#🗣|artisan-support-feedback
1 messages · Page 6 of 1
/DreamA girl with wings
This may be related to the Showdown being broken (empty), but when I look at my dream_stats, it's only showing me as having 76 Finalist Rounds, when it's actually over 1,000 and it's always shown the correct number until now. I hope that gets corrected because I want to keep increasing my Finalists number. I think the Pantheon number shown is correct.
Showdown currently borked.
@vale sable in bot channels you need to make an / and select there / dreams
than you can use posivite promt text
@mild bramble ok itested weights in bot 1
everytime i tested it an "long promt" error comes up
is that normal ?
and ok i must admin it,weights are very cool @silk acorn
Long prompt is only normal if you're using a long prompt, each weight is an extra token.
weights being extra tokens is not how weights work in any other implementation I've seen
Suggestion: how about adding a button to say that both of @hushed skiff 's outputs are good or both are bad?
Got cartoon boobs for no reason and self reported
dreamer caught horny on main, forced to be whacked with a newspaper
If it were helpful I would not have the question
This is Vanripper's artwork
This is what happens when the phrase is part of a prompt
Prompt: "girl, vanripper"
Adding "vanripper" inexplicably adds super rad improvements to any output
What I'm trying to do is figure out why
The idea has been floated already and the devs didn't want it at this time for the bot channels
Ok, got it
the ai does not have access to every artist's art style ever. that's why loras exist
Yes, correct. Does not answer the question I am asking. Your answer would be applicable if I asked "why doesn't it do this" as opposed to why it's doing what it is doing.
So is the randomizer beoken
Broken*?
im old and can probably google it, but i have no idea WTF a vanripper is
It's a person
Thus the mystery
What exactly is the system doing to interpret what it is?
It's a good puzzle.
Yeah
Scroll up, he has a distinct cartoon style, but the system makes a pretty particular aesthetic that's almost the opposite. I'm curious as to why, because frankly the results are rad. It's like an easy magic prompt improved
Improver*
Seriously, try it with whatever prompt
frontal exposure made it into the #1100484581037195384
why are the two images so similar now 😭
there are things that exist in other languages, but also if you write something it doesn't understand it usually does something random. like if you type "jfghdsgjhdgreg"
It seems like the model(s) for the bots just changed, and now the outputs have become dramatically less diverse, even on different seeds. What’s going on and why?
Should be fixed now
Showdown borked again.
Hi everybody i have a question, can we reuse a seed to get almost the same result. I have my seed that i want and when I add it at the en of the prompt, there is another seed added at the en by the bot.
This is a cool bot, I like it much, made my own too which is not based on stable diffusion tho, keep the work ongoing :)
He is made of cheese!
so why are the gpu on cooldown, its telling me to vote, but there is no images in showdown?
He's grilled cheese Obama
You can also vote in any of the bot channels.
Showdown is booked again
We're currently having some technical difficulties; thank you for letting us know!
ok
Why is the bot always so obsessed with giving this character 3 legs
lets see
Is it a glitch of the anti NSFW efforts? Like oh hey, instead of just not generating this wrong, I'll cover it up with a leg
can you send an example here as well as the prompt
prompt:Anime, vanripper, official art, polychromatic, sharp focus, overdetail, bold lines, high contrast, hdr, idol, petite, short purple hair, tilted crown, wearing only reflective silver idol dress that has sukeban aesthetic embroidered detached blue sleeves, high heeled combat boots, commanding her gang and winning battle against a huge machine, by Studio TRIGGER aspect:16:9 negative_prompt:3d, black clothing, headband, headphones, tiara, jacket, puff sleeves, kimono, comic, shorts, armor, pants, sketch, bodysuit
i would take, shorts pants and bodysuit out of the neg
if that does not work, add upskirt to the neg
If I don't, it will force shorts or pants
Oh right, I guess I copied the wrong one? Upskirt is in the negative
The three legs thing is consistent enough
i mean, when you have a sightline to her sensitive areas, and ask for no pants shorts or bodysuit, it might be getting the wrong idea
Surely no one wears dresses ever
with sdxl you can usually get away with no negative prompt at all though
I find its just that smart
Yeah I've had to go hard on the negative prompts with thisb
throwing a huge negative just confuses it
This* in particular. It's fascinating really
Here watch, I'll do it without them
Literally any or all of the things I specifically ask it not to do
something, but yeah for some reason this kind of clothing + setting + anime just does not want to work
I'm going to vote until it does
I'm going to vote until it does
i mean, if you want it that bad then sure
but you could do 10 other ideas between now and when it gets updated
doing more variety is better for the ai's growth than spamming one prompt
I do, they usually come out fairly simply
Once one picture element works well enough I move on to the next
While true, most of these I can manually edit to get there or use as init images.
I like contributing. I also enjoy digging in there and figuring things out. Still trying to figure out why "vanripper" adds so much extra zazz
i told you but you wouldn't listen lol
I was trying a path next to a stream and now a path crossing a stream. The bot seems to have trouble with the concept.
@distant niche not perfect, but works more often than not: 35mm photo. a (single:2.0) pathway bifurcates and crosses a stream in the woods (at a right angle:2.0)
that was a bit better thx
You could also try adding weights to (bifurcates:2.0) and (crosses:2.0) to force it some more.
Not at all
Three legs thing again
It's super consistent
SDXL: "I think you want LEG, here have a EXTRA LEG"
You can always remove it with a generative fill bot.
Well, yes, I plan on editing a ton of these some time in the future
I just want to keep pushing here though
I love testing things
Likewise. Sometimes it’s better to get the exact desired results from the art bot.
Well that's what all the voting will help with
I mean, when I look at my stuff vs a lot of other people's... Yeah I'm out there on the edge
Something else I've noticed
"winning battle against a huge machine" is interpreted one of two ways
Defeating the machine
Or winning a battle... While against - as in physically rested upon - a machine
That's language for you.
The "Showdown" section is not working. Every now and then the message inviting you to participate in the voting blocks my generation but going to the appropriate section to vote, the screen only offers the "Pantheon" ranking.
It's not working, it hasn't been working for 1/2 a day or so. Devs are aware, they're squashing bugs. See above.
More murder tripod
The three legs thing is funny 😂
She's a cyborg so
Guess her parts order got messed up
I'm just imagining the head of a hard-won independent city-state having to head into battle with three legs because someone read something wrong 🤣
Here's a prompt for you, see what you can do with this: (colorful:.08) fashion sketch, stabilo/watercolor. minimalist, photorealistic, masterpiece, isometric 3d, flat shading, (thick lines:2.0), (plain background:2.0). (cyborg princess assasin, full bodyshot. (umezu remix))
Interesting. Can you break down the parenthesis thing for me? I'm still in the process of understanding that sine the one reddit post
I'll check it out when I get home later btw. I do this in bouts of idle time when the mood strikes
(parameter:weight) a weight can be from 0.0 to 2.0, which applies the effect at the given strength or emphasis.
What's umezu remix?
manga artist.
Ah
Have you found that the weights and parentheses work for the Discord bot as well as they do in Automatic1111? Or is it just a placebo effect of sorts?
From what I've seen other people doing it it seems to work according to what I've seen through other systems
Try it yourself, they work great. (cat: dog:2.0). then vary the weight.
That's a slightly different syntax from the above prompt. It means blend a cat and a dog and apply full throttle to the cat parameter.
