#💬｜general-chat

1 messages · Page 74 of 1

potent jungle
#

That surprisingly all made sense. I haven’t shared or spoken to others about my SD hobby until today, figured I’d be far behind :p

still glacier
#

About latent space. To keep it extra simple (if any AI scientists read this.... spare me please) the usual stable diffusion workflow goes like that :
"prompt in actual english words" -> [CLIP phase] -> "token" (math values) -> [VAE encoding] -> [latent space neural network complex math iteration 1] -> [latent space neural network complex math iteration 2] -> ..... -> iteration xxx -> [VAE decoding] -> "pixel values"/image

#

When doing hires.fix it goes :
prompt in actual english words" -> [CLIP phase] -> "token" (math values) -> [VAE encoding] -> [latent space neural network complex math iteration 1] -> [latent space iteration xxx for pass 1 at low res] -> [latent space iteration xxx for pass 2 at high res] -> [VAE decoding] -> "pixel values"/image

#

That won't be in the test. Don't worry

potent jungle
#

I switched to Comfy and like how I can visually see the flow, makes it a bit easier to understand.

still glacier
#

basically what it means is when moving from pass1 to pass 2 we're staying in latent space. Which is good because it's fast and it avoid the loss of information caused by VAE decoding

still glacier
potent jungle
#

Is there ever reason to use any mode besides Bicubic?

still glacier
#

I'd say it's an aesthetic choice.

#

If you play video games. Think of it kinda like the different Anti Aliasing options. AAx2, AAx4, FXAA, SMAA, etc

#

each has its own strength and weakness.

potent jungle
#

I started using bicubic heavily as it seemed to produce the most accurate highres/upscaling, I really haven't played around with it since.

gilded anvil
#

@potent jungle Thanks. I managed to do it in automatic.

silent elm
#

Hey guys, sorry to interrupt you… I am pretty new and exploring the amazing landscape of stable and its ai. When I choose an image from my files with controlnet, it doesn’t use this as orientation - how can I change that ? Sorry again guys and thank you so much

potent jungle
still glacier
silent elm
still glacier
#

we all start from scratch at some points

forest trout
still glacier
#

thanks for the correction

#

I was double checking :p, I'm never using the bots channels

carmine steeple
#

Anyone get this error with SD XL?

#

OSError: [WinError 193] %1 is not a valid Win32 application. Error loading "X:\ai\ComfyUI_windows_portable\python_embeded\lib\site-packages\torch\lib\nvfuser_codegen.dll" or one of its dependencies.

forest trout
still glacier
silent elm
carmine steeple
#

Thanks

silent elm
forest trout
#

Just keep the original image sfw since they display the original image.

silent elm
forest trout
#

Wait a second for the bot to generate the results.

silent elm
#

Wooooow now I understood thanks man !!!

silent elm
forest trout
#

Depends on the controlnet model. I am guessing this controlnet model is either canny edge detection or it's a depth model (most likely this one) so you would not be able to use it to create variations of your given image.

#

The controlnet depth model creates a 3d depth map of your image and then feeds it into stable diffusion to diffuse an image that confirms to that depth map.

silent elm
#

Appreciate it a lot ! Thank you, was a huge help

forest trout
#

Well maybe it's a little drier than I was expecting but it explains what it's doing.

gaunt pendant
#

Yo how do u use the bot

surreal perchBOT
#

@gaunt pendant

FAQ: How do I generate images? Is there a bot on the server?

Currently, there is a public bot on the server that generates images available as a research beta for SDXL, you can find the current status of the bot in #1047610792226340935. There are plenty of ways to use Stable Diffusion such as the official https://dreamstudio.ai/ website or by running Stable Diffusion locally using your own hardware - check out #1080946152318443610⁠ for more details! You can also stop by #1025467151206854736 for any issues you experience while using DreamStudio or #🤝｜tech-support for any problems you encounter while installing it locally!

still glacier
#

Meh the faq message isn't that helpful actually when it comes to how to use the bot.... Check the bot channels pinned message for more information

potent jungle
#

From how it was explained above, HighResFix uses the model to make multiple passes of the base resolution image it first generates, upscaler then takes the highres image and upscales that completed image to an even higher resolution.

#

High-res is using a latent image, upscaler is using a completed image.

floral umbra
#

@still glacierHoi, do you know how i acquired the gamer role?

light rover
#

hi

cobalt crag
#

What is Lora?

plucky shoal
#

Lora is that thing which is trained usually on 1 specific thing, star wars uniforms, somebody particular. Or tilt shift for example.

silk jay
cosmic fox
#

I'm curious - why is the bot set to generate square images at 1048px x 1048px? Isn't it supposed to be 1024?

exotic sand
#

I am new to learning stable diffusion
should I go with automatic1111 or comfy?

#

or learn both😂💯

obtuse nest
#

Anyone does model training commissions?

thick drift
#

Hello everyone,
Has anyone had experience with generating flyers/posters using prompts that have actual text in them (like a lineup)? I can't seem to get the text to look normal in the generated outputs; it all comes out resembling ancient Egyptian hieroglyphics.

timid herald
timid herald
thick drift
near carbon
#

hi

high finch
#

Anyone have pfp like my they can send me?
I want to see different pfp I can use

upbeat tangle
#

hello what is sdxl? If you know a video that will help me or can you help me how to download and use it? now i have sd installed

trail lion
upbeat tangle
#

I think I found the download, but can we only produce realistic pictures with this model?

trail lion
#

Figured 😂. It's the latest model developed by stabilityai, it's a text to image model

upbeat tangle
#

I want to create landscape and anime pictures. When the sdxl is loaded, is it possible to use anime models in addition to it?

trail lion
gloomy trail
#

hello! is there any resources to teach deploy the stable diffusion on local machine? thanks!

trail lion
gloomy trail
#

okay thanks again

trail lion
#

Windows only, if you need Linux there are others out there

upbeat tangle
trail lion
#

Curious how people land on this discord and don't even know about the model, like how did you find the discord in the first place?

upbeat tangle
#

thanks for your help

floral umbra
#

Lol, what's the significance of the voice channels if there's jsut gonna be one person alone in each HAhaa Of the weeks i've been active here after SDXL came out, i've wanted to hop in one that has 3-4 people in it, but i've only ever seen many in them, but one in each

trail lion
#

Good question, the release had a bunch of people, I can't imagine wanting to join one at random personally

static lynx
#

is this for those who use SD with python working or like anything, cause i use a1111

lofty niche
#

GM

potent kayak
#

Is legal to monetize SDXL tool?

wanton nacelle
#

The bot does not seem to understand the concept of a tower that you can’t see the top of, no matter what I type. I’m trying to get something that’s really tall and disappears into the clouds, but it just keeps terminating the object in the image at a really short height.

ivory grove
#

Had no idea there was a discord, yaay for discovery

gleaming garnet
#

Hi everyone, i have a question:
im trying to recreate the process on the video below, and i was wondering if there are any free alternatives to midjourney where i could create my avatar with a similar level of quality
https://www.instagram.com/reel/CujzvcXuMGO/

potent spire
#

barely anyone and if then still not active

sinful marlin
#

Hey guys, I've got a really weird issue. Haven't used SD in a while now, and now it suddenly became REALLY slow to the point of being unusable. Additionally, the Dream Vinfinity that I had as my base model.ckpt fails to load now. Nothing in my setup changed, though I did update my AUTOMATIC UI afterward, but that didn't fix it. Anyone had any similar issue? Any pointers on what to do?

fervent thunder
#

I had a similar issue and reinstalling fixed it for me

wind kernel
#

with lora embeds, where do I download them to in the stable diff file structure. on civitai I see people using <loraembed> but not really much other info on how its used

solemn blade
#

Where do i go to create an image and how?

floral umbra
floral umbra
#

As I noted to knaapje, you do not want ckpt. It's insecure. Read my reply to him 2 comments up :)

#

This will use runpod which I've also used. And you can rent all from simple 3090 for what.. 37 cents was it per hour, to highest end Quadro 80GB a100 workstation card. Which I used thinky but costs 1 dollar 75 per hour iirc

solemn blade
#

@floral umbra I created one. Unfortunatly their is no ai platform that can create images that are able to communicate with textile machinery.

placid vortex
#

Where do you guys find models for Stable Diffusion?

solemn blade
#

Anyone into the following

I am a sock manufacturer.
I am looking to have AI create image Files from images given by customers and utilize AI image creativity.

Image files created are used to transmit data to a machine to engage functions.
Data transmission or Data signal designators to the machine are represented by the RGB colors located in the file. Machine capability is limited. RGB colors that can be in the file must be limited. Currently Ai image generators use shading, gradient, etc.. in creating images. you also can not designate the image size more specifically image size in Pixels.

Example
168 pixels wide 400 pixels height.

