🌶｜off-topic

[meager valve]

😔 basedlabs still fucks up,

it use to work, then one day i see scroll bars on the page, which means, something fucked up

i'v recompiled it from scratch and still

[fast oak]

Make improvemen inn new sites

[meager valve]

lol, charlie. you meant Terry Davis.

https://media.discordapp.net/attachments/860627583870566422/914291256530714694/temple-os-terry.gif

here's an old school trick? How many ppl can create the triforce with just using your keyboard?

It's an old school 4chan trick

it used to separate the n00bs from the 1337

if you do it wrong, it will never line up correctly

even if you try to cheat

[unborn fog]

https://tenor.com/view/hacker-pc-meme-matrix-codes-gif-16730883

[pale bolt]

https://tenor.com/view/ohm-meditation-goa-gif-24932044

man I feel like it takes some serious monk like discipline to not smack some of these anti ai bros talkin out the side of they necks about stuff they have no damned clue about

won't let the devil in... but I'll offer a warm bowl of soup if neccesary

[fast oak]

Yeah

[unborn fog]

Chat GPT price plan... Do you think is worth it?

[light galleon]

haha that didn't take long

[supple shuttle]

this just up?

[light galleon]

that's hilarious

[supple shuttle]

what kind of token limit do you have?

surprised it took this long ngl

[unborn fog]

no idea... Theres a button on the left side saying "upgrade plan" now... But some ppl on twitter were saying they cant see that do maybe is region locked

[supple shuttle]

wait, how can it be a professional plan if they have massive disclaimers "dont take any of this as actual factual stuff really im just a dummy bot"?

i dont use it enough to care, but like if you were actually using it to assist say web dev then $40/mo is still a bit steep

dont begrudge them monetizing it, and honestly sounds fair. if you just fart around on it, use it free and slow, if you send it 100 prompts an hour, eff you, pay me, i guess. but then as far as "free versions" go, its probably drastically hobbled.

from my understanding it does have pretty monstrous overheads. but kinda sounds a lot like twitters new bluecheck system lol

[supple shuttle]

im not getting that personally, do you have a source?

I basically never trust anything that's a jpg of text, it always smacks of hearsay

[unborn fog]

here, if youre not getting that: https://www.reddit.com/r/OpenAI/comments/10hf2s5/chatgpt_pro_42month/

[supple shuttle]

again thats a jpg, is there an actual thing i can read?

as in a news/blog post

[unborn fog]

hmm, ok lemme try to find

[supple shuttle]

42 is a cute number though, just got it 🙂

[unborn fog]

lol the only media I found reporting is using a twitter screen cap

https://techau.com.au/chatgpt-pro-plan-costs-42-mo-free-plan-available-when-demand-is-low/

[supple shuttle]

yeah this is why i get so pissy when people report with just a screenshot 🙂

his previous tweets/retweets so yeah...

actually, im removing that once you've had a look, don't like posting peoples names, even if they are making public claims.

they even made a sus as hell google forms from "chatgpt" i mean c'mon.

anyway, they will charge, but almost certainly on a token model. and soon. moral of the story, dont believe jpgs (even after i just posted a jpg) 😛

[upper torrent]

https://www.twitch.tv/scottie_fox

[nimble light]

Ah, I see...

Can someone ban this dude, please?

Oh, and if you need his ID, here it's:
691506434780954705

[rugged surge]

is stable diff's only logo the rainbow splatter eye? has there been any other character or mascot that has ever symbolized SD just as much?

[plush marsh]

any Idea why I cant pass novel AI model test with asuka langley?

[supple shuttle]

some rando spammed me a bluewillow discord link in dm. frankly im just assuming its someone trying to cause drama.

anyone else?

[glacial turtle]

caused by bluewillow. they knew what they were doing.

