#💬|general-chat
1 messages · Page 31 of 1
Like ((((jeans shorts)))) would be the same as jeans shorts:1.4,
If you go on 1.5 or above you really force it to be in the image
any particular advice for something like if you want cow print clothing but adding "cow print clothing" makes the model generate cow ears+horns+cow bell even when you have all of those set in the negative prompts? XD
When you want to hold something in your hand in drawings, artificial intelligence usually cannot
do you have any advice?
I used cow print after the clothes tags and it worked normaly
Use the tag in the back of your words
it seems a lot of this depends on model. i've had things that i wanted sitting on a table end up in the persons hands when i didnt want it but then they refuse to have it in their hands when you want it XD
So like "jacket fur cheetah pattern"?
instead of cheetach fur jacket?
Exactly
Hmmm
I'll fiddle with that and various weights too. perhaps adding extra weights to the negatives as well
lol
Or like bikini, cow print, or Jacket, snake pattern,
i had an issue a few days ago with bunny ears. trying to make an actual bunny anime character and it'd like.. be a bunny outfit/bunyn headband. or it'd put them on their butt or just poking upside down from the top of the photo
I wanted a bunny morph but only got the bunny ears + tail and skimpy suit
yeah, a furry model would be nice. if i went to make one myself i'd probably find a specific character with alot of art to train one off from, literally for that 1 character, seems like a good way to prpactice
I had the same problem when I tried to attach wings. the wing was coming out at the feet of the character
There are a lot of these models linked on rentry
What model do you use for realistic landscape drawings?
Maybe Dreamlike photoreal
what's reentry?
A website
There was a big list of models + links to download
But a lot of nsfw stuff and Models you needed torrent for
civitai also has a lot of stuff
that's gone ow. you have to go to github and use their API to do stuff and i have no idea how to do that
No?
I just uploaded an embedding to civitai
No github involved
And I downloaded something from it this morning
alsi i read that it's really easy to embed malitious code into .ckpt files so was told to only DL them off huggingface or civitai
idk then. i tried making an account and logging in and whenever i do it just gives me a "hey! use this instead" and gives me a link to the git API and refuses to let me make an account
And civitai doens't have github stuff that I know of
Weird
maybe it wasn't that site
i saw a photorealistic model called like "hassan" something on it i wanted
Yep hassan blend
but i went to getit and it required an account to DL it
then it refused to let me make one
Weird
see?
when i click "submit" that link pops up and i cant do anything else
figured the site got axed or something
weird...
Idk why they would do that though
or maybe it doesn't like my VPN
naaaaw. already tried disabling all my other security/privacy things. wont be disabling my VPN
Alrighty then
yeeh. if a site won't allow VPN use it's not worth using, imo
Hmm. how high do the numbers go marking something as a priority?
1.0? 3.0? 5.0?
i set gold colored eyes as 2.0 and now i'm getting a bunch of blue eyes
it's been a year so where i can generate images now? dreamstudio website? i hear midjourny v7 is awesome
Im new here, but i was wondering the same
damn theyre up to v7 already? did they have another recent update
and I think the bot is like Ben Affleck, its Gone, Baby Gone. Maybe when/if stable ever actually releases a new model instead of just getting left in the dust by everyone else, they'll bring the bot back
remember when emad said there were like so many models being cooked? it was like we were gonna be drowning in them. hard to drown in a dried out puddle
just feeling a salty for no reason, btw, i'm sure some good stuffs is coming
what's the difference between a LORA. checkpoint and a textural inversion?
and a hypernetwork
guys https://www.twitch.tv/watchmeforever theres a 24/7 real time ai seinfield with text to speech and animation lmaooo
do you guys typically keep a .txt file with all the trigger words for your various embeddings and stuff?
hello
Should I be downloading pruned models always? What is the difference exactly? Thanks
I installed everything. my video card amd i get this error RuntimeError: "LayerNormKernelImpl" for "Half" is not implemented
please help me
from what i was told unless you really plan to train or make embeddings using the model yourself and only intend to use it to generate images the pruned model is typically fine. but imo if you have all the space in the world just get the bigger one.
unless you're sure you wont ever want to do that stuff
Yea okay thank you!
you should go ask that in the troubleshooting channel
Anybody know what software that endless ai sienfeld channel uses to animate?
you have to rename the first model you DL to "model" when you put it in your models
How unsafe can checkpoints get?
just dl the safetensor version if you can
❤️
So is there any rhyme or reason behind if a vae file goes int he same folder with the model or if it goes in the vae folder? and do all of them need to be renamed to match the same thing?
i see a lot that are like 123.ckpt but then the vae is like .abd.vae
and then sometimes they tell you to rename them and put them in the model folder and sometimes into .vae
i don't get it
Hey y'all. I'm new
what chat do i generate in?
it'd be cool to have an inspiration channel for peopel to post/brainstorm ideas or inspirations for new models, embeddings, etc..
I know they can probably fit into existing channels now, but may be worth its own channel
I'm thinking about making a wes anderson style embedding or model for his famous cinematography palette
Must I have a copy of a necessary vae to make a model that requires a vae work, even if they are the same file, just with a different file name?
yo what is the difference between inpainting and finetuning
Do lora and embedding stay loaded up and work when you trigger them via prompt? or do you have to tell them to load? cant seem to see anywhere to do the later
once you generate an image, scroll down to the bottom and under the prompt it will say what embedding was used etc
Can anyone tell me the difference between inpainting and finetuning? Is finetuning using Lora?
If a vae file ends with .vae.pt it goes into the same folder as models and had to be renamed matching the model name.
Example.ckpt
Example.vae.pt
Then the webui automaticly picks the right vae if you select the model.
If the vae file ends with .ckpt or Safetensor it goes into the vae folder and has to be manually selecter as vae in vae dropdown menu
my hero XD
Loras are loaded but had to manually be applied trough the trigger word.
The third button under generate is for selecting embeds, loras, hns, etc
Yes make a copy of the vae and rename it to match the merged model
Inpainting is just an Option in img2img where you edit an image like changing eye color etc.
Finetune means making loras, dreambooth, embeds and other stuff that is applied as Additional help for better images, like specific styles
is there some sort of easy way to call what trigger words are used by each or do i just need to write them down/remember them?
For embeds its the name of the embed. But for loras with multiple words you have to remember them :/
For loras it also depends if your using the inbuild lora support or the Additional-networks extension
some have multiple keywords so you need to write them down if you plan on using each
in the settings is making it store some of the model in RAM good? speeds things up/takes it away from VRAM a bit? something?
Its just for faster model switching, if you store the models on ssd you wont need it
Also Wouldnt recommend with 16gb RAM or less
i have 32
would be nice if the thing didn't throw out an error and stop proccessing if it runs out of memory. instead just using the memory it has to do what it can. idk.