168 represents the 168 needles that are in the cylinder of the machine.
400 represents how many courses are in the sock. or how many times the cylinder has rotated picking up different colored yarn at its yarn intake points.

RGB colors in the file are used by technicians to designate fixed yarn takeup points on the machine.

Transmitting the data to the machine is not what I am looking for. I am just looking to create images

south flame
#

Me and two other educators and prompt engineers are putting on a set of masterclasses, paid prompt engineers.

Instructors:

  • Harris Terry - 30-year educator from USA, works as a prompt engineer with a startup and has given lectures and masterclasses on various AI topics including Midjourney with more than 60 attendees.
    https://aidreams.tech/

  • Geoffrey Mollet - Prompt engineer, web developer, and AI social media specialist from France with videos with 38m+ views.
    https://linktr.ee/singularitydiffusion

  • Tanvir Hafiz - YouTube AI educator and tech specialist from Bangladesh, AI freelancer for corporate clients and video production.
    https://www.youtube.com/@TanvirsTechTalk

This course will be Sundays at Noon EST, starting August 27th via Zoom.
The dates are:
Aug 27
Sept 3, 10, 17, 24
Oct 1

#

Course Summary
Session 1 - Basics

  • txt2img settings, models, prompting, styles, consistent characters, upscaling

Session 2 - Images and Inpainting

  • img2img settings, inpainting basics, inpainting for upscaling, Adetailer use

Session 3 - Deforum

  • installation, basic features, scheduler use and frames, movement controls, prompts, Init and Final tab settings

Session 4 - ControlNet

  • installation, model download, basics of UI, types of models & uses, usage in txt2img and img2img

Session 5 - Textual Inversions & LORAs

  • basics of Textural Inversions & Negative Embeddings, LORAs and various uses (characters/styles/hair/accents)

Session 6 - Additional Extensions

  • installation and use of:
  • ROOP - allows face swapping
  • Inpaint Anything - faster inpainting alternative
  • Tiled Diffusion / Multi region prompt - different prompts for different areas of the screen
  • Tag Autocomplete - see popular terms and use wildcards
  • Cutoff - allow separation of colors and descriptors by using commas
#

If for some reason a class is postponed due to an unexpected emergency or technical difficulty, the following Sunday will continue where the last session left off.

Upon sign up, you will be emailed a single-use link to join the class Discord server, and will receive a Zoom link for the class before the first session. Sessions will be recorded and available via Discord for members. There are no refunds for this course, and distribution of any Discord or Zoom links will get you removed from the course without refund immediately.

You are expected to have some sort of Stable Diffusion software (A1111, SDNext) installed before sessions begin if you want to follow along and try what you are watching. The Discord server and YouTube have install videos to walk you through the process, and the instructors are happy to help you during non-lecture time. By signing up, you agree not to distribute or share videos or materials from the course.

Let's become masters at Stable Diffusion!
https://sowl.co/s/1HsPX

#

Sample pics

#

thx for reading 🙂

earnest lichen
#

@solemn blade
Example
168 pixels wide 400 pixels height.

Step 1 - have a prompt template, that will create images in a style you're happy with
Step 2 - Have a setup that creates images locally (SD1.5, SDXL, or a custom checkpoint) or pay for an API key, and use that to create your images (clipdrop, or similar services)
Step 3 - Use that to generate images with with 50:21 aspect ratio (cause that can later be resized to 168:400px)
Step 4 - Take those images, and send them to your custom script which then:
• matches colors to the closest color available in your machine, based on a modifyable template.
• has a few filters added, so that contrast gets adjusted, so it still looks good on socks with a low resolution (though this can be skipped, if your prompt is good enough)
• resizes images to your compatible resolution, so that it matches your machine input

earnest lichen
south flame
#

No, I had no idea where to put this, and guessed. If need be, mods can take down. We are tying to help new prompters, and it really makes us almost no money, I don't know if you read the post, and did not think it was a problem on such a huge Discord. @earnest lichen

#

I could speak to a mod, or if they wish can delete it and I apologize for any inconvience.

earnest lichen
#

just get a quick ok from @bleak matrix , so it's in the clear

south flame
#

I teach AI and am looking to help new prompter

#

omg there are like 10 sunny names ROFL

#

oh you tagged ok thx

earnest lichen
#

will probably be fine ^^

south flame
#

OK, if I broke any rules I apologize. Should I wait around a few?

#

to speak with Sunny?

earnest lichen
#

will be fine XD worst it gets removed

#

just leave it for now 🙂

south flame
#

fair enough!

#

ok thank you for the info Caith

bleak matrix
#

Let me take some time to review; thank you!

floral umbra
gleaming garnet
crimson belfry
#

i love SDXL i just have no computer to run it properly atm

vestal dew
#

then how do you love it?

#

It is hard to use. If I make a prompt to give me a good character, it ignores things like clothing and situation. If I make a very short prompt to get a generic character, then I get control over clothing and situation. It's hard to get both, often impossible.

snow cedar
#

Could somebody tell me if StableDreamer bot is capable of private messaging or are we required to sit in a bot channel to generate images?

vestal dew
#

I think you have to be in the channel

cyan temple
#

anybody know how to get Digitigrade leg style robot prompt?

proper sparrow
#

where is the fps change settings

#

is it the dynamic_thresh?

finite cloak
finite cloak
proper sparrow
#

just want to know the fps setttings ..how can i change fps of video in gui settings

supple coral
#

any tips on how to solve RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!.? I am not a computer expert

normal pagoda
#

maybe you have all your tensors spread out on multiple drives

supple coral
#

is there a way to check or solve without being programmer? and more important is it just error msg or really stopping because i can generated image ok

#

not fast but ok

warm junco
supple coral
sullen fractal
#

далбоебы бля ахахах

supple coral
signal trellis
#

does anyone know of any good discord servers for ai voice model stuff? like is there something like Unstable Diffusion but for ai voice models? plz @ me if you do, thanks in advance

potent spire
lavish lake
fervent thunder
#

I think jiafei is cringe

alpine shuttle
#

Serious question if I have the hardware is MJ better than SD?

trail lion
#

sounds like a subjective topic

alpine shuttle
#

Just curious

trail lion
#

literally never tried midjourney ever

#

not even curious about it, tbh, so that disqualifies me

civic light
#

In StableSwarm, when running ComfyUI, using the Load Image node results in a "400 - Bad Request" when trying to load any image other than example.png--even after complete re-install. No errors or diagnostic info in cmd. Running locally on Win 10 Pro with RTX 3090 FE. Anyone?

rotund notch
#

Hi, I found a 3090 and a A6000 at a similar good price, which one would you choose? They look similar to me

potent spire
south flame
#

MJ is great for one-offs, things you don't need perfect consistency for. SD I use, as I need to correct pictures with inpainting, use different models, use ControlNet for poses... MJ IMHO is more beginner software. For example, the seminar I give on MJ is 3 hours and teaches you... basically 100% of the program. My SD seminar, on the other hand, is 6 hours and really just gets you from install to almost intermediate, there are so many features.

Really, MJ for me is total fail because of no inpainting. No correcting? Can't use it for my job lol

#

@alpine shuttle

potent spire
#

Thats why i asked the user questions above

spice plank
#

There a way to speed up model loading?

normal pagoda
#

asking myself that as well... and I've got a VAE installed as well

south flame
#

You can set how many models to cache in settings

#

also, using --medvram with this helps

kind shale
#

Mmm heyo, is there any place that I can find models to download ? Kinda new here D:

south flame
kind shale
#

Thankies

south flame
#

my pleasure

hushed herald
#

Hi has someone else seen that the tensors are cleared with the new clora workflows in comfy? I need to reload the base model every time

pale latch
#

it's a feature now

#

oops i went to get the link and they only mentioned it's coming quite soon this week

south flame
#

I will have to see how well that works, as I often use PS and minipaint to go back and forth with inpainting, ie paint part of a sky with crap blue, denoise at 0.2

#

can you load a pic into MJ back?

pale latch
#

anywyas. no big deal. sdxl is free and is better

south flame
#

lol true

pale latch
#

MJ does lots. i wouldn't use it but they are staying competitive

south flame
#

if it had more models it might interest me

pale latch
#

what we're doing with revision now, MJ users have been doing all year

#

it was like the one thing i considered paying to play with

south flame
#

lol, just canceled it today how ironic

potent spire
#

SD has its target audience, MJ theirs

#

With interchange ofc

#

Not a lot of people are willing to jump to SD even as it is free

potent spire
potent jungle
#

MJ is much more user friendly. Different target audiences really. One is casual, one allows for more control.

old ruin
#

yo i have question

#

Is it possible to link an image on stable difussion ?

south flame
#

Its really not all that hard, once you understand you have to work a little harder

#

I just am not into making 100 iterations and picking out 3 that are ok lol, corrections and ControlNet are a must for me

#

I used MJ a bunch, but I found it lacking in making things like my comics, and for my job no way I could possibly use MJ

#

but i see different target audiences

faint loom
#

sup homies

little citrus
#

MJ seems to be more for those who want to make something that looks pretty or interesting and that fits a general theme. SD is more for extreme control. SD also allows for bypassing MJ filters.

faint loom
#

i mean with custom models and checkpoints, loras, custom scripting, multiple passes

#

isnt mj just a product?

little citrus
#

MJ is SaaS.