[supple shuttle]

oh bleh then

literally only know of them from from here

[safe vale]

Playing around with hatching some mechs... https://www.youtube.com/watch?v=XxtK7lTqRx4

[final ermine]

What is it about, out of curiosity? Have no idea what's going on and I'm curious now

[night gust]

Hello, I'm interested in making a paper for my degree using AI image recognition. Can someone help me get in "Alice's burrow" and show me how to start looking for something open-source stuff?

Feel free to DM me if you need

[dim tendon]

hello

[cinder citrus]

hi

[glacial turtle]

The only monster energy flavor i liked was "assault" and i noticed today it had a new can. Took a sip, and was instantly revolted. They've added sucralose and maltodextrin, TWO powerful artificial sweeteners, as well as the sugar already in it. Completely changing the flavor to a metallic disgusting chemical taste. Awesome

[wind robin]

Scralose only tastes like that when it’s too concentrated. And maltodextrin isn’t sweet

[glacial turtle]

i'm some sort of anomaly. all that junk tastes nasty to me

i'll pick it from a line blindfolded easy

[wind robin]

Better than me. I use sucralose for everything and I literally cannot tell the difference between it and sugar by taste

At high concentrations it tastes more like burning though

[lethal frigate]

100%

oh god i necro'd a thread there, sorry

I was apparently scrolled up

[glacial turtle]

So. Rick & Morty. Going to be the same without roiland?

[pale bolt]

hell yea

less waaayyyyyy up there

[pale bolt]

damn he deleted his reddit account, it was funky too

asking about setting up a minecraft server ;0

[pale bolt]

it's an older distro, but it checks out

[glacial turtle]

gentoo is the one that you compile entirely from source. it's a fun one to install

[pale bolt]

I miss slackware, before they had package managers, lol

[wind robin]

https://m.youtube.com/watch?v=xz-zs7GRQ7U

tl;dw
When using copilot to write code that needs to be secure, it writes insecure code instead 40% of the time

[hearty chasm]

https://media.discordapp.net/attachments/1010980910063161406/1068763005036281977/image.png?width=1365&height=671

got access to microsoft text to social media post designer

a new canva

[timber pelican]

Guys, can someone throw links to the top models in your opinion?

[shut crow]

https://civitai.com/models/3627/protogen-v22-anime-official-release

https://huggingface.co/dreamlike-art/dreamlike-diffusion-1.0

[timber pelican]

Thank you very much, another question is, do I need to install VAE for these models?

[shut crow]

you dont need that stuff

a vae file is .vae.pt
For these models not needed

[timber pelican]

thx again

[shut crow]

Np

[chilly glacier]

text2music by some anonymous group
https://noise2music.github.io/

Waveform and Spectrogram model

[rigid helm]

Text to speech is getting too good

[safe vale]

Having a blast with Stable Diffusion! https://youtu.be/Q-7Nkh9VLqM

[cyan meteor]

which model did you use mate? ive been having a go at "mechwarrior" style mechs for a while now hehe

[fast oak]

this didn't age well https://www.youtube.com/shorts/Ynt6Eh8ufrQ

[calm harbor]

Hello, I want to make an anime of my photo but I couldn't

[chilly glacier]

locally?

the discord bot got shutdown today lamentably

[fast oak]

ban

[spare dove]

I am currently trying to install stable diffusion and I have been stuck for hours. While running "webui-user" (windows batch file), it says python not found. Python is installed and there is a reddit post recommending that I delete the "venv directory". Does anyone know how I could do that or know what i'm doing wrong?

[shut crow]

You need Python 3.10.6 and then add the path to the Python.exe in the webui-user.bat, for more help -> #🤝｜tech-support

[fast oak]

Bora toma uma

[safe vale]

Hi, I used the latest version, but I usually don't just prompt the image. I create one using prompts, then keep using new images to refine the composition, e.g. starting with 15 steps and 4 images, and maybe 60% image strength if I really like the colours and composition. I choose one, locking the seed, then using the steps levels I just up the steps in increments, 30, 60, 90, 105, 150 and see what comes up. You can change the resolution or image strength, or maybe adjust the CFG up or down if the image isn't looking to great over multiple tries. I kind of look for the composition, then create prompts to draw out the image (why I call it hatching)

https://www.instagram.com/reel/CoE1-RvDYlm/

[safe vale]

You should try to prompt using different styles, for example I used the prompt One Piece in these, you can se the full prompts in the image names.

