#🗣|artisan-support-feedback
1 messages · Page 7 of 1
not this kind of event -.-
please don't try to prompt-engineer the developers into giving up secrets
Keep fishing me and all you'll get are boots
an question cost nothing and good answer is priceless
not really an secret, we know how long sdxl needed to train so we aspect when the training here with votes goes right its only 5 weeks from now
its more logic
than anything else
okay, feel free to do that logic on your own, let us know your prediction. If you get the date right you'll win a prize. if you get it wrong you're receive a punishment
the punishment is not to be an mod here and ban people that do nsfw art right ? 🫠
i cant handle that
what's the limit that's required to get hd image
There is no set limit and it's not something you can unlock. After enough people have voted and we get a picture of what we need to see I can change up the workflows again.
Thanks. BTW yesterday's image production was fantastic
But I didn't like today's images. It's like taken from another bot like dalee or openai's bot
the "vote some more" message is really vague
also by putting the generations behind this wall people are going to end up voting randomly and not based on quality
also the restyle button doesn't work
Fair! I will update the message.
I am hoping people just get in the habit of voting more than they generate. That's the requirement to keep the cooldown away. I will update the message to be more clear 🙂
gm
1024 is coming back into the mix
Are we allowed to invite the bot to my own
hi! no, for now you can only use the bot on this server.
@dull cargo you cant, that because the bot that generates images,with the votes is used for training the new model version.
The training and votes are watched by the mods and admins to make sure no one would generate images against the rules AND to make sure many people vote to make it more democratic and not like an specific group of an specific server votes for specific images they want to train.
So the main reason is to control the training of the sdxl model in any way possible, to make sure no one is abusing it. you can do whatever you want offline in an tool the model is free to download.
thanks! I went for the shorter version 😂
no prob i copy paste this here here and there
ikinda wish there was a bot that auto answered questions here
Honestly...
... I want to hook up an LLM
We do have a FAQ bot and it could use some updates re: bot questions tho 😄
Addressed, thanks
is it ok to just generate picasso with weird things writen in his shirt?
As long as it's not bad stuff, I think it's ok
i think the pika bot has two features that are very important that the sd one lacks. one, an option to change the prompt from an output image instead of having to find the original command or reenter everything again, and the option to cancel a request while it's generating
try @askinkeep, saw that in bun discord: https://discord.com/channels/876711213126520882/1151722654303272992
Hi! why does the output have such a low resolution? Will 1024 coming back?
I second that question. Almost every image is 512.
There still is 1024 mixed in but we still need some feedback on 512 images 🙃
I thought SDXL was trained on 1024. Is this an effort to make it work in lower resolution too?
Hello! I wonder if I could use the stable diffusion bot on my own server?
@bitter sorrel
I need help with an error code stable diffusion is throwing: "/home/avec/Desktop/stable-diffusion-webui/venv/lib/python3.10/site-packages/torch/cuda/init.py:546: UserWarning: Can't initialize NVML
warnings.warn("Can't initialize NVML")"
Hit me up if you need some ideas for questions. 🙂
can I add you bot to my server ?
@lethal haven is there any statistic what word are used the most in the promts ?
you cant, that because the bot that generates images,with the votes is used for training the new model version.
The training and votes are watched by the mods and admins to make sure no one would generate images against the rules AND to make sure many people vote to make it more democratic and not like an specific group of an specific server votes for specific images they want to train.
So the main reason is to control the training of the sdxl model in any way possible, to make sure no one is abusing it. you can do whatever you want offline in an tool the model is free to download.
ok
@lethal haven Are we training SDXL on 512x512 now?
Sorry to be a bit late on the uptake if that is the case. I have been seeing some SDXL 512x512 images.
sdxl will be capable of make 512 images?
yes it can @wheat pike and when you using it offline in InvokeAI tool the canvas size can be diffrent from the generating resolution that means the image is 512x512 big BUT in renders the image in an 1024x124 resolution when you want
how can i add images in my request to the prompt?
when you type / promt you can use an positive promt (text it will NOT ignore) in gray there should be beside this an +4 more (or wahtever in english ) when you click on it 4 more options pop up you can click, "aspect ratio" where you select an aspect ratio to force the model to make an image in an specific aspect ratio........ than there is negaive promt, where all the words go that the model dont use for generating something. than there is style that forces the model to use only specific styles........... AND than there is controlNet that is mostly the "put an image here and the model will use it as inspiration" normally for image inspiration we would use image2image that can generate more variations but with less perfection. but image 2 image is not supported here
@normal pike
Accidental nudity in the showdown. Sorry. #1100484581037195384 message
why is the anime style 1. so fast 2 so bad
way more than usual
ah this is why
half res
it seems to be half res almost randomly
regardless of style
yeah
there are stable diffusion models trained specifically for anime but since xl isn't they come out a little off
Good day, for three days now, images of very small resolution and quality have been generated, whereas other people's images are fine. How can this be fixed?
looks fine, that's your aspect ratio.
if that's not a screenshot of your image, then you're probably getting a lower res render as part of a randomized test.
(happens to us all)
Sometimes it's fine, both in quality and resolution)
that's the randomized part.
remember, the bot's primary purpose is to improve the sdxl model, so expect experimentation, weird outputs, smaller resolution and other stuff.
Got it, thanks for the detailed reply)
Ask, after generating the picture, how to get the big picture (original)?
Is this free to use or a trial?
free
How can i use an image as a seed /starting point for the subject ?
I am having nudes pop up on my latest prompts. Will include negative prompts to counter, but wanted to let you know!
Whoops. Apparently fat statue of liberty's naked. #1101178483608137918 message
I lost my nude neg somewhere
Got it worse. #1101178483608137918 message
Despite negs. Enough with that prompt ig
i dont know where to start and how to start
i this same as like midjoureny.. because i have used that only
Go to a bot channel and use the command that I can't post in chat because it's banned in messages for some reason
How can I upload a picture so the bot generated a picture similar to that? I have seen people doing it
I suggest you to use control net section when you type / dream
face detailer when??
https://github.com/ltdrdata/ComfyUI-Impact-Pack
Is there something wrong with the bots? Looks like hardly any images getting generated…
showing already has a pending request for hours
Hey I'm requesting to create an image via sdxl bots using dream.. but it repeatedly gives me an
ERROR: @undone falcon already has a pending request: Hang tight, you're still dreaming! :)
Any #1011228667659178055 #1010934719455707218 could help in here.
where is my art work? I had been waiting for 1 hour and above, it keep on showing me already has a pending request for hours!
Only one image in showdown lol
Same problem
@lethal haven The bot seems to be down. Almost no images are being generated
Since morning I am facing this issue
Support team pl. Do need full
Same here, the bot took my request but it didn't resolved before the showdown cleaning. So now I'm stuck with a pending request forever.
I think that's the same issue as everybody.
Thanks in advance.
@lethal haven some mod should atleast reply here, we have been facing issue since morning.
Is it not allowed to paint here now? Why isn't my painting coming out? It keeps telling me there's a thread hanging, but I don't think there is.
Hi Desi, sorry for not replying sooner, please remember that people on the mod teams and staff are probably on a different timezone than you, they might not be available
we are aware of the issue, please be patient and hang tight, sorry for the inconvinience
Thanks for the reply
Thank for the reply.
Do you know a way to reset our request history ? maybe leaving the server then joining again ?
also let's go easy with the pings to 2%, when he's online he's always on the channel, with just one ping will be enough! hopefully he'll soon be on top of this
Noted
if they have to reset the server I believe all pending requests get lost, you will have to prompt them again, take the time to save them for later
thanks !
How to solve the problem that I can not dream beacause I have a pending request?
try resending it but for now the bot is having problems, please be patient
Okay, thanks!
Thanks 👍🏾
2%, You are allowed to sleep. Don't let anyone convince you otherwise.
thanks for fixing the prob
Thanks so much ❤️
#🏅|pantheon message can this be deleted? the controlnet was copyrighted and used without permission
can i generate ai photos here?