#

SD can be run locally or as a SaaS product.

faint loom
#

some folks struggle with accuracy and realism but with the proper setup in a backend program like comfyui you can do anything

#

the sdxl models by themselves are wonky, but there are many ways to work around (or with) them

#

like checkpoints, loras, controlnets

little citrus
#

Base models are intended to do almost everything but not be perfect at anything. Fine-tuning (LoRA, TI, etc.) is intended to train the model to be better at certain things.

#

Checkpoint is just a fancy way to say base model.

potent spire
#

@south flame part of them uses it in company with Firefly and Photoshop tho

#

for me personally, i dont want to pay for MJ anymore

#

and SD doesnt suit my needs and workflow either

faint loom
#

how?

potent spire
#

wdym how? 😄

faint loom
#

i think SD is limited only by the user

#

waiting for the day when text, voice, and image generators merge into one

#

eventually image generators will have to reach the point where they become text generators as well

#

once they learn their alphabets

potent spire
#

nah, they all are limited compared to doing it yourself but i prefer Firefly (especially in the future) and SD could have been a company for FF for me but especially atm its not worth it for me

#

plus i have enough to play around with customizing my art softwares, i dont need the art generator that is serving primarily a idea and reference generator to be unnecessarily time and/or learn consuming as well

#

well learn not so much because i have to learn to code etc. anyway

faint loom
#

not sure what you are trying to do, but SD supports masking and controlnets for advanced usage, as well as clip functionality for advanced merging/blending

#

you don't need to learn code if you use a program like comfyui

#

just have to learn node workflow

#

i picked it up in a day

#

though i have experience working with nodes from my blender days

potent spire
#

yeah but for my workflow and workspace SD is very inefficient, especially with current specs as well and on top of that i already have the tools or better said softwares that give me the maximum possible freedom and control at the cost that i have to learn art and editing skills myself to reach that point

faint loom
#

that's fair and admirable that you're willing to improve your own artistry

potent spire
#

thats where Firefly fits me by far the best since i also happen to do my stuff in Adobe ecosystem when it comes to 2D especially

faint loom
#

just be aware that ai can always suppliment your endeavors

potent spire
#

im not saying SD isnt the best for some ppl tho

#

ofc

#

it does

#

it really does

faint loom
#

no matter how great of an artist you are

potent spire
#

i love AI

faint loom
#

i think it works best for artists really

#

not meant to replace traditional methods but to provide a new way of enhancing and dreaming up new possibilities

potent spire
#

well skilled artists can maximise the potential of generative AI the most

faint loom
#

I'm planning on wiring up my photoshop edits and pen sketches to SDXL and see where that goes

potent spire
#

SD suits well certain people, but for me it doesnt work well with my pipeline. And then my specs make it even worse to use SD seriously as a side tool

faint loom
#

they just released some kind of experimental control model that turns line art into fully colored/shaded illustrations

#

though if you're clever enough you can figure out how to do all of that yourself

potent spire
#

oh, you mean sketch to image?

faint loom
#

not sure what it's called but it was just released (this past week?) along with a handful of other tools

potent spire
#

tried that with Alpaca, didnt try it in SD. Firefly will have it as well. That feature is amazing.

faint loom
#

including two new control nets

potent spire
#

controlnet itself is amazing

faint loom
#

yes it's basically an automatic photoshop tool

potent spire
#

there are a plenty of tools im waiting for Firefly to release, some of which are already existing in SD or Alpaca plugin for PS themselves and some arent known anywhere as of now

#

interesting times are coming

#

even Autodesk brings now again some generative AI stuff

#

the only one remaining is Maxon to do it as well

faint loom
#

i mean we went from disco diffusion to midjourney beta to dalle2 to early SD to modern midjourney to SDXL

#

all more or less within a year

#

😁

potent spire
#

dunno about you brother, but for me it started with that DALL-E mini xD

#

Crayion

faint loom
#

i had dalle2 access around that time

#

and i was deep into some other generative notebooks

#

like disco diffusion

potent spire
#

nice

#

so a veteran 😛 😄

faint loom
#

I've been doing this since image generation was a blurry and indistinguishable mess of pixels

potent spire
#

u are using Blender?

faint loom
#

you'd ask for a bird and it would give you a pizza

potent spire
#

lmao true

#

and all the abominations of creatures

faint loom
#

i used blender for 3D art unrelated to ai

potent spire
#

Blender has that SD plugin

faint loom
#

i didn't know that

potent spire
#

even for Maya one is developed

faint loom
#

but i use comfyui, which is a node based program like blender

potent spire
#

for some reason Autodesk brings some sick generative AI for Maya but 3ds Max (which i use) isnt mentioned...at least not yet

faint loom
#

it's limited only by how clever you can be with the nodes

#

and has custom scripting for advanced usage

potent spire
#

i can imagine that although im not deep in programming and scripting area yet

#

but i hopefully will one day

#

because i will need it

faint loom
#

other people will develop the tools you need by then

trail lion
faint loom
#

if you want them now you're gunna have to be inventive

#

i mean the tools already exist you just need to assemble them

trail lion
#

I'll take open source any day, gimp over photoshop, etc

#

professionals often dont have a choice, but as a hobbiest

faint loom
#

photodemon is cool too

#

and inkscape

#

blender also does just about everything these days

#

including photo editing

potent spire
trail lion
#

blender and gimp, they are amazing really amazing

potent spire
undone garden
#

What's holding back the release of inpainting xl models?

trail lion
#

you cant inpaint them? I wasnt aware

potent jungle
#

I am already inpainting? Though not via controlnet.

trail lion
#

ok so you are asking about inpainting controlnet loras

#

or something similar

latent quarry
#

Hi

faint loom
#

there are many ways to inpaint

#

if you're looking for a program to do it for you rather than assemble the pieces yourself, I've seen a couple paid services

untold ocean
faint loom
#

ah yeah that was the other one

#

i loved the "animated" frame compilations

#

made it feel alive

sleek owl
#

Is there a way for me to input an image of a item and have it used in the image i want to generate? like I upload a picture of a pen that i want to be used by the character writing a note?

violet arrow
#

Hello everyone! the latest update to my Deforum-Helper app brings some new features. Your control options have expanded, now encompassing camera FOV, near and far. Plus, you can fine-tune the Deforum diffusion strength. I'd love to hear your thoughts on any additional settings you believe should be included in the editor.
I've also introduced a bar visualization in the 3D live mode for more effective control and monitoring as you work. Your feedback on these updates is highly valued!
Enjoy and thanks for using my app 🤘
https://deforum-helper.vercel.app

dusky stream
lunar pier
#

hey

#

any one around ?

#

can some one explain the :1.5 in prompts

#

and how i would take advatage of that in otehr prompts i make ?

light rover
#

hi

dire mulch
#

hi

plucky shoal
#

if it is (snow:1.5) it put more attention on word snow and therefore higher probability to have snow in picture. It can also be used that way ((((snow))))
Not sure if you mean this, but probably yes. @lunar pier

lunar pier
fading prism
#

Can you please help me?

#

I am trying to get prompt to work but it doesn't

lavish lake
sleek owl
#

Hi, can i combine both an image of mine and a prompt to create a new image or ideally an image using the pen of my first image?

last parcel
#

I tried to make "Rinoa Heartilly nude" but Dreamstudio gave me a message saying "Something isn't quite right with your prompts". Are NSFW prompts prohibited or something?

novel carbon
#

can i use this bot for nsfw prompt?

undone marsh
#

Does anyone know how to change the pose of a character image without changing his/her outlook and style? Tried on Automattic 1111 img2img and controlnet but it doesn't work. It changes completely the outlook

turbid whale
#

Yall trying to use public bots for NSFW images... 👀

undone garden
sick ingot
#

tiger

lavish lake
south flame
potent spire
#

But not bad at all

south flame
#

Please, I love suggestions and need inpainting for my job, any suggestions?

trail lion
south flame
#

@potent spire said there were other inpainting extensions, personally i just use regular inpainting, or Inpaint Anything

#

I also use Adetailer

#

but I use inpainting for work

trail lion
#

ah ok, not sure, those are what I use

potent spire
#

Or Alpaca plugin if you have PS at all

#

Oh my bad, it doesnt really have advanced inpaint yet

normal pagoda
#

How soon? nyaThink

valid parrot
#

Hi, who can suggest what models and promts are better to use to make 360 landscape panorama like Blockade Labs did? They're actually very good but now after they made subscription it's only 15 generations/per month if free (and 48$ for "unlim" is not that small amount).

rain hamlet
#

hi, i am creating a lora adapter on top of stablecode-instruct and am running into a specific. is this the right place to find help?

pale latch
#

That was 1.4

#

1.0 was the 10th iirc

rain hamlet
valid parrot
#

To make 360 view (so you can sphere in Unity, put that as texture on it) add camera in sphere and "kinda" make 3d view. Like certain (CENSORED) Labs did.

rain hamlet
hallow plaza
#

is a1111 working with sdxl now?

lofty sage
#

anyone know what the best driver for the 4080 for training loras?

pale latch
#

i think automatic had support for it during the beta period even if you used the dev branch

hallow plaza
pale latch
# hallow plaza Whats the difference between the base and the refiner model?

tbh, i don't know. The refiner feels like a step child. it was the original version of SDXL but then they changed directions but decided to keep that original model as a second stage.

