#💬｜general-chat

1 messages · Page 52 of 1

astral goblet
#

New ui is a bit buggy though. i don't like it doesn't view images like before. clicking them goes to next now

#

send to tab doesn't work

fervent thunder
#

I like the new version myself

astral goblet
#

few glitches. i kept a copy of the old incase they annoy me too much

fervent thunder
#

i like having the style filters

#

if it had like face upgrades id be set lol

proud nova
astral goblet
#

i think the style capabilities are going to be very utilized by refiners

astral goblet
fervent thunder
#

surprised its 512 but maybe later versions will be bigger

proud nova
astral goblet
#

still training. we might just be testing a base 512 for a 768 refinement

proud nova
#

But I need to rewrite it each start

fervent thunder
#

neon punk is a cool filter

astral goblet
#

classified info incoming ||i showed a girl i'm potentially going to date, this stuff. made a few remixes of some selfies. dudes. it's a total in. i'm not going to show it cause i'm big into consent and discord not getting universal permissions for photos i value, but, turning her into a plastic barbie, a disney cartoon, and just a few remixes, totally hooked her in. Bro tips! ||

#

its like poetry level bait, but i can do it instead

warm junco
astral goblet
#

the novelty of it might wear off in a few months so strike while it's hot

#

various workflows always.. usually controlnet for pic remixes. sometimes i use the image to intiate sometimes i just cnet with t2i

#

i remix lots of friend's pics to great degree, but i had very positive reactions from a cute girl in our "getting to know each other" phase

warm junco
#

nice 😄

astral goblet
#

i need to do some serious SD work though. My dwarf fortress goals. Here i go looking for rock photos again

fervent thunder
#

thinking wake me up when we'll be able to do 3d models and character animations

astral goblet
#

you can right now!!! *with work

fervent thunder
#

thinking with AI

astral goblet
#

i dont wanna work i just wanna bang on ai all day

fervent thunder
#

it's not just "I don't wanna" - it's more of - I can't draw anything, not even saying about 3d.
It'll take me more time to learn this skill then for AI to start generating it lol

#

I'll learn how to fix issues on gens, but that's about it

astral goblet
#

yeah me too. i'm a great technical artist but if i don't have a very rigid technical creative process, i cant visualize an end product. i can see memories, but my imagination lacks sight. head got knocked around as a kid and i've got partial "aphantasia" the inability to see images in a mind.

#

"just practice" i've heard but iv'e got piles of sketchbooks and the visualizing process is still there. my technique is alot of muscle memory and style development. free form instead of directed

#

sucks for drawing next frames of animation because i'd have to see it before i got there, if i'm not tracing that's impossible for me

#

it's not ADD but it's certainly an attention problem

fervent thunder
#

I can visualise, I just can't draw.
I practiced a bit on tablet and got few interesting images, but it's just needs aloooot of time to both

  • learn the skill
  • draw each thing
#

additional details is like whole another thing too, might take 2x time then initial drawing

astral goblet
#

yeah. we each have our own struggles. a lot of people insist "just practice" but SD ai tools in general, really bring new opportunity to us all

#

this stuff is neat for me especially because my imagination is often just, story telling and descriptions in my head.

#

i can see what i come up with now

fervent thunder
#

i wonder what diffuser xl uses

#

doesnt let you select any

astral goblet
#

i sometimes admit that i hear voices in my head lol. in the right contexts. it's like a chorus of fast talking but they're all my voice so it's not mental

#

vanilla built to purpose at this point in time likely

random elk
#

Are there any trainings on scripts to generate writing that looks good?

fervent thunder
#

even if it's not normal thing, I didn't think it was bad, it helped me look at things from different perspectives

astral goblet
#

oh yeah! there are definately arguments haha

#

but its always me

fervent thunder
#

well, brain \ human mind is a weird thing

random elk
#

is there a model for game tilemaps? the tiling option doesn't seem to help as much as I hoped.

fervent thunder
#

idk if there's anything better

astral goblet
astral goblet
random elk
#

I can't find any models that have been shared. I've found some awesome models that haven't been shared, but none that have.

lunar quail
#

Hey folks, not sure where to start with this. I'm trying to get something pretty specific with the generator and I don't know how to get it or how to find a model for it. I'm looking to generate some ink blot style images, don't know if links are allowed but the art style I'd like to be similar to is Kim Diaz Holm on YouTube. How can I go about finding out a model that has been trained on anything similar?

past edge
#

Aaaaaand the XL 2.2 model turned out to be garbitch. What a suprise amirite? Totally unpredictable.

#

why do they even waste their time making these if they know its gonna be shit

fervent thunder
#

Give your mom back her phone

past edge
#

actually I take it back. They un-nerfed celebs and boobies. That's more than we can ask for.

grave moss
#

I use chat gpt for coding and just text. Does Stability ai have something similar I can run on my machine or is it only image based products?

Thanks

warm junco
alpine bear
grave moss
grave moss
near sluice
#

Hi Frens, What programs/processes/workflows are you liking for 1) AI face replacing into portraits and 2) editing portraits?

For example, 1) have a fashion photo and want to put my face into it with AI, and 2) make a photo of me look like a certain genre like vintage Hollywood glamour.. doesn't have to be the same program/app.

#1 is the priority.

I'm a mid-advanced used and generally work in Stable Diffusion, Midjourney, and Photoshop. Happy to use those and try new things.

digital matrix
#

Hi all, can you tell me please , which Prompts to apply to make Stable Diffusion change the interior? Not drastically, but like a new renovation and new furniture?

alpine bear
digital matrix
candid star
#

anyone know of any good models for stickers? MJ makes insanely good stickers but I'm yet to find a good model based off SD

inner iron
grim frigate
#

Is there an alternative for stable diffusion that I can use purely online? Right now I'm using automatic1111 web ui, some custom models, a LoRa and multi controlnet but my gpu is such a noob and I'm can't even generate an image and keeps saying 'Cuda out of memory'. I'm already using --medvram and low vram option in controlnet but still no luck.

So back to my question, is there an online stable diffusion that let me use my model of choosing, a lora and controlnet? Even if its paid its ok.

tight dust
#

does anybody here have elevenlabs clone stuff

alpine bear
bleak wolf
#

Have been out of the scene for a bit, coming back but still many things I haven't caught up to especially on the training side.

I have heard LoRA is very good now, however my dataset is 27K~ images, with possibly even more to come soon.

Original Sci-Fi Diffusion v1.0 was a finetuned version of SD 1.5 trained on my dataset for 2 epochs.

Since I am going to base v2.0 off of SD v2.0 i need to consider more stuff.

What would be the best approach to this? LoRA? Finetune? Something else?

alpine bear
grim frigate
primal marten
#

Hi! Silly question but what's the difference between using parentheses like ((this)) vs using them like (this:1.5)?

#

I thought both increase the weight? Is it the exact same if I use ((object)) and if I use ((object:1.1))?

#

Thanks in advance

sage shale
#

Using a numerical value is usually used as you can be more precise than just parentheses, you can go into even smaller decimals if you want to

broken smelt
#

the real difference is that you can highlight the word and ctrl+upArrow or downArrow in automatic1111 and it will go up and down .1 in weight starting from (:1)

primal marten
#

Ohh that makes sense. Thanks guys!

fervent thunder
#

Hi!

#

Is there any way I can use SD on Android?

nova lily
# fervent thunder Hi!

Well If possible, I imagine it would be similar to running it off another persons computer in the browser

#

Or what happens if you just open the local url on a device?

vast ingot
#

you can yes, but like CaL said, through another computer

#

2 solutions :

#

1 : run on your computer using the "--listen" option : that will share it on your home WIFI/LAN
2 : run in the cloud in google colab : it opens it to the full web

#

you can also use that --share option from the local install and open to internet too

#

but if using --share, don't forget to set up a password/login to prevent anyone from connecting to it

wary condor
#

where can I share AI art of TV show characters?

vast ingot
cerulean juniper
#

Hi everyone

vast ingot
cerulean juniper
vast ingot
#

new to image generation ? what made you find us ?