I was just asking because I was testing it and comparing it to an average prompt with just keywords and commas, the difference can be seen at times. Just wanted to make sure that it works.
So you'd expect to get 100% cat in that instance.
Only one way to make sure and that's to try it yourself 🙂
I’ll try it again then. 🙂
Hello, does anyone know how to generate a 9:16 image?
using the bot?
if so then just type in your image prompt, then press tab (if you're on desktop) or click/tap on the right from prompt box and select the aspect option. from there you'll be given a whole range of aspect ratios, 9:16 is one of them so pick that one.
i love you!
thanks!
np
Eats up so much token space with weights tho omg
Looks like showdown just fell over again.
Nope
You can't
mmm ok thx, that's sad
ok why is showdown now broke every day ? are there to many users using the bot ? are tzhe servers at their limit ? i mean that would be an good sign, but bad for the training part.
It'll be fixed up soon. Voting in the main bot channels still works fine 🙂 Perhaps there will be some fun improvements coming too
so the normal voting is enough to train it ? and the showdown vote is made behind the scenes ?
restyle and resize many times use the same parameter as before, this is an time wasting bug,
would be nice to have some new styles like "fabric","ink","tattoo","mosaic","carved","Double exposure "Vector" and "Fractal"
test
ok weird
I was just saying that this is a helpful way to get your prompt back after the bot goes on cooldown if you accidentally send it
What I typed always vanishes so fast
ok every time i generate an face that is far away from the camera i notice that at least the right or left side if the face is ok but the other is not...... i know SD is generating in 8x8 pixel blocks but this cant be an generating issue....... so is here an censor function active in the last 5 steps ?
here an example of the two very different face parts :
this cant be an generation fail because of the resolution of the face, because the left side is very good, would it be an typical resolution fail, both parts would be ugly or both would be good
Can you vote for more than one image at a time in the
Pantheon
Not a B, but just more than one in general
you can vote for one from each group of two photos
guy is asking for gross stuff here... #1100170312106127410 message
The three legs thing seems really consistent with any kind of dress, not just the character I keep making
is there a way to manually specify the seed in channels with bot, using the dream command?
Nope 😀
nude on pantheon, im not able to report it #🏅|pantheon message
self-reporting a few topless mermaids in #1100170514670039070
Thanks! You can always go to apps>report, or use ⚠️ btw!
I wasn't sure how or whether that would count as a strike against me for lewd behavior... I don't know if the apps>report is anon
If you end up accidentally doing something, just let us know!
nude in showdown
Ehh this just made pantheon 2nd runner up
here you can cross pictures or generate based on pictures in? Like in midjorni
showdown is down
It will be intermittent for the time being
Vote more in the bot channels if there are no showdown for now
There's not a huge amount of voting happening right now, so not surprising. I propose making generation contingent upon voting. Some sort of ratio.
No.
Yeah, like 2 votes for pantheon is kinda ridiculous, don’t knowif the counter is off or if people stopped voting after the SDXL1.0 release. Probably need an new announcement on what’s next before we get better numbers more consistent
Tipp for everyone who lost their promt because of the showdown warning, go to the bot channel where you lost it,click in the chat window, press STRG+Z, there your old promt text
I assume English isn't your native language/ keyboard language. For those with an English keyboard, I think he means ctrl + z
I suspected as much
😒
How do I do that on my iPhone?
Showdown broken. 😢
@rocky vessel @wintry siren #🏅|pantheon message pantheon NSFW
ty !
Hi! Does anyone know why all my work from this day just disappears?
No NSFW, copy, or anything like that.....Everything disappears from my channel, mentions, messages, etc...😱
hey! what do you mean from your channel? channels as the bots or chat channels?
if your DMs and messages dissapear might be discord issues unrelated to the server, try the CTRL+R trick
If you mean the generations that is intended. If you wish to save your outputs you can use the DM buttons to easily send yourself an image you want to keep, otherwise it will be removed unless it gets voted to the Pantheon
Hmmmmm.... 
noooot yeeeet... 🙃
that's ok, knowing that comfyui is there for me
oh hmmm resize seems to have a bug
Repeat request for prompt to be added as exif data. Just the prompt and negative prompt.
I think the bot is down again. I have a feeling it will be off and on for a while with updates, but am unsure
Enjoy the rodeo 🙂
@lethal haven i think the bot crashed, or you're just tuning some settings
Hopefully better now
I can look into putting some basic EXIF back, no promises
and here I thought people would be happy to have ControlNet in the bot 🙃
So sick you added controlnet
Seems like one of the nets have some weird patterns though, seen here:
#1100170312106127410 message
#1100170312106127410 message
Yeah those are my gens funny enough 😄
Might be randomizing some parameters a biiiit too hard, will keep an eye 😄
It happened once on mine and once on yours
Yep touche
i cant vote in showdown
Will there ever be restrictions on generations that used ControlNet when it comes to the showdown/pantheon?
or message
@lethal haven looks like the bot crashed again 🙄
Yeah it's restarting now. Handling edge cases for images.
Never seen alpha channels supported in images before this just turned up: #1101178553900478464 message
This is an init image for control nets from the user
Oh, then it's just rendering the control image. Interesting.
Ah! Sorry. Yeap, bot-9. I don`t understand, really. 😦
Any pictures you make through bots 1-10 will disappear after about 60-90 minutes. If you want to keep them for yourself, you need to either DM them or save them by clicking on them and choosing save image
anyone else not able to produce any images?
Hey when I put in an image it doesn’t seem to output anything
And we're down again! (showdown)
Down?
showDOWN
I saved but cannot find them on my mobile phone memory. There are files named download.. but i cannot open them
They should be saved as .png files. If you have a photo app on your phone, you should be able to use that to see them. Otherwise, you may need to transfer them to a computer.
When I click save as for the image you posted, it is a .png. Possibly your phone doesn't recognize that format.
The images are there but not visible and not with png estension...
Changing the estension in png they become visible. For some reason the png estension has been cur
Do you have an iphone, or Android?
Android samsung A8
I was going to suggest that if they didn't show up on your computer.... but since that works on your phone, problem solved
I have to rename all and add png. That s ok for recovering old images but how can i avoid the trincation of the extension png for the future?
can we set the weight of the image we use controlnet?
It looks like we can... use parentheses and number between 0 and 2. example: (sunrays:1.2)
oh, wait did you mean weight of the image used?
I have no idea if we can or not
That's beyond my knowledge base. Hopefully someone else can help
i mean how much the init image gets to control the overall result of final image.
Hi! DM'd images download correctly as png on the computer, I guess this is something related to the discord android app
not to our server or bots
I understand that when filename is too long it is truncated and the png is truncated too leading to not recognizable file. File is saved the with name download-# without estension
If png is readded then ia visible again as an image file. This hapoen only if you ask to include prompt description
great! you found the solution 🤘
I like it. I eas wandering if i am too old for discord.... 67🙂🤣
Don't worry about you, worry about your computer. 😉
Or in your case your phone, but if you're here it's seemingly up to the job.
I tried ControlNet on bot#1. I’m not sure what type of model it is using (depth, canny, openpose?) but it doesn’t seem to have that much control over the final image. The end image doesn’t mirror the control image all that well. Any thoughts on how to make it work better?