I honestly don't even think it's important for end results. It just complicates things for little benefit. Especially if you're upscaling, all the refined details it adds are just smoothed right out.

#

The Super Stage event for the SDXL 1.0 release, the devs have a big round table discussion about all the ways its not needed

#

i dont understand the refiner at all. It's purpose seems elusive

broken cave
pale latch
#

I dont know why. Great results are had without it. MJ would probably rather maintain one model than two.

valid parrot
vestal dew
pale latch
#

most people aren't using controlnet workflows in comfy either.

for most reasons, a1111 works. i think the extension got updated by now too.

vestal dew
#

Not sure why not.

#

Controlnet is very good. It''s still only limited static posing, but we had a big step backwards when SDXL released

#

static posing as in it helps get images with objects and people posed in what looks like a static pose they could hold, rather than a frame of action from a real action

cosmic drift
#

Hi i have a question. Can i use stable diffusion model in my app / platform

rain hamlet
pale latch
#

sdxl is a leap

cosmic drift
#

Hi

vestal dew
pale latch
vestal dew
#

It's not. We lost the control of controlnet, the ability to pose somewhat, to copy certain images in color and composition. We still aren't back to that with all the popular UI's. It's been a step back.

potent spire
#

So MJ finally has inpaint too

potent jungle
#

SDXL is definitely a leap forward. It just doesn’t have all the tools 1.5 had months to acquire.

#

1.5 is a modified Toyota with a custom engine. SDXL is a stock corvette straight off the lot.

pale latch
#

I kind of think that when people just refuse to learn and use the tools that are available, choosing to wait for someone else to make it available to them in their preferred form, that's more of an individual preference dillema than any kind of industry wide step backs.

potent jungle
#

It’s open source. Your basically going to lose familiar functionality between versions every time. The development of tools isn’t some centralized project.

pale latch
#

On day 1 of sdxl, i saw people putting guidance onto a 1.5 latent, passing that off to the xl model after x number of steps, and cooking it into it's final form that way. A little janky but pretty great results still. Of course, i try to find that now and it's buried in the feed.

wicked sonnet
#

Is stable diffusion unstable with memory? I just got a bluescreen cause my memory filled up

#

sd 1.5

wicked sonnet
potent jungle
wicked sonnet
#

I mean, the program

#

sdxl is a model?

potent jungle
#

Yes, whatever interface you use is up to you.

wicked sonnet
#

I suppose hmm

potent jungle
#

I use comfyUI but I recommend Auto1111 for beginners.

wicked sonnet
#

well I'm having issues with Auto1111 but I don't want to drop the tools it gives me

trim magnet
wicked sonnet
#

I see

potent jungle
#

Auto1111 very recently got some extensions to help with SDXL, but as I don’t use it, I don’t have more exact info.

harsh thunder
#

i feel like everytime i generate images for a while then close stable diffusion, i lose more space than i should

potent jungle
#

? I mean, my workflow saves three images per generation at high resolution. So hard drive space is eaten up, but I have so much I don’t notice.

wicked sonnet
#

windows can be slow at updating space changes

sonic hemlock
#

I am having problems with my SDXL install. Is there a Tech Support channel somewhere on this server where I can ask my tech install questions? Thanks in advance.

dim oxide
#

question... does the model loaded changes CLIP or DeepBoorou interrogation?

fervent thunder
#

ik 😎😎😎

dark plinth
#

this might be a beginner question, but is it possible to run the "dreamshaper" model in deforum stable diffusion?

#

I managed with warpfusion so I just assumed I would be able to in deforum too...?

vestal dew
urban dome
#

Is there a way to use the Portrait Depth Estimation feature, like in the Clip Drop API, through an existing web UI?

chilly vector
#

Why do I feel ashamed of using stable diffusion after being a few minutes with some acquaintances?

vestal dew
#

huh?

trim magnet
#

blud was doing a public event generating sexy girls 😭 💀

potent spire
#

So MJ inpaint is kinda like DALL E

hallow plaza
#

How do i use the bots to generate some images

vestal dew
#

I tutor. When the students are working on a question, I tell them I'm staring at their hands to make sure they're really humans.

rotund notch
#

where are the cards of all the bitcoin miners?

#

I see nothing on the market

#

best price for a 3090 is $700, is this a joke? like the housing market? will never crash?

trim magnet
#

buy them while u can,now with the ai craze it and when ai goes more mainstream prices will go up again

broken cave
#

@ $525

ivory grove
valid widget
#

heya, I have the --no-half problem and all that stuff about nansexception

the thing is that all the tutorials to solve it are from the files folder and none of that has helped me because I use colab

manic harness
#

Hi Everyone,
I recently started contributing to open-source projects, and I'm very eager to begin contributing to Stable . Any guidance or assistance would be greatly appreciated as I embark on this new journey. Thank you in advance for your support and insights!

royal harness
#

Hi someone now, the size of the SDXL model (in Go) and the amount of ram memory required to run it ?

lavish lake
rotund notch
#

Ya

summer dagger
#

I am not sure where to start looking for someone to do some NSFW commission work, any suggestions?

turbid whale
undone garden
#

Guys, I run automatic1111 with --xformers --api --medvram --no-half, on a 4070ti and get 1it/s on a batch of 4 512x512, that's not normal right?

turbid whale
#

Medvram on 4070ti ??

plucky shoal
#

shouldnt be normal. What model are you using?

undone garden
fossil phoenix
#

@undone garden I use it on a 3060ti 12gb with only --xformers and --api and I do around 6.25t/s

#

but not using no half.. so

elfin pebble
#

hi guys, can someone explain to me how SD works? do I buy credits to generate the artwork? or is there a plan?

wild steppe
karmic brook
dark plinth
#

Am I able to have the controlnet options in stablediffusion just like in warpdiffusion? I am running on Google Colab Pro; not local

#

might be a stupid question....

last sand
#

hey there, what resolution should my images have to train a model in sdxl?, I don't use this since v1.5 and I don't know if it's still 512x512 catthink

burnt granite
#

Is there anywhere I can find proper documentaion over the different options present in Stable Diffusion?
Or is that just not something available atm

kind basin
#

Why it's infinite loading

#

stability chat

long thunder
karmic brook
plain wharf
#

Is there a link to the recording?

red ermine
#

In what channel can I make the bot generate my ideas?

plain wharf
cobalt nimbus
#

Hey, I´m totally new to this. Where do I start to understand how creating these images work?

glass yarrow
fervent thunder
pseudo jetty
#

Does anyone know any web based Stable Diffusion image generators? someones is asking on another discord but i dont know of any

daring flame
#

I'm actually doing web development, logo designing and I would love to help you out is anybody interested DM me

eternal geode
ripe pagoda
vestal dew
#

A111 and SDNext, let's help them get Controlnet working, do what's good for the community

muted moon
#

Heyo :)

#

I've had struggles with getting SD to add scars on my characters. Any hints on how to solve that?

ivory grove
#

Are there any guides out there to setting up AITemplate?

indigo ermine
#

How do I stop getting notifications from the "events" channel in this discord? I've already set the notifications to none, but I still get them anyway…

open zinc
#

can i use stab

trail lion
open zinc
#

whoops

red ermine
#

is it possible to add the dream bot on my own ds server?

burnt granite
# trail lion Options for what? Which UI?

A1111. Honestly a good documentation would be cool but asking for too much. But I wanted to ask about the inpaint mask related options. Mask Blur. Masked Content. Inpaint Area

lone shell
#

give @delicate oxide back owner

open zinc
#

can i run stable diffusion with 4gb of ram?

trail lion
open zinc
#

Guess ill never know

burnt granite
#

is the alpha mask option new?

