947753 and #1080946152318443610

[jade fjord]

I have been out of the loop for a little while, just busy with life stuff and coming back I realize that I have forgotten a lot of the keywords for models that I had downloaded previously. I am using A1111 is there anything that is good for maintaining those keywords?

[sick grail]

What's the best way to run automatik1111 not on your own machine? Like google colab or aws or something like that.

[ebon gazelle]

Hei folks! I get an odd error messege when try to make a hypernetwork

I am trying to train my net but it just ggets an error messege

odd

[peak portal]

if you don't share the error message, no one can help. but a fair warning, the hypernetwork code was stolen from NovelAI and it's unlikely that the developer who implemented it really understood it

[fervent thunder]

colab is incredibly pricey compared to few alternatives; aws nor google cloud isn't suitable for average use, because they are kind of whole package cloud computing where each separate part will cost you fortunes (cpu, memory, gpu, bandwidth, storage). RunPod is by far the the best option available with built-in support for stable diffusion. In colab, you pay average of 0,20 cents/hour for T4, which is ancient and slow, whereas in RunPod you get 4090 for as cheap as 0,50 cents/hour. https://blog.runpod.io/stable-diffusion-ui-on-runpod/

[meager nest]

Is hassan model still god tier or is there better ones for realism these days?

I mostly mess with semi realism, but I was curious

[steep wing]

Apparently I won a giveaway raffle, not sure how to go about claiming that

[celest flame]

Can inpainting work to change someone's outfit?

And location

[broken smelt]

probably

[random lion]

I've got this issue where everything I generate seems to be drawn. As in nothin looks like a render or photoreal even if those are the keywords I use?

[alpine bear]

you could try putting sketch, monochrome, lineart, unfinished in the negatives

[fervent thunder]

how did you draw that gray box to mention the prompts?

[alpine bear]

` x3 at the front of your text, then again at the end of your text

[lyric zephyr]

Hi, just learning about LoRAs, which one makes beautiful symmetrical faces and also not sure if the Hash is the same as Trigger word ?

[fervent thunder]

test test test

[fervent thunder]

ahh nice thanks

[alpine bear]

:D

ye

[fervent thunder]

use two of them to get the text like this instead so that you don't get new rows each time :P

[alpine bear]

oh neat very neat

[mortal sandal]

Is there a group or channel that's appropriate for discussing training models? I've hit some walls with my training and don't know which discord page is best

[fervent thunder]

<p>this thing is actually more for coding than anything.</p>
<div class="tada!">like this! :P </div>

[alpine bear]

(\_/)
(o.o)
(")_(")```
No it is for bunny

[fervent thunder]

if you want coloring, then write the name of programming languages then you can write \`\`\`html (without the \ if they are visible) and then make a new row to write the cool colored code :P

[broken smelt]

#🔧｜finetune

[mighty ether]

Im looking to buy a graphics card for stable diffusion, does anyone have a recommendation for best bang for the buck and also easiest time for compatible and setup with stable diffusion and other ai models?

[fervent thunder]

hi just a silly question : what does the stability society role (black nick) mean?

[haughty saddle]

I have been getting cuda out of memory error suddenly, I'm running the same reso and all as before but suddenly started getting the error, please help, very new to stable diffusion, thank you.

I used to generate images with the same set up before

[plucky shoal]

ask in #🤝｜tech-support channel. Depends on many things.
And post there your errors from terminal when error happened. @haughty saddle

[deep cave]

Hey anytips to filter out nsfw from civitai? navigating that site is painful

[thorn pilot]

disable mature content in the account settings page

[deep cave]

Ohhh someone mentioned that but I did not realise it was tied to the account, I haven't made one yet, just downloading thanks

[hardy dove]

Guys is there any guide to install SD?

[warm junco]

For a local install:
https://www.youtube.com/watch?v=VXEyhM3Djqg

[sinful pine]

Tryna use SD to generate Anime characters however most models i find seem to be focused on Females, nsfw and the sort. Anyone got a more fighting/serious focused model in anime style?

[fervent thunder]

most models can do both nsfw \ sfw , read description , they usually state if it's made only for nsfw

if not - just add nsfw into negative

you can add dtails you don't want to see into negative too

[peak portal]

that only works if it was trained well

[bitter laurel]

good morning, all! I haven't done any training in several months. Is the best still Dreambooth? Running a 12GB VRAM

[broken gull]

this is my list

• With
  • Nothing

print("Hello, World!")

[fervent thunder]

Instructions unclear, wrote to moderator instead

[peak portal]

depends on your desired outcome

[pseudo finch]

Will the animations feature be available on the restful API in the future?

[fringe current]

hi

[alpine bear]

Hi!