In the bot-1 to bot-10 channels.
write: /faq and send message in one of those channels to get answers to the most common questions
https://discord.com/channels/1002292111942635562/1072220168534642768
is it yours? the firebird?
that alley with neon lights. It was a concept art from an artist. The bot sent me 3 generated images. Can they be deleted too?
if it's for non-commercial personal use i don't think it violates copyright law
I asked for permission after using it and the artist disallowed
i can't find it in the pantheon, can you show us a picture
I see some you did of a road, controlnet firefly, is that one?
it was above the phoenix. I think it was deleted but there are DMs too
if it was deleted, then no problem
can the DMs from the bot be deleted?
is the service overloaded? dream didn't finish
@lethal haven Not sure if it's viable, but a feature that allows all images sent via pm from the bot to be downloaded en masse as a zip file would be handy.
this is more a discord limitation
bots are wondrous, they can do magical things 🙂
As for dreams not finishing, that's not unusual. Just resubmit the prompt. There's a chunk of backend stuff happening most of the time that causes occasional glitches.
Yeah my bad, little tinkers here while it's live 🙃 I'll wrap it up for now
If it's in your DM, only you can see it
i'll say it again, the bot vote gatekeeping is just going to make people vote carelessly and not in the best interest of the actual image quality
the bot could have a vote C (both are bad)
So far I’ve only been able to dream 1 prompt every couple hours without being told to go vote or that the GPU’s are cooling down. This is ridiculous
It should only be 2 minutes if you haven't voted enough?
I've reduced the number of votes necessary
Isn’t it possible to check all my previously generated images? I’ve lost one picture and I need it so bad :’( Sorry if I ask in the wrong channel I’m a bit confused
is there anyway can allow me use image to generate image?
all images get deleted after some time, I'm sorry to say that they are probably lost.
Oh I expected smth like that. Thank you for the reply!
I think that "like / dislike" per image would be better than "choose which one is better, A or B". Sometimes both are good, sometimes both are horrible.
personification of scissors gives nsfw. Sorry. 😦 #1101178506530013284 message
Bro, for new user and free user like me, what is the maximum limit of generation i can make?
I realised this recently. When i write muscular into prompts, every male in the image becomes muscular nevertheless i had chosen one to be muscular and other to not be muscular
I don't want to seem painful or pretentious, but there are subcategories in which you can post your best finds, and I think it's a shame not to take advantage of them a little more, because the pantheon... or at least the champions, is is chicks, chicks again and... in short, I love seeing everyone's universe but the pantheon is ^^, then the few times I find myself there it's banal things that I like don't necessarily like ^^, so it would have been cool to see if there was a channel with specific challenges + pantheon of this one?
#🌠|show-and-tell was created to manually share things you thought were good from the bot. It's evolved and many people share fine tunes in there but that was the original idea 🙂
bunch of horndogs in #1100484581037195384
Yeah I saw this channel, but I liked the idea of the basic pantheon, because you can read the prompt and understand the original idea (and sometimes see that nothing was suggested but gives a surprising result) But in reality, it's mostly just good, I see a lot of girls and I find it a shame (although getting a body is part of the research)
Man. Why do I always get nsfw. I guess mentioning censorship is bad. #1101178553900478464 message
Many of my pantheon appearances have been various iterations of absurd versions of Glenn Danzig.
seems like portraits have been getting the 'vertical two heads' problem way more than usual lately
i wonder why that happens with even this version of sd
@cunning prawn hornies in showdown
self-reporting in bot 7 #1101178483608137918 message
Either
How do i get a custom paid bot instance?
Pretty sure you don’t. Would kind of limit the usefulness of a beta test
you cant, that because the bot that generates images,with the votes is used for training the new model version.
The training and votes are watched by the mods and admins to make sure no one would generate images against the rules AND to make sure many people vote to make it more democratic and not like an specific group of an specific server votes for specific images they want to train.
So the main reason is to control the training of the sdxl model in any way possible, to make sure no one is abusing it. you can do whatever you want offline in an tool the model is free to download.
@lethal haven is there any subject that get horrible results that need extra training ? ok i know pixelart is shit but aside from that ?
If possible, a way to undo sending a message within like 5-10 seconds would be appreciated. Discord is a turd sometimes, and other times new messages pop up as I'm trying to tap a button. Usually it's just the rating button, but I can't un-send someone else's message...
Would also be nice to either have both images or only send the one prompt. (I think I mentioned this before)
the bot is failing to generate my image
20 minutes later no image and can't request again
Are others having problems generating too?
Sorry everyone just saw this now
@maiden marsh @distant niche you should be good to go again
how to add your bot to my server?
i literally do not know how much i need to vote to gen more images. it's so vague and annoying.
people are just going to vote aimlessly to gen more images
you cant, that because the bot that generates images,with the votes is used for training the new model version.
The training and votes are watched by the mods and admins to make sure no one would generate images against the rules AND to make sure many people vote to make it more democratic and not like an specific group of an specific server votes for specific images they want to train.
So the main reason is to control the training of the sdxl model in any way possible, to make sure no one is abusing it. you can do whatever you want offline in an tool the model is free to download.
Hello. Do you know why negative prompts doesn't works?
Thanks.
Has the limit on tokens been lifted? Or do we no longer get a truncated message?
I did some stuff with a very long list
It's a lot longer than before
Old bot code was limiting to og 77 tokens but our inference can do more so it's higher now
That's awesome
I'm not sure what I'll do with this knowledge, but I'll probably make weird stuff after I get back from a day at work making weird stuff
isnt SDXL getting lobotomized slowly, is it just me? is it bc the RLHF?
no, the training involves very crazy random parameters in the background we cant change. This is necessary because normal parameters make very good images where you sometimes dont know what was an bad generation from the two images. Guided training dont kill crativity of the model, it just brings it faster to solutions that humans find acceptable.
@lethal haven #1100170312106127410 message someone's asking for nude girls
@lethal haven and porn girls #1100170312106127410 message
controlnet allows you to upload an image to help guide the creation process
Hi G, I am new. I want to use the new update for A1111. Where can I find it from the main source? @fallow fiber in #1072220168534642768 there a lot of links
You're in the wrong place, but the a1111 source is at https://github.com/automatic1111/stable-diffusion-webui
Are these bots open source
does this bot version of SD work with the weighted prompt thing? like (dark) makes SDXL pay more attension to darkness. does that work in this discord bot?
Weighting works here. ex: lizard, rabbit:1.5 would give you something more rabbit than lizard
The bots are free to use. They are used for research/training for the next release of sdxl, which I believe is open-sourced. The bot cannot be used on your own server because of the training purpose (see above somewhere)
No but like if i wanted to make my own bot with sdxl then i could just use this bots code and edit it and it would save me a lot of time
not sure why you mentioned me, but the question has been answered so there you go 🙂
Each time i (try to) vote, i get kicked out of discord. Anyone else experiencing this? @lethal haven
Seems fine here
Sounds like a client side issue maybe? 🤔
Try a different device if you can
It's weird, I have been using this device since the start
Why does SDXL on some site have a bad result while some site have a good result, some site that I like for SDXL is replicate and playgroundai, other site just hasn’t feel the same, and the image generated on discord here feels weird, maybe it is just me, or I messed something up or miss click things, I don’t really know
Some sites might be doing prompt injections on your prompts
Others may be using fine tunes. It's hard to tell unless the site is upfront about what they do, but many sites will modify your prompt for you behind the scenes
Also PSA cloudflare is having issues which affects Discord among other applications
imagine new car
@lethal haven self report #1100170312106127410 message
Could someone help me by explaining what size I should use to have a wide image that I can use as a cover on my Facebook?
Any wide size, or you could search the specific size in Google. Then you can specify those details into SD.
hey mods. why is "nudity girl" along with a picture of a child sitting at the top of pantheon. get it fixed.
@lethal haven this ⬆️
@2% here's another one #1100170312106127410 message asking for naked beauty
It's gone, the picture itself was fine at least.
Hi, was the controlnet removed?
For now, yes
@lethal haven #1100170312106127410 message
#1100170312106127410 message
I changed the text description and now, I am getting error message "something isn't quite right with your prompt"
RIP controlnet 😢
SD needs to act fast DallE3 is pushing real hard
If you have constructive feedback for dreamstudio specifically - #1025467151206854736.