#

and I saw a mention of a model made for sd2.0 but in a 512 base - so like, can sd2.0 different from sdxl?

open zinc
#

im goin back to youtube tutorials

grand pasture
#

whats the best SD model to train a lora of a humanoid with a non-standard head/face

mystic musk
#

The bots don't allow for negative prompts, do they?

swift sigil
#

if any getting error running stable diffiusion just use "--skip-torch-cuda-test --precision full --no-half --no-progressbar-hiding --opt-channelslast"

scenic compass
#

hey anyone here can helo me?

#

i'l tryint os install a controlnet model but can't find the download button on the huggingface page 🥲

fossil phoenix
#

Okay, so, generally... why does stable diffusion refuse to generate scuba tanks? 1.5, XL... none of them do. lol

fossil phoenix
vestal dew
#

they do know negative

#

tab from your prompt, type "negative", you'll get choices, and one is negative prompt. Then you can type the negative prompt in that new box

river herald
#

Ive made two Textual inversions, now how do I get them both in a single render together? still dodgy?

last parcel
karmic brook
last parcel
karmic brook
last parcel
turbid whale
vestal dew
last parcel
# vestal dew Use SD

How? Do I switch the model from SDXL to Stable Diffusion? Or do I need to go to another site?

vestal dew
#

You have to run it locally

#

the sites that let you create images are set so you can't change it

#

they aren't going to let people make NSFW - porn - stuff

#

NSFW is a bad name. If you worked at Coke a Cola, a Pepsi logo would be NSFW

last parcel
vestal dew
#

You need a good GPU. Do you have it?

#

that's to the best UI, SDNext

copper plover
#

Hi, I have downloaded SDXL and have ComfyUI set up, but I want to create some kind of digital art based of my own portrait like the following: https://imgur.com/a/T1XCXsw

Is it possible? It should still look like me, but have some of the digital art looks to it (like multiple colors etc) and not have a background (or solid color background I can easily remove in photoshop)

placid vortex
#

Do you guys support Easy Diffusion?

mystic musk
stable lantern
#

I would like to ask you, why I use controlnet to select my local image, will always generate no image, and then in the generated command can also see "width:0 height:1", regardless of whether I manually select the image ratio is so, the complete command I generated is "/dream prompt:UHD,highly detailed,professional,bokeh,chain saw man,senior high school student,asian,fighting_stance,white shirt,sneakers,tie,black jeans,blonde hair,tokyo,street,cinematic angle,(((full body))),rain style:Comic Book width:0 Height: 1 seed: 192892105 ", I downloaded to the local image, the url is https://i02piccdn.sogoucdn.com/a77f145ac4c1576a

violet umbra
#

Good morning ☀️😃

paper bane
#

is it possible to add this bot to my own server or not yet?

vestal dew
outer lynx
#

unfortunately, same here, did you manage to sort it out?

south flame
vestal dew
outer lynx
#

Invoices? Anyone getting invoices from Stability.ai please guys?

tepid orchid
#

newbie here, where can I find a document or tutorial that show meaning of setting value? Like strength, CFG, noise. Thank you.

civic yarrow
#

Hi. Are there channel moderators here?

unkempt pond
#

I am looking the whole time for faqs where i can find the general prompting rules/infos. But under ressources and faqs its just so general. I want to upload a picture and the the picture should be modified by SDXL. Can you help me? 🙂

cerulean kraken
#

ugh, civitai is always having problems :/

turbid whale
eternal pendant
#

is there a way to see what kinda added words are being add when you add a style using the bot?

crimson belfry
#

Yes! Stability does it again!

slim raptor
#

Is there a setting between lowvram and medvram?

crimson belfry
#

im not even sure if i can run sdxl local on my laptop

languid pendant
#

Its been a year already? God damn

crimson belfry
#

i know lol

languid pendant
#

Here i was thinking its been like 7 months

tall gorge
#

its still absolutely wild to me how its only been 1 year since SD was released. Its ridiculous how quickly its been improving. The world of AI is just built different

slim raptor
#

I tried sdxl but 30min for a 1024x768 pic is ridiculous on lowvram mode

crimson belfry
#

yeah that wont do, you might as well just use the disco

#

i wish they added loras

languid pendant
slim raptor
#

Text to vid looks wonky now

tall gorge
#

yeah its still kinda mid

slim raptor
#

Maybe in 2 years

tall gorge
#

but its getting there

#

2 YEARS

#

lol

#

thats the equivalent of a decade in the world of ai

slim raptor
#

I remember seeing report about bad looking ai text to img pic some 1-2 years ago so I thought 2years would be a fair assumption

crimson belfry
#

Text to video looks good on Pika but its not something to be jumping up and down about yet

slim raptor
#

It feels like a dream now, not very sharp or detailed

#

Like how you can see stuff in your mind but the details break ddown the closer you get

crimson belfry
#

its getting close

slim raptor
#

I hope nvidia makes some cheap AI card for general consumers

crimson belfry
#

We are moving at lightspeed when it comes to ai

#

i look forward to many of the online platforms only because we do not have the tech available at a reasonable price

tall gorge
languid pendant
#

I really would like to see technology to be able to rewatch our dreams

tall gorge
#

amd is catching up to nvidia fast in terms of software support, any other competitors will be very close after them

crimson belfry
#

im using an amd laptop but i heard that they can not run well atm

slim raptor
tall gorge
#

nvidia just has a massive lead

crimson belfry
#

its all about who controls the chips

#

now we see why the big uproar came about a few years ago when Trump was in office.

slim raptor
tall gorge
#

idk faster and more memory is more important for the most part when it comes to AI stuff

slim raptor
#

My air cooled ipad pro is already 2x more powerful than my workstation laptop from 2016

crimson belfry
#

Does anyone here use the power of ai chat to help fine tune prompts?

#

i dont see many

#

Many are underestimating the power of them

slim raptor
#

I used bing once

crimson belfry
#

bing is ok i think they use chatgpt right?

tall gorge
#

one thing i saw people do is use claude which has a massive 100k context window, dump in a file with a bunch of prompts and then ask away

slim raptor
languid pendant
slim raptor
#

Try bing, its free

languid pendant
#

Niccce okay. Thanks

#

I just tried that and its just giving me a copy pasted list of tips on how to structure my prompt. Not actually helping me

crimson belfry
#

chatGPT or claude

#

stability chat is coming soon from what i believe

languid pendant
#

Whats claude?

vestal dew
#

stability has beluga 2

wild steppe
halcyon hollow
#

Hello everyone
How are you doing

tall gorge
#

quite good my fellow

untold vigil
#

wanted to flag this open letter from generative AI artists to policymakers, calling for them to have a seat at the policy table too -- read, sign, and share (privately) if you like: -- feel free to ping me with questions: https://forms.gle/5gHAUeyPztQKBA5G6

potent spire
turbid whale
untold vigil
trim magnet
potent spire
#

and the representative of Stability AI was at the hearing about 2 weeks ago or so?

#

and btw. your letter is set to fail imo because actual artists actually used AI before generative AI was a thing

cobalt crag
#

Wait what

#

I thought Stable Diffusion existing for so long like many years???

#

It's just 1 year old?...

tough bough
#

The perspective of people using a technology is no less important than the people making it. You wouldn’t make traffic laws by only consulting car makers.

cobalt crag
#

Did I just witness the event of humanity history?

trim magnet
potent spire
#

and come one, why do those AI "artists" always blaming the corporations, its them that are aligning with companies like Stability AI itself

#

plus THEY are the problem as well

#

come on*

potent spire
spice plank
#

How do I turn off getting the red 1 in the corner whenever a random discord server wants to @ everyone? I turned off notifications and @pings but it doesnt stop the red 1

potent spire
#

its also up to the providers like Artstation etc. to take a stance

trim magnet
potent spire
#

yes

#

but it can make a significant difference

#

its not about the "the cat is out of the bag" crap, because the technology wont just disappear, but there are definitive things that can be done to at least reduce the damage

tough bough
#

I don’t think draconian laws on model training are going to reduce the damage at all. If anything, they will make it worse. If AI is actually going to make manual art obsolete (which it won’t), you want the models as open as possible so that people can adapt instead of die.

desert apex
#

Anyone know a way to make specific words 'legible' in a prompt for a logo?