#

nice little place around here for sharing AI art and tricks

cerulean juniper
#

I'm no longer a beginner, I accidentally stumbled upon you on the Internet and I became interested

cerulean juniper
vast ingot
#

sorry I got to step out for a sec

vast ingot
cerulean juniper
#

thank you

obsidian meadow
#

Hi, My Automatic1111 got an update a few min. ago, but it's not working ok, does anyone else got the same problem, models en HyperNetwork not loading on top? Anyone?

whole owl
#

Hello there!

vast ingot
#

hello and welcome around

whole owl
#

How's your day going?

vast ingot
#

quite good, preparing for a meeting, what about you ?

#

what brings you to this crowd ? You're already making AI pictures, or about to open the door to that world ?

whole owl
whole owl
vast ingot
astral widget
#

Unfortunately, I could not find it in this Discord and also in an internet search. Maybe you can help me: Based on an uploaded photo, I would like to insert objects into it in a correct perspective, without changing the original image. In a further step I would like to isolate them. I would also like to render backgrounds that match the given image in perspective exactly. What is important in all this is: Somehow I would like to get stable diffusion to recognise and adapt to the perspective. I would also be super grateful for tips on sources of the kind.

whole owl
vast ingot
#

multiple example uses, but towards the end, they compose multiple pictures together

obsidian meadow
#

Latest automatic1111 messing up the webui

vast ingot
#

yep

#

lots of complaints since 2 days ago

fervent thunder
astral widget
inner iron
#

maybe with an own modell especially for this one, an mask from photoshop and comfui, but im a beginner and hae the same problem XD

winged gull
#

hey there

#

i am trying to install a locon/lora lycoris add on

#

but i keep getting errors

#

i can generate images ok using non lycoris loras

#

my installed extensions are :
LyCORIS
a1111-sd-webui-locon
sd-webui-additional-networks
sd-webui-controlnet
any ideas?

astral widget
# inner iron maybe with an own modell especially for this one, an mask from photoshop and com...

I am also a beginner and am trying to find out if there are any methods or AIs that seriously simplify my filmmaking workflow. I work a lot with green screen and have always had to spend ages finding images of objects that match the perspective of my source material in order to incorporate them. It would be great to find methods to tell the model, for example, to place on the table in front of the speaker, for example, a cash register system and in the background other rows of supermarket tills in the correct perspective orientation, which I can then insert into the source material (video) as a still image using alpha channel or at least keying. These objects can then be interacted with and do not need to be interacted with. But it should not take as long as my usual procedure.

broken gull
vast ingot
#

SD (for shot) is the main brick

#

that video sure does shows a full wall of use of that brick

#

but it's SD

#

just as much as AUTOMATIC1111 is SD

broken gull
#

I highly recommend InvokeAI for in/out- painting sadly it doesn't support Controlnet yet

vast ingot
#

we say "stable diffusion" and think it's only 1 thing at first, because coming from DallE or MJ, there is no distinction there

#

but Stable diffusion, only the brick, is hard as hell to use correctly, and has a lot of limits

#

people have built upon it quite a lot

broken gull
#

Stable Diffusion is the name of the main model no? So it is only 1 thing.

vast ingot
#

and different tools bring different possibilies

vast ingot
broken gull
#

Aha gotcha

#

https://apps.apple.com/us/app/puppetry/id1671248086

Also since you're in films I think this will be useful, haven't tried it myself but from the 1 result it does a good job at animating faces. Although it ain't SD though, it's a separate app.@astral widget

#

And IOS only

fervent thunder
astral widget
#

OK. This really looks very very complicated, time-consuming and like it requires a lot of prior knowledge. Don't get me wrong: I'm not looking for the magic all-in-one-click solution (it will probably take a while until AI is ready), I would just like to find out if there are any like-minded people who are interested in similar use cases. And to assess whether it is worthwhile, measured in terms of the result, to familiarise oneself with it - always linked to the variable of time and time savings.

broken gull
fervent thunder
#

DON't UPDATE a1111 today
Update breaks model choice and additional quick settings if you have them.

fervent thunder
#

When I last used it - they didn't have it and it made invoke extremely annoying to work with

broken gull
fervent thunder
#

yea...img gallery isn't same thing...

#

idk, for me it's not comfortable at all, maybe you can get used to it, but nah, I'd better go on painthua or even use outpaint extention , both more comfortable to use

#

just cause of a single button =.=

#

When I work with canvas - I want to be able to manipulate image I'm looking at - on canvas.
Why do I have to go into gallery and find everything in gallery, when list itself on canvas too?
I just don't get it thinking

wary condor
#

im literally running out of space because of all the models I have downloaded

cerulean juniper
#

for some reason I have very little space left on the c drive

#

when I tried to install torch 2

vast ingot
cerulean juniper
#

yeahmonke

mellow meteor
warm junco
cerulean juniper
#

wow 10 gb clear

wooden hemlock
#

Hi how can I create ai here with writing text, I mean where is the address of channels

vast ingot
#

hey and welcome around

#

but right now, there is no bot doing that sorry

cerulean juniper
surreal perchBOT
#

@wooden hemlock

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

Currently, there is no bot on the server that generates images. However, there are plenty of other ways such as the official https://beta.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!

wooden hemlock
hoary river
#

Hey, how can I find a model from an image I saw? One of the images I downloaded has a hash [60b07ee7] and I would really like to find an download this model to play around with

vast ingot
#

this is not a model present on civitAI, so no idea, sorry

#

(civitAI lets you search by hash)

hoary river
#

Ok thanks! What is the latest SD version? Is it still 2.1?

proud nova
#

New webui is getting fixed real quick 🙂

vast ingot
#

it is ?

#

that's really good to hear !

#

lots of people wern't happy ^^

fervent thunder
#

new web ui?

vast ingot
#

lots of new webUIs going around, but I think they were talking about the last automatic1111 update that went for torch2, broke a lot of things, and is getting fixed by the minute

#

but if you're looking for new UIs, check out ComfyUI 😄

proud nova
#

Yes. Half of the stuff that was broken is now working correctly. Even my extensions work almost as with the older version

#

Only thing that still needs to be fixed is sending image from openpose to controlnet

hoary river
#

Oh thats whats going on. I updated my sd webui and it is pretty broken :/

proud nova
#

It still is

#

But hey, I think in maybe 2 - 3 days, it will be fixed

fervent thunder
#

I'm on nixos so I'll have to

#

Idk

#

Figure something out

#

Might use the same nix shell I use for a1111

hoary river
#

Was looking at invokeai. Is that a good ui to use?

proud nova
#

Does anyone has some experience with generating hand of the viewer holding something in the image? Like giving something to the character generated on the picture

hoary river
#

Yes

proud nova
#

Like POV. Y'know what I mean..

hoary river
#

Yeah its a bit tricky.

#

I mostly get it done by adding stuff like pov perspective and then make sure to add a bunch of negatives to remove the character from screen

#

You also need to make sure you aren't describing the character too much

#

Because if you add stuff like "female, solo, english dress" and so on it will think that is your subject

proud nova
#

I mean, I want to keep the character in the image. Just want to add arm coming "from behind" the picture giving something to it

hoary river
#

But its probably the hardest thing I've tried doing with SD to date so goodluck

proud nova
#

Yeah. Ig. Thx

hoary river
#

You can get lucky but its really hard to do pov out of the box. Try looking for models trained on the concept or maybe train an embedding for it

livid flicker
#

is there anyone else having issues with the Dreamstudio? My credits are gone and the API is not accessible

proud nova
fervent thunder
# proud nova Does anyone has some experience with generating hand of the viewer holding somet...

You can inpaint area until you'll get result you want.
https://www.painthua.com/ or invoke AI can help with that.

