#🗣|artisan-support-feedback
1 messages · Page 1 of 1 (latest)
First message 😎
re-dreaming takes ages or doesnt work? 😄
While you guys first start getting it going the backend needs to spin up, so some generations may be dropped until it catches up
Alright 😉
WALTER WHITE
i recognise that prompt...
when SD was in it's beta stage everyone generated Walter white
Lol
well dalle couldn't really do it could it?
no
why can't I see anything in #🏅|pantheon
my first image i made was Walter white
@static crater
me too
same for #1100484581037195384
i cant see it
Hmmm
also showdown too
oh that's important
still no
no
20 min passed and my reDreaming still not up 😦
👍
You should be unlocked to dream again if that's the case
is this visible? #🏅|pantheon message
yes
yup
So new messages are probably fine
ah so newer messages will be visible moving forward!
Lets see when the next cycle happens, should be all good 🙂
yey
so what happened exactly?
Channel permissions issue it seems, and didn't retroactively apply
perms be permin
#1100484581037195384 is still empty
Suspect when the next batch of generations move to #1100484581037195384 and #🏅|pantheon they will be fine
Yeah it hasn't refreshed yet
ok
Game works in 30min cycles where the top 50 images from the bot- channels get moved to #1100484581037195384 , and the top rated image from #1100484581037195384 moves to #🏅|pantheon
glad they still have this emoji
If you want an image to stick around forever, DM it to yourself or pray it makes it to the #🏅|pantheon 😄
Which one we have to vote, the most cool looking or the one who follow the prompt the best?
Try to vote for what follows the prompt but also the better image too
It's really up to you
Ok, cool, thanks
make sure you all vote after generating and don't pass them by! that way contenders can appear in #1100484581037195384
you have to probably restart your discord
@wintry siren I restarted discord and all messages in this channel disappeared
and still no pinned message
I think you should enable chat history
Try now, it was disabled
Oh hi fruit!


a couple of my prompts haven't gone through, or at least are taking an extremely long time to
It's almost cycle time...
Yeah some are timing out, apologies
Not much we can do about that as it's the backend spinning up
We have a new #🏅|pantheon winner!
ok I can see the messages now
#1100484581037195384 should be repopulating shortly
why did the images disappear so quickly from #1100484581037195384
They were actually there for 30min
I didn't have time to vote on all of them
if we didn't have the permission issue
aha
There we go!
whew that's a lot 👀
#1100484581037195384 is limited to 25 pairs (or supposed to be)
two of my images made it lets go!
the top voted images
Question for all of you - would it be annoying if for the 2 minutes before the #1100484581037195384 winner is selected everyone has to cooldown from generating images and can only vote?
This would encourage more voting and give the bot a bit of a break to catch up between cycles
that would be ok
sure why not
#1100484581037195384 Interaction failed.
Does retrying work?
it may happen occasionally, try again?
or is it consistent?
Works now thank you
my favorite image is the George Washington rat
Now vote for the human one. #1100484581037195384 message
I already voted for the robot
😦
A true American would vote for the RAT!
Vote for the meme one pls #1100484581037195384 message
"sa meme meme a a meme a meme athba meme is a.a meme" is the best one imo
Lol
the most American rat that ever lived! #1100484581037195384 message
stable could have been at least on an level of MJ5, if not better
this seems a lot better than if
Nah
the current if upscaler is bad
IF can make good images but need a long and good prompt to generate what you want
especially for humans
and the upscaler is trash
IF can do text better, but for art stuff it's meh imo
agreed
It's more of a realistic image generator
when's the next #🏅|pantheon image
Unfortunately a solid 27min away because the bot had a hiccup and the timer reset
OOF
I will improve on that 🙏
it's ok
Pardon the dust folks
Yep
ok everyone there will be a small hiccup as we are going to shorten the length of the cycle to 15min! You may lose some generations in progress, I apologize in advance
It's ok
and we're back
How do we get onto the pantheon
Get to voted image on #1100484581037195384 I believe
I don't know if showdown is ment to update now with the previous batch though
Ok thanks
I like the different emoji which comes up after you vote, nice touch
how come there is nothing in showdown now?
It's a bit delayed, I will work on improving that
It should happen in the next couple minutes though
does it go images are generated -> 30 mins later top generated images go into showdown -> 30 mins later top image in showndown goes to pantheon -> resests?
ok ! it's extremely late but #1100484581037195384 is up for voting! You guys have like 3min so please make it count
I will work on improving it now
oh man you guys are making some amazing stuff!!
everyones had a lot of time to practice
Sheesh. 3 minutes? Aight.
It already went
ok so I'm going to make a change quickly here, I'm not sure 100% it's going to work, so please bare with me if the cycle breaks for the next round 😅
Congrats abecar for winning the 3 minute challenge lol.
How to give it negative prompt ? And, can we share the images on social medias eventually ?
The negative prompt can be added as an argument (just select it from the list after you type in your prompt)
Share away as you wish~
Great thanks 💖
Youre more than welcome to share it 😄
After this cycle we have a reboot and there may be issues - apologies in advance if so 😅
Due to a change in the cache structure, it might not be safe to redream or restyle images from before the restart, so if you're reading this please refrain from doing that until the cycle after the restart.
Sorry again for the inconveniences and thank you so much for your feedback and early use!
Up to 3 images will enter the Pantheon moving forward as well per cycle
anyone knows why I can't dream a "spritesheet"? content blocked?
Not sure about that, will check!
There's gonna be a few ups and downs incoming
ok that looks... ok
Lets do a full cycle and see how it goes!
hmmm still only picking one
ok that should be the last reboot for awhile
and sorry for the early cycle
There may be a couple dupes in the next cycle as well, but smooth sailing afterwards. Thanks again everyone!
Quick maintenance
can we have a pinned message for what the command is to use the bot?
WELCOME TO THE SDXL RESEARCH BETA! 
This research preview aims to gather RLHF (Reinforcement Learning Human Feedback) to refine our latest Stable Diffusion model, SDXL.
Your feedback on voting for generations in #1100170312106127410 to #1101178553900478464 will contribute towards the latest Stable Diffusion model!
How to use the bot and start dreaming:
Type
/dreamand enter your prompt in any of the bot channels #1100170312106127410 and above. There are additional parameters you can adjust to your preference!The bot will generate two images, labeled A and B, and present them to you. Take a few moments to examine both images and decide which one you like better! You can also vote on other user generations.
Images will then be periodically sent into the public channel #1100484581037195384. Users can vote on these images, also labeled A and B. Once voting period is up, the highest voted image(s) will be sent into the showcase channel #🏅|pantheon and be forever immortalized!
Additionally, by participating in the A vs B voting, you're helping to train models by providing valuable feedback on which images are preferred by humans. While being a fun activity, you're also helping to improve the quality of the images generated by the bot in the future!
You can also use the /dream_stats command to get a summary your personal results!
Discord Bot Terms of Service: https://stability.ai/discord-tos
Will there be a special role for the accounts of the winners in the pantheon?
No extra roles as it's for showcasing the winners, sort of like #1072233397746929755 but for bot gens ^^
Thoughts on the cooldown? Is it ok?
it works
gives time for voting
I didn't notice the cool down but I did just start prompting about 3 minutes ago.