[fervent thunder]

I don't know how to ask this without sounding like a shitty person so I'm just going, but I wanted to get some Nazi imagery as some concept art but most online SD sites automatically refuse that word as a prompt. Is there anything I can do?

I'm willing to spend money but I cant really run locally.

[alpine bear]

"I'm trying to generate some historical imagery, but specific keywords are blocked on the online generation sites I'm using. Aside from running locally, are any premium sites that don't employ such restrictions?"

• how to not sound like shite person

To answer your question, I think local install is the only route. You might consider paying someone else who has a local install to generate your specific images.

[fervent thunder]

hello, hoping if someone can help me install stable diffusion. I followed all the steps but cant get the actual model to open

[broken smelt]

will probably have more active people in #🤝｜tech-support to help you out with that

[fervent thunder]

thanks

[gentle umbra]

Hey everyone! Do you know a website, where I can try out different custom models? I know mage space already. Are there more sites like this one? Thanks in advance!

[regal beacon]

hey has anyone here used CharTurner by MouseWrites?

https://civitai.com/models/3036/charturner-character-turnaround-helper

if so has anyone figured out how to inpaint

I'm new to stable diffusion

[past delta]

Hello

How to use this platform?

[warm junco]

Hello, check #1072220168534642768 and #1080946152318443610

[past delta]

Thank you very much

[bleak matrix]

Good morning, everyone! How are you today?

[bleak matrix]

You can find various models in #1047197565365538826

[forest aspen]

guys any idea is the Stable Diffusion XL model will be avalible for download/finetuning?

[wild steppe]

@winter island General chat is not the right place to post this, use #1092446741984444416

[fervent thunder]

Buds, why does my inpainting redo the things that I painted already. Eg, now I'm working on lips when the sunglasses turned out the way I already wanted them. But it ends up painting both again.

[bleak matrix]

Make sure that your mask is set to only change the area of the image that you have selected. (Check the edge of your mask, too.)

[fervent thunder]

Yes, thank you. I do that, but it just redoes what is no longer masked. But clearly it's the original mask

Maybe it's the seed setting?

[vast vale]

hi - question. is it possible to train a LoRA using a mac? im having trouble finding the resources to describe how its done or if its possible

[peak portal]

sounds like a bug

[dull urchin]

where can I find any stable diffusion nsfw art communities?

[winter pike]

unstable diffusion

[fervent thunder]

Hey Guys -- just a quick question -- I have stable diffusion installed on my pc -- works great -- im curious -- is there a chatbot thats out there that I can host privately?

[fervent thunder]

hey guys , i need a help , how can i use the ia on a music to make images based on lirycs?

[fervent thunder]

can we get a role like "EventSub" or something, instead of pinging everyone, it's annoying

nvm, disabled it

[mild dock]

#1072240143521554592 wiener reveal

EVERYONE! WIENER REVEAL!!!1111

[royal field]

Hey guys, how to stop stable diffusion generating multiple people when the image size is increased?

[halcyon shale]

good day everyone

[tulip yoke]

10bit generation when

My monitor has va panel 10bit native but hdr looks so shit due to va panel

[fervent thunder]

use high res fix

another option will be - to use model trained on higher resolutions, but when base is 512x512 - some images will be broken anyway

[frank spruce]

Do I need to download or install any files once I have Automatic1111 running normally in order to use negative embeddings like EasyNegative and badhandv4 or are they built in?

nvm u gotta download the PT file from civitai or hugging face and put into embeddings folder, i been genning hundreds of images not realizing i ahve files missing

[frank spruce]

But, where can I get some decent POSITIVE embeddings? things like highres, best quality, masterpiece, etc

[warm junco]

Pls post this in #🤝｜tech-support

Seems like a Connection issue

[zenith citrus]

hello anyone using Dreambooth here?

[mortal lava]

Hey there, is it against the rules if I want someone to make me a AI generated photo of myself and pay for it?

[thorny wing]

Does anyone know how many potential digits there are in a seed? Trying to figure out the number so I can create a method in JS to generate a random seed.

[bleak matrix]

You can find more about finetuning in #🔧｜finetune

[fresh crest]

Hi. I've spent the last hour or so looking at guides but I'm stuck at the very beginning. I can't launch SD more than once. I'm on linux. I run start.sh, it works fine. The second time I run start.sh I get an error

anyon can let me know where to find help?

[warm jasper]

I have updated a1111 after a long time, with the new xformers, torch and that tomesd, I got the unbelievable: an it/s measurement instead of s/it

[peak portal]

@bleak matrix hows the recovery going

[tribal bison]

hi all, pls help me asap

I'm new, so, how long does deforum take to create a video? let's say, 10 sec

[broken smelt]

a1111 does that for you if you put -1 as the seed number

[tulip yoke]

That doesnt fix it

[tulip yoke]

Depends on gpu

Depends on length

Fpd

Fps

Steps

[deft ivy]

hi all, im getting this error for stable diffusion, any suggestions?

Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:38:17) [MSC v.1932 32 bit (Intel)]
Version: v1.3.0
Commit hash: 20ae71faa8ef035c31aa3a410b707d792c8203a3
Installing torch and torchvision
Looking in indexes: https://pypi.org/simple, https://download.pytorch.org/whl/cu118
ERROR: Could not find a version that satisfies the requirement torch==2.0.1 (from versions: none)
ERROR: No matching distribution found for torch==2.0.1
Traceback (most recent call last):
File "C:\ai\stable-diffusion-webui\launch.py", line 38, in <module>
main()
File "C:\ai\stable-diffusion-webui\launch.py", line 29, in main
prepare_environment()
File "C:\ai\stable-diffusion-webui\modules\launch_utils.py", line 254, in prepare_environment
run(f'"{python}" -m {torch_command}', "Installing torch and torchvision", "Couldn't install torch", live=True)
File "C:\ai\stable-diffusion-webui\modules\launch_utils.py", line 101, in run
raise RuntimeError("\n".join(error_bits))
RuntimeError: Couldn't install torch.
Command: "C:\ai\stable-diffusion-webui\venv\Scripts\python.exe" -m pip install torch==2.0.1 torchvision==0.15.2 --extra-index-url https://download.pytorch.org/whl/cu118
Error code: 1
Press any key to continue . . .

[blazing socket]

Is there a good model for pixel art? Want to make mods for Stardew valley but no talent with art at all. I can't imagine I can't just train a model on all the pngs of the game either

[turbid whale]

https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/5485

[turbid whale]

Checkpoints and LORAs
https://civitai.com/?id=10706&slug=luisap-pixel-art-lora&query=pixel&view=feed

[fervent thunder]

How can I add my own photo?

Is there such an option or channel?

[turbid whale]

Also: sd-pixel extension that allows you to downscale.
https://github.com/Leodotpy/sd-pixel

[turbid whale]

Use img2img?

[fervent thunder]

What is this, channel?

[deft ivy]

still gave the same errror...

[fervent thunder]

I found it but it gives error

[turbid whale]

Did you:
Delete the venv directory

[deft ivy]

yes

[lapis blade]

Hey does anybody know of a Colab where I can do multi-concept lora training? I was using Linaqruf/kohya-trainer but I think it's broken now because all of the safetensors I make with it have NaN errors

[deft ivy]

odd, i deleted venv once and and didnt work, and did it again now guess it's doing something 🤷

[neat bronze]

Is anyone here using TensorRT?

[sharp condor]

Is there any server to create nsfw?

[ebon brook]

I want to use OpenPose and ContorlNet, but I don't want to buy a new PC. Where easiest to host online and access remotely? I'm debating runpod

[ebon brook]

search for unstable diffusion discord

[thorny sentinel]

Is there any open source project to create a talking avatar from your own images?

[broken gull]

If you find one feel free to tag me with it :p

[sour garnet]

Hi, I'm Joel. I lead the AI team at https://bifrost.ai (3D worlds, synthetic data, Gen AI, simulation). Organizing a tiny Gen AI meetup in Singapore this month, my place - if you're in Singapore, come chat and meet others building in the space!

Devs, researchers, builders, founders, investors welcome! If you'd like to give a short talk/demo, let me know too 👾

RSVP @ https://partiful.com/e/GLFrZgXSM2ycvWCqaHPr

[peak portal]

thats an a1111 bug

[fervent thunder]

Made some art for sale

https://www.etsy.com/listing/1479126962/robotic-human-cyborgs-20-transparent?ref=listings_manager_grid

let me know if you guys like it

[lapis blade]

would checkpoints work better?

[peak portal]

probably not. you have to upgrade diffusers to the git version

you might want to get off pytorch 1.13 and move to 2.0.1 and uninstall xformers if you haven't already

that'll fix issues too.

[neat cypress]

generative fill arrived to stable diffusion via controlnet https://www.youtube.com/watch?v=ot5GkaxHPzk

[bitter olive]

can someone explain me how to get high resolution images with SD? I usually get a lot of noise into the images 😅

[rose adder]

heyy

is stable diffusion bad for architecture i can't get a good city or thing like that

[potent spire]

definitely not at the level of Adobe Firefly/generative fill but workable for those in need here i guess.

especially those that do editing work in Affinity Photo or somewhere else

[warm junco]

Try out the hires fix in txt2img
Set denois to 0.5,
hires steps to 10,
upscale by 2 with esrgan4x+

[bitter olive]

thank you I'll tryi t...but question: isn't higher denois better? 🤔

[warm junco]

If you use latent upscalers you need to be over 0.5
If you use esrgan based upscalers you need to be lower then 0.5

[bitter olive]

interesting, thank you...still learning haha

[digital eagle]

ey guys, one question, anyone knows why promt is showing me ' The application has not responded ' ?