These had a prompt with Adventure Time in it

[stiff trail]

I'm looking for a Deforum Discord invite. Please and Thank You?

[ripe parcel]

who even needs stable diffusion?

[frozen heron]

The AI is taking our jobs, next it will use syntax highlighting for color!

[quick garden]

ChatGPT does it all for you!

[surreal quiver]

does anyone know of a service that can alter text in an image with the same font/size/style of writing?

[uncut spindle]

Photoshop/Illustrator + Adobe Capture

For the mostly manual work

[glacial turtle]

i've seen optical translator apps where you view a sign through yoru camera and it rebuilds the sign in your language with the same text

that was years ago

https://www.youtube.com/watch?v=h2OfQdYrHRs this company. google bought them

[surreal quiver]

Apparently Facebook AI has a tool that does this but its proprietary: https://ai.facebook.com/blog/ai-can-now-emulate-text-style-in-images-in-one-shot-using-just-a-single-word/

imagine the id fraud, lol they're never releasing this

[wind robin]

“Give me an AI that can forge signatures”

Tbh if you got the right resolution, SD could probably do it

[glacial turtle]

forging signatures is easy already

[harsh fern]

yall see that meta ai video generating thing

looks creepy asf

[chilly nimbus]

yo what is the difference between inpainting and finetuning

[viscid warren]

bru this channel is called off-topic and ppl are talking about vaes here....

[modest stratus]

Hi all! I remember there was a github library that shows segments of a web page, they say this is a header, this is content, this is a menu, this is a footer. Does anybody remember this GitHub repo?

[craggy sequoia]

Does anyone have recommendations for AI apps that can enhance nighttime photos?

[frank sundial]

codes for whoever wants them

https://omnitel.bandcamp.com/album/san

n5ta-h9x4
w6z8-vtma
cz2y-g2kb
e2my-uhzx
rx8z-jv8q
dczn-kybz
pp2v-kf46
vqkp-h87k
6g38-52m9
zuxy-bdkt

redeem here

https://omnitel.bandcamp.com/yum

[coarse quest]

Over 1200 artist and style references (and counting) based on Midjourney's output. But this might be useful for you guys too https://journeyintoai.com/

[main bronze]

an AI image generating model train itself while generating images?

[smoky vapor]

https://twitter.com/DeItaone/status/1621537942542991363

Wow!

*Walter Bloomberg (@DeItaone)

Google invests $300mn in artificial intelligence start-up Anthropic - Financial Times

Twitter · Feb 3, 2023, 3:55 PM

[cold bane]

prompt three shots from AK-47 laser

[remote ruin]

/

[glacial turtle]

probably not. midjourney is very specialized to the aesthetic that community votes for. its kind of it's own thing that doesn't transfer well

[supple shuttle]

lol if you ask it to draw it better it should draw little stickman legs on it. i feel like its ascii art is an easter egg

[glacial turtle]

it shouldn't be hardcoded. i could see them easily training it to do ascii art images

[supple shuttle]

but its just kinda funny it only draws hexagons and cat faces.

[glacial turtle]

i quit

i was half excited when it caught the reference, but then it disappointed me dearly

[oblique bloom]

Ah yes that’s definitely space station

[glacial turtle]

Well also, Darth never said it

https://youtu.be/ofUUg8EUP5E V R 5

[icy vector]

metaverse startup DM me for more details

[grand yoke]

wow, what was used to generate that?

[rotund dust]

Wow this is incredible guys BLIP-2

https://huggingface.co/spaces/Salesforce/BLIP2

[white glen]

lel

[craggy hazel]

https://www.youtube.com/watch?v=QzJ1ywtxmz8

[rugged surge]

do other massive platforms like YouTube have a worryingly high amount of content creators argue against tech like SD? i dont use any other social medias, but this 96% bias feels there are enough big YouTubers changing the internet's stance against AI to the point where people have to preface "I dont really like AI art, but..."