To expand on that idea - you can even just add something to the image manually, by hand and put it where you want, add it to prompt and let AI figure out how to make it work.
It doesn't always work good and will need lots of gens to actually make it look decent, but still.
Basic idea here is to inpaint like 20 times - see results, find closest to what you want - repeat.

copper shuttle
#

The first thing

fervent thunder
#

I was doing it before, it's kinda fun, but...takes ALOT of time

proud nova
fervent thunder
#

You'll need a tool which allows you to generate lots of inpainting results, basic a1111 can do that technically, but won't be as easy or comfortable to use

Edit:
and more importantly , you need inpainting version of same model you used to gen to keep same style.

spark phoenix
#

Hi guys, i have a big question. How can i create type and lettering with stable diffusion? i'm trying to do a bachelor thesis about lettering but stable diffusion has a problem with letters because he doesn't recognize it. So guys, does exists a tool that can create letters and typography?

fervent thunder
drowsy bear
#

how do i generate imgae???

surreal perchBOT
#

@drowsy bear

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

Currently, there is no bot on the server that generates images. However, there are plenty of other ways such as the official https://beta.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!

astral mulch
#

Is anyone having trouble logging in or seeing their api tokens in the legacy UI?

spark phoenix
fervent thunder
modern crater
#

Guys, what happen to civitia?
How long has it down?

fervent thunder
#

wow...

spark phoenix
polar crystal
#

Automatic1111 webui not properly working??? Any reason why this is happening

inner iron
inner iron
cedar garden
#

Ello

proud nova
#

But it's getting fixed

#

It's much better than yesterday already

#

So ig in 2 - 3 days, everything will be working correctly as before

limpid holly
#

Hello all, just joined is #1047197565365538826 down atm? Having issues getting it to display any results.

inner iron
#

civiai is down it seems, maybe a ddos or the website update a few days ago will cause the problems

#

just be patient, i think their on it at the moment

past edge
#

Protip: never update something that works until the update is confirmed to work by the chumps who updated early.

inner iron
#

in their discord they post following JustMaier — heute um 17:15 Uhr
@ here The site is currently down. Should be back shortly. Thanks for your patience

crimson magnet
#

Does anyone here have experience with Topaz Gigapixel or know if it's faster to upscale than img2img? I've been playing with deforum in auto1111 and upscaling with batch img2img on a 3060 TI. The results are fine but I'm curious if Gigapixel would be a faster route.

past edge
#

Some people here say that their models are outdated, but it is still my upscaler of choice

crimson magnet
#

Yeah for solo images I'm happy with auto1111 options, but waiting for ~1-2 hours on a 2x upscale of 120 frames is a bit painful lol. I'd imagine what I really want is their video software but $250 is more than I want to be spending so I figure batch upscale would be the best of both.

glacial siren
#

Is there a website that shows all the companies that are using Stable Diffusion to build their own model?

past edge
#

Topaz also isn't super fast. All I know is it doesn't take an eternity

crimson magnet
#

Fair enough. May just rent a beefy gpu on runpod or rundiffusion or something once I have something I know for sure I want to upscale. I'm still in the 'every other thing looks super cool and I totally need to see how it looks high res' phase of deforum so really I should be more selective lol.

silent gust
#

Hey, has anyone generated images of swords using any model? I cant seem to achieve what I want even though it sounds like it should be the simplest prompt ever

limpid holly
crystal sierra
#

wow is SD2.2 ever bad at colors.

limpid holly
crystal sierra
#

I can't get anything more than pastels out of it. I'm asking for bright and bold colors and it's just... bad

grim nimbus
#

hey, is Textual Inversion the best way to control which character is in a figure? I'm using stable diffusion to create an animation and need to select which character appears in a figure, and keep the character consistent

rare stratus
#

I wanted to ask you experts, I have a small brand of jams, I would like to use stable diffusion to not only create an image to put my jars in photos, but also to make images with different styles

vast ingot
vast ingot
grim nimbus
vast ingot
grim nimbus
astral goblet
fervent thunder
#

How can I log into dream studios it won't even let me log in

#

Did the site get updated

astral goblet
#

I saw the lora training thread yesterday an i saw allusions in there to using lora for video. I wonder if training lora on video clips of a character, with some clever captioning, may get that consistent effect people want in motion

slim raptor
#

since civitai is down i cant go read the description of anything v4.0, is it a model trained to make anime girl? i typed in land scape descriptions and it kept throwing anime girls at me

astral goblet
#

v4 wasn't made by the original author. It was made by a troll who didn't like that it was more popular than his merges, so he named his newest merge "anything v4"

slim raptor
astral goblet
charred pumice
#

Yep I am having trouble logging in myself

astral goblet
rare stratus
#

@vast ingot i try ti installa locali but my PC Is very old, how to male online?

astral goblet
slim raptor
astral goblet
foggy atlas
#

Hello. I’m facing a difficulty in training my model. I want a model that, in img2img, can give any portrait I input with a brown right eye and a blue left eye.

What type of training should I do? lora or dreambooth?

alpine bear
astral goblet
#

you don't need a model for that. just mask the eye and inpaint "brown eye"

#

"highly detailed brown eye" heh .. kind of sounds like a nasty prompt

foggy atlas
#

Thank you! Ive trained lora (around 2500it) but it didnt show affect.

#

The dreambooth model i have give ok result… but weirdly when i use small resolution img, it stopped working

astral goblet
#

i think theres even attention mask extensions. like , you tell the extension eye, and it'll automatically mask the eyes in a photo. then prompt wiht a lora for heterochromia (the name of two eye colors), and you can set it up to automate batches

foggy atlas
#

Ive never heard of that extension. That can help a lot! Thank you

astral goblet
#

i'm sure theres a tool that can be purposed to this task. AI can detect eye shapes i'm sure of it

foggy atlas
#

Off topic, but do you have any recommendations for cartoon/anime style model that doesnt mess up my img2img…. Basically like corridor crew without the training my own model part

astral goblet
#

lol no i don't . av3 maybe but yeah, not to into those styles atm. lots of gooders in #1047197565365538826 though

foggy atlas
#

Thank you. Ill have a look around

fervent thunder
#

no xl public release yet? lol

#

Bruh why can't I use dream studios

#

Where is the support

vast ingot
#

when 1.5

fervent thunder
vast ingot
#

and for DS, the team is on it currently though

fervent thunder
#

😢

vast ingot
#

it broke a leg it seems, sorry

fervent thunder
#

They updated to that new site a few weeks ago too

#

Then went back

vast ingot
#

it's the "ds next" version I believe, that new redesign ? the preview that is currently happening mostly in #1084896022368624640

#

I really like the new interface personally

fervent thunder
#

Will it make the same images ? Tho

vast ingot
#

it does yeah, there are the same main options

fervent thunder
#

Steps and stuff

#

Ok

#

Gfx

vast ingot
#

I do have access to both, some things have been strealined or been turned on by default/hidden, but it does work the same

#

some nice features on top, like styling or variations

fervent thunder
#

Cool

#

When do you think it will be back up live

vast ingot
#

honestly ? TwoDukes emoted on my message there #💬｜general-chat message
and I saw multiple developpers responding in the channels in the DreamStudio category, as well as the moderator channels

#

so I would say this seems like the current absolute priority

#

as to when, hard to say, I worked in IT for long and it's not always easy to tell

#

but i'm sure the news will flow soon at least

merry pendant
#

where would I go to give feedback on the SDXL model

astral goblet
vast ingot
#

this is urgent

astral goblet
#

😺

vast ingot
#

🐱

wild steppe
vast ingot
astral goblet
#

its the alternate start when you hold l+r and the cat's tail

vast ingot
fervent thunder
#

dream studio isnt working for me

vast ingot
#

it seems a little slow but all right for me right now

fervent thunder
#

yeah it kicked in for me now

fervent thunder
grim nimbus
#

hey is there anyway I can use code (python or bash command) to generate images?

vast ingot
vast ingot
#

if you use the official one, it works mainly in command line, and gives scripts showing how to do it in python too

#

but all other tools, like invokeAI or AUTOMATIC also provide access

#

automatic has 2 ways, python scripts directly, or API calls

#

invokeAI has a CLI that works quite well too

grim nimbus
#

thanks! I'm using webui (otherwise it's straightforwad) do you know where I can get some references on how to use python scripts or API calls to generate images from AUTOMATIC webui?

vast ingot
# grim nimbus thanks! I'm using webui (otherwise it's straightforwad) do you know where I can ...