Yea need to test something today with a commandline arg
if you can think of anyway to do that or lower VRAM usage even without x-whatevers i'd be interested.
so what's up with SAI, they hyped up more communication etc but been silent ever since
nuh those models suck
derivative bullshit
well happy you're happy
Unlimited merge compatibility
Yea thats why i merge
k dude
Oh i misread SAI with NAI xD sry
So yea SAI has been silent, hope they release something new like Deepfloyd soon
does there happen to be an add on that'll make the thing generate one batch after another until you tell it to stop?
like a repeat button. would be nice if you just want it to churn out a lot of stuff while you're away
Batch Count, but if you want to test multiple Prompts then there is an extension
But i cant find it rn
when i go to slep the last few nights just for fun i set batch count to 100 ( it stops there) batch size to 2 , dump as many prompts as i can even think of into it and then use model keyword extention to set the multiple keywords to "random" and it generates a lot of neat stuff sometimes
^
hey everyone, would love to get some peoples opinion on my Stable Diffusion Deform project
experimenting with AI to try automate as much of the workflow as possible to create an educational video series. So the visuals are stable diffusion, voice is elvenlabs, script is written by chatgpt
guys, what does the fix do and how is it used?
"the fix"?
i think they mean highres. fix
I saw that they use fix like highres in civitai how is it used
very well done
it's just to upscale a smaller photo soon as it finishes generating it
can use it to do other things to it AFTER it generates the smaller image then upscales
not sure exactly what the point of it is since it still uses a ton of VRAM to upscale it
thanks! A video on the Gladiators is next
thanks!
can SD make use of running 2 non-SLI GPUs?
say i had 2 3060s. could it use both GPUs at once?
even though they dont really 'do' SLI anymore
have help me get icon and something avatar art, im newbie
SLI wouldn't have helped anyways. You can run two instances of SD, there's a project to make that more seamless somewhere. Webui itself does not allow you to use more than one GPU.
not SLI
but thanks for the info
I meant SLI or no SLI makes no difference here.
Here's the project for multi GPU batch inference. https://github.com/rbbrdckybk/dream-factory
Can we choose how much space something will occupy in the picture?
like the space a character occupies in a landscape painting
Is there an ai music generator with a function similar to our img2img function that i can generate cover music with? Tried Riffusion but it doesn't seem to have that option
Exploring the latent space as usual. And you? 🙂
Well, I am awake and my eyes are open, so that's a start, bahahaha!
It's the usual morning of shove all my meds in my face, hiding under this blanket, and thinking, "Ah! It's cold!"
How's your exploring going?
well right now I'm doing experiments using some new extensions and techniques to influence the diffusion process - stuff like dynamic thresholding and continuous prompt interpolation.
also beta testing some new TI embeddings and see how versatile they are and what else you can make with them by mixing different embeddings. So... lots of fun 😉
Sounds like you are enjoying yourself and I, 110% approve
stamp
It always amazes me seeing how many awesome creators are here
Do you have a particular theme you're going for?
guys in which text channel i can render my art?
I have always multiple projects going on - multiple ideas or themes. If you test a TI embedding, model or any kind of fine-tuning I first try to test the strength and distinct style it was made for - or the idea behind it.
After that I try to mix in different styles, combine it with other fine-tunings and prompts and see how versatile and flexible it is.
The great thing about this is that you never know what is going to happen. You might combine something that was intended for something completely different and create something totally surprisingly new.
With this technology and tools this occurs much more often compared to anything else I've worked on before. So the experimentation phase with different themes and mixing fine-tunings is super exciting.
if anyone knows then tell me
Right now you sadly can't create images of your own on the discord. The bot has currently been shut down.
1/ The Government's 'text and data mining' exception to UK copyright law would have left those in creative industries without legal control over their work, should it have been used by an AI platform.
I therefore held a debate today to highlight the importance of this issue.
I am extremely pleased to say that in response, the Minister confirmed to me that these plans will be halted, pending further consultation.
This is a huge win for millions employed in creative spaces across the UK, and I look forward to seeing the updated proposals.
You mean post your art?
That's the great part of AI--there's a lot of unexpected creative nature to it.
Things you might not have thought of
Oops why?
And you can translate to so many different areas and aspects and areas of art, which is really fun, and exciting. It's also a challenge to see what you can mix, too!
I don't know why. Maybe it will be back at some point. Here's the announcement from a couple of days ago: <#📣|announcements message>
Thanks for information
Yeah. I think one of the most interesting aspects to me is that I learned new things about my own creativity because you can work and explore so fluidly between different worlds, themes and styles. It's really really inspiring.
In Automatic, if I put a url to an image in the txt2img prompt would it actually pic up the image or just the words in the url?
I tend to be hands-on when it comes to traditional art (physical mediums.) So playing with clay and then just being like, hey, SD, let's talk about meshing that clay with some paper mache to see what would happen, it allows for a lot less time/effort/etc beforehand. Plus, it's just fun to see what you can mix!
this sounds really exciting and also very inspiring - mixing physical and digital since I'm pretty much exclusively working with digital tools. I guess excluding playing an instrument which controls digital instruments again ;-).
I don't work with my hands as much as I used to, both from my ailments, but also because of space, etc. But I think I've always been better with them. It may simply be from my problems with eyesight, but being able to touch a medium, to feel its weight, circumference, etc, helps me understand it better. I'm also a pianist, and though I haven't played in a long while, the knowledge is still there because I started when I was very young. But there's nothing wrong with digital mediums haha or using them. There's so many options these days.
I'm just better with traditional mediums rip
is openjourney v4 better
which gpu better? rtx 3050 or rtx 2060
Hi , I installed stable diffusion 2.1 but the images I make are ugly . I saw that we could teach the software to make better ones but I don't really understand how
i used openjourney model and it produces more stable images
Hi! You can try some custom models made by the community
https://civitai.com/
I guess not.
From the stats I'd say 3050, more VRAM.
Yes, I can totally understand. I think the most important part is that you still explore your creativity when you can - doesn't matter how often or what kind of medium you prefer over time. Sometimes the environment you are in makes you try something new and this may open new ways of thinking and inspiration.
Maybe a loss to AI art in the UK?
Thanks
well hello
I like that open-ended thinking---take inspiration from whatever is around you. Always observe, always take the chance to be creative! Honestly, having a screen to draw on really helped me a lot when it came to drawing. It made me feel much more "connected," and pens have gotten so much lighter, and more comfortable, so you can really do a lot these days.
guys i am starting huge project which i think might be bigger than one man can accomplish. basically i am planning to create huge fight/pose dataset from old movies like "seven samurai", "drunken master", "way of the dragon" so if you're willing to help, let me know.