Parameters and models are randomized. Hit reDream to try again with the same image
Hey when I put in an image it doesn’t seem to output anything
Hey when I put in an image it doesn’t seem to output anything
lol blurred out duck
Sorry, that duck is just too hot to handle!
the control net feature makes totally different images that show things that are mentioned in the negative promt
dream prompt: droneview on single mountain,high detailed,sharp,asset,far,intricate details,zoom out,SNES, style: Digital Art negative_prompt: caotic,asymetric,noice,blur,bloom,shiny,oversaturated,strange,chaos,trees,city,buildings,docks,animals,drone,helicopter,plane,flying,clouds,hills,field,river,water,waterfall,moon,sun,roads
plus
makes citys,spacestations and so on
@lethal haven
Question for CN : are we supposed to input preprocessed images or regular images? (or either and its auto detected?)
Regular images at the moment
Preprocessed ones may not work well
Especially since models are being randomized, if you submit lineart you'll probably get a depth model and that won't work too well
so we just use regular images and hope for the best ?
Yep!
oki doki 😒
#1101178553900478464 @mild bramble barely sfw
I tried the same prompt with and without input image, I can't see any difference
is it even working?
oh there is something up
its says image-init now instead of controlnet, does this mean we have img2img now on here 😲😃
ok fix incoming for the controlnet images, should hopefully also get better results than before for odd size images
and no - no img2img for now 🙃
@2% thanks keep us informed, not going to bother the servers till then !
calm down
Rules mention the following
Rule #4
No NSFW Content
Excessive gore/blood/nudity or anything that exists to suggest sex including sexualized poses. No illicit imagery. This includes NSFW content involving or depicting underage characters such as schoolgirls, lolicon, etc.
Would it be ok to have a massive panel in some openspace displaying a woman in lingerie ? Unless you're working at victoria's secret probably not.
If you still feel like there's been some abuse and/or want more explanations please feel free to open a support ticket.
Not a fight nor personal, just server rules. Nudity plays a part in that. It isn't about what's "normal" to you. These are the server and Discord rules we must maintain. If you want to turn it into a fight and or be rude and snippy about it, please take a step away and leave.
Never okay to talk down on someone and fight when I've been more than nice enough, warned you etc
Okay, have a bit of time away, please.
When you are actually ready to move past it and not be so condescending, you're welcome back
@2% CN is now working, but for some reason every other image is very dark / very light, like a clear pattern that something is going on with offset noise or something, it can't be random its always one of the two images
mahn we really need img2img in here 😢😢
So I'm trying to get this thing to give me a heavily pregnant woman in a starfleet uniform and I have no idea how to make it any further along than this
Is this more along the lines of what you are looking for?
Yep
I did end up getting what I wanted after several more iterations
i mean ideally it would be a bit bigger than that but I'll take it
put the pregnant part of the prompt closer to the beginning, or use weights on that part of the prompt
how do weights work
like, quotation marks
or some combination of asterisks?
because the entire first half of the prompt is a lot of words to say "this woman is thin and has a very large belly" so I cant frontload it anymore
use parentheses and number between 0 and 2. example: (sunrays:1.2)
or as I used it for the image above, ((9 month pregnant):1.8),
the higher the number the more its weighted?
yes
lol that's not ok either, report them
nsfw prompt in #1100484581037195384
uh oh
Can we get Controlnet working properly. I installed DW control net and it not working at all. Also control net crash/breaks when doing img2img. Please fix asap.
how can i get a lion standing up in human like posture
Please teach stable what a Shinto priest looks like. I have tried hard, no way to get one in any AI.
@snow rune Only the depth controlnet (the big one) released at the end of last week works, every other just destroy image quality
ohh that makes sense. When are they fixing the others?
"They", thats vague, all currently released controlnet are coming from three different people, just fine tuning on their own. The most serious is from huggingface (depth), others are more like single people project using few rented GPUs.
There is no official controlnet like for SD1.5 where most of them came from the same person/team
Stability AI is working on their own
you can test here
but I fear their might be trouble with SDXL & CN because even here the results aren't that great
so either we just need more time, or new CN architecture
I don't know how to explain but it looks like the images either follow the input but produce a "poor" image (by poor I mean grainy, artifacts, not following the prompt style like everything turning into a bad artwork, even more messed up hands, weird too bright/too dark images)
lowering the CN strength eventually remove that effect but then the image just vaguely follow the input
Depth one from huggingface is the least affected by this
but even then, the image quality is lower (and thats with a workflow with the refiner)
When using controlnet, I've noticed that usually one image is ok, and the other is barely there. Is that how it is supposed to be for now?
I think they are using two models
one for each image
and one of them is just messed up
or thats just setting
I get a darker image out of sync with the prompt/style
Some files aren't being named properly, they're getting uploaded without extensions
After download, it's certainly a valid file, but discord isn't showing it because it lacks extension
Please help, I believe stable foundation gives 25 free attempts, I can’t generate any thing using / Dream
remove the line-break from your prompt
Please someone help. I am new to most ways of using Stable Diffusion. How to use ControlNet option? I don't have detailed information about this option and how to use it for img2img in bots.
you can't use img2Img
when your prompt the last option should be controlnet, then you can drag & drop an image
Unfortunately, link doesn't work.
Why?
Sorry, but can you explain in detail? I also use the last option, but every image I upload gives me an error after that. The error message is: file upload failed!
DIMO, possible reasons are: the file is too big; Your internet connection timed out; the aspect ratio is wrong (doubtful); the file is the wrong file-type; or the bot is having trouble; or some reason I didn't think of.
yesterday the bot went wrong
regarding CN
they had to reboot it
maybe thats the same thing today
AR doesn't matter they just crop (but I suggest using good AR so can make the cropping yourself instead of random)
There's like 10 nudes in showdown @wintry siren @static crater
got them ty
This seems to happen with cut-and-pasted prompts
Funny, I've seen it work for some folks in the bot channel. My attempts failed, however.
What types of inputs work besides a depth map?
you don't input depthmap
you input a regular image
they apply a random preprocessor and the matching CN model
two of them (most likely different from my experience)
for the two results imgs
Mine had no similarity at all, will try a non illustration next.
oh yeah they are just bad
look at bot 5
I just made one
one is a total mess
other barely ressemble
(and is messy too)
the CN model are not on point now
I hope those are just the "tiny" models
and the beefy"
are better
Yeah not super great yet
released canny and openpose are trash so far
only the depth one from huggingface is decent
This was okay. Maybe with a detailed prompt it could be improved, including some negatives.
this is the behavior I get with canny models
they turn everything in those kind of artworky
even if you go full raw, photo, nikon in the prompt and anime/artwork/drawing in negatives
@vernal charm
This is what I made to compare with SD1.5
obviously SD 1.5 fails at generating at 1024p
but the style is respected (prompted for flat 2d anime, masterpiece)
in SDXL the style is just weird
and even if I prompt about photo etc
Looks like aliasing/antialiasing issue almost
Suggestion: if you're gonna send both prompts, why not send both images too? I usually don't care, but sometimes I want both of them, especially if it's comparing two images with the same prompt.
Is there a way to see all your generated images at one place? Like midjourney?
Samraat, Not exactly. after 60-90 minutes your prompts and images are erased unless they make it into the pantheon. If you want to save your images, you need to dm them to yourself or clict on the image and choose to save it.