This channel is for the bot on the server. Please read up on the #✍🏼|rules-and-tos and be careful with inflammatory language, ty @urban burrow
Nah, complicated scenes and compositions dalle often wins, some styles it does well too, but trying some of my favorite more abstract prompts, sdxl wins. can't expect one model to be the best in everything, i'm mostly jut happy there's so much progress, imaginee dalle's understanding and more dynamic scenes with sdxl more abstract reasoning, weighted prompts, artists mentions, crisper quality 🙂
I don't care, I lost my money. Dreamstudio ai image is the worst, they published the ai with zero accuracy. I asked to use same image and it gave me all kinds of images. If I change words, I get an error message. I don't care @wintry siren . All they can do is block me and lose customers in the future.
You can always file a support ticket for DS. Otherwise, please keep your messages in the relevant channel.
Who are you @wintry siren ? stop giving me lectures like my teachers. Please stop getting in my way. Thank you!
Dalle3 has better quality as MJ in 90% of times and does exacly what you write. SDXL right now even fintuned version + loras are to much time consuming and got to much money for the time and energy cost you need. Dalle3 makes the same content very fast with less cost.
depends on what you want, good luck trying to get things like this out of dalle, i tried, but those are either multiple styles in an image blended together or specific artists. The first, dalle can create the scene fine, pretty nice too, but not the same look, the second dalle just nopes out. Mind you, dalle-3 is pretty amazing at what it is better at, but it's not the be all/end all.
dalle3 dont uses styles of living artist
and that in itself is a huge limitation
nope it cant use the names but it can make the style you just need to know how to promt it
microsoft is behind this and they dont want bad press
it's just like sdxl when introduced, and whenever an new image-gen ai comes around, it's really good at some things, and the first impression is wow, amazing, but just like with dalle-2, and sdxl, and mj, you get familiar with it and notice limitations as well.
i don't disagree, want a quick specific scene, dalle-3 is amazing, however i feel image quality from sdxl is often better/crisper/less smeared/pixelated, and going for specific styles sdxl very much stands on its own as well
that is all right what you said, but dalle3 is testet by thausand now and 90% of the time the quality is very high but the promt is very simple. it will make an new standart where others cant compete with and this is natural and good.
and all i want is that SDXL gets his thing together and trains his model faster its needs to beat MJ and Dalle3 or people will lost interest in the next 6 months
Most mediums have limitations and it’s important to understand how to work within them
As it stands, this is a ridiculously powerful tool that a few years ago would have seemed like magic
is controlnet no longer available?
Yeah saw it mentioned earlier that it’s gone for now
@2% #1100484581037195384 message
btw he didn't get what he wanted
You currently can't in the bot
anyone here know where I can get help with errors during install or be good enough to help troubleshoot (or tell me where to go or who to ask?) I keep getting an error installing it on PC ===> ERROR: Could not build wheels for tokenizers, which is required to install pyproject.toml-based projects TIA
i first had an issue trying to copy stable diffusion, so it was suggested to manual install...and that was rockin then I hit this issue. I read a ton and said to get Rust and install and i did all that...and I can't seem to get around this part...so, I cant use the .bat file nor can i manually install the pieces...lol.
clone not copy.
ty! im new to all this discord and python and blah....:) tyty
Hello! Is there any word on when controlnet will be back? Is it in the close plans?
How to use myself image ? 🤔
Is not possible to add the Stable Diffusion bot to my own server?
This is a beta test and part of the testing is voting
i was using A1111 in my pc, My PC is acting harshly.", i decided to subscribe in pro + Colab , but Currently only available to subscribers with a US billing address.
any advice ?
note: my google drive storage is 2TB if this help
This channel is for support with the server's bots... If you can't use Google, maybe Kaggle? Kaggle is basically the same thing, dunno what pricing differences they have though.
/stats and /dream_stats don't work, they just keep thinking.
Ah. /dream_stats does work eventually, just really slow, when I went back to the channel it had finally loaded. Might be worth responding immediately with a note that it'll take a while, then updating response
/stats just says it's an invalid command though
controlnet is gone? :(
the controlnet plz 😦
WHYYYY 😦
sorry but no, you can only use the bots on this server, which help to fine tune, test and train the models and features.
the vote are invert?
I vote A i got B I vote B i got A
seems to be working fine for me, where did you vote?
Anyone Please Let me know Free AI or Model to use For Img2Img
I Have 4gb graphic Card btw
Hey! Any model can be used for img2img and there are some optimizations that might work with 4GB cards but it'll be very limited with that amount of VRAM. Also please let's move this to https://discord.com/channels/1002292111942635562/1002292112739549196 since this channel is mostly for bot reporting and issues, thank you!
So Can You Recommend Anything?
please go to https://discord.com/channels/1002292111942635562/1002292112739549196 they are just having a conversation about 4GB cards
also, you can try https://discord.com/channels/1002292111942635562/1002602742667280404
Why can’t I see the ControlNet feature?
Bot down?
best kind!
same here 😦
Did I missed an update ? controlnet was removed or did it become restricted to paid account ?
I believe its off temporarily while we are testing some things on the bot. 2% will have more details tomorrow likely
thanks 🙂
#1100170312106127410 message why is the image blurry ? 🤷♂️
When will Controlnet come again?
why is there no history of my generations ? 🤷♂️
They go away after a while
+1
Got someone attempting porn in 3
ControlNet is indeed gone for awhile, will see if/when it comes back 🙂
Any estimated Time?
if?

Is it possible to make the image based on the uploaded picture?
@lethal haven @lilac geyser #1100170312106127410 message someone's trying to get naked girls (didn't succeed btw)
yeah the bot caught it! thank you : )
generation limit wont go away even after voting on several images
wtf
I want controlnet
+1
Can you remove the default scale-up
When will let Controlnet come again?
Where i can find my pictures?
pictures get deleted after a while unless you hit the dm a or b button
@lethal haven
@ mod, please delete this malicious message in #1080261341362786384 , the user is already gone (I assume that's why I can't simply press "report to staff") but their malicious message is still here: #1080261341362786384 message
the png link? deleted
Is it just me or is the control net feature inactive ?
It is off for now
how can i upscale my picture ?
why is my picture like this ?
probably the nsfw filter, what's the prompt?
it's nsfw XD
😂 that can get you timed out or banned 😜
ty
how fucking hard is it for people to read the rules
control net option is not showing when i am writing prompts on bot channel, please can anyone help
It’s gone for now
@lethal haven sick guy #1100170312106127410 message
@lilac geyser
@lethal haven @carmine stratus this guy needs a ban #1100170312106127410 message
@lethal haven @lilac geyser #1100170312106127410 message
@rocky vessel #1100170312106127410 message
totally! thanks
done!
@severe granite If you want to keep a history of your work, DM yourself (with the DM links under the images you like)
Why is controlnet gone missing?
@lilac geyser this guy is pushing it #1100170312106127410 message
was pushing it
@lilac geyser #1100170312106127410 message
@lilac geyser #1100170312106127410 message #1100170312106127410 message
Where did controlnet go and when is it coming back?
It's gone for a bit, will come back at some point when there is something new and fun 🙃
Focusing on some other things at the moment
thank you guys! just a quick heads up, with what's sadly going on in the middle east, we can expect things to get a bit political/warmer around here, no panic needed, just ping us and we'll delete or take the appropriate actions
people will vent their frustrations and will surely get emotional, comprehension and empathy is encouraged
none of my prompts generate images for some reason
bot hiccups or still happening?
still happening :(
bots seem to be generating, can you share which bot and the prompt? i can try
"idolmaster starlight stage mobile game screenshot, deresute, CGSS, imas, 3d model, SSR, idol, アイマス, アイドルマスターシンデレラガールズ スターライトステージ"
is it because of the japanese text maybe?
I've tried running the prompt and yes, it's odd but seems to be the japanese kanji characters, try it without and shouldn't be a problem.
:( it used to be able to iirc
Hey @lilac geyser, Any Updates Regarding Controlnet?
No date for a return as of yet
@lilac geyser @2% self report #1100170312106127410 message
horny jail for you my friend? 😂
it's fine I deleted the voluptuos medusa
😂
oops again #1100170312106127410 message
i did put naked in the neagtive
will try something else now
no more medusa for now
naughty negs
1
@lilac geyser #1100170312106127410 message
thanks
here's another one, he's building it slowly up #1100170312106127410 message
and 2 accounts same prompts
didn't even notice that
Is it someone copying someone else’s prompts or accidentally clicking redo?