some yes. most simply :
the scripts and extensions are your best bet here. there are lots of them already, some are pre installed on your UI too and you can add more in the "Extension" tab.
most of those are quite easy to read and copy/modify to understand how it works, and you can then expose some params of your script, and show those as fields in the UI. Quite nice https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Custom-Scripts

#

as for the API, you can add the "--api" flag to your launcher (the same way you add --xformers) and it will expose an API on localhost:yourport/docs

#

it's a rest API that gives all features of the UI, but that lacks documentation

obsidian tapir
#

model embeddings room doesn't work on me, how to solved that?

vast ingot
#

try restarting discord maybe ? it should be accessible to everyone, we don't have restrictions on it

hoary river
#

Is the ui fixed yet?

hushed tangle
#

Who cares about challenges

#

Give model

fervent thunder
#

yeah drop that model lol

#

smeagol voice we wants it its our birthday

upbeat tangle
#

hey guys, i cant select model

#

its normal?

slow chasm
#

hello everyone

#

there its someone here that can help me make some html request to some api ?

fervent thunder
#

I'm uploading a video of my experience

vast ingot
fervent thunder
#

Login attempt

vast ingot
#

I saw, I still can't really say more than those tips, cleaning cache, other browser, #1025467151206854736 , I have no more on that that could help you

fervent thunder
#

I think it works

#

Was buggin

#

How long will the legacy site be live for

upbeat tangle
#

@vast ingot hello i have problem with ui do you know anything?

vast ingot
#

I haven't updated in fear of breaking it recently

#

it was broken for so many people

#

right now I'm making my first pictures of the day at almost midnight after a day of moderating so I'll let the tech crew help you

fervent thunder
#

Bad authorisation string

#

How to fix

astral goblet
meager eagle
#

I want to make a story book, Just for fun, What model do you recommend for childrens books

#

I want a nice and consistant style

ancient bear
#

hey guys is there a way to automatically generate pictures? so i can just start it once and leave and when i come back i can look at the results in the output folder?

fervent thunder
#

you can right click on generate and select run forever or something like that

vast ingot
#

yep, exactly that

ancient bear
grim nimbus
warm jasper
#

Is there a way to do tiled inpaint, to be able to inpaint images larger than what the GPU can handle?

vast ingot
fervent thunder
warm jasper
#

I'm installing InvokeAI. This may take a while

warm jasper
#

How many times the installer has to download Torch? My internet plan is limited 😰

warm jasper
#

It seems InvokeAI doesn't work on 4 GB VRAM, even 256x256 was too much. RIP

fervent thunder
stuck sun
#

A1111 is pretty broken right now

#

not possible to load a training model in dreambooth

vestal dew
#

what?

#

I wonder if there's anything new in the world of hands and feet fixing. Ive been hours with controlnet with no luck\

fervent thunder
#

just loaded invoke AI first time since last month ago...it got no updates since then o_O wow

vestal dew
#

The article only shows a few images. Some of the images showing off MJ's "fixed hands" have people with three fingers and a thumb and five fingers and a thumb.

This is typical of AI claims. Those were the best of hundreds of images, picked to show off MJ. They couldn't find five images with perfect hands for their article about perfect hands.

That's telling.

#

Also, the prompt was supposedly “a color digital camera photo of two Caucasian politicians shaking hands and smiling outside of the White House", and they got a great likeness of HW Bush - BTW, with his wedding ring skin colored and part of his finger.

Solved the problem indeed.

vestal dew
fervent thunder
#

wont see fix there i think until more models are release in higher resolution with more parameter to pair with it

vestal dew
#

OK

#

Today the police came to my house - they examined the contents of my drive for the images I've generated. A detective said that all the extra fingers and limbs and toes etc., the mangled hands and feet, if those images were to get out into the public, people would go insane from seeing them.

fervent thunder
vestal dew
#

whats the update on negative?

#

I have to look at hundreds and thousands and gigabytes of these mangled images. I can take it. When I look into the abyss, the abyss blinks. However, most people can't take this. It's maddening.

fervent thunder
vestal dew
#

these images are the most horrible things I've seen, and I've seen people's heads severed in accidents

#

Well, I'll give that one a try

fervent thunder
#

Hands are still broken, but I'm tkinking about it like "if there's 1 image out of 100 where it can help - I'll still take it"

vestal dew
#

I agree

#

I have DreamArtistBadHandNeg

#

but it didn't seem to help

#

badhands5 is 2 months old

fervent thunder
#

yea

vestal dew
#

that's not new

fervent thunder
#

thinking I think new model does hands better too

#

but still bad at holding things, so yea...

vestal dew
#

2.2 you mean?

#

2.1 I meant

#

Oh my goodness, 2.1 makes a tinker's bollux out of my prompt.

fervent thunder
#

SDXL I mean, not released yet, except for tests on DS

astral pivot
#

can someone extract a lora for me, my pc keeps running out of memory

vestal dew
rigid ember
#

Going back through old questions and advice. Is there a particular reasoning/logic as to why you feel amd is best? I've got some opinions in a comp building discord that intel is better for multicore use in ai art generation. But I've been hearing both sides with conflicting info Dx

vestal dew
#

what else does it say in there?

livid kayak
#

how to make original characters? i want a fusion of sepiroth and cloud in final fantasy?

fervent thunder
vestal dew
vestal skiff
#

can anyone help me?
i installed stable deffusion on my pc including python 3.10.6 and when i click onwebui-user.bat it says

No module named pip
Launch unsuccessful...

what can i do?

surreal perchBOT
#
FAQ: What is Stability AI?

Our vibrant communities consist of experts, leaders and partners across the globe. They are developing cutting-edge open AI models for Image, Language, Audio, Video, 3D and Biology.. AI by the people, for the people. Learn more here stability

hoary river
#

Is the ui fixed yet?

astral goblet
#

why-you-always-lying.gif

proud nova
hoary river
#

Ah okie thanks. I guess its going to take a while. Without knowing I pulled the latest yesterday, whats the git commands to revert and go back so I can at least do some generations?

proud nova
#

Generating finally works again for me. Doesn't it for you?

#

And to return, I think you should use command git checkout a9eab236d7e8afa4d6205127904a385b2c43bb24 in SD directory

#

Not sure how it'll work. I am dummy at python

plush haven
#

Just a thought, but with depth & normal maps (generated from controlnet) essentially being a simplified/approximation in 3D space. Technically we could transpose the view in 3D space, generate another image and continue around the object and map it as a 3 dimensional object.

vestal dew
#

maybe

#

but I cant even get a decent pose uising control net

hoary river
#

Looking at the models on civitai.com it doesn't look like a lot of people have a handle on SD 2.1 yet. So far most of what I've seen is inferior tot SD 1.5. Will 2.1 become easier to manipulate on newer versions and that's what everyone is waiting for?

mellow meteor
gaunt tinsel
#

When someone is making a series of photos that are only slightly different. how are they doing that?

compact quiver
#

Does anyone know of a Discord server with an image AI bot you can use?

fervent thunder
#

Look it up online

compact quiver
#

I am online,

fervent thunder
#

whats this server

vast ingot
#

this is the Stable Fundation server. mostly about Stable diffusion, an AI that makes pictures.

surreal perchBOT
fervent thunder
#

ok

#

thx

cyan marten
#

Hello I'm new to the server

bleak matrix
#

Good morning, everyone! How are we all today?

vast ingot
#

Hello new, I'm Guizmus, welcome.

cyan marten
#

Thank you for welcoming me 🙂

#

Oh you're a community mod is there a channel i can ask some questions

#

or is it allowed to ask questions here?

bleak matrix
cyan marten
#

oki i'll take a look at those channels first thank you for letting me know

bleak matrix
#

But you can ask here, too

#

Np!