I’m up to 25k images on my phone
It’s like I sit down to eat and the next thing you know there’s 30 images of Cthulhu eating spaghetti
Is it really too much ti ask to put her hands out of sight
BARBERS
His palms are sweaty, knees weak, arms are heavy
There’s vomit on his sweater already, Cthulhu’s spaghetti
fetish shit
At least there’s no tentacles
But like make her eat spaghetti with Cthulhu or something
fetish can be many things
@/spatio posting things with "fetish" including babies in showdown
There is “cameltoe” in showdown
Not sure how to react to showdown pics to flag it on my phone
you can't report showdown posts, it's really dumb
Control net is pretty great but when scrolling through pics I often get snagged in the voting pause
We need an report function even for showdowns pls @lethal haven
how about read the rules lol
good morning everyone! thank you for the reports and suggestions! my only thing would be to try to keep this channel on civil terms and without any personal issues/beef.
reporting now works only on the bots content or text channels, for showdown or pantheon at this time just try to ping any of us here with a link and we'll deal with it.
Thanks to you for this info 👍
@storm badge we support any kind of artistic expression unless it's violating our t.o.s or rules clearly written here> https://discord.com/channels/1002292111942635562/1040299965647437905 specifically about rule#4, nsfw content is not allowed so please bear that in mind when generating images. thank you.
How to ImageToImage in bots sections
no image to image yet only ControlNet2Image with help of text promt
@solemn kernel
but its a little random so the training for the model is harder
why did they blur this image so hard
nudity likely
the prompt literally asked for a dress
filter sometimes picks up on own things and can blur, despite prompt
is not always perfect
I just wish the filter instead of directly blurring the image just slapped a spoiler over it
Your entire history on this server is women with pregnancies, and seems to be very fetish esque
Think you've been told and warned before for posting stuff skating along the lines of blatant fetish material, some of which other users have reported.
Please adhere to warnings by staff and adhere to #✍🏼|rules-and-tos for future notice.
I have not been warned about anything in fact but okay
I originally joined for space stuff but learned that this thing is egregiously bad at making it realistic tbh
Its just i have seen few dropped image and typed their prompt
get her jade!
you have been warned and timed out 3 times. please follow the https://discord.com/channels/1002292111942635562/1040299965647437905
really sick guy over here #1100170312106127410 message @wintry siren @lilac geyser
@wintry siren @rocky vessel @lilac geyser please.. > #1100170312106127410 message
got them ty
@wintry siren @rocky vessel djiezus #1100484581037195384 message
okies
got it tyyy
yw
this is not picard
and that is not a bucket hat
this thing somehow followed none of my prompt
looks like him to me
Modify your prompt:
"Patrick Stewart" or "Patrick Stewart Picard" or even "Patrick Stewart as Captain Jean-Luc Picard from Star Trek the next generation".
For the hat, specify it more and it will show up. If you don't get what your prompting for, google some alternative words for the same thing Other name for bucket hat gave me Fisherman hat, So try: "Kangol bucket fisherman hat"
Either there's some bots auto-voting in #10 or, well, I don't know what else to think. Surely no one would consciously vote for censored renders...
(check the current showdown)
I second this observation by @mild bramble there is some activity going on in #10 that is very odd. We would both like this to be investigated as nobody should be using third party software or hacks to get an advantage over the other users here. If its a mistake by us then so be it but we would rather bring this to the attention of mods and other users whom may be impacted by this.
Seen it happened before, think it could be random images chosen during low-activity periods.
Could just be a troll, almost no one seems to bother to vote, when i'm active in a channel and vote on gens there, suddenly plenty end in showdown
Are the bots presently borked?
I guess it's maintenance time
Controlnet is not working 😦
I agree
I would like to ask you, why I use controlnet to select my local image, will always generate no image, and then in the generated command can also see "width:0 height:1", regardless of whether I manually select the image ratio is so, the complete command I generated is "/dream prompt:UHD,highly detailed,professional,bokeh,chain saw man,senior high school student,asian,fighting_stance,white shirt,sneakers,tie,black jeans,blonde hair,tokyo,street,cinematic angle,(((full body))),rain style:Comic Book width:0 Height: 1 seed: 192892105 ", I downloaded to the local image, the url is https://i02piccdn.sogoucdn.com/a77f145ac4c1576a
Hey folks! Could anyone point me in the right direction?
🏁 I'm working on a project to honor teachers. My idea is to turn them into kids using AI, aiming for a super realistic result. Ideally, I'm looking for a free tool to make this happen. Where can I start? Thanks a bunch!
@2% Looks like you changed the output format from png to webp. Fine by me. Next, exif data for the prompt/negative prompt? 🙂
Quality of the images has gone down.
hi there, how may I dream:prompt from a private server?
You can't use SDXL from a private server yet.
hornyposting got in showdown again
is it possible to generate img2img using this bot?
not yet :(
I’m really digging how powerful controlnet is
I typed glasses. You see what did it show?!
LOL wow wtf
it probably created glaes or something 🤓, bad word filter strikes again
bot is insanely slow, high traffic maybe? 😅
it's taking ages to generate, interactions are failing too often.
Should be back to normal soon 🙂
bot is super fast now 🚀 🚀 🚀
Is there anyway to add the StableDreamer bot to your own Discord?
not yet
Sorry, but what is the point of self-reporting images if you are put on time-out for it?
Used this prompt
photo of a woman laying down looking back at the camera, shot on kodak portra 400, aspect:3:2
Bot generated a nude, I reported it and was about to go in here to flag it – but was already put on time-out.
thank you for self-reporting inappropriate content. that particular nude was already reported before you did it. looking at the previous history, we found a couple more reports and "on the edge" cases (girl in bed with feet to the wall, etc.), the t/o was a consequence of those. please try to stick to the t.o.s and rules regarding nsfw content on the bots.
Thanks for the explanation. Since I was not aiming for nudes it's kind of a lottery on what the bot will generate no? (wanted to try out weird angles of the face, since it has/and is a bit of a challenge; cartwheeling, hanging upside-down, laying down etc)
Not a big deal really, I rather get timed-out on (what I consider) harsh ruling, than having the channels filled up with thirsty generations. Was bc I self-reported just before I got banned that had me questioning the logic – I'll be quicker on my report finger and go back to generating manatees 😉
no worries, most of your content is OK and good! it's just that we have to be strict with the rules regarding nudes/violence/nsfw. keep up the good work!
using negative prompts can help, just to steer clear and be extra sure
Usually do if I think there’s a risk of getting unwanted nudity (like certain artists, photographers, or anything with transparency).
Didn’t think ”woman laying down” would automatically be interpreted as something nsfw - hopefully such a bias is something stability can/will address.
If that stuff is reason for a timeout there sure are some double standards, sorry to be harsh, but seeing the average content prompted for and then giving timeouts for "on the edge" like this cases is bizarre
yeah the ai needs the anti-horny bonk
Why can't I save my photos anymore?
@keen nebula What seems to be the issue? Images are now in webp format instead of png, but should be saveable just the same.
@mild brambleI create the images but when I save them on my hd they do not open😭 😭
So you have the files? What are you trying to open them with?
I recommend XnView.. that'll handle them nicely. https://www.xnview.com/en/xnviewmp/#downloads
@mild brambleoh thank you so much🥰🥰🥰
anyone can help me with my generation,, idk what to do to it so its gonna generate exactly what i want
Original prompt I created a long time ago with this bot:
VS
When i tried to recreate it by feeding it to the AI with the new feature:
VS
When i tried to recreate it from just the prompt:
Why do both the new images look so bad? Did it change something in the AI's engine?