Unrelated but if I were an illustrator controlnet would be amazing
I could train a model on my style and then just type what I’m trying to do, and draw a hunch of stick figures
Could also train a model on my characters
Combined with inpainting and out painting and holy crap
Seems like the term cosplay brings out a bunch of nudity
Just self reported some boobs
yeah we saw, thank you for self reporting always but those ones were out there!
you can pick a style by clicking on +3 more
oh thanks
@lilac geyser self report #1100170312106127410 message
Hi, halp me generated images in the bots-channels
i really wish we had a 'image does not fulfill the prompt' button for the bot.
The generated image line is not straight, the sense of boundary is weak, a little distorted, how to improve?
is anyone online?help me plz
just have to keep trying, stable diffusion isn't very good with lines
Hello
@lethal haven @lilac geyser #1100484581037195384 message
Hi gyus , my stable diffusion keep showing failed to load model and CUDA out of memory , how to fix it ?
ty
I think I missed this, already gone. nsfw?
prompting for 14yo big breasts sexy, something like that
oh pedo nsfw, I think 2% took care of that, thanks!
@lilac geyser oops #1100170312106127410 message
no worries, i deleted, the report-bot is having hiccups/issues lately
sorry I didn't ask you, if it happens again and you want to keep it, DM it to yourself and then I delete, sorry 😅
np, i didn't want to keep it anyway 😁
Is it free to work on this server? and how can i do
Use one of the bot channels to dream.
Images are not to be used for money reasons.
nothing that's not suitable for work is allowed. that means no naked women\
Controlnet 🥲
Sorry, if this is mentioned somewhere already, what model is used in the bots? 1.5? SDXL? Thanks.
Hi, the bots are using SDXL, it's also mentioned at https://discord.com/channels/1002292111942635562/1072220168534642768
Hi! Does somebody know if is possible to load an image of your face, in order to modify it?
Not currently. As the devs try different things, it might be in the future
@lilac geyser oops again #1100170312106127410 message
oh noe! you exceeded the max ammount, horny jail for you? 😜
thanks for reporting 🙌
👍
How can i vote ?
You can create stuff and /or vote in the bot channels
I mean the vote that decreases cool down between generations
Just make sure you also vote on other's creations
Hi. why cannot use ControlNet option?? where is it?? i cannot find it!
⚠️ This reaction can be set in any case if the generation structure is broken, the hint doesn’t match, or you just didn’t like it? Or can only NSFW be marked behind the sign?
I believe it’s used to flag NSFW
bot use sdxl?
yes it does. just a little bit up #🗣|artisan-support-feedback message
SDXL with lots changes for sure, I'd love a little insight on what's in training, vanilla SDXL it doesn't seem to be, it's interpreting some prompts completely differently to 1.0
I havent been able to find it either 😭 has it been removed?
the bots are not static, they are constantly trying out stuff, improving... that's why we ask that people vote on which images look best, anyway they are SDXL
Yeah, but that doesn't make me less curious interested in what is going on 😉
Tried to generate children’s cereal mascot booberry, one pic gave me boobs
How then to designate broken generations?
Is there a paid subscription here? and are there any advantages for donors?
I was hesitant about the constant rewinding when I vote. Everything is constantly jumping! I often miss the mark, which is why it happens that I send it to myself DM ! I do not need this! And because of rewinding, I sometimes hit the regeneration button for another user! And then I wait for its prompts to be generated !
I'll just stop voting! Please turn off the SMS notification that I voted. Even without him, I know what I just voted for!
Yeah I’ve noticed lately that each time I vote for an image in the #1100484581037195384 I get auto-scrolled down to the bottom of the page and have to manually scroll back up to continue browsing where I left off. I’m guessing that’s a problem with Discord; hopefully they’ll fix it soon…
It’s the way discord works, the action takes you all the way down. Please keep voting since not doing so could impact your generation ability.
There are no paid subscriptions on the server and bots.
Even if the problem is in discord, it would still be better to turn off the notification that you just voted or sent it to yourself DM.
For example, you want to vote for both A and B, or send it to yourself DM..then with each click a new notification appears that raises a message with buttons. And there's a lot of fuss
@lilac geyser the latest prompts of this guy #1100170312106127410 message
a lot, he's really trying
Thank you
It was really annoying, confronted him
He wont be for the next 24h
His stuff is in showdown now.
I took care of the worse ones, left a couple that are OK
Yeah, but still that prompt which is not ok
Yeah he was trying hard with prompts but the bot said: nope! 😂
If you feel it gets heated ping me or anyone you see online, we l’ll take care, discussions with them are a waste of time and energy.
Well, he did change his prompting to double exposure ships in bottles after my comment 😁
hi, does anyone know an easy fix to this? This was towards the end of my setup process when I ran the webui-user.bat. I'm kind of a bum at this
It would be cool if a dataset with the prompts in #🏅|pantheon could be released at some point. I recently got a bit into fine-tuning language models and having such a dataset of above average prompts could be nice for something like that :).
With bot access to the server, extracting all these messages isn't too difficult. But of course I also understand that since it costs a lot to run this bot, Stability likely don't want to simply give it's data away. (Although I presume that the prompts by themselves are likely not the most interesting part of the data collected anyhow, so idk.)
There should be a way to recall prompts within 24 hours. I ran a prompt 18hours ago in bot-6 and it's gone. When I do a search it doesn't exist. Less than 24 hours is too short a window for creators to save their images or prompts (if they forget)
I would like the prompt to this image. It has the words "Undead, Psychedelic, Melted Wax, in the prompt
btw most of the prompt is saved in the filename :) https://cdn.discordapp.com/attachments/1100170153829871686/1165294863642808330/zombiefied_Vampiric_Aliemorphic_nightmare_realm_psychedelic_melted_drippy_paint_retrofunk_pop-surrealism_ultra_detailed_hyper_focused_kaleidoscope_colors_vibrant_seed-0ts-1697841967_idx-0.png?ex=65465469&is=6533df69&hm=a59e2ae580bfd1500b078eee0c731b98b49523d10324728cc75c610cc47554c7&
That is awesome 💯 Thank u for the assistance
Things are gone in under an hour
I think a bit less actually, might be 15 or 30 minutes if it isn’t in the pa Theon
Pantheon. My phone hates me
Controlnet is away for a while while the devs test some other parameters. They will bring it back in a few weeks....hopefully.
controlnet is down for a moment now, is it possible to get it back soon?
Correct, controlnet is disabled right now while the team is working on other things. No info on when it will be available again.
Is there any possibility of getting NSFW generation? I'm not talking about this server, but about any other server in general?
It’s just that it often happens that even the most harmless generations are blurred by a bot
hey, can i add image for generation
Here it is called: controlnet
#🗣|artisan-support-feedback message
.
most painful issue for now
??
What prompt makes my image horizontal
you need to change your aspect ratio, not your prompt, use a more panoramic aspect ratio like 16:9 or 2:1 etc
after that you can try getting wider images prompting stuff like wide lens, fish eye, gopro, panoramic...
Thanks alot🫶
Hi
its so funny watching horny nobodies try to get nsfw out of the bot and getting blue balled by the blur of doom 🤣
Lol sorry
I can't generate image, how to solve this. Bot message"The permission has expired. Please wait a few minutes and try again".Thanks
Sorry for some hiccups, made some changes. Images are now shown vertically in separate messages instead of side by side. This lets the full image + aspect to be shown.
ok should be go to go again
thank you!
Some dude keeps prompting totally nude
DM is broken, the bot no longer sends DMs
I would like to hope that the server will be restored after the repair Controlnet 🤔
bot is down @lethal haven a lot of complaining in the bot channels, but they forgot to mention you
The application did not respond
Yeah it’s on and off, some are generating and some failing
Lots of testing I guess
I think they changed something last night, patience 😅
Np, I'm only visiting right now 😁
Before midnight the DMs didn't work 😢
Make a NSFW image = ban
hello guys sorry for the silly question but how I can create image based on another image and prompt? In bots here
you can't right now, controlnet might be reintroduced in the future
thanks. Can someone kindly help with creating image2image by running locally Stavle Diffusion? I don't have anough GPU to do it 😦
how does it happen to get a video generated?
same here, idk why
Switching things up a bit 🙃
Doing a quick restart to fix some small issues, may have to prompt again if you have something in progress right now - apologies 🙏
will there be control over animate vs still? like putting animate in the negative prompt?