[weak vessel]

Hi guys ! Is there a way to automate outpainting on photograms, keeping the seed with a constant result ? For a 2’ video with medium shots that I want to turn it large shots. Did you test this workflow already ?

[neat cypress]

it is just new i think it will get better constantly

also you can test on different models and different samplers etc

[potent spire]

Adobe one too but yeah

in 1 year we will see again how much they all improved

or if something comes in between

let alone few years

[neat cypress]

how much adobe cost atm to use it? @potent spire

[potent spire]

Firefly web is free because its basically a test-/playground, for Photoshop it depends ofc

i made a deal with them and pay 35€/month for the whole CC suite

in september it will go back to normal price tho so i can make another deal probably

[untold moss]

i stepped back for about 2 months.

Im back now.

What did I miss?

[warm jasper]

I configured the remote access, now I can toast my laptop remotely!

[edgy canopy]

Hello, i'v a question, how do i uninstall an extension?

[broken smelt]

you can just delete the folder from your extensions folder

[edgy canopy]

Oh, thanks 🙂

[vivid isle]

So does anybody know which AI is being used to make the voices for these deepfake songs, specifically which ai uses the model from this video for instance https://www.youtube.com/watch?v=fGklID_hoS8

https://huggingface.co/QuickWick/Music-AI-Voices/tree/main/Freddie Mercury 300k

That's the model, but what program is it for?

Is it a free program?

Are any of them free?

[idle mauve]

if anyone is looking for prompts or inspiration to use stable diffusion, i just spent a couple of months building a word https://victorjuliani.notion.site/Flumenara-f391b6ee99bb4340bbc636493eb5a006?pvs=25, and would love if people illustrated parts of it... just @ me if you can make art based on this world 😄

[vivid isle]

Should I forward this to #🤝｜tech-support instead of here?

[golden tartan]

https://styledrop.github.io/

New Google Paper.

[vivid crest]

Is Stable Diffusion an LLM ?

My understanding is that LLMs are more concerned with text generation and not multimedia.

By that understanding StableLM is LLM and not SD.

Asking this because we had a discussion at work that anything that accepts natural language as input could be classified as LLM? And the argument is that, hence, StableDiffusion is an LLM ?

[barren prism]

hi，stable can provide the api to create videos?

we want to use stable make video

[neat condor]

okay gang time for my usual training power share i do on my free time : have 20-30 mins so imma train ur dataset on my 3090 for free. if u have crappy gpu or no gpu at all, feel free to jump in.

[ocean crag]

Is it better to train for a celebrity a smaller set of images like say 40? or does 160 images have better training? or will large set of images will cause over training? When would you know when to choose which number? Any advice or guidance would be greatly appreciated

Additionally, will 100 steps per image be good?

[knotty turtle]

Have you guys experimented with removing the background from your training images? If so, what affect does it have?

[proud crest]

"How should I start to make videos with the SD?"

[tidal lichen]

What are you using to train? kohya, dreambooth, etc?

[hard plinth]

ay, small world. Also would like to know this

[grand pollen]

Hi All,

Incoming noob question, as I only got into AI art about 2 weeks ago, first by getting early access to Leonardo AI (cloud platform), and then in the last maybe 5 days installing Automatic1111 and SD.

I've been doing a lot of reading up on everything, and am making some good images, with different models, lora's etc, though I'm learning all the time, and my prompt work needs improving.

In terms of training, I've managed to install Kohya_SS GUI today, after much pain, as it doesn't "just work" after running setup.bat, and I had to mess about installing latest xformer, and also updating the PyTorch that was installed by default to use the cuda one.

I've trained successfully on a few images, but I'm wonder if anyone has any tips on how I should use the "Max resolution" field in the "Dreambooth LORA" tab?

The model I'm using as my base to train on, is RPGv2, which is itself a trained 512 x 768 model. I assume that my best approach here is to set that field to the same, so that my Lora is trained at the same res?

My source dataset are all portrait orientation, and bigger than the training resolution

Advice appreciated.

Thx 🙂

[unborn idol]

I cannot use Philip K. DICK for prompt? Like, what?

[naive wigeon]

Stupid question. Is there an easy way for me to see the past images I’ve generated in one place?

[bleak matrix]

You can find more information about training in #🔧｜finetune -- I suggest that you start there! Since you're new to SD, you can learn more about prompting in #📝｜prompting-help

[bleak matrix]

Do you mean here or locally?

[grand pollen]

Thx Sunny!! 🙂 Much appreciated.