@sinful belfry Plenty has changed, backend parameters are constantly being tweaked and optimized, the model's being updated. If you're looking for consistent results, you probably want to use clipdrop.
The whole purpose of the bots on the server is to help train and improve the model, so results may vary from one day to the next. 2% is in charge of the controls 🙂
I'll surely check out clipdrop. Thanks!
webp makes me throw up
what is controlnet for
drop a picture in control net and the bot will base the generated image on that picture
The bots are hanging again.
can u help me? how can i add the bot to my server?
you cant, that because the bot with the votes is used for training the new model version. the training and votes are watched by the mods and admins to make shure 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. @clear basin
ahh i see .. thank you very much.. trying some time to figure it out..
no prob
but you can use the latest version offline
sdxl 1.0 is free
you can use it in many ai image tools
@clear basin
i mainly use bluewillow , can i use sdxl 1.0 in prompt ?
?
bluevillow is like midjourney
i dont think they have an command to use sdxl in their channels
i think you can use online SD model sites
wait
i think they have sdxl too
and its cheaper than midjourney
you can use it full online
in free version also without any limit
i checked the free version has sdxl
@clear basin
thats great! thanks
im using bluewillow cause i use the faceswap app through discord.. i cant find similar enywhere else
at least you can test out sdxl without using an offline tool
yes true
nsfw content needs the pro account there
ah an tipp, when you make an image there or here with sdxl you can force the model to take some of your promt parts more serious than others. as an example you want an cat and not anything other you write
(cat:2.0) and when you want an cat but it can make somethimes other stuff (cat:1.0) 0.1 and so on also works. with this you can scale how much of what stuff will show up in the image.
@clear basin
this gives more emphasize to the cat ? by typing it in right this (cat:2) ?
got confused ;p
i think 2.0 is the maximum and 0.1 is the lowest
it also works in negative promt too
so 2.0 will garante 100% its an cat
0.1 will make more random things when it comes to generate the cat in the image
another thing
(BLEND cat and wolf):1.9) will make something that mixes this two BUT everthing over 1.0 will make it more wolf and everything under 1.0 make it more cat
i was against this weight methods first because i was thinking its harming the random generation of images here and the training compared to that, but the devs say it doesnt matter
this can be made with the blend technique
something that dont exist
wow amazing
when we train sdxl here we make sure that every new version is more precise that means with the help of the community the new versions of SDXL can generate the image people want without generating 10-40 until one is good
thats why i help here
and thausand of others too

?
Hello, why my prompts are keep blocking by your bots/servers?
whatz you blocked ? what warning is shown ?
when the showdon starts in the showdown channel we all can only vote there until its done than the bots can generate again
Consider execute the "/dream" command first, then enter prompt 🙂
@lethal haven as a mac user i'm very disappointed about the decision (again) to save the images as webp, i can save them but there's no preview so what's the point?
this happens to everyone, it's kind of random and the only thing you have to do is resend it
This happens when your message goes through as a regular message instead of a slash command
It's due to new line characters in the copy paste
Thanks, that was driving me insane
You need to hit delete and the. Space on each line
If that makes any sense
Or at least that’s what I’ve done when copy pasting to make it work
Is this bot free?
Yes.
Wait for a little bit; if your dream doesn't finish, you'll see a message. Then you can try again. To learn more about how it works, check the pins.
Is there any way to use this bot in my server?
A Dream started 15 minutes ago hasn't finished and it says "still dreaming" when I try again. There are very few dreams appearing in the bot channels compared to normal, so the system must be slow or jammed up right now
Same
Hello everything is fine? Do you know if something has changed in the generation? All my requests are not returning either an error or an image....
same!
@lethal haven is it possible the bot is experiencing issues?
Oh no I'll take a look!
Hi guys, im getting error about ur still dreaming. How can i stop dream and ask for new dreams?
Quick answer, you can't
F.
I don't seem to be able to find my dreams in the bot channels when I search for them by "from: user" . Also, if I submit a prompt, leave for a bit and come back, its near impossible for me to locate it. Can anyone provide tips on how to manage this?
I cannot dream, the server says I have a pending request but it is not true. How can I reset?
Thanks for letting us know.
DM yourself the images you would like to keep! To easily search for yourself, you can use the mentions function and search for your username.
OK there are some novel issues on the backend and some dreams are unable to complete. Looking into the best solution. Stay tuned and apologies!
Yes it is, I´m having the same messages
It started to work now. Thank you
No problem!
search mention instead of from
bot still borked? :(
having over a two minute wait time for an image, still waiting
anyone here?
Most of the staff is usually asleep this time of day. You could try pinging a stablestar or community guide or moderator. Check their status on the right sidebar.
Don't ping the devs until after you've tried the others first
For bot related issues please feel free to ping me.
showdown with NSFW image
I'm usually available
If there's an image, please report it, as this is the easiest way for us to see it. You can always ping me, too. Even if I'm not available, I'll eventually see it
You can just link to the message if it's in showdown--no need to search for it; I just mean if you happen to see it--and I'll take care of it.
next time i copy the link and post it here
normally i see something NSFW in bot channels i use the ! emoji
Sweet! Ty! That's an easy way to do it. You can use apps>report to staff as well, ofc. As always, you can search for any image via mentions if you wanted to report something and forget/it gets lost.
i know i know
#1100484581037195384 message @rocky vessel nsfw
⚠️ pantheon
hey so i tried restyling my art but the ai taking so long so i couldnt generate a dream, i tried rejoining and it didnt work
its been more than 7 minutes
hi! sorry to hear that, sometimes the bot has too many simultaneous requests, try re-dreaming whenever that happens. just hit the restyle button again.
Hi looking into this and doing a restart ATM
I think Controlnet is not working @lethal haven
It's a little strict on input image censoring right now
So unclothed people will blur the input resulting in blurred outputs - I'll make a little adjustment soon
(Along with better quality and speed of ControlNet requets)
Thank you!
Done
It should be more forgiving on the input (but not the output) and the quality of the ControlNet results should be much better

Corrected the aspect ratio/size of ControlNet requests - everything was being scaled down and 1:1, apologies
Over the last 1/2 hour the bots aren’t showing my prompts as images. Instead they are showing me files, which I have to download and append .jpg to the file name. Then I can see the images. Why is it doing this instead of just showing me the images?
which model is used when starting prompt with "Dream"? " Stable Diffusion XL"? How to see the model list here in discord?
yes, the bots are using sdxl (with controlnet option), there are no model selection or lists implemented at the moment.
smiling faces have been pretty terrifying with this recent model
There's quite a few things being tested at the moment, so expect some oddities 🙂 Also expect some really good stuff too.
If you're doing unstructured landscapes, sdxl can do them any size up to at least 3072x1344. If there's no structure, there's nothing to mess up.
How can I use this as an image input for the bot
Controlnet
is there a way to invite the bot to another server? a bing chat query suggested that this is possible but did not give instructions
no, they are keeping it here for training.
okay fair enough thanks
my last few prompts have given me some odd images. Technically it's right, except it stopped being photographic
I'd imagine it will sort itself out with time
From the bot?
Because there is definitely a chance to get some images that are gonna be cooked and looking a bit weird right now 🙃 but there's a method to the madness
Yes, direct from bot-5. I'll just keep plugging away with whatever prompt to help sort it out.