It does a pretty good job at coming up with the unique images, but specific words requested from the prompt are not legible. It might not be possible but I thought I'd ask 🙂

potent spire
tough bough
#

Draconian laws just mean that the largest IP holders replace their own artists with AI reaping massive rewards in throughput and efficiency. Everyone else is left out in the cold.

trim magnet
#

they can try to ban it , just like when they tried to ban torrent and software piracy, are both of those things dead?

potent spire
tough bough
#

Artists still lose. Artists lose more. Everyone else also loses.

potent spire
#

its about regulating

#

both from state as well as sites

undone garden
#

I trained my first decent Lora today but something strange happened. It started great, I got perfect portraits with no prompt at all apart from the Lora, batches of 8 were 90% workable or so, it stayed consistent even with longer prompts and inpainting. Then I messed around with different Loras and models, different prompts, and one hour later when I went back to trying that Lora it was as of it was broken

#

A no prompt batch of 8 gave me botched faces, full body shots, weird positions, a wall a cat and a bus

trim magnet
wary belfry
tough bough
#

I think the current system where training isn’t subject to copyright restrictions but the outputs are also uncopyrightable is quite good. It limits the ability of AI to compete using scraped data, but it ensures easy access to it as a tool.

undone garden
#

If I forced it through prompts I started getting decent portraits but nowhere near the quality

spice plank
wary belfry
trim magnet
undone garden
spice plank
#

If your taking about the court case that happened a day ago or so, they rules that the AI itself cant copyright it.

tough bough
wary belfry
#

because you din't make a " great picture of a landscape", you just gave it into production by a programm. You didn't make it

trim magnet
#

if i print it and then paint on top of it while recording the whole process i guess you could win in court

spice plank
wary belfry
untold vigil
#

appreciate the back and forth, and if the letter ain't right for you, nbd. I'll just add there are tons of precedents of grassroots efforts making a difference in internet policy - e.g. net neutrality, and stopping harmful copyright legislation like SOPA.

tough bough
potent spire
tough bough
#

But why would you want to?

potent spire
#

i just saw, Artstation finally at least made the filter available to exclude AI generated stuff from their marketplace

trim magnet
spice plank
#

Wonder if a lora is copyrightable, i realize the licenses and such might get in the way but its technically your own creation, you chose how to train it.

potent spire
wary belfry
tough bough
spice plank
#

Maybe if it its way over trained to only create 1 image? 😆

trim magnet
outer crescent
#

Well the issue with things like suddenly making copyright/training regulations after the major companies have done it is making fees and various hoops to jump through then becomes anti competitive for the more open source AI models, for big companies they could even campaign hard for draconian laws because anything like that would just be an overhead to them like an electric bill, but also a handy barrier to keep out anyone else.

tough bough
tough bough
outer crescent
#

I thought by now we'd have celeb likeness's being officially licensed in loras and the like

tough bough
#

We do not want AI to be monopolized by only the existing IP holders. That is a great way to get all the downsides of the technology and none of the potential upsides.

outer crescent
# tough bough Exactly this.

exhibit a for me of this kind of "big company does good thing for really bad reason" is amazon campaigning for an increase in a minimum wage. I'm all for an increase in minimum wages for example, but they are because 1) thats pennies to them 2) thats a whole lot of money to mom and pop stores.

potent spire
potent spire
trim magnet
outer crescent
#

This is the thing, sometimes a good can be done because it consolidates someone elses grip on something.

tough bough
# potent spire in what way exactly?

What do you see as the downsides of the use of AI? I almost guarantee that a few corporations monopolizing them do not remove most of them. It certainly constrains other people’s ability to benefit.

outer crescent
# trim magnet even if u guys pass that law,it will probably benefit our ai companies in asia,s...

and thats another issue. its why you have people against reducing emissions to the point of denying its an issue. its fine to say in one half of the world "nah this is actually a bit bad, luckily we're in a good place now." but lol if you expect other nations to just lag behind you because YOU decided to add in regulations or say industrialization is a terrible idea. you just cant expect other countries to give up advantages.

trim magnet
#

yea while u kill each other over dumb things there will always be other ppl ready to pick up the scraps

outer crescent
#

and think about political campaigns. theres like 15 billion with a B put into advertising with the media by politicians. if you can essentially bid for the AI's favour, everyone else is screwed.

potent spire
tough bough
#

For example, artists losing their jobs. That will still happen because the companies will no longer have to hire artists, nor anyone big enough to license the models. Even independent artists lose under a glut of mass produced work. If the AI is relatively freely available, anyone can add it to a workflow and compete if that is what is necessary.

tough bough
potent spire
#

artists will only lose their jobs when AI becomes superintelligent

#

above human level

#

anc conscious

#

and

tough bough
#

I agree, but some will lose them before that. It is a good example for illustration, but if you have other issues, state them.

potent spire
#

you cant make AI do certain things that all the elite level artists can do unless that AI gets to human level and above

outer crescent
#

thats literally one of the biggest issues the writers strike is over.

potent spire
#

@tough bough one of my issues with generative AI is more like what people do and are potentially or actually allowed to do with it. Generative AI as technology isnt something i want to get rid off

tough bough
potent spire
#

for example people spamming generative AI on artists platforms

#

yes, i dont consider AI art people as artists

#

and never will unless certain circumstances

outer crescent
# tough bough Not really. It is over streaming residuals. AI just got tacked on at the end.

I said one of. 🙂 But part of the residuals hollywood accounting issue is breaking a job for 1 fulltime person into 10 day gigs with no major contribution. if you can replace even 1/3 of that with AI thats a lot of real jobs gone and turns artists into zero hour contract workers hanging around outside supermarkets for someone to pick them up in a van in the mornings for work. 🙂

tough bough
potent spire
#

but yes

#

it took Epic Games long enough to filter AI stuff on their platform

#

what i think should be done by the policy makers is to let people have to pay licenses for LLM for example

#

just like Sam Altman proposed

turbid whale
#

Pay 2 Win

tough bough
#

It would have to be a totalitarian state, more likely.

Anyway, you would be as likely to see that under a monopolistic outcome as not. If the platform decides that they like their AI, it will be all you see with no recourse for seeing anything else.

potent spire
#

i would rather have a monopoly by certain companies/corporations at this point

tough bough
#

That falls foul of original position fallacy. You are just as likely to get an authoritarian state that disagrees with you and suppresses your opinions as one that agrees and elevates them.

trim magnet
#

does chargin ppl for software stopped software piracy?

potent spire
#

reality is more complex ofc

outer crescent
potent spire
outer crescent
#

Netflix Technical Director, AI/ML - R&D Tech Lab <--- look this up lol. Hell, apply if it floats your boat. But saying that AI is just something thats tacked on and not an existential threat is silly.

crimson belfry
#

@wild steppe Wow! I totally overlooked this! The main thing that makes me sad about ChatGpt and Claude is the censorship, I can not even make a cheetah hunting a gazelle because it is deemed as harmful.

outer crescent
#

it'll probably makes games better and movies worse for a long time. like cgi initially 🙂

potent spire
#

generative AI?

crimson belfry
#

As soon as I am home I will be trying stability chat, I would love to test out how my prompts run on there.

#

I really feel like many underestimate the power of the chat bot and its ability to be an extension of your brain personally

#

Each person working with and training their bot will have it specifically tuned for them and their needs.

outer crescent
crimson belfry
#

When the LLMs want to be humanized and you can not ask the AI to create a prompt based around a cheetah hunting a gazelle we have an issue with how the algorithms are coded. To even consider that a roadblock is not a very good look on the behalf of the company itself. @outer crescent I understand what you mean but this is logical for any ai chat that wants to have long term success.

#

If this is the case we are not being scaled up, We are being hindered and will be moving backwards or wherever the coders of this algorithm will allow you to go.

#

The success of AI coexistence with humanity will teeter on our ability to be free while using this technology.

pale latch
outer crescent
lean laurel
#

Man, crazy to think that Stable Foundation has only 1 year of life

fervent thunder
#

Hello, May I ask how do I optimize my 4090 on windows 10 to get as many it/s as possible?

#

been playing with it for over 1 week now and still getting 5 it/s

#

my laptop with 3060 mobile gets close to that

signal goblet
#

What are my options on amd windows if I have a ton of ram but not enough vram for comfyui controlnet

tulip thunder
#

hello, I need some help plz

#

is there a way to tell the AI to not touch a specific area in case of doing image to image ?

for example: not changing or touching the head even slightly, not changing it at all, but the rest is ok to be changed

or not changing/touching the sword

like fixing a specfic area of the image

#

you might call it "reverse inpating"; changing everything except a specifc area

broken smelt
#

ya thats still inpainting

tulip thunder
#

wait, really ?

broken smelt
#

oh wait i misunderstood

#

you could probably crop your image and then outpaint

#

outpainting is exactly what you described

tulip thunder
#

hmm, I don't want the image to drastically change

#

in case of image to image, I want to fix/lock (making it not change) a specifc area, but the rest can be changed slightly according to how much I set "image strength" (%)

soft notch
#

yo guys, i am new to this stuff, while installing that web-ui.bat file, my pc crashes. Is it a problem with my files or is my pc weak to handle it. (there are other issues sometimes as well where it says git bash could not be foudn etc.)( PC specs, i7 3770k, 12 gbs of ram, gtx 1650 4 gb)

fervent thunder
#

lmao can we start a petition to bring stable diffusion to geforce now

cobalt crag
#

Do I need to be expert at english in order to type some prompt for my checkpoint?