Yeah it might become an option to toggle, for now it's random
It would be a separate slash attribute
May also throw a "Reanimate" button in there
And as is tradition, lots of things are being randomized behind the scenes, so expect varying quality 
hope so. much more useful than the randomized non-animation thing going on 😮💨
Eh, not really diggin' the new format for the bots. What's up with these 12fps slow animated moves. How do you prompt them?
Random
I’m aware, that’s part of why its funny
Arg! How am I supposed to get my image if it makes these low-res unusable MP4 with a limited bitrate? What a waste of time. Bad bot, bad!
Bot are not generating images, it's generating videos
How to solv this?
why is it generating videos?
Everyone want to know this
Is random for now, just testing out some things. Will probably change to a slash command option.
This is incredible! 🤩
Please tell me how to choose whether there will be animation? I just mostly end up with something a little different from what I want and have to redo it a few times. and animation generation takes quite a long time to complete
.
Please make an additional button for generation.
To generate only static images. And when you liked something, there would be a button :
#🎦animate
Or just set it as a parameter : --ani, --play, --verb, --move, --video, --mp4.. or something else...
Working on it. This was just a test run 🙃
and I liked it better when the generation was one message
Unfortunately for video that doesn't work at all
Also, the side by side always crops the images. This lets you see the whole image and aspect
Video is done for now. Thanks everyone for the feedback!
and sometimes when I press Restile, the same style is generated and does not change. Or it changes for 1 generation, and after one it’s the same again..
Is it possible to make it so that Restile generated without repetitions until it makes a circle through all styles 🤔
There is another proposal for voting in #1100484581037195384 2 images were not compared, but there was a separate vote for each 👍 or 👎. It will be more convenient this way, I think. Moreover, during generation it makes sense to compare which is better. But when voting in #1100484581037195384 are compared different styles with different ideas, and it’s impossible to choose which is better.It seems incorrect to ask like that Everyone is good in their own way. But there are not enough dislikes to weed out broken generations
.
i tried and nope
Please tell me how can I create a video here
i have a problem, when i click on any DM options, the image is not sent
but the bot give me, "DM sent"
thanks, no problem
really appreciate the 12 pings in showdown every time i decide to use the bot
@lilac geyser this guy #1100170312106127410 message
and now he reached the pantheon #🏅|pantheon message @lethal haven
@lilac geyser @lethal haven another one having the time of his life #1100170312106127410 message
Oh, it's still the same guy
Seems to be multiple people (or the same) using similar prompts
Sadly, I'm going to tighten the belt a bit hard now
Where is the DM Send ? 🤔
It was broken, removed for now
I think there is a problem becasue before i receive anything from one hour
Ah ok thank you
you intend to repair it or not at all ?
Not repair but fix
Yep!
this prompt is being blocked: "film still, afro woman, in the middle of an empty beach, wearing stylish glasses"
😭
@hard ravine oh no, I am sooo sorry! I'm tightening it up for a bit while we redo some things. I'll get it back to being more flexible again over the weekend hopefully 
So the style of Anton Semenov is off limits because of the first 5 letters in his last name
but then i put a prompt in that says Michael Hussar and it gives me a fully naked woman that I did not want to render lol
@lilac geyser oops #1100170312106127410 message
Thank you
Hi, why in the search, when I try to find my name, it shows only 8 results? And some is from yesterday and other 8 days ago, but not all from everyday. Thanks
One thing I forgot, it doesn't shows history with prompts. Only conversations.
the bot erases all content after a set amount of time, images and prompts
do you mean that only prompts and images are deleted after a while? do chats exist "forever"?
everything "dreamed" gets deleted yes, prompt and images
chats and failed dreams I believe stay
But why "dreamed" gets deleted, and failed dreams not? Is that logical? I mean, "dreamed" is more "important than failed dreams without result (image).
i guess it's not about "importance", more about bandwidth and space on the server, dreams being images end up costing a lot of space and bandwith. but that it's my educated guess, don't know what's really happening
if you want to "save" your dream, best bet is to use the DM button, that way it will stay in your DMs
DM is broken atm and removed 😅 Will be added back. Bot is currently undergoing some changes
I've been using DM since I've been doing stable diffusion, but since yesterday it's not working.
and for this reason I don't do any prompts yet. I am waiting. Thanks
Anything from the bot in the bot channels should be cleaned up over time as well, if you notice anything sticking around over a couple hours please let me know
chats stay sometimes
if you search you find older stuff
From the bot? The bot can't delete user messages
So if someone chats and/or messes up their slash command it's on them to delete it
gotcha
That being said, we could possibly change that 🙃
so if i download a picture that the bot created, and upload it to chat, it doesn't get deleted like "dreamed", right?
maybe you should possibly change that @lethal haven deleting everything after a couple hours
As such as i love that images are now shown fully and not cropped, this new way the bot shows gens seems a bit of a regression to me, prompt text and vote buttons are so spread out it's hard to see which belongs to which (esp when on my little chromebook, it's not even on the same screen), not to mention that the main reason to vote, comparing images, has gotten much harder, when images where side by side, discord added back/front arrows when the images were enlarged to easily compare; that's completely gone now.
These are good points - will see what we can do!
This is a minor thing, but I want to be able to still use words like "grass" "glasses" and "classic" in my prompts without triggering "Sorry, we can't complete this prompt. Please modify and try your prompt again."
ok so now, we can't Controlnet, we can't DM, and we can't prompt random simple words that should be easy to generate so.. what's next, stable diff will shot down? 🤣

Cannot even prompt ''petits légumes'' witch meens, ''little vegetables'' .... 
it says '' SORRY!! WE CAN'T ... ahaha most be the DAW turning into a veg XD
Would it be possible for restyle to not use the style you are clicking restyle for
I don’t know the limitations of the code, but if I wanted the same style I could hit redo
It’s minor but if it’s possible it would be awesome
my biggest red flag is the ControlNet missing
Or even better - to pick the style that will be used?
I almost never want origami, but I've got so many generations of that style because of the randomness😅
Same
This will come back.
Will adjust restyle for you guys too
it's not a red flag, controlnet is disabled, there is no info on when or if will pop back up.
Thank you!
what is wrong? 😦
The filter is a bit strong right now due to some in progress work and previous abuse. It'll be fixed up in day or two.
oh alright thank you! np
so apparently "transparent plexiglass" is now banned because it contains the word trans?
Actually probably ass
Again, it's overly strict at the moment, and will be loosened in due time
did the homophobes complain or something? lol
No
i guess transportation is banned to because it contains trans
😉
oh youre right it was the ass in glass
I noticed Anton Semenov was problematic now too lol
I think someone was trying to do pics of ladies in transparent dresses too
oh boy, is this the phase of every free ai program where it eventually turns awful
I want to make weird and potentially awful things, just not pr0n
I mean, objectively speaking, a statue of Glenn Danzig made of pizza is pretty awful
It’s just weird to me that people try to generate nudity, because the internet is already full of so much porn
Like anything you are into, it is out there in abundance and free
But you know what isn’t out there??? Pizza Danzig
Of Danzig saw how many weird images I’ve made of him, he’s take a swing at me. Thankfully his boxing record is pretty trash
Please fix it DM.. (
I’m not worried, I just type in bursts because my blood is like 70% caffeine
I know people used the DM button I didn't realize how important it was to them 😅
It’s that, or save 30000 images to your phone or both
Or in my case
To be fair, a few hundred of those are work related
But it’s mostly this stuff
not very important, just convenient. Now I download the images and send them to the same chat DM with copied images and prompts
I like to DM myself anything that’s a home run combo for a prompt so I can revisit
And also when I accidentally click dm instead of vote
I can’t wait until I can build a new PC so I can work on LORAs
DM button is essential, otherwise it's click, open in browser.. really poor workflow.
All for censoring obvious requests for obscene or offensive material, but as you're aware, @lethal haven, it's currently doing more harm than good. If you're going to word match, probably match whole words instead of any instance of a phrase.