#

There are a lot of guides and members here, so please don't hesitate to reach out. If you need technical support, though, I would ask in #🤝｜tech-support

cyan marten
#

Oh i'm a pm seeking for a way to be a part of the AI

#

so its more of a "generated image" issue more than techincal

vast ingot
#

part of the AI ? what do you mean by that ?
there are quite some ways you can participate, this is an active community with quite some projects

shadow birch
#

does anyone use automatic 1111 on google colab?

fervent thunder
livid kayak
warm trail
#

Hi all ! I’m new here and would love to try an generate my first stable diffusion image. Is there a channel for prompts here ?

vast ingot
#

sorry but not currently no
the website has some nice features and free credits though, and you can also install it, but we don't have picture bots currently

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

Currently, there is no bot on the server that generates images. However, there are plenty of other ways such as the official https://beta.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!

stuck summit
#

Hi, can someone help me to understand better how to use stable diffusion? Where can i use it? is there a bot? Need to install on my pc to exploit the gpu?

vast ingot
#

well, I answered one of the questions just now but I got good news on the other sides

#

yes you can install it

#

you can also run it in the cloud in a notebook like google colab, would your computer not be powerful enough

prime parrot
#

.

vast ingot
#

it requires minimum 4GB VRAM on your graphics card, and is quite a lot easier to use if you are on NVIDIA

stuck summit
#

But i'm trying to understand which way is better to do what i want to do

vast ingot
#

what are you trying to do then ? ^^

stuck summit
vast ingot
#

the different distributions available in #1080946152318443610 and elsewhere are fairly customizable and can handle most task, if you spend the time setting it up for it

stuck summit
# vast ingot what are you trying to do then ? ^^

I need to use it to create artworks whenever i want and in some cases nsfw, but for now i can't find a easy way to do that. In particular, i've seen some artist doing pretty good work manteining the same styles, so i want to learn and try by myself on this technology

vast ingot
#

lots of things but first of all, yes, Stable Diffusion can handle this

#

NSFW means you'll need to install at home, most website have filters outhere currently

stuck summit
#

yes, the artists i mentioned are using Stable Diffusion, thats why i'm here, i want to know how to use it and what is the better way to do that now

vast ingot
#

it will depend a lot on the type of style in the artwork, but the first thing will be to find the good model to use

#

asking the community around here, or checking civitAI for example

stuck summit
#

so firt thing first i need to install it on my pc?

vast ingot
#

that would be my first tip yeah. or use it in google colab, but if you got the computer for it, why not

warm junco
vast ingot
#

go look at the #1080946152318443610 , choose the right tool for your need.
I would recommand either :
InvokeAI for the unified canvas feature and the top tier UI
Automatic1111 for the whole lot of extensions and the correct UI
ComfyUI for the node approach to image making, the most versatile workflow, but a lot more technical at first glance.

stuck summit
#

Rtx 2060 should work very well i think

vast ingot
#

should work ok

#

you will be able to do most things

#

6GB VRAM

#

benchmark of the cards for SD

stuck summit
#

are there any kind of guides to install it? i'm pretty confused ahaha sorry

vast ingot
#

no problem

#

any idea on what one you'd prefer to try out ? that will depend on that

#

the most used currently is automatic1111

stuck summit
vast ingot
#

I'll show you the uis

grim nimbus
#

hey is it possible to setup SD webui on my server and access to it on my local device (they're not in a same network)

vast ingot
#

you run it on the server, using either "--share" or "--listen" option

#

--share opens up a portal through gradio.com to relay your UI

#

--listen opens up port for the console to listen to calls, so you can access it using the IP of your distance device if you forward the ports

#

in both case, there exist an option to set a user/pass --gradio-auth username:password,

grim nimbus
#

thank you!

fervent thunder
grim nimbus
vast ingot
fervent thunder
#

don't you need gradio args for it too?

#

Looking at my collab args, it has --gradio-img2img-tool``````--gradio-queue``````--share``````--gradio-auth {username}:{password}

zenith igloo
#

i want to make a lora of an anime character i like, how can i do this?

maiden bough
zenith igloo
maiden bough
#

there are a lot of good dreambooth tutorials out there

astral goblet
#

good place to start is to develop photo collecting skills. you'll wanna start saving and organziing all your photos you like. once you got good photo sets, like 10-20 images of whatever you wanna train, you can start thinking of processing those for training

#

i found the hardest part of getting started wasn't the technical settings, but just figuring out image captioning and collecting data sets.

zenith igloo
maiden bough
#

dreambooth works on a few as 6

astral goblet
#

pick another to train for practice!! xD

#

yeah i say 10-20 but you could go lower if you want. for science

glacial siren
#

Anyone know of any text-to-3D programs?

worldly sequoia
#

I get "Login to view models" forever

#

But I already logged in

astral goblet
worldly sequoia
#

Looks like https://grpc.stability.ai/gooseai.DashboardService/GetMe hangs forever.

astral goblet
#

goose

#

theres the problem right there. theres a goose in yer ai

#

the old site gives me error "bad authorization string"

mellow whale
#

Just starting to learn the ropes of SD today, was using MJ for so long, it is just so different, lol

oblique osprey
#

Yeah. Can’t get to things. Right now image generation via api seems broken too

mellow whale
#

With unlimited freedom, it is just too easy to make something NSFW

oblique osprey
astral goblet
#

people generally haven't kept their hedonistic tendencies in check for a while now, and AI is like a giant bowl of high grade opium on that philosophical approach to life

proud nova
#

New webui seems fixed

#

With some minor bugs, but it's functional as before

kind kraken
#

For nsfw image generation, what is the most elite model to use?

proud nova
#

another one....

#

Look up Civitai. You can find (unfortuanetly...) lots of them here. Just create an account

fervent thunder
#

I think you'll have to enable nsfw in settings too now

proud nova
#

of course

kind kraken
#

Say i wanted to create my own model for nsfw in stable diffusion, all i need to do is to give it training data ? @proud nova @fervent thunder

proud nova
#

ugh...

#

Training models is harder than you think

fervent thunder
#

no idea about training, I don't do that

proud nova
#

For decent model it requires at least 1000 images.

#

I also don't train models

#

I trained LoRAs, but models are tabu to me

vast ingot
#

training can be hard for sure yeah

#

no need for 1000 images though, you can train a model with 15 on dreambooth just as much as a lora, gotta start small

proud nova
#

Really? Only 15 images?!

#

I thought that dataset of this size is enough for LoRAs but for whole models?

kind kraken
#

@proud nova

proud nova
#

Users can train/upload their models/LoRAs and you can download them

#

So yeah

#

It's almost like huggingface

trail pecan
#

Stability ai are moving in the right direction. They keep getting better and more creative. It is immediately clear that they do not want to compete, but to surpass themselves yesterday. We respect this approach🤝. Stable Diffusion v2.2.2 XL best current model! Keep it up and wish you continued development!👍🔥💛

neon cave
#

how does it work?

#

like what are the advantages of Google colab

#

is it just like stable horde? do you gain anything from hosting on stable horde?

mellow knot
#

iterations per second is just the amount of steps a GPU can do in one second right

#

or are iterations and steps different things

ancient bear
#

hey guys, just a question: if i set the width and height to high, it wouldnt work and the console tells me that cuda is out of memory. how can i fix that?

latent hawk
mellow knot
winter pike
#

i wonder how the final SDXL model will look

#

i can finally generate homer simpson and other 2d characters

fervent thunder
#

xformers, medvram and lowvram has consequences tho.

Xformers makes results incinsistent, so you won't get same exact image with same exact prompt
medvram will make gen a bit slower
lowvram make gen alot slower.

steep wing
#

Definitely use Xformers, the benefits heavily outweigh the consequences mentioned here

humble imp
#

🫥

uneven forge
#

Anyone experienced with prompt tags to write ? using existing image to generate new one

#

pretty new to this xD

astral goblet
# proud nova Really? Only 15 images?!

more can get more quality variations and results, but you can get really good generations out of less. i've done training as high as 70 image sets tops so far

#

trying to develop a workflow right now, drilling it on smaller sets. i'm looking to build some bigger data sets soon, but it's no problem to start small. i've learned tons

stuck summit
#

hi, how can i create good realistic 3D/anime girls?