I mean my model is only based on 1.5, but if you like dark fantasy, horror fantasy, horror sci-fi, a blend of these, regular anime, gothic anime, angelcore, devilcore etc And if you like characters that know how to hold swords (most of the time) and have five fingers (65% of the time) https://civitai.com/models/6209/dark-gemini-v20
but you should be able to get nice results
I'm gonna start working on 2.0 or 2.1 if I can get them to work
I have already installed 3 models. I put them in the Embeddings folder. I have one in .pt and two in .safetenders . but I don't know where I should put those being in .safetenders
True :). I use my wacom or tablet when I'm doing sketch img2img experiments. It's great.
made this tutorial video for promot engineering. You could use this format and reverse engineer for stable diffusion
https://youtu.be/HqS3bKZ2ovk
I think yes . Pfiouuu that's so hard XD
it's not hard
remember to press the button next to the model dropdown menu, the refresh button
after putting the models in the folder
What one do you have? I had a DTF-521 for ages that I used. I got a XP-Pen and it works great, though the driver is a mite buggy at times.
I also have a little Samsung tablet, which version escapes me at this particular second, but it's great for on the go.
if you ever go for another budget one @bleak matrix I recommend Huion
I have a Huion Kamvas 16 2021, the precision is so good, the pressure levels, the colors are alright but not HDR10+ with 1000nit. It's a budget tab I bought for 250 dollars, but it functions like it is worth at least twice
I've heard really good things about Huion
I thought about them, too
I wish I had USB-C with DP
cause then I only need to connect one usb c from it and one to the power
now I need that collection of cables
That's one complaint I've heard
The XP-Pen was really easy to set up
Considering what I had before with VGA
🤣 '
It's a solid improvement
I still use a very very old Wacom from like 15 years ago - the ones where the pen tips didn't break after some use.
Hooray! Lol. So like my DTF-521
Hahahahahaha
I still have all my old ones, though I don't use them.
@bleak matrix oh it's not difficult. Just pop in hdmi, power and usb to computer
and done
Yeah, I have lots of convertible cables. I've just heard that people have had trouble with cables, including with XP-Pen.
they should just move the training to Israel where the courts have said copyrighted materials can be used for training
Thankfully, mine worked.......
UNLIKE THE 22" HD Cintiq. That didn't work. Had to return it.
And move every employee there too? 😛
just move the datacenter there for the training
employees can stay where they are
in any case it will be an arms race, and the countries which will allow training on copyright will win, unlikely the UK will forgo the business.
don't think the courts will take that well, but it's an idea i suppose
not illegal I think
if you have few questions ask away, for extensive from zero - to hero tutoring however i charge
Dm check
Which model do you use for artistic landscape paintings?
and how i can do upscale?
Hi there 🫡where is the newbie room?
What's Openjourney? Leaked Midjourney v4?
Solid thread and reply to the whole (OMG we found a duplicate in the data research)
https://www.reddit.com/r/sdforall/comments/10rb18z/stable_diffusion_emitting_images_its_trained_on/
Can I get Dreambooth support here? Or is that another server entirely? 
Hi, does anyone have the InCaseArt hypernetwork, or incase a3 (i think it's the same) ? It's not on civitai anymore :/
Hi
Is there a way I can use it with 4gb of vram?
Guys I have amd graphic with 8gb of vram , can i use AI generation image with this. ?
What's Openjourney? Leaked Midjourney v4? Or just a troll model?
where do you see it?
How can we fix minor errors in the resulting picture? For example, in my work, the character has strange pieces of clothing in his hands. how can i fix this from img2img?
@warm junco hey
do you know model for logo
and how to use
TEXTUAL INVERSION files
Como faz pra se inscrever
Im pretty sure this is just someone who pulled gens from mj and made a model from it. if it was an actual leaked model itd prob not be there for legwell reasons
Is there a free stablediffusion api that I can use with nodejs etc?
That makes sense. Civitai wouldn't dare to risk it if it was really leaked
Then again it's hard to prove either way
Como faz pra fazer inscrição
Apparently the UK government is backing down from their changes to the law that would better protect AI research: https://completemusicupdate.com/article/uk-music-welcomes-government-decision-to-abandon-new-data-mining-copyright-exception/ https://twitter.com/sarahjolney1/status/1620855093515157505
This may signal that the UK government is going to start targeting Stability AI and other related companies, on behalf of the luddites and corporate middlemen like the MPAA, RIAA, and other similar organizations
Does anyone know if brackets for neg prompt still work in positive prompt box in auto1111?
Technocrat Karen and pudgy face minister are living in a bubble that doesn't exist.
Just the reference to the uniqueness of british and other parts of the UK creativity... these people...
They started taking up music middle men money and they cover it up as having some sort of need to protect the "uniqueness"
this is actually great because now, clean music libs will have to be made and the same people that pushed for this will get recket by their own devices
And in the US, apparently the anti-AI activists now have a meeting with AOC's office (She's a prominent elected official with a ton of power) : https://twitter.com/emishson/status/1620924684236898304
that last one... If you been affected by ...-AI utilization in the arts, television, and production
imagine the eye rape going with no gov oversight
these people run to daddy state the moment they understood it's game over
panic is never a good state to be in
that's why we don't have cars today, because all the horse and carriage producers stuck together back then
And now we have way too many cars.
anyone have any idea how well SD works running on a Quadro GPU?
can't wait for shots of AOC sniffing by a fence with her arms behind her back, think of the artists
You think that she's going to join forces with Eshoo and demand a ban of AI content and the destruction of existing models?
I don't know enough about her to say whether this meeting will go anywhere, or if we are going to have to fight her as well
I have no clue what these people are capable of doing actually, depends on the understanding they have and how much money is going to be thrown at them really
what are they going to do, use boots on the ground to enter every neets basement that downloaded something from the web
it's impractical at this point to go all in
We really need pro-AI artists to be meeting with politicians to combat the anti-AI activists, and we shouldn't wait until we are fighting a defensive fight
We need to be proactive
bottom line, anything that makes the art market money laundering biz look even more transparent is going to get push back from elites
at that point all bets are off
it also depends on what courts this is going to be tested
but if any of these people get info on the matter, they will save themselves from making a spectable
so I'm counting on a lot of train wrecks
you wouldn't download a car would ya!
just the fact that campaign existed, tells a tale of what people are willing to try to sway opinion towards something that will protect their established interests
we may even find out that ai is a gateway drug to bank robbery or something
hey, dont know a good model for logos sry,
But Textual Inversion (Embeddings) ending with .pt go into the Embeddings folder, then you can select them by using their name as tag or press the third button under Generate.
It is been a week since I first tried stable diffusion
You know what they say about being born too late to explore the world and too early to explore the universe
have you started feeling the urges to rob art galleries yet ?
No I am just glad I lived through the birth of AI
heh, me too
Like useful practical application of AI
what a ride
This is just the beginning unfortunately
I saw the introduction of home computing, the internet, now this
dude, life comes at you fast
I remember browsing the web on a modem connection and my family fighting with me to shut off the internet so they can use the landline
I can't even imagine what a ride it was for my grandparents to watch it since tv to new media where everyone had a mobile on their pocket
I understood that reference
yea a rectangular slab display with touch screen that connects to anyone in the world
with insane 4g internet speeds
all wireless
suddenly the phone wasn't stuck to a cable, shazam mofo
I mean, next big thing is space exploration, if we don't get slaughtered by ai or something
maybe this is the big filter
dunno
I am genuinely sad I will miss out on this
I honestly do not want to grow old nor die
well, if you hold out couple of decades you may get your wish too
for better and for worse
there's no "or" in that scenario
Your health worsens with age, makes everything less fun.