#

I am not good at describe some people like policeman in usa usually do

soft notch
tough vine
#

does anyone know why when I want to download an image I get it in webb format and not in png format?

slim raptor
#

Ex: a man sitting on a bench the beach at noon could be written as:” 1man, sitting on bench, beach, noon,”

slim raptor
tough vine
neat elm
#

anyone know what's the difference control-Lora and controlnet sdxl 1.0?

warm junco
warm junco
fervent thunder
#

Hi, I want to buy large amounts of credit, it only allows for $1000 increments. That's a no-go. Who can I speak to about this?

#

hi is there any way to generate img2img in this discord bot

ivory grove
novel dew
#

(I hope this is the appropriate channel, there's a lot) Hey everyone, I've gotten to the point where it's about time to upgrade my PC again, and I'm wanting to take SD into consideration for my parts. Specifically for the GPUs I'm wondering a bit about what would be sensible. 3090s (24GB) are going for 650-700 euros used in my area right now, looking around a bit that seems like a pretty good bang for buck and I'm not that scared of used cards. Any opinions on that?

languid helm
#

hello everyone

#

is rtx 4090 good for stable diffusion ?

#

i have budget of buying rtx 4090 but is there any other good choice ?

#

i am going to install this gpu in my server and primary intention is deep learning and no gaming at all

knotty turtle
#

AUTOMATIC1111 1.6.0-RC just released
Has a crap-ton of new features.
Anybody tried it yet?

signal wagon
languid helm
#

then wouldnt 2 3060 be enough ?

cobalt tulip
#

Hi guys how are you all?

#

I have question

#

why my generated picture is not clear? like if i create picture it will come up uncomplete like his/her hands are uncomplete facelook uncomplete or strange, i don't why its happend. anyone knows?

ornate dock
#

hi guys, is it possible to generate seamless images through the bot?

plucky shoal
#

@cobalt tulip cant it be missing VAE?

cobalt tulip
plucky shoal
#

check what VAE is, or post your picture, what issue it has.

polar totem
#

What is the best model at the moment to generate UI / UX designs?

true mauve
#

what is the best img2img model for colorizing pictures?

slim raptor
#

it saves everything you generate

true mauve
#

guys why is sd 1.5 inpainting not giving nsfw?

slim raptor
true mauve
#

I mean I had an sfw image and wanted outpaint the clothes but it just puts another set on clothes

slim raptor
#

what is your model name and your prompt

#

some model is censored

true mauve
#

Epic diffusion,anything diffusion,stable diffusion all inpainting

#

Used stable horde

slim raptor
#

if all of those can do nsfw, just type in naked or something

true mauve
#

my prompt:Naked,human body,no clothes

slim raptor
#

try (naked:1.5)

true mauve
#

Cant I am on horde

#

Maybe using a lora?

slim raptor
#

dude, take a look at civitai, most of them are just nsfw

true mauve
#

ye

slim raptor
true mauve
#

let me

slim raptor
#

done yet?

#

installing

true mauve
#

a1111 is not even worth using now

#

COmfy is just better

slim raptor
#

how so, how is a1111 bad

true mauve
#

automatic 1111 is so bad man

#

after the sdxl integration update they broke it

#

Welcome to the world of Generative AI

#

Use colab lamo

slim raptor
#

my pc cant do SDXL fast enough, 30 min for a 1024x768

true mauve
#

Have fun

slim raptor
#

i7 6820hq, 16gb ram, quadro m3000m 4gb vram, very old by pc standard

true mauve
#

Use google colab

#

they offer a 16gb gpu for free

#

scroll down

#

and click the "open in colab button"

slim raptor
#

does google colab even allow nsfw

#

you just need 4gb vram to run locally, what gpu do you have

true mauve
#

Basically the processing would be done in google's supercomputers

#

also make sure once you are in the webui

#

I dont reccomend comfy to beginners

true mauve
#

click the play icon on

slim raptor
#

looks like it have 8gb vram?

true mauve
#

Bro why is sdxl so hesitant to generating nsfw!

slim raptor
#

amd naming scheme is confusing

#

my bad

vale steeple
#

alright back

vale steeple
steel pebble
#

Ive been able to install a standalone version of SDXL in the cloud with the text to video extension on a dedicated cloud GPU (a6000 in my case). ive also been able to integrate this with twitter's api and automate a process. if anyone is trying to do anything similar, reach out and i can help.

slim raptor
#

post your result here man

true mauve
#

I have seen stability.ai took a page out of midjourney's book
and made sdxl less realistic and more dependent on cgi
They really want to beat midjourney

ivory iris
#

Hello everyone. Could anyone please tell me if there's any way to upload loads of pictures of me, make a model of me, and prompt the model so I can generate ultrarealistic pictures of me? If so, please tell me how.

slim raptor
true mauve
#

runwayml offers 1 free plan

#

to make a model on yourself

ivory iris
#

What's the best way to access Stable Diffusion that allows you to use a lora online?

true mauve
ivory iris
bleak matrix
#

Good mornin', everyone! How are we all today?

potent spire
#

How about you?

bleak matrix
#

Haha, but doing better

potent spire
#

I didnt see the sting coming

#

I thought a vene is "dead" lol

bleak matrix
#

Wasps are not on my friends lists

potent spire
#

My colleague nearly died here because of one

#

He was allergic

bleak matrix
#

R i p allergies

potent spire
#

Sting right on throat

#

He was suffocating

bleak matrix
#

I don't think I am--results pending

#

But oof, epipens needed

undone garden
#

Is there a limit to how many images you should use to train loras? Because I'm about to go all in on this mf with thousands of faces autocropped from various videos

torpid crest
#

Hello, does anybody know the steps to get automatic1111 to work with a self-signed certificate, and then be available through a No-IP dynamic dns domain and have SSL?
I need to know how to properly generate the certificate, how to configure the COMMANDLINE_ARGS for it to work.
I have attempted it a couple of times, I get certificate errors, automatic1111 straight up crashes, it says that my dns name does not match "localhost", etc.

queen yarrow
#

it is normal that generated images can only be downloaded in webp ?

torpid crest
#

Oh and you're running automatic1111 right?

queen yarrow
#

okok i'm going to automatic1111 i'll come back if have more questions, thanks for help !

crimson belfry
#

hello everyone

crimson belfry
tulip thunder
torpid crest
rotund wharf
#

is there a way to point towards a directory for my models

#

i have them all saved in a different location

#

if so please someone point me in the right direction

warm junco
#

For the webui-user.bat

lyric dock
#

any great tools to get started with stable diffusion you all recommend, either generating or training,

solar swan
#

So am watching those corridor crew videos on youtube where they make the animes using SD and a bunch of other tools, they keep mentioning "reverse noise" or something, describing it as the noise not being generated by a seed but instead being generated by the input image used for img2img. Can someone describe what that is exactly or how to use it?

lyric dock
#

face palm now i see the tools channeL:)

orchid mason
#

Does anyone have the upscale 4x-ultramix balanced?

tulip thunder
torpid crest
tulip thunder
#

is there a way to perfectly swap faces of same styles ?

like swapping human to human, or 3d to 3d, 2d to 2d etc

vague rampart
#

Just want to share a thought. When I visit civitai or huggingface and think about downloading one of the new sdxl models, I constantly think to myself "how do I know SDXL base can't do that already?". It's a good problem to have..but hard to tell without A/B comparisons.

pine ice
buoyant forge
#

Quick question. I have SDXL model in automatic1111 based on the github model provided. As well as I used the sample on clipart. It seem like clipart produces much better quality results? What could be the explenation for the difference in quality? Is the model I am running locally not working the same as the model on clipa

ivory iris
#

In Google Collab which GPU accelerator is meant to be used for image generation? TPU or T4?

civic grotto
#

Does anyone know of an existing workflow to apply hi-res fix to a number of images after the fact? I'm currently passing them through the PNG info tab in auto1111, sending the parameters / prompt to txt2img. I have comfyui but still learning.

#

trying to figure out a batch solution

clever shore
#

Is stable diffusion censored?

hollow robin
#

you have to be specific with your prompts if you want to add or avoid nsfw images

clever shore
#

It's not creating a few images ,like a car crash

#

Maybe I have to work on prompt writing

hollow robin
#

that could just be your model or your prompts. Not every model recognizes every prompt

clever shore
#

Yeah, thanks

hollow robin
#

ive seen some models have an excel spreadsheet with the prompts it recognizes, and theres a tag autocomplete extension that helps with that

clever shore
#

And what is seed?

hollow robin
#

ever played minecraft?

clever shore
#

A Lil bit

hollow robin
#

think of it like the seed of the world

clever shore
#

Idk abt that sry,

hollow robin
#

if everything including the seed stays the same, the image will generate the exact same. if you have "-1" as the seed its random within the bounds of the prompt

clever shore
#

Ohk

true mauve
true mauve
hollow robin
#

you can also keep the same seed and change the prompts to adjust the image, as long as they stay mostly the same

clever shore
#

So the more seeds , the more random img from prompt will generate

hollow robin
#

each image has its own seed

#

its either randomly generated when you generate the image or you pick a specific one(which i wouldnt recommend unless youre redoing something)

true mauve
#

For randomness try playing with the guidance scale

hollow robin
#

each image only has one seed, unique to that specific image and those prompts and model

clever shore
#

Got it guys, thankyou

true mauve
hollow robin
#

you dont really need to worry about it, id say just leave it at "-1" and let it work

clever shore
#

Ok

hollow robin
#

I'm not sure if this is a good place to ask, but does anybody know some sort of ai voice over program? Like where I enter words and it'll read it back to me.