#

do i need to change model or something like that?

warm junco
fervent thunder
#

nah, photorealistic isn't good tag to get it, you'll get game-like graphics

#

just "photo"

stuck summit
#

i've the three standard models, they seems not working well for what i want to do

astral goblet
#

depends on the model. if it's one of the photorealism models that are triggered with that word, then it kicks in. base model picks it up other ways though

#

3d render in the negative sometimes helps. i add it last if i'm having issue and it can be very dynamic

warm junco
stuck summit
strange aurora
#

Any word on AMD GPU friendly SD UI’s?

stuck summit
#

i've seen some artist around doing this very well so i'm trying to recreate

astral goblet
#

there's a directML fork of a1 webui

#

lol

stuck summit
#

sorry i don't understand well

astral goblet
#

i was telling @strange aurora and lol.. i have a policy to not dutch rudder with guys

stuck summit
#

ok ahah

stuck summit
strange aurora
astral goblet
vestal dew
stuck summit
#

and where i have to put the model file? just into the model folder?

vestal dew
astral goblet
warm junco
astral goblet
#

oh that too

#

i may actually try the dml branch, as theres lots of features included

vestal dew
astral goblet
#

don't know yet

warm junco
astral goblet
#

the repo i linked has xformers built to it

winter pike
warm junco
astral goblet
#

so the github says, but it may just fork the a1's guide verbatim

vestal dew
astral goblet
#

nope. it's not finished yet. the beta releases are on DS

strange aurora
astral goblet
#

DirectML should work on amd cards

#

i dont think it's as fast as cuda yet but it is faster than cpu mode

#

i may actually not use it turns out

vestal dew
cosmic root
#

Hey. Can anyone point me to a "Google Colab" for Stable Diffusion. I can't find one

#

i used to use one

winter pike
cosmic root
#

Isn't that for running it on your own GPU ?

astral goblet
#

Colab & Runpod & Paperspace adaptations AUTOMATIC1111 Webui and Dreambooth.

cosmic root
#

I can't handle running Stable Diffusion on my own PC i think. So i was looking for what was called a Google Colab, that allowed it to be used over the internet free

astral goblet
cosmic root
#

OH thank you!!!

#

Got it!

astral goblet
#

Mess with the best, die like the rest.

cosmic root
#

Can you use 1.4 wieght? I dont have 1.5

#

lol! thats right flow

#

Hack The Planet!

astral goblet
#

Zero Cool? Crashed fifteen hundred and seven computers in one day?!

#

the model gets uploaded to the collab when you're using it

#

oh and if you'venot watched "elementary" the holmes adaption where he's a detective in nyc, they do some great throw backs

cosmic root
#

Cool, Thanks

vestal dew
hasty bane
#

guys,does having more models in stable diffusion result in longer loading times?

vestal dew
#

no

#

they load pretty fast if you have them on an SSD

#

I think the processor also has something to do with model load time

manic coral
#

Hi folks, if you're in London, we're hosting a meetup with Emad and our CEO, Lukas Biewald on April 20th. Join us for a night of talks on LLMs, Q&As, drinks and hanging with like-minded ML engineers!

👩‍💻 We have limited seats, register here: https://london.meetup.wandb.events

icy glade
#

How do you guys see the level of stable diffusion vs Midjourney

#

Midjourney seems to have advanced so far with v5

waxen cave
#

im looking for a good toutorial on training stable diffusion on a new object

slender vault
dusk spade
#

Hello all! AMEN 💎💯

mellow whale
#

V5 is nice and all, but SD seems much better

astral goblet
#

one word, well two, but it became one. inpainting

vestal dew
#

V5 doesn't run on your machine. Someone else is in control. Hells to the naw naw naw

astral goblet
#

cause i need some lovin like i never needed loving before 🎵

mild abyss
#

hey uh

#

I tried to go on

#

thispersondoesnotexist

#

is it dead now

#

cause it aint there

mellow whale
#

I got an out of memory error when merging models, how to get around with that?

mild abyss
#

🤖

woeful spade
#

Hello, does anyone know how the function of putting images to generate variants works? I was trying to do it with some Genshin Impact images but I don't know how it works, so it generated somewhat strange images

vestal dew
vestal dew
woeful spade
vestal dew
#

therer's a slider for it in the UI

woeful spade
vestal dew
woeful spade
#

Hello again, I´m doing anime images, How can I write the prompt, so that the images that I am generating come out as accurate as possible? I'm just learning to use AI

tidal lion
#

I think you should use images that suggest the same topic in combination with control net

woeful spade
tidal lion
#

you can go to https://civitai.com/ to check the prompt of each photo and choose the one you like because it is the original model so it will be correct. and control net you can find a lot of tutorials on the net, it's quite hard for me to explain it to you

#

or more advanced you can mix models or create your own model it will look like what you want

woeful spade
worthy olive
#

whats the chat to gen arts ?

tidal lion
#

Can you ask more specific questions?

plush haven
#

if only AI was able to generate a passable side profile from a front view image

tidal lion
#

yes i think they can but they hide it because it will threaten the gaming and art industry

#

but if you try and take some time, you can still do something acceptable

plush haven
#

hmm would love to see such a tutorial if it exists. i have concept art trying to match but failing miserably for the past few days

tidal lion
#

i think they will suggest it soon or we can look forward to the technique of 3D modeling AI from nvidia :3

mellow whale
sullen marten
#

guys know anyway that i can extract multiples checkpoints to a lora ?

boreal dirge
#

what is the latest engine out rn?

#

I've been here for 10 minutes lady expect a response sometime later

vast ingot
#

Hello Ricardo then and welcome

vast ingot
boreal dirge
#

I'm using it to generate images and they either come out distorted or old

vast ingot
#

ok the best model then

#

well there is a ton of choice currently, it will depend a lot on the style you want to achieve.
The current best place to look for those would be civitAI.com, lots of choices

boreal dirge
#

have you seen midjourney?

#

what do you think they use?

vast ingot
#

yep for sure

boreal dirge
#

That realistic look

vast ingot
#

well they do use the same tech internally, Stable diffusion, but they don't say publicly their exact pipeline
My thinking is :

#

they use multiple custom trained models internaly for a start
They add parts to your prompts to call on the good styles in their models
They have some kind of post processing, like hypernetwork, to get even greater aesthetics

boreal dirge
#

I see

vast ingot
#

as a model, I'm almost certain they don't use anything public, but there are models that try to copy their style

boreal dirge
#

I understand now

boreal dirge
vast ingot
#

if you are using 1.5 still today, yeah, there is

#

I personally still love 1.5

#

but I could tip you towards those :

boreal dirge
#

I just decided to look into image generation

#

I just want to see what the best look like

vast ingot
#

Deliberate/Realbiter/Realistic vision : for realism

#

DreamShpaer, Illuminati, Dreamlike Diffusion, RPG : for art/Fantasy

boreal dirge
#

thank you

vast ingot
#

I'm not good at anime at all personally, but I know this field has tons

#

my personal love stays with Realbite currently, but I like the "realistic" look of photos it gives

boreal dirge
#

okay okay

#

thank you very much

vast ingot
#

no problem 🙂

#

have fun

boreal dirge
#

is there a website for it?

vast ingot
#

lots of models there

boreal dirge
#

okay thanks

#

https://api.stability.ai/v1/generation/stable-diffusion-v1-5/text-to-image

#

does it have an engine id?

vast ingot
#

you are using SD directly like that, in the API?