Might be kind of neat to keep track of which art styles are winning the pantheons?
I guess you had a bug with tied votes?
Love this one
Do you also have a fold?
❤️
I personally take the time to look at the prompts when voting, but a lot of others don't. But...we also are in a phase where we need the prompts to be thought of right? not sure
Yes please send YOUR feedback here in relation to your experience on rating from MJ etc 👏🏼
Stress testing any model requires prompts of all shapes and sizes.
I understand the style adding cause it's on dreamstudio, but will the wider pop? should it just be appended the prompt maybe? that would be my only input.
yep, 3
waiting for some better foldable to come out
I gotta remind myself that voting and saving are two different processes with the bot so sometimes I vote for something and forget to save it 😐
Nice! I have the exact same one and am waiting for a better foldable too, I just can't go back haha
yeah no way I'm going back to normal phones
pixel fold may be an option, but them bezels
:X
only Google can spend years developing a product that's focused on media consumption and end up putting bezels on the main attraction
Eventually, those styles won't be appended though.
They'll just be baked into the checkpoint
May the best Mona Lisa win!
only 1 mona lisa, but it's a good one
oh wait
that's the other chick
😄
surprising honestly, there were some good Lisas
Oh nooo. I have to choose. They both are good though.
Just want to thank you all for your valuable efforts, you created some amazing images!
I'd emoji you but I guess that's not allowed heh.
why are my stats all 0
There's a bug with the query where you need a pantheon win to show stats. I'll fix that up today
Your stats are there though 🙂
aha
2 of my generations on pantheon!
where'd everyone go 
Some of us have to sleep lmao
Loving this version seems a lot more coherent and it's easier to prompt in normal language.
Yeah with two words you can get an amazing image
Can everyone see showdown and pantheon?
Yep
Everyone can vote in the showdown and see the Pantheon
Are you unable to or just wondering for the entire server?
will keep improving also 😄
so please continue telling us what you like best
I like the text!
Cool, it's a lot more coherent and doesn't need the negative prompts v2 needed. The resolution and detail in the images is really good, contrast is great, seems really versatile for different styles. Just a much improved model in general... Need to test more to see how much you can steer it through prompts, it makes awesome images but would suck if they all end looking the same as Midjourney did at some point. As I said it feels really versatile but need to keep testing. For now I'm loving it, great work...
Really appreciate the feedback! Definitely trying to make the model feel good to use
how are the styles applied in the bot channels? underwater a set of other keywords related to the chosen style?
Currently yea but that is just temporary for the moment
works very well, they are probably very short prompts?
yea nothing too crazy atm haha
The model should listen decently well to prompts (I hope) so you often dont need too much to vary the style or output. Then again we are also still learning how it reacts as well as all of you. I have seen some pretty surprising creations here in just the last day
its really awesome what you all are making 
getting results is pretty effortless already, how good is it going to get? will it get to a point that you can add endless concepts and it all knows how to merge them? if you do that for long enough for text2image, maybe that'll become sentient first instead of the LLM's 
Can they react to pantheon with 🍪?
Yep
Ok, it's just there aren't many reactions, I guess it's the speed it goes at
Not sure how many people are aware of it either?
Was there a communication to the server about them being there?
In the announcement it mentions
Small quality of life improvement suggestion: add a button that lets you dm both pictures instead of one at a time.
Loving this painterly style i'm getting by default with the neon cityscapes
Very different from most generations I usually get
I think it's not picking images for the pantheon from bot 2 anymore...
bot 2 doesn't exist for me
I guess the #2 pantheon stability.ai logo I did wasn't accepted 
it was very pretty, but I understand the confusion it might give
Hi everyone, #1103708504142925824 indeed isn't cycling with the game at the moment. Will restart the bot after the next cycle to include it.
are you not able to see #1103708504142925824?
apparently I had to add it to my list of channels
it had disappeared
sorry about that, it was replaced recently to add bot-1 into the fold
aha
@static crater do you know if the previous #2 pantheon image is recoverable? or did a mod hard delete it?
pantheon images are all kept in #🏅|pantheon
do you mean a generation from the #1103708504142925824 channel?
yes, but it seems mine was removed, no; #🏅|pantheon message
oh that's odd
yes, maybe a good reason for it of course, but I'd like to retrieve it if possible
it may have been due to the restart #1103056510248026172 message
it was bot-4, pretty sure
maybe it was Emad, taking revenge for me generating clowns in the past
@lethal haven
Images made after noon in #1103708504142925824 are not tracked for the game
It will probably clean them without moving to showdown or pantheon
It's not part of the cache because the channel was not in the list to track
New ones from here on out should be fine
I think it was bot-4 though, but if it's gone, it's gone!
Only the top 50 images make it to #1100484581037195384
If they didn't have enough votes they might not make it especially if it's busier
it was #2 winner of pantheon though, don't understand how that related to the top 50 images, it already arrived in the pantheon
It got removed from pantheon you're saying?
yep
I think it was a mod that didn't like the fact it represented a logo for stability.ai
It doesn't delete from there it only deleted from bot channels and the showdown
which is fine, but would be nice to know (and if it's not a bug also of course)
Might be a bug
We've had 1 blank in the pantheon before
Was it ever there or always missing the image?
how can I help debug 😄
I haven't seen it arrive in the pantheon myself
it was gone the moment I opened the channel
Ok might be a bug I'll keep an eye
If a mod had deleted it, the entire post would be deleted, not just the embed image
yeah, makes sense, or show edited tag or something maybe?
is it possible to regenerate? is there a history of all prompts / seeds?
must be, it's also used for RLHF right?
not sure I understand the reason for the showdown. Bot I get, the model can learn the better of a prompt, but showdown compares two completely different prompts. is the goal to only learn from the pantheon?
it's iterative voting, top voted generations from bot -> showdown -> pantheon
we're gathering data from each step to improve the model
ah ok, so each step contributes, that makes sense then. thanks!
and I realize I just reworded your answer..... long day
I'm toying with using classic prompts vs natural language (and bumping up against the token limit)
it's doing really well with natural language!
I missed the stage earlier today, was it recorded?
Hi Eclairs.
hey geck! whats up
I've been working on my POW image.
alright thanks
Another one? 👀
im looking at the pow archive, .. its soo impressive what people made with sd
Yeah, I was pretty close last week, but people are too good at those.
i can't compete with them, lol
It's 25% the model, 25% the prompting. The rest is inpainting and good seed rolls.
Do the cookies in #🏅|pantheon have any purpose?
wow the latest gens are looking crazy cool
bot
alr
And new model
it's the same one you used if you were in the alpha and beta, this is just open to more people
but we'll be updating the model as we go along
I read between the lines and thought kaj was saying you guys update the model.
can you remove the stats command (not dream_stats)
@mint forum you are official with Stability now!? thats awesome. Bodes well for SDXL training toolsets. Congratulations ❤️ !!!
Any chance this could be intergrated into an API? Not the generating part but the rating feedback and stuff, for invoke and plugin for webui and stuff? Could be cool to have a pantheon that can connect to all the local ui’s and people can submit their nicest generations etc
Sometimes I feel I need a “neither” vote when neither of the options match the prompt well.