It is why I think old people are sometimes at peace with idea of death
I developed a tinnitus in my right ear because of teeth problems that affected my jaw
I will witness landing on Mars when I am bed ridden and dying of some cancer
meh
Hopefully, AI development accelerates exponentially as we reach the singularity
and the event horizon
Now that is something I would wish to see
I do not really remember the movie but it is not about it
AFAIK it means the events after singularity
Where it is impossible to predict
I know, it's just that kind of scifi horror, and that specific movie is so weirdly stuck in my head
he's actually in two of my favorite horror movies of all time
sam neill?
yep
Jurassic park?
hahaha
In The Mouth Of Madness
he's actually in a bunch of horror movies
but those two are simply the most wtf of the lot
at least for me
I am watching its trailer atm
I dunno, I got into scifi through comics
Moebius mostly
great motivation to learn french at the time
dude was a genius
could wipe his ass with any of these one trick pony artists on insta
Seems like I missed him
I also googled if they turned any of his comics into a movie with no luck
he did some marvel in the 80's
but his style isn't mainstream
he is like the fine arts of comics
Jean Giraud , his real name
there's a moebius textual inversion on civitai, not bad
meh it is old and weird of a style
sci fi is like a numerical model running indefinitely
the more advanced the more they converge on future
true look
I mean of course it looks old and boring
Personally I am in love in Deus Ex HR and MD ary style
art*
that's like a honed down final fantasy with guns slash solitary dude in sepia going round
yellow sepiaish
low-poly black and gold
I don't know much about it, gotta look it up
It is dope
a lot better than cyberpunk garbage
it is realistic sci-fi, ie. tries to predict future without a lot of fantasy involved
with a touch of Punisher
Not really, more like Noir/detective crap
You are not sure who did this to you and why and there is several to blame
and you uncover everything as you go
big adventure style
can you picture feeding your over plot to a digital assistant and having media ready in a couple of minutes based on your input
that's gonna be wild
that's holodeck starter pack territory
I am gonna input it in stable diffusion and see what I get
futuristic, noir, cybernetics, black and gold, sepia
dude
some bad ass super hero character came out, looks like he's bout to open the whoop ass can
lol i love this phrase
and true!
I've noticed a quirk with stable diffusion that when you prompt a color the color tends to bleed out into the rest of the image. is there a way you can make "blue eyes" just stay on the eyes rather than making everything a blue color pallet?
happens to me with 'pregnant' and now all the guys in gangbang are pregnant too
wholesome
paint them with a loving family for a change
One thing I'm wondering about how Stable Diffusion was trained. For each image in the training set, what was used as the text description of it? Did they just use whatever text description was on the website? Or was a curated list of keywords used? ie: a photo might have a pretty basic web description like "a photo of charlize theron" which would seem like it would provide a pretty weak set of keywords, or was something added like "woman, charlize theron, actress, blond, jeans, white shirt, smiling...etc"?
A loving family has existed since the beginning of time, this is the future
It is like how the internet was turned into a medium for social media gargbage and showing people pictures of your baby
so you're saying trans humanism is going to upgrade everyone into roasties
can't wait
everyone is already a roastie, given the chance
I never really imagined tiktok would blow up by this much
or onlyfans
billion dollar companies now
I love specific patreon meta, like 60 thirsty dudes paying someone called the ally that generates porny stuff
and it's got pretty convincing rationalizations for why ai is art on the cover page
I am kind of torn if it is real art or not at this stage
it does kind of feel it is remixing images at most of time
if you conceptualize something... turn around some object and call it a water fountain
post a campbell soup can ironically even
so... let's roll with it
I mean, the funniest from the Depp trial was the artist he had living in one of his homes that he is clearly waiting on to pass
the rich all have their Basquiats
the ride never ends
Correct, its for north korean refugees
wut? bru AI art is the greatest thing to come out this last decade
I'm officially an addict
can anyone help me undertsand how to use the bot ?
@trail wing si they are closed we cant use them ?
looks like it
lmao why cant we use bots ?
they must be fitting in the new python discombobulators
Then why are you here then?
Are the dream bots down still?
when can we be able to use them.
These snowflake artists ruin innovation.
to act like an annoying 12 year old. thats my job!
That doesn't look right to me.
is it possible to buy SD dreambot for my own server? or they all down regardless of which server they're in.
theyre all down
just use it locally

with a1111
web UI?
yh
#1047610792226340935 and #📣|announcements . The bots are down for a bit for updating.
ok thx
hello guys, i once saw on reddit a post that had a link to a pdf of like 200 pages where you could see all photo prompts effects ( like edison bulb, low light.. etc) but i cant find it anymore. does anyone had a link eventually?
A Canadian news outlet is targeting AI art, with a completely one-sided view of the SamDoesArt drama with dubious veracity: https://www.thestar.com/news/canada/2023/02/02/whose-art-is-this-really-inside-canadians-fight-against-ai.html
Apparently the SamDoesArt art guy is Canadian and is now seeking to lobby Canadian politicians to pass laws against AI art:
"Now, Lam and others have spoken to legal experts and plan to speak to politicians next month trying to stem the infringement of artists’ works, including via the hashtag #createdontscrape."
so if you use the same exact prompts/negative prompts in the same order along with the same sampling method, model, VAE, photo resolution would the resulting photo be the same?
Hi, been interested in using SD more than my limited Colab free time, especially since I've become more and more disappointed with MJ's quality.
Was wondering what would be the current best option that is hassle-free and easy to use? I don't have too much time to tinker with it and don't mind paying for a service as my PC is not good enough to quickly generate images.
Also interested in those LoRA models and Instruct Pix2Pix so a service that makes that possible would be appreciated as there's just too many to look into. Feel free to ping me or DM me if you have an answer, thanks in advance!
yo does using instruct pix2pix on smaller images loses quality? I use it on 400x400 images and its as if I get a zoom then a blur? Do I have to upscale them first?
His art style is so creepy looking to be but I'm not 100% sure why my brain hates it so much. Maybe the way he draws mouths and noses? Just can't look at his drawings for long before I need to look away lol
Some of his art is just tracing others creative works, like his Squid Game stuff being traces of screenshots from the show
is there perhaps an extention for the webUI that triggers a little sound for you so you know when the batch of renders is done?
sounds like it 'should be' simple to script and would probably be useful for a lot of people
I think there may be an option for that in the settings, or at least there used to be. Or maybe it was an extension. I can't remember, I just think I have seen something like it.
any idea what it may have been called?
where are bots?
my stuff takes about 2-3 minutes per image when i use VAE, LORA and some other stuff
what is LORA
it's another tool to add specific looks to a curent model without having to train that model specifically
Umm... I will load up sd now
specifics looks? like dreambooth?
like say i wanted to take an existing model and add a character like.... super sonico to it. i could make a LORA and then add it to the command prompts to make one of the characters look like that
you can use dreambooth to make a LORa yes
hello guys, don't want to break any rules here. i was wondering can i use this server to generate images?
bots are shutdown for now
ohh thank you!