@lethal haven Leaderboard is borked.
It's been borked since yesterday
yeah there's an overhaul in the works to address that
hopefully everyone can get their all time stats too instead of past 24/hr
@lethal haven And exif data for prompt/negative prompt? Any prospect of that happening soon? 🙂
I can see about that 🙂
style: photographic
the model seem to have wrong styles seeping through a specific style
2% commented on that above
I've been wondering about that! A large number are definitely overcooked.
Yeah I'll adjust that stuff later today probably 🙂
⚠️ pantheon
Should be no more instances of this now moving forward
whoops, reporting myself: #1100170312106127410 message
Why when I generate some images in the bot, it suddenly appears as a BIN file to download?
And when I download, you can't even see the image, you open it and there's nothing
@lethal haven
Usually when new lines are used in the prompt
Yes, but now what? how do i go back and generate again with the same prompt
If you can't go back and recover
If you rename the file you should be able to view the images ok
just add .png and open it again
Also there may be a bit of a slowdown for a bit, it will resolve
AAH, thank u ♥️
np! 😄
Hello 🙂 I have an error after prompt :
ERROR: @Seiki already has a pending request:
Hang tight, you're still dreaming! 🙂
I hope I didn't do something stupid, or send too many requests...
Tanks for your help
Nope, nothing you did 🙂 Dreams may take a couple minutes right now, but that should pass shortly
Okay tanks 🙂
@rocky vessel User (V) e i n ☥♕ is prompting for MILFs in BOt-1 over and over
what an creative name 😒
Thanks for letting us know, as always!!!
Then you probably want to delete most of the ones from Exostrike in the showdowns
hi why some images are uncensored
Is it me or is the bot going so much slower? And now the quality and differences on the images it generates varies more
These are training models, so the quality will vary from time to time.
yeah but i remember that two weeks ago or something like that it was different, did they change it or smth?
from 2% a few days ago: There's quite a few things being tested at the moment, so expect some oddities 🙂 Also expect some really good stuff too.
I want to create a man posing for a photoshoot using specific items (attached below) such as a specific shirt, pant, jacket, belt, hat etc. How to do it?
@lethal haven Minor thing, but the filename for (a) and (b) images is the same _idx0 suffix which suggests there was some intention of differentiating those to make downloading without needing to rename when you're downloading both images? Also, tickle re: exif data for prompts/negative prompts 😉
Wonky bot? Seems to be a lot of unfinished/failed ones
Looking into it
Should be good now
Thanks
sorry small hiccup, should be fine now
what is the usage limit for the bot?
There is no limit.
Dreams not finishing should hopefully clear up soon.
That's the dream!
Can I add this bot to my discord server or is that not allowed?
damn... just lost a prompt. I sent it and received a message that the GPUs are down
because that's fucking weird
lol it's not wtf, it's not my fault some people have weird fetishes with wlw relationships
I have seen far worse things being prompt here
that doesn't make them ok either jfc
it'd probably get blurred if it was straight or gay too
how many images can I request per hour or day? I got timeout ed
I guess I didn't know that kissing was nsfw
Sorry to post here but Mod-support gives “something went wrong” error:
In my main account “fuzzyrobots” which I’ve used on your server since the first beta last year, the server suddenly disappeared a couple hours ago while I was using dream in the bot-1 channel. When I try to add this server back again with the invite link on the official site, it says the link is expired. It looks like “fuzzyrobots” isn’t a member of the server now. Do you know what happened? It doesn’t say that I’m blocked and I don’t think I would be for any reason. I love participating in the showdowns and I’ve voted a huge number of times, more than double my number of generated images, and I want to continue as a member and aim for nore pantheon entries. Thank you!
never said that lol
censor bot sees two people too close together. two people = too much skin color = censor
Hii!!! I'll look into this, that's so bizarre. I can 100% vouch for your involvement and have appreciated your participation and reports!
revoked, sorry this has been an error!
Thank you two so much for getting my access back, and for the quick response!
sorry for the mix up! Lots of reports today! Thank you for being very kind about it!
/{best quality}, masterpiece, artbook, mature female, adult lady like woman, grey hair , black hair, multicolored hair, aqua eyes, bouncing breasts, wet clothes, pantyhose, hairband
wtf
anyone knows how to extend creation when the stable foundation produced the picture A and B
???
click what bottom for extending creation????
!!! are you reporting or trying to generate that here?! on the stabletweets channel there is none of that ! 😅 please mind the https://discord.com/channels/1002292111942635562/1040299965647437905 regarding nsfw
Haha my man/woman is trying to speedrun into a timeout 😂
I noticed that the following pairs of aspect ratio options provided by the bots are identical values:
5:4 / 4:3 (both 1152x896)
4:5 / 3:4 (both 896x1152)
I also noticed that none of these are true (1152x896 is neither a 5:4 or 4:3 ratio)
Ok so I realize that literally none of the pictures are drawn in the actual aspect ratio except 1:1
Ok so now I also realize that certain sizes just aren't possible in SD so the ones being used are probably the cloest approx. compatible sizes 😛
The 5:4 and 4:5 aspect ratio options should just be removed since they are redundant to the 4:3 and 3:4 though.
some sizes work better too due to the training
Is there a way to get the seed of a character I made in the bot here?
For me to remake in likeness?
@lethal haven
You can see the seed used when you generate the image. However note that we do a lot of novel testing here and you won't necessarily be able to get the same results with the released Base 1.0 XL model
at the very end of the initial response to your dream request
Hey! How do I add a picture of mine as a reference for the dream prompt?
thanks
Woah what's going on in the Showdown?
Images not displaying
@lethal haven have you seen showdown?
It seems all the files end with *.png<garbage>
no pressure! >>>>>>P>R>E>S>S>U>R>E>>>>>>>
might be some old ones in there
Will let it cycle through the mess
Yeah it's fine now. Next cycle should be 100% ok
With this lovely change by Discord all the images are displaying as 905x905 now 
Ye, it's a discord thing. They do like to mess things around. Just like it happens here as well.

Quick reminder that if you copy paste prompts from somewhere and it has a Negative, watch out that you're not putting those negatives in the positive by accident.
Strippers in the showdown. again.
It has happened to... someone I know. 😆
bot down?
Should resolve shortly
Now it's working, ty
what happened? The bot started to dream, then hours later nothing
It should be okay now?
Hi everyone!
Huge thanks to all the users participating with their votes! Although SDXL 1.0 has been out for awhile we are still doing lots of testing on new stuff behind the scenes!
To further improve the future versions of Stable Diffusion your feedback is paramount. We've implemented a small change today to help encourage more people to vote, and to deter the spammers who are only here to make inappropriate content.
Voting more than you generate in the last 24/hr is now necessary to consistently generate images back to back. There will be a 2 minute cooldown between generations (similar to the cooldown in-between rounds) for those with a low vote count.
So take a bit of time to vote on some of the other images you see in the bot channels and #1100484581037195384! You're helping make Stable Diffusion even better with your intentional votes.