That was thought of...
negative prompts are not working? did I break something?
Why was my prompt of "sad robot floating in space" content blocked?
Negative prompts are... finicky for some reason. Looking into it.
I saw some negative prompts where the key is negative: instead of negative_prompt:, but the autocomplete is negative_prompt:
could be that¿
this works #1103708504142925824 message
Ah yes sometimes it works
again, check the difference betwen how sometimes is called "negative:" and other times "negative_prompt:"
I'm going to give some feedback here - you guys are making amazingly great stuff! Keep it up! 😄
That's interesting, will definitely take a look!
the bot autocompletes as negative_prompt:
I mean... not the bot, the command
I am, thanks 🙂 luckily for me the SDXL team is filled with top talent to the brim
do our images get deleted?
If they do not make it to the pantheon, yes. If you really like a generation and want to keep it press the corresponding ✉️ button to get it DM'd to you
thanks 🙏
is the sdxl bot access limited or will it continue being free?
From past experience this won't stay up. It will eventually come down.
SDXL bot is mainly used to collect feedback from you guys so the model that gets released later will be better 🙂
i can certainly see improvments in the model allready from when it first rolled out
yea
You should be good now
SDXL is nowhere near the quality of Midjourney, even the older V4 models. Shame that's the case even after all this time to catch up
That's a good question, and you're not the first to ask. We might have a proper answer for that in the future, but for right now yeah just don't vote i guess
are there any weight modifications like brackets or "!!!"? or maybe a way to switch between two prompts each iteration like with [example1 | example2]
Btw should I also vote on image other people generate in #1100170312106127410 /2/3/4/5 (I've been doing that) or should I only vote for those generated by me and vote for others only in #1100484581037195384 ?
Also, what to do when they're both good and follow the prompt?
Definitely 🙂
Try and choose your favorite of the two 
If you really feel that neither fits and you dont like it you dont have to vote and that will be accounted for
I would say try and pick a preferred when possible though if one exists 🙂
👍 Great then I will continue to do so :)
Just curious, is it training during the pauses?
Its not actively training during this process but we are gathering the info to improve it with ourselves. Training is still manual for now haha
We will likely be updating it regularly though in this period (sometimes without warning/notice)
oh ok, thanks for the info
Anytime 😄
also a fully automated training loop for a model by community feedback does sound like a cool potential future experiment 😎
Do you all feel the bot category should continue to live up here or move down to just above the General category? (Or somewhere else)
I like it up here it's easy to find and feels organised, also imo the General category might become be a bit crowded if all these channels were put inside it.
Thanks 😊
Up, plz. Easier to find.
look bug in loop here waitting lot of hours bot continue say still dreaming!
anyone got this issue ?
I was in the loop once for ~7 minutes but I eventually got an image
thank's ok i'am block it's look more 7 h here ^^
Negative prompt still broken?
From what I understand you are randomising the settings (cfg, steps?) to gather info. Wouldn't it be useful to have a "none" or reject button when none of the images are good? Sometimes the settings seem really off.
I mean we can use re-dream so it's not a problem for us but you are not getting the feedback
Hi! It should be good in a minute or two, happens on restart sometimes.
If neither image are good, at this time just do not vote for either one. The vote isn't mandatory, and no vote is sorta a vote 🙂
This should be fixed now, hopefully for good 🙃
how do I make it add the text correctly?
kinda wish there was a way to change your vote. sometimes I vote for one that is really close to the other but then decide I actually like the other better
dreamer shows me:
Hang tight, you're still dreaming! :)```
but I am not
what happens?
Oh no ok on it
I think I had a similar problem
Quick restart
k 🙂
It's back up. If anyone gets locked out again please ping me.
Sorry, but it also tells me that I’m still dreaming, but my last dream command was about 40 minutes ago, which never finished. (Before that I also had about 5 dreams not finish but at least the error was going away after about 10 minutes and it was letting me try again)
"still dreaming" problem as well
I perceive the "voting game" as going a bit too fast, relaxing some of these timings and expiration times would be better for the mental sanity of the players involved.
Sorry about the locks, trying to get it fixed up for good!
Longer time between cycles? Longer pause for voting? More pauses for voting? What would your suggestions be? 😄
Nah, no longer pauses. Just leave it like this. We just need to adapt.😊
- longer time between cycles, so you can experiment and iterate a bit longer; the history also gets cleared way too fast. Leave it up for an hour or so, at least.
- the mandatory pauses for voting are ... a bit unnecessary, no? The reminders to vote should suffice.
I like the mandatory pauses. They slow things down which is good imo. (you could even make them longer imo)
- also, leave more time for voting, so you can properly look at the images and the prompts before they zap away.
sure, slowing down is good. But it's also quite anoying if you just want to get your result, and subsequently proceed to vote. It breaks the flow.
or at least process the queue that is already there when a pause occurs.
I'm still locked out. I'll leave it for now. Have fun all!
Sorry working on some stuff - bot will be intermittent for a short bit
Dream worked for me again - thank you!
Bot seems unresponsive.
Btw: "/dreamstats" does not work while you are waiting for an image, this isn't a huge issue but it might be nice if it did
seems to work again :)
Yep!
am I just imagining things or are a lot of the faces and necks elongated today?
World giraffe day
Hey everyone! Pretty sure the forever-dreaming issue has been resolved and you should now get a message if a dream ends uncompleted so you know you can vote again.
I get that even on 1.5 but yeah that happens strangely a lot...
Did the bot crash?
Doesn't seem to be working
🤷
@lethal haven Bot is offline, mind checking what happened?
right, it's stopped working
Anyone knows what the numbers 1265, 198 at the end of some prompts are?
Back up
I think that might be height and width
Hmm, don't think so, the proportions make no sense
Only some of them have it
what happens if we appear in #🏅|pantheon?
It stays there
like a showcase?
I think so
Not really expecting anything else
You also get higher on the /leaderboard the more you have on pantheon so there is also a sense of competition
Think of it more like a starboard. No real reward but karma.
there are default negatives that are applied to every generation
but you can also add negatives on the bot ui I believe
also wrt to /dream_stats it would be really cool if it also told you the absolute number of images rated/generated by all users.
Whoevers prompting the lyrics to black parade i see you lmao
is there a list of restricted/banned words?
@edgy tide that one with the glasses that also looks like wine glasses is A+ lol nice job
Nice update!
In messages where it tells you to vote, it links to this channel which appears as "🔒 No Access" https://canary.discord.com/channels/1002292111942635562/1101178553900478464
bot 6+ will be opened if more channels are needed, ignore the 'No Access' for now!
Hi there, I wanted to share some thoughts:
**-**Regarding the new system with the 2 cycles: If I understand correctly, the cycles are now twice as long with more images are being added at halftime?
If this is the case, I think it's a good idea (showdowns with only ~4 images are sad). However, it might be worth considering making the number of images added andor if there is a second cycle dependent on how many images are currently already in showdown. (iirc it varies quite a bit over the course of a day)
- Also having the top message give the time until the current showdown ends might be cool
**-**Another idea could be to pair up similar images in showdown more often. This could potentially make comparisons more meaningful, but idk.