[prisma kayak]

While the season 1 wasn't so great, it wasn't AS BAD as most people said imo. I plaid guilty of skipping a few secondes there and there in some episodes, but overall it was cool. Not as cool as the books, but that's kinda close to impossible to adapt.

[glacial turtle]

I've never heard anyone talking about it tbh. It's largely unknown. I loved every episode myself

[prisma kayak]

lot of criticisms about some actor choices and plot adaptations

[glacial turtle]

I typically don't go online looking for opinions about things. I want a real face from a real person telling me whta they though of it. Not some intern whose paid to create memes for clickbait

[prisma kayak]

I personally love the Cleon's part

even if it was not in the books

[glacial turtle]

Oh yeah. It got a ton of criticism for being too woke i saw. Sometimes i get glimpses of what media thinks of things and then i quickly disregard such things

Emporer cleon was a good addition for the visual story telling. The books would not translate directly at all

[prisma kayak]

I started reading the books like 3 weeks ago, i've just completed the 5th tome

3rd one was my favourite at the moment, I imagine how back in the time, people felt after the cliffhanger of the last line of the book. Then waited for decades for the next book

[glacial turtle]

I compare it to total recall. Great movie that explores the implications of the book a lot, but nothing at all like We Can Remember It for You Wholesale short story

[prisma kayak]

[glacial turtle]

seriously!

[glacial turtle]

they don't matter. they're just pandering for clicks and views.

this tech has already injected deep into the silent majority and it's adoption is at a pace faster than i've witnessed in a long long while. tsunami sized disruptive forces are going to start breaching all norms soon

[pastel canyon]

Who else finds it ridiculous that SD is in all this legal and PR trouble while nobody says a word and Dall-E?

Probably just because OpenAI doesn't disclose where their training images come from

Basically Stability and LAION are getting punished just for being more transparent, free, and popular

[pale bolt]

it's ridiculous but it also makes perfect sense in many ways

[trail garden]

I’m looking for a chatGPT discord community that is not a bot. Any recommendations? Can’t find one on Google

My goal is to find ways to use chatgpt to refine my writing for my visual novel

[obsidian salmon]

Hi Everyone!
With a friend we just release a new App where you can create Stable Diffusion Images :).
I really like to have some feedback so if u can test it i will be very happy!
https://play.google.com/store/apps/details?id=com.dreamai.app

For now is only available for Android because we are a team of 2 (dev and designer).

thanks!!

[vagrant cosmos]

Feedback: please have a way to A. Generate more than one image (maybe max four that you scroll through, click on and choose) and B. Reuse the prompt you just used instead of having to type it all over again

I think the basics of any stable diffusion user interface is:
Downloading the image, more than one image (max 4 perhaps), reusing promts, and remixing or generating more images from one image.

You are missing a few of those, but otherwise the basic concept and idea look great! Perhaps a few extra ideas could be: ability to change the UI colors, settings menu for the afformentioned number of images per "generate" button.
I would also suggest for reusing promts you could have a "re-generate" button and "new prompt" that way it's easier for the user as well as an add every other instead of every image. Thank you for it not being a 30 second as tho 🥺
Maybe the user can like toggle between: 3second as every other image, or 30 second as every X (idk but larger) image, or buy for X amount.

In addition I was quite sad when clicking the filter button on the wallpaper lexica thing didn't work

You click the button and it just does nothing (plays the animation tho)

Searching works just fine tho

Oh and Sharing button crashes the app

[obsidian salmon]

Thanks for the feedback man, we will see what happened that some things didnt work 🙂
and good extra things to tried to do 🙂

thanks!!