Thanks again to everyone who participates! 🙇
Ah I see, I always try to participate as best as possible in the showdown votes, those in channel prompt also and in pantheon, alas, I really only vote under two conditions, The quality that it can bring to the algorithm is an exemplary face, and because I love it, so... it's seriously subjective xd
I'm not coming to complain, ladies and gentlemen, far from it, but suddenly... despite an incalculable number of votes that I have cast (only on what I like), well, I can't vote more than what already exists, and spam the vote well.. not great either ^^, so, well I have to wait 2 minutes, at worst not serious, but it's the possibility of redoing thanks to the vote which actually doesn't work ^^
Unfortunately for the first 24 hours if you were really active generating and not voting much there will be a bit of a catch up. Take some time to vote in the showdown on new images or other people's bot channel images, it shouldn't take too long to catch back up
To stay on top of the cooldown you literally just need to vote on your own generations + 1 other technically. It's not too much. If your own generations don't deserve a vote in your opinion then there's tons of other people's gens you can compare 😄
I am 100% open to feedback on this if it needs some adjusting. After you've overcome the backlog let me know how it goes.
To keep you informed, I anticipated your response, I voted in the space of 30 minutes, more than 20 times I think, and it was only on a case by case basis for 2 minutes ^^, I remain invested in all cases, if there is any change I will keep you informed! Thanks for your help!
Yeah, we certainly do not want people spam voting without intent, that will not be good and hurts everyone in the end.
Thank you for the feedback! I will also check your stats to ensure there isn't any bugs with the new system.
What if I'm unhappy with both results? A "neither" vote would help by allowing me to vote them both down.
Ah it was actually looking at previous 48 hours - I reduced it to 24 hours
We had something like this before but this actually caused even more confusion and questioning by users of when to use it... 😅 So we decided against it
So if you're unhappy with your gens feel free to skip voting on them and instead just look for someone else's to vote on
or at the very least pick the better of the two
I'm afraid spam voters will promote them randomly anyway... But we'll see!
Same. I don't want to annoy users so I did not implement a cooldown between votes, I was thinking maybe 3 or 4 seconds to reduce spam voting... but I really want to balance this with a good user experience.
What did you call the neither-good option? Just curious.
Reducing spam wasn't stupid but if it exceeds 10 seconds it becomes absurd, less well useless, and what's more there isn't enough flow at times that it would become irritating to wait for the person who has a minute of less than you ^^
I always wondered if voting was a tool for you, to understand what works best (and least) or for people's pleasure? Because if it's a tool, why not do a vote + argument? for example "Good physical proportion" or "Corresponding shadow" yeah I know, we're only on discord, but you never know ^^, this could make it an interaction + a tool and an assurance that people don't vote too much for just spam.
I made it a
button actually 🙃
Ah, that would have confused me! 😂. A red "neither" button would be cool.
I'll put it back on the table with this recent change in mind
But then... what purpose will you serve in the equation? I imagine the system is actually calling for us, isn't it? If we join my idea just above, why not also do the opposite? for example, shown that there is a problem with lighting? Certainly everyone is aware of this but the "ignored" button well, everyone will go spam in fact, so as not to give credit what
Now, why would there be no credit via voting? A user who votes my image A for example, allows me to speed up my waiting time bar.
Yes I have the answer, the famous group of friends who spam each other, alas, there are flaws everywhere, but if the idea crossed you I'm curious to know what it is.
@vernal charm What I mean is that you must have a use, and the "ignored" or "nothing" vote is of no use, except that you clicked on Nothing ^^
I'm sure the trick is to extract some measure of a model's success through the simplest possible means. The low level details (WHY did the user vote good or bad) are irrelevant noise statistically (hopefully!)
Also the more cumbersome voting becomes the less likely people will want to participate.
This.
Tricky balancing act!
I understand, it's really a shame, I would have really enjoyed being critical, with a real analysis, but I understand, you need something quick, it's normal
how much impact does voting have on the development of sdxl?
In fact, when I think about it, this is where the subscription formula becomes the solution ^^, everyone pays attention to their credits, but take advantage!

the bot is repeating images
I, for one, do not like the subscription method. Even if it's a number of free gens per month without subscription, it will eventually turn into pay to play only. At that point, a lot of interested, but money poor, people will be unable to take part in this experiment. Then that puts new and upcoming tech out of the reach of potential rising stars.
It was only a stupid observation of course, if the will of SDXL is to be free, good for us, it's unique, but it can solve the problem of spam between vote and creation
So I'm coming back to be annoying once again, but despite my votes, still no possibility of redoing without the waiting time, is it me who misunderstood?
Same prob for me. but votes are counted from today or our stats since our arrival?
Looks like votes are way up in Pantheon! Maybe the days of 1- and 2-vote winners are now past.
okay!
@split drift You're about 40 votes away. You have voted 115 times in the last 24 hours but you have generated 150 times
im so far! 😢
@noble vigil You are about 70 away
You have voted 218 times but have generated 280 times
no pls, i dont have the card ( but its my idea ) 😄
😄 Lucky it's just last 24/hours...
I prefer sdxl to midjourney
But need premium plan 🤪 , voting is not representative of quality, as it often takes a lot of generation for a correct result. how to judge if the prompt is correct from what it thinks it wants
Sry Bad english 😅
the stats only show your last 24 hours voting or all time?
maybe my english doesn't make sense lol
All
When you do your user stats it's all time
Leaderboard in showdown is 24/hr
Perhaps with this new system I can add a 24hr_stat command
It would be very useful indeed
Would it be too resource intensive to provide generated-vs-voted numbers in feedback messages (like the "GPU cooling down" replies)?
can always cache it and not be "real-time", I'll see! 🙂

I hate when i type out something long and get “it’s voting time”
So if any of you knows how to recover that on an iPhone I’m all ears
DMs should be forever queue Diamond commercial
I find that the bot is very slow, it was faster before, that's normal ?
I always copy the prompt before sending the mesaage
I need to get into the habbit of doing that
@lethal haven Excellent work on the vote to generate feature. Another suggestion: once the voting period has expired for images, perhaps remove the voting buttons...
@prime bear Arrow up after you've sent the message should keep the prompt alive.
Arrow up?
The opposite of arrow down.
Not to sound like an old person, but where is the arrow up on an iPhone?
I used to be with ‘it’, but then they changed what ‘it’ was. Now what I’m with isn’t ‘it’ anymore and what’s ‘it’ seems weird and scary. It’ll happen to you!
Grandpa Simpson
I have no idea where it is, but when you find it, use it 🙂
Can anyone write a prompt allowing a generated image to be as faithful as possible to the image put in the prompt?
The new 2 minute limit is flawed. It claims to remove the limit if I vote more, but despite voting for images (showdown and in generation channels) it still forces me to wait 2 minutes. If I am not mistaken, then this should be fixed.
Works for me. I suspect it may be calculating over a 24h period, so you may have more voting to do.
And while @lethal haven hasn't shared the algorithm, it's probably based on a ratio of generated images to votes. You can always use the /dream_stats command and see where you stand.
It's simply vote more than you generate within the last 24/hours. That means if you vote for all your own generations and just vote for 1 additional person you can generate freely.
Ideally you'd vote a bit more than that though 😜
More intentional votes also means even better images will make it to the pantheon as well!
@prime bear click on the ”dream” in blue after your username
If it just keeps loading a blank, go to ”showdown” and then back to the bot-channel and retry
It might be worth weighting the votes so that you need at least 50% of your votes to not be for your own generation to limit restrictions, you know, to encourage voting for other peoples' work. So a vote for self = .5 points, and you require points = number of generations to generate freely. It might also be worth considering a vote in showdown quotient, so that a certain percentage of total votes need to be cast for showdown in order to qualify for unfettered renders.