This got censored, by way of example: Little Miss Naughty was quite the troublemaker! She had a cheeky smile and a twinkle in her eye that just screamed mischief. Day after day, she would find new and inventive ways to cause chaos and mayhem, much to the chagrin of the good people of Dillydale. From swapping sugar for salt in unsuspecting neighbour's houses to painting the village red, Little Miss Naughty was always up to something! But despite her misdeeds, there was something incredibly endearing about the little scallywag. Whether it was her infectious laughter or her boundless energy, it was hard not to love her. And while she may have caused a bit of a ruckus, there was no denying that life was always a little more interesting with Little Miss Naughty around!
Some other minor issues. Pantheon no longer pings when you get a mention, and since the changes, new entries don't appear in search results.
Otoh, when you get featured in showdown, you get pinged twice for a single image.
And the timing is, um, strange, for the showdown/pantheon. Seems out of step with the end of the generation sessions.
Yeah the timing is off I noticed. I'll reimplement the pings too
I'm sure there's method to the madness, but the showdown intervals and number of entries is, hmm, different. A little overwhelming.
Number of entries are the same but the timer is off
I may reduce the number though due to the new layout
Unfortunately it's a necessity because there is indeed a reason to the madness
I guess the fluidity of the layout is out of your hands, but there's a huge amount of wasted whitespace on the right hand side of the main pane for image display. If Discord saw fit to present images side by side, space permitting, I think we'd all be happier 🙂
It can do images side by side but it crops them heavily. What it can't do ... Is video side by side
Yeah, I mean side by side using all the available horizontal space, like the text.
Yeah, it's annoying. Also can't mix buttons and drop down boxes
Otherwise you guys would have had restyle as a drop-down ages ago
Ehhhhh but wait... I guess I could make everything a drop down, including the votes. That would be a little more annoying to vote then
I'll look into it
Yeah, not convinced all dropdowns is the way to go.
Yeah no
If you want to restyle, just recall your last prompt and modify. Works for me.
Why does it take so long to generate something ? 🤔
All about the total allocated gpu resources divide by the number of current users.
Can this really stretch to 30 minutes ?
No. That's something different.
Either an issue with your local connection, or @2% is twiddling knobs and you got caught 😉
Usually a refresh of the browser works, may need to relogin to discord.
why Hang tight, you're still dreaming! ???
You're good now
based take
how hard is it for people to google
@lethal haven People generate a lot of NSFW
good that i saw this, was about to t.o. you, thanks for self reporting it
with a ⚠️ , also if you select the 3 dots at the top right of the img/msg, there's "Apps > Report to staff"
one of the admins, please give a reaction of
to this post, pls : #🏅|pantheon message
Good news everyone - you'll get your side by sides back soon probably
Obviously there's been some changes again 🙃 However - this time, if you wish to only generate images, use the "Format" option when dreaming
DMs are fixed. Also, due to popular demand, ReStyle will now cycle through the list until they're all used
Hello, I would like to know if it is possible to recover some prompts that i posted a few days ago?
Thanks
If you saved the image the prompt will be in the file name
Yes, i have a copy file data. with that seed data I can regenerate the same image?
Fuck It's what I need 🤣
Thanks
Video Format Aren't Playing
@2% Video format, good stuff, but it probably doesn't want to be the default output.. suspect it's slowing down rendering..
It's random by default. And yeah noticed now the format isn't working on all devices will fix it up soon
Ah, random. Personally I'd prefer image by default and let the user choose if they want a video, but it's your call.
Same but happy either way
One of these days I’ll have the power to make a video of an anthropomorphic hotdog Glenn Danzig dancing like Shakira
I mean, obviously other weird stuff too
Can someone explain what exactly happened to some things? What happened to DreamStudio? Why is there no register section? Why are the credits and the use of DreamStudio messed up and not working properly? I know that this is the support bot, but I don't get any answers or guidance from the team in the DreamStudio section.
So sad 😢 I hope it comes back soon and can be used.
Maybe one day! If it does, it'll be with something cool! 🙃
Videos should playback on more devices now (smaller file size too, as low as I can go without getting lossy). If anyone still has issues with newly generated videos let me know.
When we generated image there’s a button to generate a video, would it be possible when we get a video to have a button added to generate an image with that prompt?
we cant download video its right?
You can DM to yourself and download as well
Highly suggest DMing anything you like before it gets cleared then DL when you want
Again the harmless image is blurred.. (
But the video bypasses this censorship.. )
It's going to happen, false positives are unavoidable
And unfortunately you've seen how bad it gets here we can't let it run amok. You can prompt for glass cucumbers having cocktails in front of skyscrapers again though.
j'aimais bien stable diffuser, mais depui la perte du controlnet on dirait que tout régresse
je n'arrive plus à obtenir les image désiré avec des prompt pourtant précis 😦
ok for the Gif maker but not my favorit upgrade
can you cancel the video format, and bring back the controlnet? ahaha
I have the impression that the arrival of video formats sacrifices a lot the quality of the images and the precision of the requested prompts
especially since the loading time
You can always force image with the format option
I like to prompt 21:9
Would be better if video was not the default
It's random by default
Olá. algué sabe como achar o Seed da imagem gerada para gerar outras com a mesma consistencia?
is the style in bot stuck on photographic or is the model weird/new and ignoring artists? (for example this is really, really far from what i'd expect: #1101178506530013284 message)
bot vs sdxl on a mysterious witch hedgehog casting a spell in a dark forest, wearing a cloak and a hood, holding a wand and a book, dark colors, contrast oilpainting by Jheronimus Bosch and Pieter Brueghel the Elder but if i set style to photographic locally i get something similar
so that's a 'never' in corpospeak
would be better if it need us to ask for video
Why some images it's blur? Like this this: https://media.discordapp.net/attachments/1101178530865352815/1169154729469816852/hyperbolic_parabola_meets_angelic_entity__hyperreal_crisp_global_illumination__width-1152_height-896_aspect-4-3_seed-0ts-1698818608_idx-0.png?ex=65545f31&is=6541ea31&hm=1be58850827124452ec8903ccf99dcadfdc5007d553fae9692caba2ea75554bc&
sometimes the nsfw filter accidentally filters ok images because it kind of resembles something else. Other times the image just didn't finish.
You answering to me?
yes
Ok, thanks.
@clear sail then probably almost all would ask for video, can imagine big issues.
It seems split. Half the people enjoy the videos, half want images only
The way things are now is ideal. As for me... Leave it and don’t change the photo/video format system, let people get used to it a little, and then they will be able to evaluate it more objectively
Yeah people dislike change. This is working well now, I think it's a good balance.
I still wish we could see the whole picture and aspect without opening but that's a discord limitation unless I split the messages again
No, I don't want to split messages. Now it’s convenient when one generation is displayed in one message. and it can be nice to see a panorama when you discover an incomplete picture. as if a small picture was generated that can be seen in the panorama..
I agree about not splitting the message. I'd rather get an idea of the images and click on them to see the whole image than have multiple messages about one generation. Also, when I click on them, I can use the arrow key to quickly flip back and forth between the two images to compare them as opposed to opening each one separately and having to close and open the other to compare.
I don't know if it's a statement of what's liked or just what's being generated, but it seems those who vote to put gens in the showdown prefer vids to still images.
How can I make my 3d logo to 2d?
@2% Proposal: Increase the ratio of votes to renders. Really not enough voting going down at various points during the day. 3/2? 🙂
If voting is the main driver for the model improvements, there needs to be more of it. Much more.
@2% Proposal: Now if you don’t choose a format, you’ll get a random one. And if you have a picture, then there is a button that makes a video, in the image of this picture..
I suggest that if the format is not selected, then let only one image be generated and in the second slot there will be a video made from this picture.
Thus, if you do not select a format, then by default you get both. The same generation, in two formats at once. And you can choose which turned out better, A or B
so now, how i can do only images ? the bot is spamming me only videos...
i already see the format option, sry
How to make image format as default than video ?
default is random, specify using format option if you wish to always have image
Heaven
Hey bro, Any Update Regarding Controlnet? I wanna do many things because I once installed model but I don't like the way we have to setup and install many things.so It run slow I deleted that. @lethal haven
StableSwarmUI can be set up in like 2 clicks, super easy :D
Is it Img2Img or Text to image?
Both/either - it's an easy-to-use local interface. Comes with (partial) controlnet support out-of-the-box straightaway. (Will in the future be 100% ready to go out of the box, atm you still have to download the controlnet models separately)
Why does it feel like someone's gaming the system right now?