[bleak matrix]

You can learn more about training/images in #🔧｜finetune There's a lot that goes into the types of images you should use/train with. You want to use the best images that you can find. You can also find more info in #1080946152318443610 as well

[bleak matrix]

np! You can also see #1080946152318443610 for more info, too

[grand fox]

Question can .safetensor contain malware?

[real apex]

hey, looking for someone that has used stable diffusion a lot and can help me with a question. if you can help me solve the question ill give you some crypto

[knotty turtle]

Dreambooth

[tidal lichen]

With 10GB? Which specific script/repo/ui/etc?

If you please 😅

[knotty turtle]

Haha. I know, right. I just use Dream Look. ( https://dreamlook.ai/ )

[tidal lichen]

Thanks! I'll check it! I've tried a half-dozen trainers in the last 2 days and nothing runs 😖
Or it does, with massive errors...

Oh wait, you pay money 😱 😱 😱

How many tokens do you use in one training?

I guess $10 would be worth it at this point 😆 but the whole thing is to run it all locally

[knotty turtle]

Haha. I know. I'm a diehard local guy. I spent literal months trying workarounds to get it to work on my card (I'm an artist, not a technical person). Soo much time wasted. Finally had an awakening moment when I realized how many hundreds of hours I wasted, realizing I could just pay someone else to do it for me for hardly anything. Decided to try it out and have had incredible results. A bonus is that the training takes like 8 minutes. But, yes, I'm with you. I want to eventually make it work locally. I'm saving up for an RTX 5090 Ti SupaTitan Xtreme

[tidal lichen]

👍
Well written.
Are you able to choose which model you want to train? .safetensors files? 2.1?

[knotty turtle]

Yep. They have both a simpleton version for people like me, and then an advanced option if you know what you're doing to actually tweak things.

I think right now they only have three models available. I've been using realistic vision

[copper crystal]

word filters have their downsides but they're a necessity because people keep trying to use commercial services to generate toxic material. so a filter is applied. we've all seen how 4chan users attack AI demonstrations in the past. this is just how things are now. for good reason really.

4chan fucked around and found out

[tidal lichen]

oooooo

Yeah, I can get it running on my machine with the standard/base/default/SD models, but I want to train on some specific models.

[copper crystal]

ii mean, i'm a diehard local guy too, but technical user galore. I grew up in teh command prompt. Wouldn't call myself an advanced linux user but i have touched the sky when i compiled gentoo from source and got it to launch.

Even after all those roads walked, i am still thinking about using a cloud service for training larger datasets. if only to not turn my room into a boiler room during the heat waves

[hasty sorrel]

General question, has there been any talk about a potential software for stable diffusion in the near future, if not down the line?

[fervent thunder]

Hey guys how can we use weights

[broken smelt]

ctrl up and down on a word

[hasty hornet]

if you're using a1111 - (word:1.1), (two words:1.2), (two words, another thing:1.3)

[vagrant kernel]

hello, where should be --disable-nan-check added?

[broken smelt]

idk if u truly want to use that but edit the webui-user.bat and it will be after the = in set COMMANDLINE_ARGS=

if youre getting nans in your error log, sweeping it under the rug and pretending it doesnt exist isnt really a fix lol

[finite sable]

is there something in stable difusion like the generative fill of photoshop but without the guidelines restrictions?

[hasty hornet]

You mean like...outpaint?

I don't remember what generetive fill does, it's weird name

[finite sable]

you select the area you want and you tell the AI to generate on that

[hasty hornet]

oh, it's inpaint, yea automatic1111 can do it, invokeAI can do it

[finite sable]

Does it have any restrictions?

[hasty hornet]

a1111 + https://www.painthua.com/ can do both outpaint and inpaint iirc

[hasty hornet]

if you are doing it locally - no

everything photoshop can do can be done with SD tools too, usually better, photoshop didn't invent anything new...at least so far.

[finite sable]

is it harder to use than photoshop?

[hasty hornet]

I guess, but harder in a good way - you have more options \ more control

if you use a1111 at least so

[finite sable]

also when i said generate i meant generate things on an already existing image

not something like midjourney

[hasty hornet]

yea , that's inpaint

[finite sable]

is there a tutorial so i can know how to use stable difusion on inpaint?

[hasty hornet]

should be somewhere on youtube, idk

[vivid isle]

I guess I am going to go ahead and ask #🤝｜tech-support then

[safe echo]

Sharing with you all a project that I made (inspired from RoomGPT) which let you create your own AI generated design, print on a shirt and ship at home.
Feel free to DM me for more infos, feedbacks or simply have a chat with me!
https://www.teesgenerator.com

[peak portal]

you mean StableStudio?