Hey guys! Curious what y’all think of my SD educational video series: https://youtu.be/oX7ZGeFwWyo
Using deform for the visuals
hi guys i have a question
is it possible to run stable diffusion locally from an intel iris plus graphics gpu
is changing the seed the only way to setup parallel batches?
that is, it appears batch size only increments the seed. it's not like we can setup a parallel batch for different cfgs.. or different samplers
What do ; (semicolon) and __ (underscore) mean when writing prompts?
O
Fuck dis shit
It depends on what interface your using. I use automatic 1111 and it doesn't use ; or __ that I know of. So what is the front end your using?
Damn a whole ass minute to send that shi
What's the difference between "stable diffusion" and "deforum stable diffusion?"
so, whats will you guys do after Ai art is outlawed?
any suggestions?
maybe take up crochet
is it going to be outlawed?
It might, sooner or later
I'm thinking on maybe Boxing, I would punch random people on the street but that's also illegal so...
Stable Diffusion are the standart generation image. Deforum Stable Diffusion are a extension to create videos using frame generated by SD.
I imagine that it'll just be moved underground, with a particular vengeful focus on training the styles of those who made it illegal
Seems unlikely, but politicians can be pretty stupid and the luddites screech pretty loudly
Wouldn't be the first time a government did something really stupid
i mean i definitely see where some people are coming from
i'm a digital artist, and i see ai images as more of a novelty than anything
so if someone copied my art style via ai, i'd be like "oh, that's interesting"
but i understand that most people don't feel that way
i also don't sell or buy commissions, so i have less stake in the matter
and as for whether it's theft, i honestly don't know, but if an artist says "don't train ai to copy my style", i think the right thing to do is accept their wishes
i'd be disappointed if ai image generation was outlawed, especially because it's clear that the technology has potential that can be furthered, but i'd move on with my life
ultimately for me it's just something to do when i'm bored
also i don't think that it should be used to replace traditional art in animation industry or the like, because it just doesn't work
like, you can tell when something is ai generated
at the end of the day, i'm just a stupid girl on the internet so what do i know
People are always so over the top dramatic
but Mark Zuckerburg just announced he wants to focus on generative art, so it's not going anywhere
Style wise I actually find it interesting that we can spot which models were used to make the art. Even in the neural style transfer days, models had their own unique styles like an artist would
when SD voice cloning model
how do they do the video? https://www.twitch.tv/watchmeforever
Hey, any tutorials on how to resize to be able to use as banner for youtube?
Any idea why I have trouble getting "groups" of people in some Models?
probably just limitation of the model atm
i feel like even with 2 people your accuracy is going to drop, generally speaking
Mmm am I blocked?
nope, so I need to ask where I can use the dream command, but the question gets blocked by the mod for some reason..
Hello anyone! Does anyone know how to create an artwork with a person with a sword and a shield?
'it just doesn't work'
yet
The developers announced that the bots were closed forever.
does anybody have that image on the subreddit that explains how ai images are generated in a very simple diagram
noob question but how do i use a .pt embedding/
save the file to the embeddings folder, then in the prompt box the keyword is usually the same as the filename, but check on civitai, it will list it there
Hello!
quick question, where do you guys get good prompts from? I usually use propmthero but the site is mostly dead at this point
Okay. I e we oils love to learn about all the embedding and also where I can try out prompts. Where can I do that?
Basically is there a walkthrough that can help me here?
Hi ^^
I'm new in generating images with AI, so when I generate something the result is a green image, someone knows how to fix it?
Why is my "/dream prompt" not working?
guys you think which gpu should i buy? rtx 3050 rtx 3060 rx 6700 ( no xt) i want render and play games with ultra graphics. i dont know can rx do rtx?
Anyone using deforum notebook here? quick question xD was working fine last night .. ive come on today and it isnt working .. ive deleted it all and started again but i get an error on Path Setup
can anyone help 🙂
hey guys, can anyone point me to the approiate channel where i can discuss a project built on stable diff i'm working on? I've hit a little roadblock and figured it's best to get some input straight from the source
would #1011228667659178055 be best?
never buy 3050, as for 3060 no it can't run ultra, maybe on 1080p and if it's the most optimized and not very demanding game. rx 6700 is slightly faster but not by any record measure. 3060 ti is probably more likely to run ultra on 1080p (not on 1440p) and since 3070 and 3070ti dropped the ram size to 8gb it limits what it can do with SD. Though it can do most things it needs to do. 3060 ti has the added benefit of having more vram to play around with, even if it will do things slower than the 3070 or 3070 ti
But if you buy a 3050 you might as well be buying a 1070, cause uh..it's faster
the 1070 that is
but my 3070 ti can run 1440p games at either high or ultra, like 2nd highest or highest. But if I get a really really demanding game, I turn on DLSS
actually I have it on either way
@young ivy 3060 Ti also has 8GB
RTX 3060 has 12GB and is the recommended budget GPU for AI stuff for that reason.
3060 is the budget gpu for AI yes but he asked about playing games with ultra graphics
I don't think a 3060 can do that
I will play games in 1080p
Yeah, this is a channel about AI, not gaming. 😄
and never activate raytracing
it is but it's also a general chat and he wanted it for both
so he's in the right place
i will use gpu for stable diffusion to 😄
I'm stuck between rtx 3060 and rx 6700 right now. rx 6700 can rtx on ultra? I don't know much about GPUs.
If you want SD and you're on Windows, AMD is not an option.
You can probably play on high settings with DLSS and I'm not sure if rx 6700 can run ultra but it cant run SD
- webui on Windows.
so rtx 3060 is best option
high settings with DLSS with a 3060, at 1080p it's not gonna be noticably different from ultra at this point since high has surpassed how much a 1080p can display in its ppi
and colors, bitrate etc
if you want to see ultra, you need 1440p
so high isn't a bad resolution
i think i will buy rtx 3060 then i have ryzen 3500x it will enough?
i am using gtx 1650 right now
16
ram a bit on the low side if you wanna get serious about SD
Easily good enough to start out with.
if you can grab another 16gb (of same brand and type as you have) then you'll open up more options
yeah but not get serious with
now it takes 1 2 hours to render with gtx 1650 it's terrible
Render what?
sd photos
at what resolution? O.o
1280 720
Even your CPU can render faster than that I think. Something is wrong then.
with hires fix?
yes
it takes 1-2 hours per render? or half hour? or 12 hours?
1-2
like per picture?
yes
something is wrong then yeah
It shouldn't be that slow.
have you done a pip install -r requirements.txt and pip install -r requirements_versions.txt and updated to latest version of ui? Assuming you use automatic
Which UI are you using and with which parameters?
no we're asking what the UI is called, since you're using it, you must know
i dont know how can i check?