**-**Lastly, I was curious if there is any specific reason why reactions are disabled in pantheon?
Overall, I appreciate the effort you're putting into this, the application is fun to use and I'm looking forward to the data gathered further improving SDXL :).
really appreciate the feedback ❤️
What does styles actually do. For example, photographic usually zooms in and fantasy art often adds a female character. Does it add to the prompt, and if so is there anywhere I can see what?
Tbh it's not clear to me if it's ok to vote for more than just one image in Pantheon?
its says vote for all of your favourite generations so i presume that includes more than one
and do you mean showdown?
yes, showdown
Yes you can (are encouraged) vote for one image from each pairing. The more images you compare the more the dataset grows :)
i noticed it is possible, but i stopped doing it because it's unclear to me
that's what i'm doing in the bot channels
look in showdown channel description
i didn't notice that, thanks!
When it tells you that it is time to take a break, it would be good if that message included a retry button to just click and submit again the same interrupted generation... sometimes I'm writing a prompt and then finish it only to get it lost in limbo
Yes, please....
Quick tipp if you are on the computer you can press the "up"-button and get the prompt right back.
Also feels like there are too many break, or too long, feels like I'm interrupted a lot more than before.
The pauses are a bit longer at the moment. Could stretch the rounds out to twice as long, that was the original plan as well.
After this pantheon cycle I will restart with double the time between cycles.
Due to the way the timers currently work if the game is busy it'll be an extra min or two cooldown. If it's too much we can reduce it again 🙂
Please keep this feedback coming! Invaluable 🙏 I really want to make this as fun as possible
Question about neither. I pick it when i can't decide because they both follow the prompt and i like both of them. If they're both no good i pick none. Is that how it's supposed to be used? @lethal haven
It's the other way around I think i.e:
#1103056510248026172 message
Only use
if neither image is worthy of a vote.
I want to be sure, it's not clear (to me).
😅 This is why we were not going to put a "Neither" option in, due to that confusion. It should be for when both images are "bad".
We may change it to 👎 A and 👎 B but not too sure yet 
Once again, that feedback is great! Perhaps we just have to reword it/change the emoji/etc.
Oops 😅
The "neither" option is also a difference to the http://pickapic.io/ site which users might be familiar with. The setup there is rather similar and the "neither" option works more or less as Eloyse assumed it did.
SadCat will be making a sad departure today! Overall feedback has been that it mucks up the UI flow and it's purpose vague - we will be back to don't vote for either if you feel neither deserves a vote 🙂
Tip for anyone who's creating (seamless) tiles, you can test them in your browser at https://www.pycheung.com/checker/
is it impossible to use seed?
i mean, sometimes when I want to reSize or reStyle, I want to edit this particular seed, I dont want everything to change
We had it like that before, the issue is it ends up causing duplicates easily 😦
right click- apps- report to staff
Any tips on getting interesting backgrounds? They're usually just Grey for me
sad
in what context
When I'm doing simple prompts of objects
Actually it happens in complex prompts also
can you show an example?
Like my one in the showdown
Most of the images have very simple backgrounds
aspic aspic catgirl translucent echos to infinity 45mm food photography Terry Richardson y2k plastic hell hell angel Satan satanic aspic casino slot machine fisher-price water pipeline hail satan oil rig on fire
that prompt is a bit all over the place
I'd start my simplyinfing it
I like my word salad
legit, but it's evidently not working well
I'd like to help but I can't understand what's the intent in the prompt
I have no intent
I like pushing the ai to it's limits and seeing what comes out
In 1.5 it would have elaborate backgrounds
1.5 is a different architecture
But in this one it seems to think I'm describing an object
SDXL is designed to understand the user
you'd have a better time at giving it a proper instruction, like "portrait of a cat girl" and adding random words
words like crazy colors, etc would probably help
I'm trying to keep my word salad but add phrases to get it to think up a background
Adding "landscape photo of" is helping
Even looking through the pantheon I'm seeing a lot of images with very detailed subjects and sparsely populated backgrounds
Obviously prompting is different than other versions but it seems harder to get it to fill in a good background
Yeah I don't see that personally
plenty of images with good background detail
etc
keep playing with it, you got some cool generations going
They need to pay me to prompt instead slinging Javascript
reSize stopped to work
Btw @spice sparrow that first picture of yours in #1100484581037195384 (or that style in general) is really cool imo, also imo it would work perfectly for this weeks "picture of the week" event #⭐|pow-info.
wow, thank you
😳
never thought about this event, but posted (hope in correct chat)
I was working on my submission and actually trying to get something similar in style XD. So seeing your images I thought I'd point you towards the challenge :).
😍
thanks
used resize button 3 times, got the same 3:2
Oh that'll happen, it's random

Discussing changing it to be the same seed, but then it would only do 2:3 and then 3:2 and no longer work at all (because all combinations are done)
The current setup randomizes seed too so it can keep doing it and not get duplicates
isnt it easy to add check func to start from the begining again?
looks like iam cursed, + 3 in a row
If you want to control the dimensions in an easy way, just (click) arrow up for your last prompt (as long as the channel is not cleared) and change it
yw, i got the tip from somebody else
why almost every time when my prompt contains nothing about persons, in Fantasy Art i get a woman 
Some feedback / requests:
- Regarding resize, it may be nicer to have three buttons for the proportions (landscape, square, portrait). (Unless you are planning to add more proportions)
- Would be nice to be able to DM ourselves images from the Showdown
- Not too sure about this one, and not sure if it's even possible to implement, but having a special vote that counts more (like 2 or 3 points) once per showdown may help highlight really good images and also break the uniformity in the votes (often the best ones are all tied up), this may help with that.
oh, and the model is amazing, I'm super surprised by how versatile it is! beautiful...
What's the best way to make text? in been putting mine between quotation marks
likely that Fantasy art is a bit skewed towards females, you should try to construct your own prompt with fantasy themes and see what happens
a sign with the word "SDXL" usually works for me
but it may take a few tries
For example: a sign for a movie poster “Me too Island” Kevin Spacey Bill Cosby and Charlie Sheen?
you can try, longer words will likely require more tries from my experience
Thanks!
Is it just me or something wrong? I'm unable to do anything. Can't vote, can't generate.
Oh wait, #1103056510248026172
Yeah sorry bot will be offline for a short bit - apologies!
no worries, np 😊
What do the styles actually do? Like do they add things to the prompt or use different models or what
I would like to know also. If it was injecting into the prompt it would probably effect the token limit and it doesn't seem to. I assumed it was either a separate model or something like a lora
I’m also curious, because the different styles greatly affect the content of the scene and composition, not just the appearance. I’ve found that prompting for photos while negative-promoting against comic, anime, etc. and selecting a style like Fantasy Art can often get a photorealistic scene of more interesting things happening, compared to the Photo style which is usually just facial close-ups. So maybe the styles are different sets of training images of different kinds of scenes.
I’m very impressed with SDXL’s text generation. It seems comparable to the Cronos (I think) blend that used to be on NeuralBlender, which until now handled text better than anything else (besides Imagen), and I was sad that it was removed from that site. Now, SDXL’s image capabilities which of course blow away the old NeuralBlender, combined with its text capability, have a lot of potential for graphic design inspiration.
For comparison, I tried the equivalent “Midjourney” movie poster prompt in Midjourney. As of 5.1, the text is not even close (although they’ve said they’re working on it for a future version).