We'll see, I don't want it to be too restrictive! 🙂
Can toi add "redo edit promot"?
Use arrow up for the previous prompt. (while the input field has focus)
If you're using a crudphone like Fnord, you may need an alternative keyboard app that supports arrows.
Yes its for phone , its for that i propose re-edit button
so when you dont vote you need to wait 2min until you can make an another image......... that ok BUT can the message pls say that we need to vote in the showdown to acces generating ??? @lethal haven
and we need an "both are bad" button....... somethimes both images should not exist
this would also help to train this model faster
am i the only one that is getting the chats deleted? I wanna post a picture in show and tell but it's gone, and then minutes after it comes back, I don't understand
do you have anything written that might trigger nsfw filters? maybe a link to an external URL?
if it is just an image and sfw, maybe the bot is having hiccups, just please be patient and try again 😎
I mean that the chat is gone, not the pictures, the pictures are perfectly normal
the autobot deletes messages in the chat that contain nsfw/links and such, that's why I asked. Also be mindful that the images on the bots get deleted after a while, if you wan't to save them please DM and then "save image"
the chat show and tell should be there but it's not, at least on my pc
@lilac geyser #1100170312106127410 message
How can I check my Creation/vote ratio? To know where I am, without waiting for me to be in the showdown leaderboards..
Yeah I have to create an additional command to check that
yep, can i have command?
Curious to know if there is a command to modify your prompt without having to retype everything, to be able to change the aspect ratio more simply, the style, for example
Sorry if a lot of things are repeated, it's not the will that I have, but I'm not 100% English-speaking and have to read thousands of messages above, I admit it's boring ^^, I should rather consulted the faqs, but are they up to date?
just click arrow up to get your prompt back end then it's easy to adapt
Adapt my friend
is there a ratio of requested images : votes to avoid the 2 minutes block?
just vote more than you generate within the last 24/hr
Can be bot channels or showdown
please work on the writing part of the bot too, it does not create the writing and words in the images properly. That needs a big update.
we train the model by voting , just vote for the images that make text good and after millions of votes the version 1.1 will make it better
thats not an joke
Dear friends, please forgive me for my bad English.
Over the past week, the quality of the drawings has deteriorated significantly in terms of detail, and the AI has become worse at understanding words. What's happened? And how long will this last?
can we generate 2 images then the 2min pause, please, at least that 😭
doing one then wait is.. painful
VOTE
i do, images on the bot i use, or should i vote on other bots too?
vote for some other people's stuff too and it should help
ayt
I didn’t know where to post it to get people’s attention. There are some bugs with textures.
And I confirm the words of other people, the bot no longer understands words, ignores many requests in the drawing
Don't worry it will change again soon 🙂
All part of the fun. 2% is continually tweaking (some) parameters to test and improve. Soon he'll be adding exif data for prompts and negative prompts 🙂
@lethal haven A slightly different take on the voting / generation ratios. In channels, .5 points when voting for self, 1 point when voting for others. In showdown, 1 point voting for self, 2 points when voting for others. Should encourage voting in the showdown without making voting too onerous.
current system does incentivize people to vote haphazardly for other people just to rack up gen points.
quality tradeoff for quantity.
Why do I keep showing up: ERROR: @craggy rose already has a pending request: Hang tight, you're still dreaming! :)
I tend to vote for images with shorter, simpler prompts. Reading through a long, complex prompt (to judge whether it was effective in generating the resulting image) is a bit of a bother, especially when the bot channels are busy and auto-scrolling rapidly with new entries.
there are three parts that are trained here
1.shows the images really what the text wants
2.what images of the two has less artifacts
3.is the image boring or interesting
and i think number 2 needs the fix first,than 1 and at last 3
that also why i chose sometimes very simple things to see how good the model handles it now
and i cant wait any longer for version 1.1
Errors.... ERROR: : Must wait 2 minutes between generations - to remove this limit, vote some more!...... Why
so you can vote! you help the server and training, just vote on a couple images that you like and you should be good
Doesn't discord strip EXIF/iptc data?
everything here is for free and with votes just to train the new model versions.
If you find it's lasting longer than 2 minutes please let me know
Will it always be free?
The purpose of the bot is to assist in testing various parameters and experimental techniques to help train future SD models. Voting is crucial for this to be a success. The bot is constantly changing and will continue to evolve. Nothing is set in stone.
@lethal haven Oh no! 😮 Your dream didn't finish. 😕 messages all over the bot channels nothing gets generated
getting the same
I'm also seeing this, please be patient my friends, while you do please vote! 😎
Oops #1101178506530013284
#1101178506530013284 message accidental nudity. Guess I should neg the nsfw
no worries, thank you for reporting it
I'll just leave my music alone for now
Could use a cancel button for the bot. Even if it disappears after 2 seconds, it would be nice; I've accidentally tapped for a new image that appears below mine a bunch of times now.
^ this fr
Also sometimes I get like, origami 2x in a row when I wanted anything other than origami
as an example, I am not directing specific hate towards origami, I think it looks awesome, it's just not what I'm usually going for
Is there a trick to making a draft image more important than the prompt text?
You mean for controlnet?
You set the controlnet strength, but that's not available on the bots
Are there any stats available about how often certain aspects ratios get used?
Time for something different tonight
Is there a problem with the bot at the moment? The quality of the images has been very poor lately.
There is no problem - this is intentional 🙃
I'll throw the good back in too - don't forget to vote! 🙂
No problem, I’ll vote for the best quality images when I get them. 😉
May I ask why the bot is being nerfed?
Just curious.
It's not intentionally "nerfed" per say - testing some 512 stuff
Remember, the point of this bot is experimentation! 🔬 🧪 🥼
Got it.
@lethal haven someone's asking for gay porn, it's in the showdown now #1100484581037195384 message
Can anyone recommend a 3D design website?
I gotchu
I got that one
All good
Alright - thanks for your patience everyone re: the 512 results. We're back to 1024 and Upscales only for a bit 😉
so much better 🥳
That's not to say it won't come back! 
all for the good 😅
well, something came back 😆 #1100484581037195384 message
@lethal haven better check this out #1100484581037195384 message
@lethal haven this guy has some weird prompting #1080261341362786384 message
we're watching yes, what a convoluted way to make nsfw
thank you for the heads up @fallow fiber
he must be desparate
yw
I made an image cmd for my own channel bot. I made controlnet model selectable as a 2nd option… if only model or image is used, bot sends a message saying both are required and returns the command. Can also specify if input is a map (module=none)
Just a suggestion to get away from random controlnet assignment
Man, the photography is incredible.
I've been getting a lot of 'your dream did not finish' lately.
Would it be possible to either have the notification stick around longer or to have it ping us the same as it does when our dream is done?
Oh no! 😮 Your dream didn't finish. 😕
Same
It's random by design intentionally.
I can add a ping.
Restarted, should be better now.
@lethal haven Sometimes bot gives an image that is not HD. I mean It used to provide image in 1344x??? but now its 674x???.
This is intended. Please continue to use and vote! It won't be permanent
is there an date when sdxl 1.1 will come out ? is it this month ? or can you only say this year ? @lethal haven

I personally do not know. I am but a mere bot slinger and inference engineer
ok......... do you know at least an big event that is planned in the next months ?
lmao