Don't you know what files you downloaded and installed?
i installed python: 3.10.6 and git
this
i dowlanded here
okay
now they can better help you out :P I can't cause I'm about to throw up
good luck
#image a H red designer logo, vector logo, strong brand, web agency
thanks!
has anyone found an okay ai that works to make simple 3d objects from images?
What do you use to launch the program?
are you sure its not doing something silly and using cpu hardware to do it?
also how many steps
I use 20 or 40 steps, I checked in task manager it is using gpu
webui-user.bat
Try editing that file. Change the line with "set COMMANDLINE_ARGS=" so that it says "set COMMANDLINE_ARGS=--precision full --no-half --lowvram" and try it again.
okay thank you
what should i put for 3060ti?
Should be fine without anything there.
If you can use half you should.
Those commands disable features that are very useful, but only work on more modern GPUs.
For Gaming i would Recommend the RX 6700XT, for SD i would reccommend RTX 3060
Do you think 3060 is enough for gaming? The 6700 xt is expensive for me. I can get 6700
depends on the games but for Full-HD gaming it will be enough, Cyberpunk 60 fps on med-high settings i think
Which of the normal rx 6700 and rtx 3060 do you think is more suitable for gaming?
i cant afford rx 6700 xt now
ive found a rx 6700xt that only is 20€ more expensive
in my country 20€ is exp 
@upbeat tangle you can watch this comparison: But its not FullHD its WQHD, so with Full HD you even have more FPS
https://www.youtube.com/watch?v=Zx4dYnrNJOs
@upbeat tangle But before you buy anything check if your current hardware isn't enough for now. Did you try SD with the parameters I gave you?
rx 6700 xt not compatible with sd? Or is it bad compared to rtx?
It takes about 30 minutes at 1280 720 resolution. but it happens in 5 10 minutes at 512 or 720 pixels
AMD has no RTX, RTX is from Nvidia,
Also AMD Cards are not as good supported for SD but that will change over time
I was very undecided. i think i will buy rtx because i want to be able to use sd and rtx
RTX is unuseable on 3060 xD you then have to Upscale with DLSS to get the massive fps lose back
And if you replace the "--lowvram" with "--medvram"? That should make it faster for low resolutions, until your VRAM is full.
@upbeat tangle go with the RTX3060 it has 12gb vram for SD, RTX if you realy want and Game Performance is enough
oh that was bad. so what other card supports sd, rtx and gaming?
faster than before thank you, still slow due to low vram
nah it is useable but not like you may prefer. You cant go High Graphics + RTX on, you then need DLSS. 3060 is your best bet
yea go for the commands occam said, also add --xformers it can reduce vram usage
Low VRAM doesn't make it slow. At some point it will just throw an error and fail, when the VRAM is full. Try removing the "--medvram" and check which resolutions are still possible.
And yeah, I forgot about xformers. That also helps with VRAM.
When you say rtx, are you talking about ray tracing or rtx 3060?
raytraycing
If I have an image that I like but want to change the colour of the object, how can I do that within Automatic? Is it a case of covering that object in inpainting and typing a new colour?
then it goes into #1025467151206854736
covering eyes for example and type only blue eyes as prompt
i play with dlss then. Longevity is important to me, ability to play new games for a long time
ok done
Its a car, I am trying to change it from orange to blue but it keeps being orange. I put the orange in negative and it made it red.
Doesnt seem to understand metallic blue.
yea then go for the RTX3060 it has DLSS and will with help for RTX and demanding games too
did you select the inpaint model?
you can ask here too, but for ds there is the specific channel 🙂
bruh
...no 
Changed to inpainting and now its still showing as red. I have added red, orange, green, not blue to negative prompt but its still showing as red...weird.
just in case, I have problems with ds in-painting, and instead of having what I want I get a weird blob out from nowhere
Hmm did you mask the whole car ?
Yeah
I wonder if DeepFloyd may be delayed after this research article showing that Imagen models may have enough dataset memorization to be useful for lawsuits against AI art model creators: https://arxiv.org/abs/2301.13188 The researchers talk about how it could have implications for lawsuits and legal battles.
those people are grifters trying to be relevant, the worst kind of academia related ass clowns
the main clown on that paper posted a video on how to do it on GPT-2
ah ty!
Alguém do Brasil que possa me ajudar não consigo ativar o bot
Is there a way to blend two images like mid journey?
there is an extension also with (chicken : cat:15) without the space
hello where can i ask for support ?
for technical stuff there is #🤝|tech-support for other stuff there are other channels
okay thanks
sure depends on the question
i need a black pyramid with golden details
I was actually thinking of blending the faces of my two women I find really attractive and see if I'd like that blended face more than the two
will we ever see another model from the team while this lawsuit is ongoing?
I'd be surprised if the lawsuit impacted development before any kind of judgment is made. that's an issue for legal.
has anyone gotten a FileNotFoundError when attempting to generate an image based on txt2image prompt?
Hi. Where do I go in the discord to open an SD account?
I've had very good success with groups with Real Vis1.2, but 1.3 and other models its legit impossible.
Hi guys, could someone tell me where I can generate an image?
@dense fiber dreamstudio.ai requires only little effort (sign up using Discord for example) and is the official site for generating images using stable diffusion.
Nice, thank you, 🦾
if you have a good gpu you can use it locally too, using invokeAI
ok, i'll try that later
thanks
What about the bot? Is the update going to be developed for a month? Do you really want to add new features? And the fact that it was closed, is it forever or not? They just didn't write why it was closed. Of course, I don't believe in the trial, because I consider it a lie. If there is a real reason, people respond.
guys anyone has kohya ss working for you? if yes could you give me the version of xformers, pytorch and cuda u have installed?
Hmm. general question. if you take a model can you toss it into a subfolder inside the correct folder? or does it need to be directly in that folder? like can i take anything 4.5 and toss it into a folder inside models so i can then put a txt file and some photos for better info on it?
if i can that'd sure make categorizing stuff easier.
Hello everyone could someone help me on how to install stable diffusion on the Mac?
what channel is right to ask for a good resource on how to install SD on my PC with github, i went though all the steps using the resource i found and couldn't get it to work after installation ? thx
tech support. or just check the pins for the tutorials
I can send you a link
@mental ether start here:
https://m.youtube.com/watch?v=VXEyhM3Djqg
Is it possible to do VERY limited changes to character’s faces while running background embeddings?