I think they said they are adding keywords to the prompt, and they plan to turn it into Loras later.
@lethal haven, got a suggestion / question... I noticed something I'm doing which may be worth discussing. From what I understand only a sub-set of the images peoplle vote for go to the the showdown (is this based on votes only, or aesthethic rating, or something else?).
The issue is that sometimes I avoid voting on images not because they are both bad, but because I don't want them to be in the showdown. I normally start iterating an idea and adding or removing parts of the prompt to see what works, this means at the beginning the images generated may be ok but not what I'm after, and since I will keep iterating on the idea I don't want to spam the showdown with images of an "unfinished idea"... hope this makes sense? One solution may be to have a button to send an image to the showdown, this would allow anyone that sees an interesting image to send it to the showdown, but not necessarily every image gets there.
I think it may make a better game, but not sure if this is a good change for the learning process... but we may end with a better selection of images in the showdown.
Or should I just vote everything?
About 20% of the time this prompt triggers a “content blocked on this server” error, and I’m sure it contains nothing objectionable: A warning sign with the text “Warning! Handsome Dude Ahead!”. The words “Warning! Handsome Dude Ahead”, written on the road warning sign, traffic sign that says “Warning! Handsome Dude Ahead!”, in front of a handsome dude standing in the road. —— Most of the time, the same prompt runs ok. The error has occasionally happened with other prompts that contain no phrases or words that could be considered rule-breaking. I hope that I’m not bothering the admins by triggering the error with innocent prompts such as these. Does it happen to anyone else?
The people need to get hip to little jelly bean men
thank you so much, this is awesome 🤟
Hi! The Showdown is based on top votes. While we are in restricted access, and especially during quiet times, you can definitely see a lot of your own stuff go up. I'm hoping once the bot is opened to more users this will no longer be an issue as the handful of people using the bot now will not be the only ones participating and voting on images 🙂
I'll try to see what's happening here - the message you're referring to I don't think is from the bot itself, so I can follow up with the server admins to see what's going on
@frigid pebble you have some cool ass prompts
hah, thanks!
having lots of fun just trying ideas
it's nice to see what everyone else is doing, learning lots as well
are artist names included in the data set?
many work
ow nice!!, then i'll try some 😄
it didn't recognise starwars in my test
yeah you can only try, also the styles are sometimes not too accurate but still have effect
some of the star wars characters are there, not sure if it would work as a style
can we use the style thing for any modifier or is it a set list?
it's a set list, I think it adds some keywords to the prompt
hmm okey then I'll just stick to prompting
the styles work really well (mostly) saves time
Ye there just not what i want them to be. I prefere other style's but that may happen when we fine tune the model
Is this model more similar to deepfloyt and should we there for change our prompting style?
Don't think you need to change much, to me it feels a lot like normal SD, maybe a bit smarter, understands more and the images are much more detailed and consistent. haven't used deep floyd so can't compare to it. Also you don't need so many negatives as in 2.x
I don't use that many negatives in 2.x it confuses the model. But it's true this has a better understanding and it can reproduce text
not sure why but sdxl thinks all llamas need to have at least 2-4 mouths. i haven't been able to generate one with a single mouth. https://media.discordapp.net/attachments/1100170312106127410/1106730508337020979/detailed_photo_of_a_anthropomorphized_llama_wearing_a_black_hoodie_sweater_male_features_hood_up_over_head_at_night_in_a_city_alleyway_style-Photographic_seed-3041888724ts-1683935512_idx-0.png?width=671&height=671
Are all the bots dead right now? Getting "The application did not respond. Only you can see this message." on multiple #bot- channels.
Was hoping to generate new PowerPoint slide images for a talk today 😭 . Oh well, my older SDXL images are still awesome so I'll just keep those.
@lethal haven
Back up now
Negative prompt is still not working?
“Application did not respond” for dream, and “Interaction failed” for voting
@lethal haven bot's down again
Back up
Can I ask you a question?
Good morning all, is the bot down for everyone else 😁
yes 😩
back up.. 😦
are nagatives on XL actually helping with quality?
not sure, I almost never use them, and haven't seen anyone using them much
would be interesting to compare with/without, but I don't feel they are needed like in 2.x
I think they mentioned they are adding negatives automatically, but I may have misunderstood
Yes they mentioned a basic negative prompt, not sure what it was and can't find it either. I think if you need them, just use them sparsely and focus on what you really don't want to see. No guarantee they work, though.
It's hard to make comparisons at this point because they are randomizing the settings (step, and cfg) so I don't think it would be too clear if the negatives have a big impact in the image quality
negatives can be powerful with these models from my experience
Perhaps you can give us some examples?
what do you mean
I think she means an exaple for this particular model. I use them all the time in other models but not in this one, but as I said I find it hard to test the effect on this one because the settings and seed are randomized.
seems restyle only does comic, anime, and fantasy. Any others?
restyle only picks from the pool of styles you can choose yourself in the dropdown
how choose
Go into #1100170312106127410 - #1101178553900478464 and type /dream, followed by what you want to make
i know this ,but watercolor style so weak,do not like watercolor
It's not like watercolor at all, it's more like a CG oil painting
米兰
Which model is used in the bot channels to generate images? I'm trying to use the same prompts in my locally running stable diffusion but getting very bad results.
it's a new version, SDXL, still WIP and they are testing it here
the same prompts won't work as it's a different model
Wow it seems to be a huge improvement over SD 2.1
Like massive.
Where do I find the Bot layout explanation, of upscaling one of the 2 images, and what all those icons mean? PLS.
yo
has generations become free on DreamStudio also? or the shenanigan is only applicable to Discord?
how do you add the bot to your own discord?
exclusive to Discord!
awesome! ty for letting me know 😄
you don't currently haha
ahhh nooo, midjourney hates me now so i kinda need it for 3d concepts.
mj hates you? you're a madman on the run? 
the bot traffic isn't tooooo crazy, especially closer to the bottom i'd say!
so, until there's bigger updates of wider-launch, i'd say strategically gen in less traffic channels.
best of luck!
Yeah, they used to make a 20 image limit so i kinda.... made 25 bot accounts so i could keep using there amazing service lol
you're a legend hahaha
They helped so much while doing my diploma
wonder how long it will be until stable gets to the same quality as midjurney
a friend gave me some generated low poly mountains pics and i remade them in blender, so it isnt to rough
from stable
when i press tab i only have a negative prompt, style, and aspect ratio options, i also only have 4 style options, is this normal? or is there a way to see more settings?
that is normal on the bot, for now it only has a few options here
can i upload an image as a reference here?
No reference images for now. Only pure text prompting atm
"/dream"
How does this work? Prompted, but nothing happened .
#🗣|artisan-support-feedback message
Follow the instructions there. There is a tiny chance the image doesnt come through and in that case you can try again 🙂
Thanks
@hushed skiff hi
What's the command to generate an image?
dream
Ok, thanks
Please refer to the pinned message for info on dreaming 📌
Can i add the bot to my private channel?
I was asking that right now, thank you.