#

I'm sorry I didn't understand that question then.... you can't take any model in there

boreal dirge
#

oh

vast ingot
#

you are limited to the ones available in there

boreal dirge
#

okay thanks i'll figure it out

vast ingot
#

what I'm talking about is if you run it on your computer or in the cloud

#

sorry

boreal dirge
#

that works just fine

#

I'll check it out

modern crater
#

Guys, how to change AUTOMATAIC webui theme to dark?

warm junco
jaunty tusk
#

guys what the f***?
Channel verification level is too high
Due to extreme demand we can't provide a free trial right now. Please /subscribe or try again tomorrow.

cerulean juniper
#

lol

cerulean juniper
#

for collab: --medvram --opt-sdp-attention --xformers --theme dark --listen --share --port 7860 --no-gradio-queue --gradio-auth "user:pass"

proud nova
#

In speed etc.

cerulean juniper
#

this fixes the issue with model selection: --no-gradio-queue

cerulean juniper
#

this is a duplicate of the webui-user file

proud nova
#

New webui is fine now

#

Some extensions still needs to be updated but most of them are working now

trail pecan
#

Say please pickapic Works on the basis Stable Diffusion XL or all 4 models of stable diffusion are involved?

torpid igloo
#

is there anyway to see what users on like pixiv are using for prompts and stuff?

warm junco
vestal dew
#

Yes, if they upload the original PNG

#

the PNG contains the meta data which includes the prompt and other info. However, most people delete the meta data before posting.

#

well, I take it back. Looks like Pixiv converts the images to JPG

modern crater
#

Hey guys, can you guys recommend me model for generating multiple haircuts of person?

fervent thunder
#

thinking probably any model will do, you just prompt hairstyle you want

modern crater
stuck summit
#

Hi, good morning, after a few tries i generated something good.

#

Now i want to ask something that i don't understand well.
Sampling mode: is possible to integrate another sampling mode?
Unified canvas: is really good but i don't understand how the mask or brush work at all
Extensions on Invoke are possible??

grim frigate
#

any tips for prompts for buildings / cityscapes? Most of the times the buildings are really jaggy or the windows are just messed up.

torpid igloo
#

i tried downloading some of the specific character models in civitai but even when using them the characters dont actually show up in prompts

warm junco
torpid igloo
warm junco
#

Loras go into models/lora folder

#

Everything under 2gb is not a model

torpid igloo
vast ingot
#

loop ?

warm junco
vast ingot
#

strange message for the bot to block, sorry

#

fractals + loop

opaque kernel
#

Hey everyone! what's the current best method for generating depth maps?

cerulean juniper
warm junco
opaque kernel
#

is there a service which has that integrated, to run it remotely?

warm junco
#

idk, but you can run sd webui remotely and use it then

grim nimbus
#

hey I'm using exact the same model and prompt/parameters as examples in civitai, but the image I get is much less colorful compared to the examples, do anyone know what may be the reason?

warm junco
gray raven
#

Any Stable Diffusion discord bot?

grim nimbus
#

Another quick question, when I tried to add --share to webui.sh, gradio always throw this error: *** Failed to connect to ec2.gradio.app:22: [Errno None] Unable to connect to port 22

#

do you know what could be the problem? thanks!

warm junco
grim nimbus
grim nimbus
warm junco
#

that tha vae gets applied

grim nimbus
#

ok thank!

warm junco
#

in needs to be the exact same name like
Example123.safetensor
Example123.vae.pt
and both in the models/stable-diffusion folder

grim nimbus
warm junco
young bay
#

Any mods on?

balmy glen
#

@rigid bridge please stop.

fervent thunder
#

Did anyone else get a dm from @visual haven

ebon inlet
#

guys i've been out of the loop on the best local installation forks for a while, is the
AUTOMATIC1111 fork a good choice for right now?

cerulean juniper
#

yeah

ebon inlet
#

beauty thx

bleak matrix
fervent thunder
bleak matrix
fervent thunder
#

does stable have any kind of chat ai?

toxic nebula
#

hello

#

are there people who train their model based on one artist and pretend to commission art as them lol

#

is this a thing that happens or is the mind of smae deep and scary

surreal perchBOT
#

@fervent thunder

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

Currently, there is no bot on the server that generates images. However, there are plenty of other ways such as the official https://beta.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!

fervent thunder
#

I'm currently a billionaire from my commisions

bleak matrix
cerulean juniper
#

C:\Users\Admin.cache

#

spent an hour to find this folder to remove the corrupted package

opal crescent
#

anyone got any idea how to register in chatgpt with virtual number?

toxic nebula
#

ok im 10% done the training for my ai model on stable diffusion

#

and the preview images look horrendous. Is this any indication as to the quality of the generations in the future

#

?

toxic nebula
waxen cave
#

this is starting to bug me how is it i can find no info on training objects for stable diffusion ? like say i want to train it to know what a tea cup os

#

every attempt making a lora for a object has failed and the ai is close but does random things to change the object

#

smar dream booth training lora or what?

stuck summit
#

Hi, can someone explain me if it's possible to change diffusion metod and where i can find this metod?

#

In particular: a path that i find requires: Duc Haiten Super Cute (find)
and use "Magespace" or "Waifudiffusion"
What are those?

fierce charm
#

I there anything in SD similar to Topaz Video AI to upscale and improve videos, without changing them

fervent thunder
grim nimbus
#

hey anyone got idea how we can translate a natural language description of a scene to a suitable prompt for stable diffusion to generate the scene?

hallow plaza
#

When using dreambooth images that are 512x768. Can I also use images that are 768x512 for my regularization images?

merry pendant
#

played aroudn with Bing AI image creator for the first time today. Shockingly good at interpreting a prompt and delivering a solid, coherent result. Also, everything it does looks like ass

#

prompts that no SD model anywhere can do, it did flawlessly. but, again, looks terrible

astral goblet
#

It's gpt4 paired up with Dall-E if i'm not mistaken. It should be good!

#

i got a notification from my bing app about how it's capable of doing image generation now, but i can't find it in the app at all

halcyon granite
#

anyone else experienced auto1111 image slideshow not working (Firefox)? The only way to fix in aware of so far is restarting the UI

astral goblet
halcyon granite
astral goblet
#

upscalers are in img2img tab now ooooo

astral goblet
gaunt laurel
#

Hello, I'm new here, I came to ask this question: How do I generate logos with Stable Diffusion (v 1.5 using Stable Horde)?

halcyon granite
astral goblet
#

outside of that professionalism, there are models specifically for graphic design and logo esque style generations

gaunt laurel
#

ah ok, dang

astral goblet
#

i'm not sure if stable hoard allows such things though

gaunt laurel
#

well, ig I will just use it for inspiration then (but idk how to make logos)

gaunt laurel
astral goblet
#

yeah it's cloud community hosted stable diffusion setups

#

i dont know if they allow you to point the hoard at any model you want to use though

gaunt laurel
#

yeah, OHHHH, you mean like custom extensions?

#

yeah, they don't allow that

#

I was just wondering if there was a way to use prompts for it

astral goblet
#

i should say, a pet peeve of mine is "do you even know?" style questions. so... yeah. chill on that attitude please.

warm junco
#

StableHorde allows Logos, they even have nfsw workers

gaunt laurel
#

no, I didn't mean it rude, but some people hear something and get it confused with something similar

astral goblet
#

Prompting tips would be "logo" (heuh) "graphic design, vector art, line art, thumbnail logo sketches, pop art" things like that maybe?

gaunt laurel
#

ok

halcyon granite
gaunt laurel
#

yeah, I may just not use SD for that since I would like it to be vector graphics, and i think it can only generate raster images

astral goblet
warm junco
gaunt laurel
#

so, what is a way for me to use a specialized model for free/cheap?

astral goblet
#

don't mean to sound like an elitist here but 100% professionalism matters

astral goblet
gaunt laurel
#

well, idk anything about SD based models

#

and they only tell you the name of models, not what they do

astral goblet
#

ooo don't forget the cherry picked screen shots lol. it's a problem. hard to know what a model is going to do. best way i've found is to download it and try a few prompts i have saved.