All the AI’s I’ve made (very small sample size) completely change the base characters
AFAIK "no". not like that, anyway. you could potentially use inpainting to do so though
you mark spots on the photo. either spots you dont want to change or spots you DO, then change settings, set batch size and run it and it'll regenerate those spots
i've done this a few times in art i really likes where people had messed up faces/extra fingers/arms to limited success. but i can only generate a new 600x800 photo every 1.5 minutes. rather than like 12 in the same time frame that i could do with a 3060Ti
@warm tiger Okay awesome. I didn’t think@it was really possible but that’s good to know
Also, I’m very new to this was me reading a lot of what you said..
no gifs on this channel. I get it 😂
i've only been doing this like 3 days now. but yeah it's a lot of fun to mess with. I think at some point I'd like to make a model using my own artwork
Yeah that would be amazing. I do a lot of stuff on my phone but SD only appears to be on desktop/Mac is that right?
afaik it can run on basically anything 'if' you set it up right
which may or may not include writing some of your own scripts to get it doing so
i've seen linux/windows/macOS install guides, using python, and a couplel other languages
Yeah I don’t know much about that. I’ve seen there’s a few apps that have stable diffusion that are pretty good
someone confirmed the opther day you can do it on your phone using google collab
“Draw things” is one
the apps probably connect to an outside server that it's being run on
which would make sense
The apps run completely locally but only have a 2gig storage limit so what the AI can create, is limited
Oh I’d love to figure that out
hmm python/pc doesn't seen to want to swap to this model
python crashes whenever i attempt to swap to it... wonder if it's an out of memory thing without it giving the warning
where do i generate prompts?
That feature has been removed from discord. You need to install Stable Diffusion on Mac/PC
@fervent thunder
Has anyone heard of a way to connect SF locally running to websockets? Been thinking about having it talk to my stream bot and have it so chatters can have it generate an image while I play games and have it show
is it running on the same machine you play games on?
There's no Dreambot any more?
search Stable Diffusion Online. You'll find some public SD sites
if it's the same machine, then you can't game at the same time it's generating images
Nope, different machine, I do 3D rendering for my job, so have a machine that is literally built for rendering, which coincidentally works great for AI too
what GPU does it have?
Four 2080 TIs, was built a couple years ago
do they work together for SD?
I use Octane render, so the more CUDA cores, the happier it is
Seems to work fine, had no issues generating lately
can you train?
Don't know how yet, still learning the basics of prompts, was just wondering about using it for some fun when I stream
I hope you figure it out
It'd be fun, already have a goofy ChatGPT thing, but I have to enter that manually
I have a GTX 970...can make images...that's about it. I've run the gamut of what I can do - make mostly profiles of people with mangled hands
When I saw this, I thought I could make good action shots - nope
Can I pay to continue using the dreambot?
they're gone
there's nobody offering them for $$$$
you can pay for that other one on Discord...can't remember the name
starts with an M I think
That one is too expensive 😛
discord didn't want people generating NSFW images, so Dreambots went away
there are many web based SD's available online. Search
Of course. So if I start looking for that, I won't find hundreds of servers offering naked pictures? 😛
Does anybody know the link to the video of Olivio Sarikas where he interviewed a female artist?
I don't find it anymore.
Well, maybe not hundreds, but definitely lots, especially anime related.
There are many, not sure about hundreds
The one's that you're talking about, use my machine?
No, they run on a server somewheres
Oh, that's pretty good. Thank you!
sure thing
Where’s the channel, where I say something and then it make it?
/deram
How do you do a prompt ?(or whats the command )
How does twitter feel about all of this ai art stuff in general
Hate ai art
Why
guys.. I want to use the AI for website ideas like I do in midjourney.. u guys have any tips ? cause I tried and I got some weird results like this https://imgur.com/a/dzYylO1
95% of artists just hear claims that "SD is collage"; "SD is trained on confidential data" then, you know, mob mentality...
All i have seen is ai art steal people's art and merges other people's art
That's what most people say
They think it's also gonna destroy their jobs
They're mostly confused with img2img and txt2img
😦 i didn't knew that amd gpus were so bat at AI stuff
its a vocal minority. the 1% rule applys here. only 1% of the participating community are making the content
I wonder if Intel cards can do it, or is it all the nvidia cuda thats the magic
everyone else hoards
What?
the silent majority is always a real thing
i got a 5700xt for very cheap and now i can't even generate a single image, i had better luck with my APU
it took 2 minutes to generate a 512x512 but it did it
damn
i think you're doing it in cpu mode then
no now i can't even generate a single image, it fails
i had a vega 64 which is older than that. it took about a minute for a 512x
torch.cuda.OutOfMemoryError: HIP out of memory. Tried to allocate 3.38 GiB (GPU 0; 7.98 GiB total capacity; 4.84 GiB already allocated; 2.88 GiB free; 5.04 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_HIP_ALLOC_CONF
and dies
oh well
it's a great gpu anyway so i don't feel too bad, but i'm sad because i bought it thinking i could use it for AI art stuff
I bet it'll be viable still.
like I read Zuckerburg said he wants to pump billions into generative art
so maybe not so hot with stable atm, but still could give you some returns in the near future idk
Is the inpaint feature broken in the lastest update of automatic1111 for someone else?
Not really a fan either but if he's dropping money into it, that at least guarantees it should contiunue on
I mean, he can hire lawyers too you know lol
its not like its blood money its just dumb people fleecing
One short prompt shows a major limitation of SD: Goku and Naruto fighting. Ive posted the image in general-images
I wonder has dalle ever updated? i mean since dalle 2
or are they all about chatgpt now
seems like dalle was the shit, now i never hear anyone mention them
course who wants to pay i guess lol
because you can't use it
Where do you input the model embedded code?
much more use with SD because it's free
yeah absolutely
still dalle got so much hype people had to be paying. like my aunt had heard of it lol
SD cannot do action. It cannot do multiple characters in once scene, especially interacting
i know they did an outpainting thing, thats the last thing i remember hearing
yeah but can any ai do that frank
yeah i was talking about dalle and just finishing my thoughts
maybe 3.0 will
is it like a parameter thing maybe
Up until a few days ago, I was generating for hours a day. Now I've stopped. It's boring generating mostly portraits with mangled hands and feet, no action
maybe just impossible until you can up that ante 🤷
i dont really generate much at all for months. im waiting for a better official model
yup
Ive generated all the mangled hand and feet portraits I can take
the realization hits when you see images people present as excellent, and those excellent images show a flaw of some kind
dall-e never been updated since the released model. but there's been work done around it that they've never publicly released. certain academic networks have access
most of what openai does hasn't been released. probably same for stabillity too
dont they have chatgpt4 coming soon too
chatgpt is useful to generate stuff you know about
well, chat gpt is based on gpt3. gpt4 is a new language model that could come
a new chat gpt would be built on that
you cant rely on it for what you dont know because it spits out false info as if it's true
you can use it as a tutor easy. feed it a text and get it to explain it back to you in terms it knows you know
yes, as long as you know about what it spits out
I use it to generate explanations for questions on ACT tests
I mean, for being like the pin-hole camera of chat ai, its pretty fucking incredible
about half of what it gives is false/bad
yeah relying on it as an oracle of truth is just delusional
don't use it like that
some simple math it botches
we cant use emojis in here?