#

ive tried some basic programs online but they tend to be pretty bad in my experience

true mauve
#

You don't really need to write long prompts

#

Unless you want absolute control over your images

#

@clever shore

clever shore
#

Yeah, I tried a few prompts with 3-4 words and it works pretty well

true mauve
#

Make sure you have 50 steps

#

And the resolution is 1024x1024

#

And don't use a1111

clever shore
#

Ok

#

I'll check it out , thanks dude

true mauve
#

Python is general language

#

Learning that would make you a ml engineer

#

But libraries like pytorch and tensorflow will

#

Numpy,panda

#

C

#

For basics

clever shore
#

I'll delete this conversation because it is not related to this group

true mauve
#

Java,cpp in my opinon are not good

tough vine
#

Does anyone know why when I generate an image in discord and when I go to download it comes out in webp format?

buoyant forge
#

Quick question. I have SDXL model in automatic1111 based on the github model provided. As well as I used the sample on clipart. It seem like clipart produces much better quality results? What could be the explenation for the difference in quality? Is the model I am running locally not working the same as the model on clipa or their playground in terms of quality.

neat elm
#

hi does anyone know how to load the civit ai lora model in python code?

#

for example i want to load this
https://civitai.com/models/112902/dreamshaper-xl10

i tried
pipe.load_lora_weights("./dreamshaperXL10_alpha2Xl10.safetensors")

but it returns
ValueError: None does not seem to be in the correct format expected by LoRA or Custom Diffusion training.

worldly lark
#

I am trying to find my dream post forever still cannot fjnd it

scenic shadow
#

Good morning
I am looking for a French community on Stable diffusion to answer some of the questions I have about using the software.
My level of English is not top, it would be easier for me to exchange in French.
Do you know of such a community?
Where does it exist on this discord?
Thanking you

true mauve
scenic shadow
#

Une communauté au sens large.
Celle-ci m'irait très bien si je peux échanger en français de temps en temps.

true mauve
scenic shadow
#

ah comme moi avec les cours d'anglais ^^

true mauve
scenic shadow
#

ok merci

true mauve
#

its a pretty quick read

#

The Clipdrop website does not provide information on whether purchasing a Clipdrop Pro subscription gives users commercial rights to the art they generate during their subscription. However, Clipdrop empowers teams of all scales to create better visual content for a fraction of the time and cost
The Pro subscription includes features such as Stable Diffusion XL, which generates high-resolution realistic images with AI, and Uncrop, which allows users to uncrop photos to any image format
The Clipdrop API also allows users to integrate best-in-class AI to their apps in minute
While Clipdrop is an AI-powered collection of tools for image editing and generation, it is unclear whether users can sell or auction the art they generate using Clipdrop Pro

#

But according to US

#

If you just modify the picture even with 1 pixel its yours to keep

#

also you can always run local for free......

#

ahem lets assume you dont have a powerful pc

#

then it wont work

#

but google offers you a 16gb vram gpu for free

#

everyday for 12 hours

west echo
true mauve
#

for traning LlaMa-2 meta already spent 6 millons

#

Openai got pumped $10billon

#

Google has plans to pump $100 billon into generative Ai

trail lion
true mauve
west echo
# true mauve $235M is peanuts

however Hugging Face is a Repository and isn't training its own models so comparisons with money spent on Dev isnt an apples vs apples comparison 🙂

true mauve
trail lion
#

Ok but I wouldn't risk my Google account over that amount of money

true mauve
#

you can always make another

#

besides runpod,paperspace,kaggle

#

besides they only ban things like detected nsfw images,nsfw chat models like pygmalion

#

because if they were to ban fucking everything

#

than colab would be point less

bleak matrix
#

Good morning, everyone! I hope you're all doing well!

true mauve
#

Good morning? man its night in asia

undone garden
#

When I'm training a Lora, I'm keeping the portraits at 512x512 but what about the full body shots? Should I resize them to something like 300x512?

#

Does it make the difference by itself as empty space or will it go apeshit

fierce pebble
#

I picked up one of those Coral USB TPU accelerators

is there any chance of running stable diffusion through it?

haughty vapor
#

hey guys, can you use temporalkit/warpfusion in comfyUI?

lunar steppe
#

Hello everyone, I can't get it worked

dapper nest
#

hello guys, I have one question. How can I integrate a Space of huggingface on my website?

tribal inlet
#

hey guys, i'm trying to use control net to upscale a image via tiles and i'm getting the "CUDA out of memory. Tried to allocate 42.00 MiB (GPU 0; 4.00 GiB total capacity; 2.48 GiB already allocated; 0 bytes free; 2.56 GiB reserved in total by PyTorch)" message even with the Low VRAM option ennabled.
I'm running in a laptop with NVIDIA GeForce 1650 with Max-Q Design.
My command line args are the following: --lowvram --xformers --no-half --disable-nan-check --upcast-sampling.

Am i missing something or my system really can't run this upscalling?

#

(i'm using the automatic1111 webui)

wary belfry
#

it has to load the control net model in addition to everything else into memory. Seems like 4GB is really too low for that. Can't you do it in two seperate steps?

tribal inlet
#

maybe google colab is a good option then?

ivory iris
#

Is it possible to control the mouth gestures of a generated image?

wary belfry
ivory iris
wary belfry
ivory iris
ivory iris
fervent thunder
#

hello

#

Anyone in here that can hop in vc to help me out real q

#

I am getting images do not match error but the images are 1 to 1 the same size?

wary belfry
fervent thunder
#

pls?

worldly fulcrum
#

Hello I would like to know if you have any advice for me to optimize this prompt
A captive galaxy floats within a sphere of red and silver-tinted glass, engulfed by an enigmatic blue mist. In the background, an apocalyptic cyberpunk world takes shape, with dragon silhouettes as well as wandering zombies adding a captivating touch. This fusion of a galaxy and post-apocalyptic horror is captured with hyper-realistic precision.

mild condor
#

which best version of python is recommended?

slim raptor
slim raptor
worldly fulcrum
slim raptor
#

SD doesn’t do compositions very well

#

just paint in where you want your object to be

#

first you make a sphere then inpaint a galaxy then some blue mist then the other stuff

worldly fulcrum
ivory iris
#

Is there any way to generate a picture of a model biting their lip? It seems to be a complex task for AI similar to what generating realistic hands used to be

serene raven
trim magnet
wispy jasper
#

Anyone have an animatediff workflow to share for comfy?

trail lion
ivory iris
trim magnet
#

something like portrait,face only,bitting own lip.

ivory iris
#

Yes, it's like a selfie so you can see the face, neck, shoulders and upper torso

trim magnet
#

upper body works too and u could add full body to negative

ivory iris
#

Could you please explain step by step? I've only been using Stable Diffusion for 2 days

trim magnet
ivory iris
#

Thanks, done.

#

Make a LoRa of the character

trim magnet
ivory iris
#

At least 20

#

You ideally want different angles and clothes of the same character

ivory iris
#

What version of Stable Diffusion are you looking to use?

jaunty sierra
#

Hi.
Is there a prompt to tell to the ai to think outside of the box ? To have some creativity and add stuff we did not necessarily think about ?

#

Please :3

true mauve
dusty ember
#

is it possible to do img2img with sdxl and comfyui?

vivid dome
#

how do I switch my pipeline in VLAD1111 from backend to the original pipeline again? I've been trying to get VLAD working again but I can't for the life of me get things to work since I tried messing around with SDXL

vestal dew
#

I thought it was in settings, select ORIGINAL

#

However, Im having an issue with VLAD that all the menus are doubled. For example in settings, it shows Original, Original and Diffusers, Diffusers

#

When I try inpaint, it gives an error

#

others claim it isn't happening to them, but I've tried fresh install and even install on another computer. Vlad right now is not working

cobalt crag
#

For stable diffusion, can I give a color prompy with HEX Code Color?

#

like...

#

#204d52 hair color, 1man, male

lilac rune
true mauve
lilac rune
#

This is a discord robot of sd-webui, which supports roop, controlnet and other plug-ins.

#

Support multiple sd-webui node deployment

true mauve
#

so it uses a cluster of gpus?