It's only available on this server
Well thankx you
is @hushed skiff open source
someone tell me, from bot-1 to bot-10, is there any difference between these bots? Does it have anything to do with the model?
the channels are identical in terms of using the bot. Just 10 of them for now to not cram everyone in 1 gen room 🙂
thanks
Why in each bot, the message always shows blocked? Always need to manually click on the display to view
How to use stable diffusion in discord
Gm gm 🥂✨ wonderful to get a piece in the pantheon 🥂 Qs are we allowed to post the work to socials from new SDXL bot? Should I tag it SDXL ?
you can and if you'd like to say it's from SDXL on the Discord that would be great
Wonderful thank you will def tag ✨
is @hushed skiff open source
👋 Hello! 😄 I would like to know when more styles will be added to the bot?
how do i flag inappropriate content? nm found it
The application isn't responding to me 
Short downtime for some maintenance! Should be back within 10min
Dope ok. Addicted to the ïmağeż
just an idea: it would be interesting to open source the bot or allow the bot on other discord servers while still piping the results back to be learned from. not sure how realistic that is as I imagine it would set your GPU's on fire but a wider audience using the voting system is never a bad thing.
ann tamaki, cyberpunk tech wear, no color, full body shot, beautiful woman, perfect body, long hair, elegant, gentle, portrait, coloring page for adults, extreme details, no black fill, line art, vector, cyberpunk feel
Style: Fantasy Art
which model
and how many numbers of steps
and how much prompt strength is used here in this discord server
I can't install stablediffusion because gfpgan never downloads
venv "D:\stable-diffusion-webui-master\stable-diffusion-webui-master\venv\Scripts\Python.exe"
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: <none>
Commit hash: <none>
Installing gfpgan
Traceback (most recent call last):
File "D:\stable-diffusion-webui-master\stable-diffusion-webui-master\launch.py", line 369, in <module>
prepare_environment()
File "D:\stable-diffusion-webui-master\stable-diffusion-webui-master\launch.py", line 277, in prepare_environment
run_pip(f"install {gfpgan_package}", "gfpgan")
File "D:\stable-diffusion-webui-master\stable-diffusion-webui-master\launch.py", line 143, in run_pip
return run(f'"{python}" -m pip {command} --prefer-binary{index_url_line}', desc=f"Installing {desc}", errdesc=f"Couldn't install {desc}", live=live)
File "D:\stable-diffusion-webui-master\stable-diffusion-webui-master\launch.py", line 111, in run
raise RuntimeError(message)
RuntimeError: Couldn't install gfpgan.
Command: "D:\stable-diffusion-webui-master\stable-diffusion-webui-master\venv\Scripts\python.exe" -m pip install git+https://github.com/TencentARC/GFPGAN.git@8d2447a2d918f8eba5a4a01463fd48e45126a379 --prefer-binary
Error code: 1
stdout: Collecting git+https://github.com/TencentARC/GFPGAN.git@8d2447a2d918f8eba5a4a01463fd48e45126a379
Cloning https://github.com/TencentARC/GFPGAN.git (to revision 8d2447a2d918f8eba5a4a01463fd48e45126a379) to c:\users\chris\appdata\local\temp\pip-req-build-_6uwl4pt
stderr: Running command git clone --filter=blob:none --quiet https://github.com/TencentARC/GFPGAN.git 'C:\Users\chris\AppData\Local\Temp\pip-req-build-_6uwl4pt'
fatal: unable to access 'https://github.com/TencentARC/GFPGAN.git/': Failed to connect to 127.0.0.1 port 7890 after 2085 ms: Couldn't connect to server
error: subprocess-exited-with-error
wrong channel chris
I think when a prompt has too many newline characters discord (at least mobile) won't show any images
There is something for sure for certain prompts, images are attached without extension and it's kinda odd
I'm going to test a bit when the gpus aren't napping
Yeah even a single newline and it breaks lol see #1080261341362786384
It's probably naming the file the same as the prompt and cutting off after the newline, so it's not getting an extension. Could just wash the input of newlines probably
hi guys i have a question about controlling camera movement, angles and rotation ... i have generated a realistic figure ... can anyone give me some suggestions ?
#🤝|tech-support meant for this channel?
@lethal haven I don't want to disturb you, but the bot is down
hello i wanted to know how to run stable diffusion on mac and also do animation in stable diffusion
can any one help me on this
The application did not respond
3AM for @lethal haven 😦
He'll jump on it soon's he can 🙂
the bot's still down?
Seems like
Back up soon
Maybe not - apologies, not on our end. Contacting who needs to be contacted
Oww, okay, hope all is good :))
O aplicativo não esta respondendo
Bot broke?
mds
Still waiting 😪😪😪
Sorry everyone out of our control at the moment
was the bot banned by discord for spam or something
not sure if you will be able to answer this question if you prefer to keep problematic things internal, but i'm just curious
Como hago para generar imágenes,?
we're not sure
We were too powerful
el bot esta desconectado temporalmente pero sin tiempo previsto para retorno... Lo unico que se sabe es que hay problemas con Discord y no con nosotros
I forgot what I did with my life before the bot. At least I get to see my winning pic in the pantheon for longer.
Whoever randomeizer is really gets the best of that lol
Maybe they're the one who broke it. 🤣
That's the move lmao
I don't want to disturb you, but the bot is down
Yes for me isnt working too
It seems bot down
And it doesnt let me to vote , it is suposed just cause bot is down?
There are no more images. We generated them all.
Will the bot return tomorrow? I'm a little anxious...
hopefully today is the day for the bot's return
Would a dev care to answer if it's possible that we could get img2img in the sdxl beta? Like the way midjourney does it would be fine. Although I do have doubts it will happen because the other sd previews didn't support it IIRC
Plus it wouldn't really support the mission of fine tuning the model so maybe another reason to doubt it
unlikely, sorry.
Ok, thanks
Will the bot be back?
hi, is it possible to install homebrew, git, SD on my external SSD and run it from there?
I have SD installed on an external SSD and working, so yes. I don't know about installing the other two externally, but my intuition says it would be more difficult than it's worth.
right now do you run SD on an external SSD with Mac?
I've done it on all 3. Mac, windows, linux. Moved the SSD between them. Thing is, all 3 machines had git installed the 'proper' way.
do you just pull the SD main file to your SSD and just cd stable ......
and how to install git on an SSD?
cause i've removed everything and git cloning lol, i think i should stop it now
i know pc can just cmd on a file where they wanted it but couldnt find a way to do it on a mac
depends what runner you're using. I'm using vlad's version of auto, and yes you just cd [SSD-name]/automatic/ then sh webui.sh (or use the webui.bat on windows)
by 'proper way' I meant on the computer, not on the drive. Don't know how to install it on there. Trying to install a commandline utility on an external drive and not just on the computer is something difficult enough that I'd personally choose to find a way to install it on the computer.
let me try this, thank you sir
Hey guys please don't discuss irrelevant stuff in this channel, this is for bot-support
I'm not able to create
Any wild conspiracy theories? Here's mine: Discord is holding the bot hostage for $$. "Nice bot you got there, it'd be a shame to see it rate limited to irrelevance"
I was wondering if allowing the bot to DM people tripped some sort of spam filter
Does a DM'd image have higher resolution than just downloading from the channel, or is the advantage more about having the prompt written out and the photo not taking up space on your phone/computer until you actually want it?
bot not working
It's the same res as what is in the channel. The channel gets cleared after 30 mins so it will get deleted forever if you don't have it dm'd to yourself. At that point it's inside of discord and discord doesn't purge images, so in theory it'll be there forever.