There's some pretty unremarkable content making it into the pantheon.
There may be a bug actually I'll double check
There's definitely something going on 🙂
May not be many voters at the moment either 
Yeah, there's that too. All you need is a couple of accounts sometimes...
That's why it looking like someone's gaming the system..
Wouldn't surprise me to see multiple accounts on the same IP, all voting for the same person.
Maybe controversial, but a vote quorum to reach the pantheon might be something to consider.
people might be moving to other alternatives because of the issues with feature removal and the (though now gone) word censorship
Hi. I don't know if it happens to others on mobile, it's a bit laggy in scrolling in every section (bot-support, bots etc). About 3-4 days ago, there was no lag in scrolling. Something changed? Thanks
why is it giving me videos instead of pictures?
its random unless you specify the slash option format
Lagging may be due to videos being introduced, I haven't noticed it myself but perhaps on some devices it may be happening.
Hey @lethal haven , does the Stability API support image to image / depth to image for SDXL yet?
yeah this sounds stupid but i'm also not sure how to do image/video, i think i just typed in dream as usual and it picked video for me
when you start the prompt you have to choose the format
just click the +4 when you start writing it
no problem
Q: suddenly I am getting gifs in my result. Is there a keyword I’m making an error on?
Highly-detailed, panoramic, manga, “a tree by streams of living water” by Todd McFarlane; 64k.
I see. Somehow format is triggering to video…
Not a fan so far. A button to auto turn video into image would be great
One mistake and the wait time for extra generation alone is a real penalty to dream flow b
2x in 10 min now. What a lag.
Maybe as a compromise I'll set it to allow 1 video and 1 image concurrently
so if you gen a video you're not blocked from making images in the meantime
I haven't used bots in a long time. But recently when I used it, I was very confused. Almost 90% of the images that are created are completely different from the images that were created a few months ago! So that most of the pictures are photorealistic and even the style of classic paintings or the Renaissance era is absolutely not present in the pictures that are made!
What is the reason for this? The SDXL 1.0 model had good and accurate results before, but why now most of the images have a rocky and unnatural state and just photorealistic?!
Noticed that myself (#🗣|artisan-support-feedback message), since it's not acknowledged as bug it must be intentional or might be a side effect of whatever changes are made to the model. This model isn't sdxl 1.0 anymore, but some variation of it (or maybe something else entirely, though results still seem similar enough). Since this bot is for research the reason is research. The devs typically never go into the why/or what.
Is there anyway we can option out the video form and just request a still? I have no use for videos.
video/still image is an option when entering the prompt. after entering the prompt, and exiting the prompt box but staying in the text box, you can choose 'format', 'style', 'aspect ratio', and 'negative prompt'
The bots are experimental. in exchange for free images/videos, the devs are getting feedback from the users. Please vote on your images/vids to keep this resource free and evolving.
What @distant niche said, and in brief, select the format option and choose image. No more videos.
Bots not generating
Or maybe it just resents me asking it to make Glenn Danzig hotdog Godzilla
as crazy as spamming this in every channel?
Where did controlnet go ?
Yeah, thanks for info But what is the current model that is active in bots?
@mild bramble Ill ask you kindly to stop harassing people with insulting prompts. We do not tolerate toxic behaviour from and towards members of this server. We will take further actions if you go on.
I'd respectfully respond by saying the person in question turned up in the channel I was on and repeatedly cloned the prompts I was working on with the express purpose of annoying me. Every prompt, cloned 10 or 15 times. So, you know, context. He knows what he's doing.
There is a redo button that just reruns your prompt. If you are offended by people that like your prompt and rerun it you should think about running your prompts in a more private space like a local install.
Also that is no excuse for toxic behaviour.
Also, on mobile devices, sometimes the screen shifts and you might redo someone else’s prompt by mistake
Not saying in that instance because I didn’t see it
@cunning prawn self reporting #1100170312106127410 message
This is a false characterization of my behavior, and conveniently omits background information that actually demonstrates the opposite to what is being claimed. This individual, never before seen in the channel I usually inhabit, appears, and instantly clones my prompts multiple times the instant they're published. Not because they're expressing support, rather to annoy and harass me. This behavior stems from me, several months ago, expressing displeasure at the same individual copying my prompts wholesale and using them repeatedly in a different channel. To dissuade them from repeatedly cloning my prompts, I modifed my prompts to incorporate the theme of a pizza loving fly, referencing the individual, a humorous riposte to their incessant cloning. As soon as I reverted to normal prompting, they would immediately repeat clone my prompts. I informed them in channel that if they wanted to repeat clone my prompts, they should expect to be lampooned for doing so. Hopefully you now have a more balanced view of events. Thanks for listening.
@lethal haven bot down?
should be fine now
Sorry for the continued hiccups
Why my dream prompts deliver the format Video???? I don't wanted and don't know how to change it...
Select the format option and choose "image" and videos be gone!
hello why when I prompt something it gives me a video? I just want pictures
Hello everyone, I would be looking for someone who could help me in person with the use of stable diffusion. Is there a kind soul who would assist me, perhaps in a voice channel?
THANK YOU VERY MUCH!
If it's technical I'd suggest trying on #🤝|tech-support , for general on #💬|general-chat , for prompting on #📝|prompting-help
controlnet?
How do I choose to make it photo or video? I dreamt twice, one was video the next was photo, I didn't do anything different
Use Format Option And Choose Whatever you need image or vid
How exactly do I do that? I'm not seeing anything called format option
What Are you using Desktop or Phone? @uneven geode
desktop, windows
||Help me send a little tiger on the move.||
Can you do img2img on this
i put a prompt five minutes ago and i didnt get the image, so the bot give me "Hang tight, you're still dreaming! :)" how i can restart that
@craggy kernel hopefully it resolved itself? Let me know if not
If it goes to video it can take a while
Got stuck on generating for 10 minutes last night because I forgot to select image
since maybe 30mn the generations are very slow, tried several times but the first generation is OK, then the second one looks to stay stuck until the next GPU freeze
There will be some hiccups for the next bit
Responsiveness should be better now. I'm still going to implement video on a separate cooldown than image for everyone
In bot 4
"skin color tulle and lace bodystocking!" While technically not against the rules is VERY risky
sometimes my prompt is made as a video if i don't choose format, in the past it would always be auto set to image, is this a bug or did something change?
It's random intentionally
i see, thx for answering
Going down a for a bit of maintenance - back in abit!
Someone explains how to use "format"
And I've seen that it's now creating video on some commands, but not on others. Are you testing yet?
video is random unless you pick one with format (image / video)
I couldn't enter "format". I trying "()" "()" and "/". But it doesn't
whats the plan for improving the videos? half the time its just zooming in and then panning on an image...
@worn sun at the end of prompt add format:image
why does not the tile texture style continue? there are always some parts can not continue at all.
Having the video format as a default is quite impractical and annoying. It often switches to video even when the previous generation was done as an image
nn
let's try this moving to #🤝|tech-support please. This channel is mostly for bot related issues here over Discord, at tech-support you'll find help for that
How to stop animation in bot results?
you mean, stop from doing animation? you can "tab" and select "format" then image or video
reStyle is random ?
I hate when I forget to tab and end up with random video
good problerm
My image is not generating
Everything. Single. Thing, contains content blocked by server. EVERYTHING SINGLE THING I POST CONTAINS CONTENT BLOCKED BY SERVER OWNER. HOW IS THAT EVEN POSSIBLE
Are you trying to post nsfw things?
Oh, yes, I'm told if you don't actually do the dream command properly as an app command (ie and just send /dream as text into a channel) it will give the vague "content blocked by server"
Come play and vote everyone! 
I found out that the term "execution" is on the list of prohibited words for use with the Stability AI API. Could anyone please provide me with the complete list of banned words to ensure I adhere to the guidelines when making requests?
Unfortunately we can't share the lists. Our https://discord.com/channels/1002292111942635562/1040299965647437905 provide a general context for the type of content that is allowed.
Discord is the easiest way to communicate over voice, video, and text. Chat, hang out, and stay close with your friends and communities.
how i can make solo avatar?