#1108055793674227782

[peak portal]

do it

[potent spire]

Nah, thats false

lol

noone beats Photoshop in its tools right now

for generative fill, yeah its in beta and depending on style MJ and SD have the upper hand in generating specific stylized images

Also Adobe doesnt allow copyrighted material to be generated

MJ and SD do

But for in- and outpainting, Firefly/generative fill are already top level before even ready for final release

[hasty hornet]

You do you, I'd prefer more control of what I'm generating instead of advantage in...what, having it inside actual graphics editer? does it have any other advantage?

[fervent thunder]

i remember generating this image

[fervent thunder]

it has been a emoji ever since

[potent spire]

its very good in adapting to the lighting, style and perspective and surroundings of the image

for the control part, what exactly makes the in- and outpainting of Stable Diffusion with whatever plugin more controlable than generative fill/Firefly/Photoshop?

the latter one is ofc suited towards people who are skilled in art and editing and those are the ones that have the most control by far

so at the end it depends on certain factors and circumstances which will make people to pick one over another (or maybe both :D)

[hasty hornet]

Inpaint:
You can outline\ draw whatever you want and change it, you control how much you want to change it, if you want it based on something else.

Like...you can take image, put it on your image and let AI cook. It won't be good in one attempt, but well, compared to no control in photoshop (aside from just prompt) - that's still something.

outpaint:
kinda same deal, I guess, mostly random on both.
Some gens bad, some decent, some good.

Difference is - photoshop don't have models to chose from.

---

Just based on what I've seen I'd use AI tools from SD\ a1111 and it's plugin set and use photoshop or well, any other editor for editing purposes.

[potent spire]

the first one is available for the web UI of Firefly and will be probably soon available in PS too if we talk about the same thing. Im not sure. If not then you may be right 😄

And i agree for the second and third part, for the quality generative fill is very powerful in outpainting.

For the model part, yes Photoshop doesnt have all the models that are available for Stable Diffusion

the thing is tho that people who are using Stable Diffusion are from what i see mostly relying fully on that SD + A1111 +ControlNet (+ eventually some other plugins) pipeline and work completely different from someone like me

[hasty hornet]

technically there's controlnet too, but never tried to use it for inpainting , it's been some time since I inpainted something

[potent spire]

the workflow is radically different

tbh i personally never used Controlnet before

[hasty hornet]

yea, probably...for those who already use photoshop that's probably a good deal, especially since most of them probably don't want to learn all this mess in a1111

[potent spire]

the key guy from Firefly team experiments with it atm so probably he has some plans with such a thing for Adobe stuff too but lets see

[potent spire]

there are several factors and things here to consider ofc

same with people who use MJ only

or MJ + SD + those stuff like A111 etc. or MJ + PS/Firefly

[north nimbus]

Do you recommend the official adobe firefly thing or the sd plugin?

[potent spire]

im biased on that one, id say Firefly from my perspective but you gotta decide what you want

[hasty hornet]

I'd use PS with a1111 if I were some kind of art guy probably, but I'm not...but I just don't see myself using PS only rn, maybe when it's more mature

I'm just doing fun little images and sometimes edit them, again, with help of AI , no manual editing

[potent spire]

generative AI plays a smaller role in my workflow which also brings me to the point that the maximum control i achieve is by doing more manual work

everything has its advantages and disadvantages

im planning to do paintover over generated images as well

but thats not my main work

[north nimbus]

oh wait for firefly you need to run it on their servers right?

[potent spire]

in my opinion i get why someone prefers the SD "standard" workflow, why someone prefers MJ and why someone prefers workflow like mine

[potent spire]

yes

[north nimbus]

damn, I can't abuse my GPU

so yeah I think I'll use the pluigjno

[potent spire]

if you want it locally you gotta use Stable Diffusion

but running locally means you need a pretty good PC

[north nimbus]

do you have the git rep for the plugin?

[potent spire]

well for PS you need a decent PC either

soo... 😄

just type SD photoshop plugin

although there is more than 1 i think, not sure anymore

there was that plugin that is bound to Dreamstudio tho which means youd pay for credits

[north nimbus]

Thanks. Do you know of any plugins like this, maybe for AE or Illustrator

[potent spire]

not really

AE is a hard one

i mean people are building AI plugins for it but its a tough one

[hasty hornet]

yea, I see advantages in all of those tools in some way...

• MJ simple and really good to just generate things
• SD for control \ options \ variety \ upscale, whatever else people do this days
• Photoshop for manual editing and since AI is already there, can do some things there too.

But tools are tools and best to use them for what's they are good at.

Generate in MJ => make it better in SD => further edit in PS.
Profit

or well, probably skip MJ part if using SD, since can't really get MJ model to process image on same model...

[potent spire]

@hasty hornet exactly 😄

i mean here and there some salt is unavoidable XD

you know, people talking about whos workflow is better etc.

but then again, it all comes down to what you want etc.