🤔
i tried to use one as a reaction and it got nuked
how simple?
funny things about SD is people tell you "do this" and "do that" to get better images, but they don't show any of their results
I haven't really done maths stuff with it, just more general questions asking it to write things and I've been blown away (again, considering its early as hell days for it)
One told me to write longer prompts, more in the negative. One told me shorter prompts, less in the negatives. I asked to see examples. Reply: ask me in three days
i always say, do what gets you great results personally. start with small prompts. add what you need. use embeds for a good time
I asked Chat GPT for an explanation to an answer about a reading question on SAT. It used quotes not in the text as proof of the answer, which it correctly identified
I say learn to draw. Then you can get exactly what you want without twisting in circles trying to get SD to do it
you really can't lump people together. people will tell you something different often, especailly since this is barely a year old tech. no one is an expert and theres a ton of meta thats just wrong and snowballing
nobody has shown a reliable way to get rid of mangled hands and feet
see my problem with drawing is that i have partial aphantasia. i can't actually see an image like good artists can. i only get glimpses and my thoughts are mostly all my other senses instead
makes drawing a giant pain in the ass for me. about a third of people are estimated to have this. maybe more
Damn thats interesting. I actually had to look the word up lol
I have mostly blind
It was fine two days ago, then woke up mostly blind
Damn. Go to to the doctor dude
Ive been to seven
ooooh
oh no. losing a sense really sucks. i'm sorry to hear it. hope you figure out whats happening
I was told it's a viral infection in the eyes, but they now only prescribe antibiotics for life or death cases. Blindness isn't life or death, so they won't give me antibiotics
but you know flowwolf, and I'm no artist, but I dont think I have that, it's another reason ai art is good to me. I mean I assume maybe ai helps
oh sorry, frank. i thought you meant you woke up kinda blind a few days ago and hadnt gotten it checked out
they gave me something to stop coughing - I wasn't coughing. They gave me something to relieve a stuffy nose - never had a stuffy nose.
ai art has helped me with conceptualizing SO much. I consider myself an artist but i have a different approach to it in the past. very technical and planned out before i go into the process
antibiotics will clear it. but they won't give it to me
thats fuckced up
when i was in college for 3d animation, i did more script work than modelling work. really great art can be made with code
what has AI conceptualized for you?
the hulk as a terminator cyborg
just standing there
sure
with mangled hands
im trying to get into training embeddings, typically what models do people use to train faces?
you got a gen for that flowwolf lol. actually sounds a fun concept
the idea has been rattling in my head for decades. i've tried getting it out a million different ways. believe me.
it was one of the original generations i tried at lol. i spent about an hour doing it. those oldies are long gone though. that drive died and i've yet to run a recovery operation on it
you've never drawn it?
i could actually see midjourney doing a good job with it if you added two jpgs of like hulk and terminator
but midjourney is 30 usd or 60 for deluxe wtf
for a year, i'd pay it
i've drawn it but it sucks, because it's like driving blind. drawing without seeing something in your mind is so hard
people often do this "just draw" but they don't get it really, its literally mind blindness
the cure for that is to draw something
ok
Here's a question
i'll just cure my life long aphantasia. got it
I pay 15 bucks for lunch at work. 30 for a year of ai would sadly be nothing even if they cut it off lol. but a month? hell no
If you rely on an AI for creative inspiration, what happens if the AI is taken away?
oh no. inspiration is free. it comes from thin air.
how would open source ai be taken away exactly?
you just said you couldn't come up with a single idea for terminator cyborg hulk for years
But SD made it for you, but you can't show it
a single image. i've got a whole 1000 year lore of ideas
single image
i thought you were smart. twisting words isn't smart
did you ever try to draw one?
hey dont argue guys, its friday night lol
lol yes bro yes. they were junk. we went through this lol
i can't draw. it is like driving while you can't see. right?
i got a certain art style but it's never what i intend. its very free hand and well, shit
ok
i can draw 100 mickeys in the exact same pose. but i can't transform him to the next frame at all, because i can't see the image thats supposed to be there. i can describe it in 100 different ways, but i can't see it before it's actualized. drawing it would only be done through muscle memory
does any AI not mangle hands?
about a 3rd of all people have this condition and can't draw new ideas since they can't be seen in the mind
people claim to have solutions - they'll show you examples in three days
and from what Ive seen of the Deep Floyd stuff, it should be better there
what the ai helps with is conceptualization. standard photoshop work can be done after that
one thing people do is photobash hands in, then ai mash it together. works 70% of the time, everytime
Conceptualize Ubermensch in SD - Superman raised by Hiter to lead the Nazi armies.
yeah and with upgrades to openclip feeding what deepfloyd and future models can do, all that innaccuracy nonsense will be forgotten like crank starting cars
upgrades that are censored to remove or hide content?
i don't really care about race supremacy so i wouldn't bothher with meth head imagery like that
seems methed out
It's a story I'm writing, so I need that image
it seems like you're writing that story while high on meth, just saying
I made it in DAZ, but getting it into fighting poses is hard because the cape won't cooperate
Hitler's fucked up dude
Ubermensch ends up killing Hitler, ending the 3rd Reich, restoring democracy in Germany
i can't tell you if the company stability is going to use meth head supremacist imagery to train their new media release. i'm going to bet.... no
it's anti supremacist imagry, anti left, anti nazi, anti hitler
yeh you definately did not lead with that
so i'm going to go ahead and take your lead, and assume it's rooted in a lot of meth
the imagein my little avatar image, whatever it's called, is Ubermensch
your good intentions are probably just delusion
you keep projecting the opposite onto what I'm writing
what you project comes from you
if knowing nazis have always done a lot of meth is projection then call me a movie theater
I just told you that Ubermemsch is anti left, anti nazi, anti hitler
yet you keep projecting nazi onto it
well like, you introduced it with "they were raised by hitler"
oh okay
Superman is a HE, not a they
yeah, maybe don't do that in your story if you don't want people assuming you're just creating propeganda for race supremacy
seems pretty methed out
You're projecting nazi, methed out, supremacy - that's all coming from you
he's literally raised by hitler so.. i'm not sure what that's supposed to bring to the story
people project what's in their minds
i mean, the author always puts something into an idea for a reason
What it brings to the story is that he's raised to do evil
by hitler
yet he does good
but wait.. he's not really evil he's good! -- literally every propeganda campaign in history
it kind of rings like "hitler was kind of good"
you can tell a tree by its fruit
just delusions of a hitler fan
he kills Hitler, ends the Nazi regime, restores democracy
tell me, have you ever really been a huge fan of dracula and all the evil he did in the tales?
you think defeating hitler is being a fan of hitler
naw i don't really care
lol knew it
when the mad Muslims invaded his country, emissaries were sent to him to demand surrender
yeah he sure showed those guys lol
He told them to remove their turbans in the presence of a king
evil warlords vs evil warlords still deserve no worship. you're a cliche though
they said they never remove their turbans in public, for it is ordered so by the prophet
he offered to assist them in keeping that command.
He had their turbans nailed to their heads
sent them back to the invading camp
the mad muslims withdrew
there's a lesson in that
that you've begun to worship the propeganda version of a warlord. cool cool