#

civit i think has a way to generate on the site with a model, but i don't like doing that. costs money i think. it's all wrapped up in legal licensing bs. the whole saas scene of SD is just bananas right now.

merry pendant
#

I've found it's a lot better to find one model that's flexible and experiment a ton with it

#

learn all the ins and outs and how to really make it shine

#

rather than hopping between 30 mixes looking for the best waifu from your one prompt

#

my $0.02 anyway

astral goblet
#

https://civitai.com/?query=design this query brought up a bucnh of good ones. see if the hoard has any that show up from thes set "designvision" in there is crap. avoid it i promise you.

gaunt laurel
#

is there any models made specifically for logos?

halcyon granite
#

the the img2img upscale someone referred too latent or have more options been added?

astral goblet
potent fox
#

I would try and train something using logos

astral goblet
#

i think it'd be good at coming up with similar logos rather than designing a new one that fits the intended purpose

#

i'd rather do the process like this. design the logo myself, put that into controlnet to finalize it

potent fox
#

Great idea!

astral goblet
#

as you can see it effectively communicates stadia's brand image

potent fox
#

That's what i call creativity 🙂

astral goblet
#

a logo is most effective when it can communicate the brand in all shapes and sizes and colors and prints

potent fox
#

I mean, controlnet gives you the power to provide anything to it, like a simple sketch or a shape, and create something beautiful from it.

grim nimbus
#

hey is there anyway to make the character looks alike across different pictures? I do not want to use LoRA or textual inversion because it will make the scene less faithful to my prompt

astral goblet
#

sometimes people get ideas that they don't try because it could waste a lot of time and be risky if it doesn't pan out

potent fox
#

That's what i find so amazing about that. You can make a simple creative thing, being it only a hand drawn figure and then the AI gives you so many inspirations to make progress into what you have in your mind.

gaunt laurel
#

I have 2GB VRAM

potent fox
#

With 2GB VRAM you can only run basic things, sadly.

gaunt laurel
#

yeah, very frustrating

#

I could use a laptop that has 8GB, if that would make a difference?

potent fox
#

8GB VRAM?

gaunt laurel
#

yup

#

not mine, but I have access to it

potent fox
#

Of course it makes sense then.

gaunt laurel
#

ok

#

how fast would it be?

#

it's specifically an RTX 2070 Max-Q

potent fox
#

Well, the 2070 is not the fastest out there, but at least better than a 2GB GPU.

#

But that model might be a good one thouugh.

gaunt laurel
#

ok, another question, how do I setup SD?

#

locally

#

and also have a censor for NSFW (not required, but would be nice for some more vague prompts)

kind kraken
#

General question for running a stable diffusion, if I have my own rig to run SD, how much money would I save in the future as opposed to running it in a cloud service ?

potent fox
gaunt laurel
#

nvm, I can't use the 8GB laptop rn :/

broken gull
gaunt laurel
#

is there a service I can use for really cheap/free that has more customization?

broken gull
gaunt laurel
#

I don't think it has much customization

astral goblet
#

myself, i don't generate nsfw outside of novel curiosity. won't say i don't at all cause that'd just be a lie. but even though its not even close to a significant part of my generations, i keep the nsfw filter off. the thing is a giant pita and it uses resources too

#

if i were lending my machine to the hoard i would turn it on

gaunt laurel
#

it seems pretty expensive

#

compared to runpod

potent fox
#

You could rent a virtual machine. Pretty accessible.

vestal dew
#

I generate NSFW by accident. Sometimes it creates naked horrors, and then the more I add something to negative, the more it does of it

fervent thunder
gaunt laurel
#

colab is EXTREMELY slow for some reason

#

took over 5 mins 1 time

#

for 1 image

fervent thunder
#

Are you sure you are using gpu?

gaunt laurel
#

I thought so

#

maybe not

#

I should try it again tbh

#

do you know of any links for that?

fervent thunder
#

idk, my collab doing good on t4

vestal dew
#

here's a puzzler. Sometimes I put in there "blonde hair" to give, obviously, blonde hair. However, the eyebrows remain black. So I put in "blonde eyebrows" into the prompt, and they get blacker. If I weight like (blonde eyebrows:1.5) the eyebrows get very thick, even blacker. What is going on there?

fervent thunder
potent fox
fervent thunder
vestal dew
fervent thunder
#

same 😄

gaunt laurel
#

then, how does upscaling work?

#

(might not use it, but just curious)

#

it seems cool

gaunt laurel
vestal dew
#

then I was teaching a group class on Sunday, in walks a kid I had seen before, but now his hair was different. I noticed his eyebrows were like the super black ones I get when I try to get blonde eyebrows, so I asked him if he was AI generated

#

I counted his fingers - four and a thumb per hand

potent fox
merry pendant
#

like, I don't want to generate dicks, but what I get as a result of SD not knowing how is infinity times worse

vestal dew
#

yes, they have sometimes what looks like a snake in place of genitals, horrors. I put "NAKED" in the negative, and sometimes it gets worse

fervent thunder
#

thinking I don't generate men, I don't get naked men, profit

gaunt laurel
#

👀 what the

#

so immature smh

gaunt laurel
vestal dew
#

I generate men and women, and get weird naked ones for both, even sometimes when I add weight to the clothing mention

fervent thunder
#

it's paid only tho, isn't it?

gaunt laurel
fervent thunder
#

yea I get weird ones too sometimes

gaunt laurel
#

but, I will try the free ones first

merry pendant
#

you'd think with 90,000 training images of cherubs from 1800s paintings they'd at least get it close, but no.

fervent thunder
#

nsfw in negative gives weird results sometimes too

#

like it can gen boobies poping out with no nipples wot

#

or maybe it wasn't nsfw negative, maybe it was just model result without nsfw in either prompt, I forgot 😄

merry pendant
#

I recommend runpod wholeheartedly btw

#

I used it until I bought a 4090

hallow plaza
#

What models are trained on 512x768? What about sd1.5?

gaunt laurel
#

How do I upscale with Automatic1111?

potent fox
broken smelt
merry pendant
#

is it better than the ultimate sd upscale script?

#

oh nm it's just a straight up upscaler

#

interesting

gaunt laurel
astral goblet
#

it's cool to see people discovering it all in big waves lately. it really does hit hard when it works

broken smelt
#

ya if you google it it should be the first hit

astral goblet
#

ever since the seam fix it is just nuts

gaunt laurel
#

ik, but what is it?

broken smelt
#

i think its also in the extensions list if you searched for it too

gaunt laurel
#

what's it do? and does it work on colab?

astral goblet
#

it breaks an image into smaller squares, scales each of those up with img2img, so you can prompt it with "photographic detail" or some that helps boost it, and stable diffuse the tile to the new size. denoising at something like 0.2 - 0.4 for consistency. less is usually better but you want enough that it adds details other than the scaling artifacts. then at the end it does a pass on all the seams of the tiles so they blend

#

it should be able to work on collab. it's just a batch job

#

i think the next upgrade for it will be able to manage what prompts go into what tiles, but that would be a lot of tedious code work and i can understand why it's not here yet

gaunt laurel
#

in simple terms? lol

astral goblet
#

it do good resolutions ez

gaunt laurel
#

ah ok

foggy atlas
#

Hello diffusion experts. I’m sure everyone has watched the Corridor Crew video on Spider verse animation. How do you lock a same noise pattern on every frame? And what does it mean having the noise focus on the face? 😦

hollow crane
hollow crane
merry pendant
#

does anyone run SD locally with a 4090? And if you do, does it ever overload your pc and kill the audio?

astral goblet
merry pendant
#

I agree that it shouldn't but it does 😦

#

diablo 4 did it too actually. any time the GPU gets massively maxed out it has a chance of cutting my audio. i think it's eating up all the interrupts

#

i hvae an external USB DAC, which might be part of the problem. I just wish there was a way to tell it 'leave 2% available always' or something

astral goblet
#

it's probably because somethings up with your system

#

what's your mother board chipset?

#

i would immediately suspect that

merry pendant
#

hmm

#

i guess i can try upgrading my bios, see what happens

astral goblet
#

USB dac absolutley sounds like an interupt problem. there may be a usb slot that's on another pcie lane

merry pendant
#

ooh also a good idea

astral goblet
#

if you've got another usb port or even better an entirely seperate usb3 header to connect that dac too.. doooo it

merry pendant
#

i ran into someone else with the same issue and he uses ASIO stuff

astral goblet
#

last time i ran into interupt problems was when i was using an active repeater on a game controller with 4 controllers playing a game.

#

heh

merry pendant
#

lol

astral goblet
#

it was like everytime he would hit a certain button too.. and we didn't figure it out for a while. why the game kept slowing to shit

astral goblet
#

i hope he didn't die