[hasty hornet]

yea

[potent spire]

not everyone for example is willing to pay monthly for Photoshop, let alone Creative Cloud itself

and on top of that Midjourney too

who btw are going to introduce a 120$ subscription

[hasty hornet]

Yea I was thinking about prices for all this mess...

22$ elevenlabs to make AI talk in a voice I want
10$ GPT to actually make AI generate words
10$ for copilot, just cause I code shit and want it, even tho it probably sucks at coding, as any other AI
20$ photosop
80$ cloud?
10$ MJ

150$ total just for ai fun, yay

well , last 3 I don't want, just threw it in the list 😄

[potent spire]

wait, whats the 80$ one?

😄

[hasty hornet]

AI scammed me, it said creative cloud is 80$

2nd attempt was 20$ 😄

[potent spire]

oh lol. Creative Cloud is 60$/month for normal pricing

[hasty hornet]

it scammed me twice then

[potent spire]

i made a deal with them and pay 35€/month until september

[hasty hornet]

oh there are different plans, I see...never read whole thing

[potent spire]

but your workflow is not affordable for everyone either 😄

or just dont want to spend money at all

[hasty hornet]

my workflow isn't affordable even for me

[potent spire]

O_O

you sold your soul

[hasty hornet]

I mean, right now I'm using only SD 😄

11labs , gpt4 and copilot is just what I want, but not so much to pay 42$ per month for it yet

[potent spire]

aaah okay

😄

understandable

i pay a shit tone of money for my softwares and actually even more for plugins for those

hardware not including ofc

[hasty hornet]

want to make AI recognition of my voice + make gpt4 answer in a voice I can make with 11 labs, so I can ask questions with voice and get an answer in a voice I will train...sounds fun

[potent spire]

interesting 😄

actually similar to one of my future projects with software development involving AI

[hasty hornet]

speech-to-text => text gen => text-to-speech basically

[hasty hornet]

What's the idea?

[ripe needle]

does anyone have the lycoris from Vincent0123?

[potent spire]

Developing a AI voicemaker that has voice of someone whose voice i have fed to it

mine for example

So when i speak i speak with that voice

Ofc for my voice id ask someone else to test

But thats later, i will do some other projects before that

Also id like to build a ChatGPT plugin for 3ds Max and later for some other programs in my workflow

[hasty hornet]

hm, I don't remember If I've seen something like that already or not...there were so many voice AI related things I went through looking for open source ones lol

[potent spire]

Many of such things are subscription based

[hasty hornet]

yea, most of them

https://github.com/neonbjb/tortoise-tts
Open source, can be trained iirc

There was something from mozila too, but I think it's kinda dead now, not supported for years

• facebook made somethign voice related iirc

[potent spire]

If i make one i wont publish it tho

Would be a private thing

[hasty hornet]

understandable

[potent spire]

For chatGPT plugin for softwares i use? I would charge for that since GPT-4 costs money

And since id be the key host it would come down to me

[warm jasper]

The Mozilla TTS thing was so hard to install I gave up back then. I think it's the coquiai/TTS the "spiritual successor", at least it looks very similar

And it does generate voice in Portuguese, but it's not very good by default

[hasty hornet]

didn't see that one

[hard plinth]

Anyone have any free ai upscalers they'd recommend?

[broken smelt]

https://upscale.wiki/wiki/Model_Database

theyre all free

cant go wrong with 4x ultrasharp

[fleet nebula]

is there any good guide for using the google collab deforum? I'm getting an error and don't really understand why

[copper crystal]

no more collab. they shut it down for free users. you can still use it for stuff but not popular scripts that see a ton of heavy free use

they intend the free side for students not prime time use

[fleet nebula]

really? I ran it yesterday without any errors but I dont know what i did diff today

[copper crystal]

they block em on a case by case basis i thin

[fleet nebula]

this is the error im getting: render_input_video() missing 1 required positional argument: 'uncond_prompts'

[fleet nebula]

oh weird :/

[copper crystal]

i just know collab is breaking free diffusion use lately. it would be a full "collab shut this down because of these reasons" message if that was the case. maybe not here though ¯_(ツ)_/¯

[pallid dune]

what's the best

.safetensors

.cpkt

or .pt

[prisma cairn]

safetensors all the way

is there any nsfw channel where i could ask which certain model it is?

[obsidian sedge]

I just shelled out for a 3060, new monitor and power supply. I don't even play the games that would really benefit from it. Unless you call this a game. Thanks Stable Diffusion!

Seriously though. I can't wait.