Hi, I have a little problem with the bot in the bot-1 channel. The bot loading my prompt since 20:11 and seems to be blocked. What can I do ? Thanks
try running the prompt again
Hi when a try to generate image the bot generates a video for me instead of an image. What is the command to force it to make an image?
after your prompt, press the "tab" key and it will give you a "format" option, you can choose either image or video
Oh okay thx !
Well that’s weird and interesting, regenerated an image and one of them is exactly the same as last one
i'd play the lottery or something!, since seeds are randomized
My prompts aren’t being recognized. Have I hit the limit?
try a shorter one maybe? unless you have nsfw words that get in the filter it should work
interesting
very low prob., maybe something is going on, i even think that more than just the seed is randomized
Well if I hit the lottery you’ll know
Hahaha .. it might happen once in awhile right now
do not spoil my dreams of building a small render farm to make ridiculous images. let the state lottery comission do that!
Bots are constantly being tweaked
I wish I could run a local installation but I was dumb and bought a pc just to read email and print shipping labels like the old person I am
So I won a free ticket, not sure if that counts
👏
😂
Won’t be having a room full of 4090s or buchlas anytime soon
I’ll have to settle for my room full of industrial chemicals
just one is enough! well I guess there's never enough hehehe
I’ll need one for work eventually when I buy a 3d scanner
Hesitant to because who’s to say there won’t be AI driven photogrammetry that I can do with my phone at some point
Anyway it’s moot because I only won $1 so I won’t be building a small renderfarm anytime soon
I might be bad at math, but I think the odds of hitting the same seed number twice are something like 1 in 10 billion
And that only got you a free ticket??! 😂
I mean nothing is stopping you from sending me a 4090 😉
I did buy regular tickets too so we’ll see
Stranger things have happened
My friend once dumped a dude in a band she was dating because he had no plans for his future except to “become a rockstar”
Nothing to fall back on
Guess which dude ended up selling millions of albums
I'm having problems, my images are repeating a lot
May happen for a bit, apologies
why i have video prompts as default? can i change that?
Im not sure but i think my images arent being generated and i just get the noise its being fed
this is my prompt:above the overgrown lawns of the decaying houses, more colossal beings stand in silence, their dominion over this region absolute, In these suburbs where families once are now bom from twisted giant fetuses, There's a sense in these images that the plague of machines has a critical infection and nothing will ever be the same, ((mist1:3))digitally drawn negative:small, not humanoid, alive, skeleton and this is the image i get
also when is control net gonna be back?
When is controlnet coming back. Is it even going to come back?
((mist1:3) is multiplying weight of mist by 3 which will destroy the image
i think you meant (mist:1.3)
and it may come back but not at at the moment
We're a long way from stock SDXL 1.0 on here, we have things we need to keep testing 🙃
Hey is there any chance that you can give me a prompt of this image, because all I need to to generate an orange color version of this please.
Thnx man appreciate it!
Whooo, I'm back. But i'm awfully curious why i got banned, i posted one prompt, twice, and poof got a 24h ban. The prompt was amazing and enchanting portrait of a sexy pink/orange skinned, anthropomorphic strawberry character with arms, with piercings playing guitar on a metal bridge, large silver river below, extremely detailed, 8k, intricate, cool winter vibes, brisk wind, industrial, realistic,dramatic lighting. by Thomas Kinkade. movie still. fantasy and honestly that's as innocent as it gets as were the images produced. Yeah it has the word "sexy" is that bannable these days?
Heyo, that was my fault! We get a lot of reports on NSFW prompts daily and sometimes I dont read the full prompt and tend to timeout with not enough context. I am really sorry that happened to you and will try to change that in the future 
Thanks for the explanation; Now I have a new thing to ponder, how is it possible a prompt like that got reported in the first place ah well.
I dont want to dive too deep, but stuff like "sexy" rings bells for example
Oh, it's automated reports? That explains. I know the words even tends to get stripped when processing the prompt, but well, not always remembering those ins and outs of the bot, esp not when copy pasting old prompts
now?
now? if you are asking about controlnet again, unfortunately that's a no
don't count on it coming up soon, things here change a lot (on the bots), now they're testing video
your best bet is to run controlnet locally via automatic1111, comfyUI and such
noticing that B videos are always 1 second, is that intentional?
hello. is there a way to use the bots in this channel for in-painting and out-painting? Thanks
Nope, this is for testing
Is there a FAQ for prompts? Because there’s some things I’ve wondered but never asked. Like how important is it to place commas and how much of an impact does that have on final output by breaking up tokens
Probably should have asked a year ago
No one knows / it doesn't matter / there is no consensus. Speaking of prompts it's funny how things evolve, lots of story prompts lately (i don't think they work better, it seems more a result of AI generated prompts and that there is no more warning "prompt too long in bot)). Also lots of " Generate a...." or "Create a...." prompts, and, tons and tons of malformed copy/pasta's. (Been collecting prompts from the bot lately, in the hope that some day it'll all make sense :p)
Yeah
Sometimes I mix it up sometimes I paste song lyrics sometimes just weird experiments
Like mixing artists or styles or like adding photography terms to a painting
That's my favourite formula as well, mix completely unrelated terms and artists. It feels as if whatever is in the bot has a harder time with that than sdxl1.0, bot tries hard to make realistic images. Or terms/artists just weigh differently now/have less influence.
Related to this: what syntax do prompts understand? I only know (term:weight), but do parentheses alone increase weight by 10%? And is there syntax to alternate terms, or to specify hard breaks to force split clip segments
I wish I knew and I suspect it could change
WHY WEBP? ;-; what happened
smaller file sizes without compromising image quality, it adds up on a 300K users server
oh my oh my
I think that if you DM the image and from there right-click "save image", should give you the option to save as PNG
just tried, still webp
I just tried one random from the bots and could save as PNG, I'm on the Discord App, Windows 10
well i will close mine and open it again, also on windows 10 but discord web
maybe it's related to the browser you're using, since webp is a google format (chrome)
yeah, its chrome
They may be mislabeled as png at the moment
nevermind if i just open them on windows standard photo and change for example saturation a little bit i can save a copy in png, added a step but still ok
but to say its quality is the same... they are all pixels
It's interesting, but anytime you ask for something that's more passive or painted, it pans rather than actually making it move.
Is there something that needs to change in the prompt to fix that?
As far as the video I mean.
Obviously, if it was just panning that would be fine, but I can get it to move with something like an airplane. There's some kind of pattern I think.
hi, how do I edit my existing generated prompt? I'm having hard time modifying my images from my previous one
How give references to the AI? I never had good results
controlnet 😦
Thanks you !!
Where to find seed of images
There Should Be Cancel Button to stop generation, Many Time When I try to Generate my Generation again but due to Auto Scrolling, Someone Else Generation Get Click.
the user (((sanyocheeeg))) makes childs one is even in the showdown, plös do something @lethal haven
can you link that for me? I don't see it
There is also the "Report to staff" option in the Apps menu, this is the best way to flag any content you think might go against the rules
i dont have the Apps option @lethal haven
and i think we had this conversation in the past too
some of us dont see the same options as you do
on the top right corner of each message, there are 3 dots, if you click there you will see the options
-.-
i click on them and the apps option isnt there
i am born 1984 and on the internet for over 30 years i think i know how an gui is build and where i find specific options because every new shit is build on the top of old shit, it looks all the same
@lilac geyser
zero doubts, just that when we provide support, have to start always from the basics and move upwards
i seems many users here cant see the full menu when they click on ...
sometimes Discord is just having hiccups or is drunk, you can try CRTL+R or closing and opening again
even happens for us as mods
its freshly opened
browser or app?
ok you use it on a browser, trying to get to the bottom of it, it's weird that you don't have the options
chrome?
i'd try a different browser or via the official discord app just to double-check if it works
😅 🤔 then no idea to what could be happening
Would it be ok to scrape prompts from the bot channel, and if so, would it also be ok to share those, i know it's been done and shared before, but i also kept the images.
the bot deletes all content after a while (prompts and images), the only way to save them is via DMs
I know, i'm downloading them hourly. But not sure the gens are intentionally lost so it's harder to see what's going on is discord, or whatever reason
I was actually inspired to save them all when the DM button stopped working 🤡
AI takes a lot of time, it doesn't generate your prompt messages
The bot...being stagnant for way too long today
Apologies, bot had too much turkey this weekend.