Nice ok I see. Thanks!
If there is a role or something to be pinged when the bot comes back online I would like to sign up lol
Me too lol
you can right click the #1047610792226340935 channel and set notifications to "all messages". then you get a ping when the next message is posted there, which will most likely be "the bot is online again"
does anyone know when the bot will work again?
only discord staff can determine when
considering how they're a bit slow in support from what i've heard, i bet it won't be up again until at least wednesday from my thoughts
oh how long will it take?
Ah, if you come back on Wednesday, I'm happy! I'm dying to use it again this is addictive!
i mean you can use clipdrop in the meantime or pickapic
Thank you, I loved the pic by pic... if you know another one, let me know!
Soon
any updates from discord
w
This is an outrage. I demand to speak to your manager!
That's going to be my first prompt when it come back online.
For some reason i dont have permission anymore to use any of the bot channels to test out SDXL!
bot is down for everyone
i dont even have the permission to type in the channel. is that the same for everyone?
Yep
Bot is currently disabled
By Discord (I think)
Not sure why though, probably spam or ratelimits?
bot's back on, but it still cannot dream
because it is not fully on yet ^^ hang tight
ah gotcha, what's going on in the background if it's ok to tell btw
i just unlocked the chans haha- we are setting up ^_^ , though the bot is not enabled currently
when the bot is up, there will be an update in #1047610792226340935
I think we are good to go 😉

YAAAAAAASSS 🎉🎉🎉
🎉 amazing that the bot is back :), very cool 😎
Btw is style:Fantasy Art working for anyone?
When trying to use it i get:
Oh no! 😮 Your dream didn't finish. 😕
Give it another try and see what unfolds! 🚀✨
Example of this:
#1100170312106127410 message
Should be ok now - for real now 😉
cool thanks for fixing it this quickly! :)
This ended up not working on mobile for some reason I think because there are so many members
Quick question, is the resize button supposed to work even when an aspect ratio has been entered to the original?
hi there, so over the last 2 to 3 hours, i generated various version of mahatma gandhi as CEO, college student, etcetera, and voted A or B for each outcome, and clicked the DM button ... but now i am not able to find those photos anywhere, neither in the SDXL bot channels, nor in my DMs .. any help would be appreciated, thanks.
Please check your permissions for sending DMs to ensure you're able to receive DMs from the bot.
Strict settings may not allow the DM to go through
Bot 7 photos aren't seperated, unable to vote for A or B @lethal haven
Looks like same goes for all bots
I just came to say the same 😄 DM-ing an image isn't working for me either
I think they're down again
Nooo! Hopefully not as long this time!
Sorry everyone, reverting the change to image format - bit more testing needed 🙃
We should have checked the #1047610792226340935 it's a 10 minute maintenance, so let's hope all goes well.
It will be back up momentarily
I'm like that guy from blink 182 who cannot dream tonight
I want to know, is this bot free or after some usage it will be paid like midjourney
The bot is for research, it will be free and once we're done it will go away
When? 🥺
No idea
sorry if this has been asked but why dont my negative prompts show in the bot message?
Are u using a style?
Ayy the bot is back, nice!
where is the list of arguments (e.g. for negative prompt?) I think I checked the pins on all channels but didnt see them
I see people using parentheses for emphasis. Do we have any official info on if they work here?
it has been confirmed that they do not work on the bot
all arguments show up next to the prompt when you try to enter them
I can paste almost anything except one thing, when I do it says oh no dream didn't finish. I idea what it could be, I thought maybe I had some sort of hidden special character w the wrong encoding but I basically retyped it and pasting still didn't work
Is there a list of negative prompts already built-in? If so, where can we see the full DEFAULT list?
It was a blocked word for sure. I do wish there was a more effective method for this considering some of the sketchy stuff I've seen allowed on here lmao
By "sketchy stuff" you come across, it helps us a lot on our end by right click -> apps -> report to staff
This helps us see what slips past filtering amongst all the generations!
is there any chance to make prompt related to the real picture from the llink?
i mean is there an option to make image to image prompt?
hi
@wintry wharf Hi! There is no img2img support at this time, sorry!
How do you use /dream?
No slash command comes up for me
dream command isn't there in any of the bot channels for me, i restarted discord too https://i.imgur.com/U6xlu0x.png
🤷
it should work like that but it dont be working like that
try again?
What happens when you just type /dream, without hitting enter? What comes up in the dropdown bar?
no dream command seems to exist for me, even typing just / and looking at all the commands
that's odd, try to rejoin and see if it changes anything
^
Would also check mobile to see if same issue to be sure
Since, perm wise you have no issues we can see
rejoin didn't, works on mobile tho
try CTRL+R on discord
how bizzare, it works on mobile. Doesn't on desktop, also the web version doesn't work either
can you use slash commands on other servers?
I'm just seeing the same set of commands on every server, doesn't look like they are working properly
then its likely your account settings
try settings > accessibility > use legacy chat input
I'll have a deep dive and see what I find
legacy chat was on apparently, I think that was it
Thanks!
sometimes the bot just completely ignores that I put a negative prompt in, it seems. Like, the negative prompt doesn't even show up in the text description when its generating and clearly is being ignored in the pictures
or the ar 16:9
please ! make the bot public!!!!
Porque o stable cria imagens as vezes borradas? Embassadas...
?
why does "stable" generate fuzzy images?
{"id":"a1c9fc0dae2b17eaf0614b9893cdfd7f","message":"Unexpected server error - if this continues to occur, share the error id (a1c9fc0dae2b17eaf0614b9893cdfd7f) with Stability.AI support for assistance","name":"server_error"}
Is there anyone can slove this question
Can the bot be added to our personal discord server?
Sorry, the bot can't be used in private servers, it's here for research and experimentation purposes only.
Some blurry images may be due to the NSFW classifier being a bit over sensitive - this is actively being worked on to be more accurate.
If you're referring to blurry textures and plastic like skin within the generations themselves...
Just trust me that it won't be like that for too much longer
I think negatives don't work if you structure it like Prompt, Style, Negative Prompt? Is that why the negatives get ignored sometimes?
oh, it just ignores negatives if you set a style. Setting an aspect doesn't ignore negative prompts
How do I get the bot to work?
how to add bot to my server?
just hop into #1100170312106127410 - #1101178553900478464 and use /dream
The bot is only available here right now
What about the Diffusion.gg bot ?
That uses / draw
are the bots 1-10 free of charge to use or are they linked to the dreamstudio credits?
they're free
use /dream, fill in your prompt then if you bring your text cursor outside of the prompting box there should be an option to change your aspect ratio, although you are limited to 3:2 and 2:3, 3:2 is the closest to 16:9 so choose that
Thx
unfortunately you cannot use a custom aspect ratio at this time, idk what the devs are doing with this bot atm
Before numbers have a command ?
yes, and your prompt
you cannot use it at this time, as they're doing some maintenance on it
It appearse that there are two stats commands whicha re supposed to do the same thing, however the second one produces an error when run (/stats).