[pallid dune]

ok thx man (i wonder why hmm)

[sage ravine]

Quick question - I'm rendering mass images in my pc and it's taking 4 hours for 768 x 768 but after zooming them they are not clear obviously so should I re upload on stable diffusion increase resolution or use hi rez before rendering which will increase the time 2x-4x

And suggest setting if ur saying hi rez fix to reduce that time

[hasty hornet]

based on my expirience hi-rez fix only makes it slower, but upside is - you don't get bugs you can get doing too high resolution on a model not trained for it, like skewed \ deformed too long things , repeating objects e.t.c

[sage ravine]

What's the use of hir rez steps? Like how many should I keep 10-20-50-100?

[bleak matrix]

Good morning, everyone! How are we all today?

[gray gazelle]

Good morning @bleak matrix

How can I contact the ceo?

Any email addresses available?

[bleak matrix]

If you need something business related, you can file a #1010934719455707218

Ticket with details, etc

[gray gazelle]

Okay thank you 🙏

[lapis sage]

yo

[bleak matrix]

Please see our rules regarding NSFW - #✍🏼｜rules-and-tos to find out more about content on the server. If you're interested in looking at models, you can check out #1047197565365538826

[charred iron]

Does anyone know can the current version of the controlnet openpose processor process foot key points or not? I have done a little bit of searching but there is not much info on google.

[sage ravine]

Dont know y But yesterday I created few picture and when I'm using same prompts and settings results are opposite to it (I created like human in space suit but now it's weird and eyes only with blended metal) I saved the prompts and all

[jolly olive]

is this discord related to STABLE DIFFUSION?

If so I need help with some errors

[gentle otter]

yes. use #🤝｜tech-support. and no need to yell.

[jolly olive]

ok

[warm jasper]

I knew I should be using something to help with the prompts. Let it generating pictures but I prompted shoes instead of boots and the results were disappointing

[sage ravine]

Help?

[hard plinth]

are these programs I download? Do I need anything else to run it? Never messed with a pth file so some insight would be great.

[warm junco]

a .pth file isnt an programm. its just a file usable by other programms to upscale an image

for example in auto1111 webui .pth files go into the models/esrgan folder

[hard plinth]

Oh I see ok. So what program would I have to download to use these upscalers?

[warm junco]

most Stable-Diffusion webuis support upscalers.
if you want a standalone programm for upscaling then you can use:
Cupscale: https://github.com/n00mkrad/cupscale
Chainner: https://github.com/chaiNNer-org/chaiNNer

[strange willow]

any one can help to get an indorsment to my first paper on arxiv for CS and im willing to share my paper with him/her

[fervent thunder]

https://youtube.com/shorts/LH7RG0M1rnY?feature=share

[flint mulch]

Hola, hay algun canal para hablar en español?

[hard plinth]

ay thanks you're the best ❤️

[unborn cave]

I see in promt of some art - lora:FINAL Robort. I can not find it. Can some one share it please( or place where it is)?

[jolly olive]

send help please #🤝｜tech-support

[quartz tartan]

hello everyone, I want to ask how to generate text on pictures that you want?

[hasty hornet]

sd is bad at doing text, unless you use controlnet and outline text for it

or well, unless you wait for SDXL to release

[copper crystal]

use deepfloyd

https://huggingface.co/spaces/DeepFloyd/IF

https://discord.com/channels/1002292111942635562/1101520089980882994

[wide tinsel]

Hello ^^

[bleak matrix]

Hello! Waves

[bleak matrix]

There are various ways, but it helps to prompt it appropriately. Things like: text "help me!" on a sign are a good way to start. You can learn more about prompting in #📝｜prompting-help or check out the #1080946152318443610 and various other parts of the server, but I do recommend checking out #1101520089980882994

[turbid whale]

Did you save the seed?

If anything you can just drag and drop the image into the UI and it'll load up all the metadata for you.

[sage ravine]

Don't know how to I always use -1

😭

Btw when I increase rez of pic from 768 to 1500+ it starts going weird so I have to use hi rez now need solution for that too

[hasty bane]

Does anyone know an image bank to sell, which has auto fill in image information?

[hasty hornet]

png info tab, put image here to get info on image parameters

[gray gazelle]

tf

weirdo

[alpine bear]

just react with a ⚠️ and the mods will see to it

[hasty hornet]

no info on release date, as far as I know, probably few more months

add another few months until actual good models based on it will release too

Why not 1.5 based custom model ?

[alpine bear]

Any reason you're looking for XL specifically? 1.5-based models have been very well-tuned for many different styles. Between one of the existing models on HF or civitai, and the various extensions like controlnet, depthmap, regional prompt etc... we can get some great stuff :V

[hasty hornet]

or you can use bot if it's current state is good enough

[alpine bear]

img-img has a vast array of workflows with current local installs

well you'll be waiting a bit - our gpu emperor has not told us anything new about XL's release date

[hasty hornet]

community made models based of 1.5 aren't worse at most aspects then SDXL

[tulip yoke]

No