Thanks @torn venture ! You're right, old stub for stats is still in there. Will get that removed 😄
@lethal haven 'naked fat girl' and sex 3 lesbians' in the showdown? 🤨
😐 Thanks for pointing that out - please report any you see. Odd that made it through the filters, will look into it ASAP
I already have reported all the stuff this guy has generated at #1101178506530013284
Thanks. Will be rebooting the bot shortly to address
These messages should all be cleared now. Appreciate the reports! 
Filter should be tighter again moving forward
I had asked in the feedback channel but why are “business” and “corporate” banned terms yet “large breasted” women prompts are not only allowed but generated?
Filter will be updating soon!
Standing by for successful de-horny-ing of the channels 😀
Oh.. also should we report things like where the image isn’t too “bad” but the prompt is thiiiiiis side of skeezy? Like “16 year old white girl in a bar” plus keywords for physicality?
yes! if the prompt is breaking any of our #✍🏼|rules-and-tos please report the message and we'll review it
Sounds good, the Discord reporting system is a bit vague but better to err on the side of caution than the other. And to echo the earlier poster, I too had seen several questionable outputs make it to the Pantheon showdowns so I’ll do my best to report them. I do think people take advantage (knowingly or not) that after midnight my time (Mountain time) the mods are fewer online.
slow bot today?
/fab
prompt for changing image size
how do I create images?
Just use the /dream command in any of the bot channels like #1100170312106127410!
is there a way that i can see what prompts other people's images used?
Should be below the images near the voting buttons...
how can i make videos using warpdifussion? any tutorial? I have stable diffusion locally, I have a very good computer
there's a way to use training images to create others images?
There is a channel for animations - https://discord.com/channels/1002292111942635562/1098959457964277831 someone may be able to help there! 🙂
There is also tech support - https://discord.com/channels/1002292111942635562/1002602742667280404
there is a fine tuning channel if that is any help! https://discord.com/channels/1002292111942635562/1026382406279770152
@lethal haven bot troubles?
If both images are nothing like what I asked for, should i vote for the one that is remotely closer or for none of them?
I’m not sure what channel the user is in, but there’s someone named “child emperor“ making some questionable prompts , and some results were in the last showdown
i was told that we should consider the prompt, but if we can't, pick the better image
also, report bad users
seems to be dead right now, i just got the same thing
ey guys, one question, anyone knows why promt is showing me ' The application has not responded ' ?
Any text I enter gives me that message
Oh, ok, thx
Hi, I know that the bot is not working now and that is why I ask the following question:
- When the bot is down, is it necessary and prudent to report it to some specific administrator or should we just wait?
If it's not already flagged in #1047610792226340935 feel free to ping a mod for it!
Ok, thanks.
Hi, for the SDXL Showdown, is it possible to add code so that you're not up against yourself? For the one that was at 11:08 (EDT) for June 2, 2023, I'm was against myself, but Fnord was in the same situation 🙂
I’ve been in the same situation a few times
Is it possible to add a bot to my private server to use it's commands?
how to use?
in a bot channel, type /dream to get started
info's in the pin! 📌
How make image aspect ratio vertical ?
There is an argument you can select after the "prompt" portion that lets you change it!
Here's how it looks on desktop:
Only in discord desktop right?
You can use it in the app/browser as well! Just make sure to check the "other" options after prompt!
I see now
Vertical is 3:2?
tropa n to conseguindo fazer o resize
Oh no! Your dream didn't finish.
Give it another try and see what unfolds! 🚀✨
Why does he say this?
I háve some : missing module pip or something Like that
May you help mé with that someone
Oh no! Your dream didn't finish.
Give it another try and see what unfolds! 🚀✨
Why does he say this?
sometimes it just fails. Other times because you've given it a filtered word or something nsfw related
Its still failing and i know that i didnt donwload anything that is NSFW realted
Currently, there is a public bot on the server that generates images available as a research beta for SDXL, you can find the current status of the bot in #1047610792226340935. There are plenty of ways to use Stable Diffusion such as the official https://dreamstudio.ai/ website or by running Stable Diffusion locally using your own hardware - check out #1080946152318443610 for more details! You can also stop by #1025467151206854736 for any issues you experience while using DreamStudio or #🤝|tech-support for any problems you encounter while installing it locally!
Why does it give me the file and not the image
Can't send screenshots but look at #🤝|tech-support
bot seems to be crashing
just came to say the same, see it in other channels now
i think we need to tag @lethal haven
@alpine hamlet if you're around, we are suffering because the bot is failing
That does bring up a question: do we tag mods, devs, "stable diffusers", guides... ????
For bot, @lethal haven will look into it. Please hang tight (:
Take your time ... I'm just avoiding work 'cause my $400 million contract is blowing up 🙂
Yeah feel free to tag me any time for bot issues. Unfortunately I was in a meeting at the time the troubles were happening but I don't think it was anything on my end specifically and seems to have resolved. If anything else happens feel free to ping me 😄
You also cannot use "kids" or "child" or "chidlren" in many cases
you can use "boy" and "girl" though, along with the age of the subject(s)
i believe this is because so that people won't create inappropriate content that has minors in it
i had some with "boy's" get shut down last night
like they always kept failing to dream?
yeah- i mean, it wasn't a big deal for me, found other words, but i couldn't see any other words that might have swung in a bad direction
not sure if there's a word filter before sending it off for the bot to dream
how to use
check the pins, also head to a bot channel, and type /dream to get started
Bot sent me files in document format but i cant open them it says unable to open how to fix that?
rename them .png
I'll squash that sometime soon, just avoid using pastes or special characters
i mean this is the wrong channel, try the #🔧|finetune channel
I see. Thanks.
you're welcome
It's still a little annoying that we can't put in "kids book of" whatever because "kids" can be misused. I totally get it
Cool. Can we also then add "big boobs" or "big breasts" to the naughty list? I mean, we're trying to keep this PG-13 and all, aren't we?
The boobs arent even big
I tried using the large boobs one it literally doesnt work
Can we use weights in the prompt when using the discord bot?
There is no per-prompt weighting on the bot for now
once the model is open sourced anyone can do what they would like with it. On here though we are keeping everything pg-13
I suppose I mean will the base model be like SD 2.0 with very little NSFW training data?
Cant talk on training data specifics yet. Just know the model is much smarter and knows people a whole heck of a lot better. This is definitely not the same model 2.0 was
minor annoyance, but lately discord stopped showing the "open in browser" when viewing an image from the bot. This only goes for images from the bot. (i've restarted the discord client, but no change)
i clicked re-style and it did the same
same for my own bot that i've coded. right click -> open link
oh. i knew this, durr. thanks
the problem is my browser now downloads the img
that's annoying, not the case with chrome, at least for me
mostly i'm annoyed because it's habit for when the client won't load images, but the web link pops them right away
Is there anyway to add the generation bot to a server like Mid journey? or is it local only to this server?
only local for here, we use it for research purposes here 🙂
Can I upload photos to stable ?
I was trying to help someone out who was typing the wrong command in one of the bot channels i told them to type / dream and a bot removed my mesage and told my content like that wasent allowed
happened to me a few times too
for some weird reason the server managers added (slash)dream to the automod filter. why is this?