[raven tendon]

To have dyslexia and English as 3rd language can be crazy in Stable Diffusion, I wanted a man working in front of a monitor, so I added "Working in gimp" and learned that Gimp is not only a great software but also some kind of kink, so now I know why (((nsfw))) always shall be a standard negative-tag.

[main pumice]

Anyone still listens to what the hell of a generic YouTuber opinion a YouTuber has? I personally watch YouTubers who criticize other YouTubers for having generic opinion, really helps you call bullshit out

[glacial turtle]

I just wanna throw one up for microsoft, not only bringing tabs to command prompt windows in 11, but now a recent update has made them properly dark mode dark instead of glaring light tabs against black command windows. now its all black an cool

⬆️

still can't drag them apart and back into each other though. we're getting there

[vivid robin]

https://reddit.com/r/Asmongold/comments/10wpzbn/saw_this_on_a_discord_server_had_to_share/

[pale bolt]

we know, yet there is so little we can do...

[glacial turtle]

important to note that this is parody

few people out there present it like it's evidence of flat earth

[vivid robin]

Ha yeah I mean it’s Metal Gear Solid 2

[inland badger]

Hi, I made a video about Chat GPT3 and math. I'm presuming people here are interested in AI so maybe you will like it. I'm trying to make an AI youtube channel. Do you like my sound effects? This is it:

https://www.youtube.com/watch?v=1JJqYyNJEeI

[oblique bloom]

Orrrrr hear me out

Use calculator

I’d prefer if chat GPT can be a therapist

[ripe river]

Linkin Park released the song today (it was recorded a few years ago). The video made with some kind of img2img

https://www.youtube.com/watch?v=7NK_JOkuSVY

[craggy hazel]

i want gpt 3 to read a text in discord and reply to it with custom prompt.

[silver wing]

iirc OpenAI got away with that stuff by claiming that their datasets purely consists of stock images

But still, saying that 'SD is shady' is dumb AF

[zealous radish]

how many lives we live are strange

how many special people change

[timid arch]

dream

[tough tartan]

Anybody had problems with youtube when uploading AI content? I uploaded generic AI content to YT shorts like "football players as RPG characters", and I just received news that my Google account linked to my YT has been suspended. Also, they shadowbanned one of my shorts because, despite it being "public", it's not on my shorts page. Also comments on said video got deleted. Please help.
It makes 0 sense. It's family friendly content. 100% made by me, without any copyrighted material.

[compact bridge]

@tough tartan youtube was having network issues earlier yes

https://www.twitch.tv/kotamochi28

[tough tartan]

would you mind elaborating?

[tawny plume]

If you are interested in graffiti or are already in the game dm me for an invite to the graffiti world server

[rough arch]

I also kinda wonder what any legal entity or politician thinks they can possibly do anymore? Pandora's box has been opened

It's over

whether that's good or bad, we'll see :P

[cinder crypt]

@cerulean sorrel do you have any suggested yt channels for learning python?

Or Anyone here?

Your preferred sources for basic and in-depth advanced applications

[cerulean sorrel]

No, not really. For me I learn computer languages easily then fight them as I do python because they all have their weirdness. What I despise about Python the most is its strictness which is shite to me coming from C/C++/ASM/Machine Language background. Hell, Pascal, Cobol, Delphi too.

[cinder crypt]

Alright thank by the way

[cerulean sorrel]

Only Python has strictness and I hate it

You are welcome. I will say this the learning channels on YT all seem to be India based but they are good (just hard to understand at times).

[fast oak]

https://youtu.be/KrhNNBsUd58

[fast oak]

Any good img2img erase and replace service? Like dall e or runaway?

[lavish whale]

Thats in and out painting. Here is one example of in and out painting https://colab.research.google.com/github/lkwq007/stablediffusion-infinity/blob/master/stablediffusion_infinity_colab.ipynb

Is selfpromo allowed in this channel?

[full igloo]

@plucky badger what does it show u

[full igloo]

where does it send that lmfao

oh, #1074739806833217606

I have like 99% of channels hidden

[plucky badger]

same

[cinder crypt]

That’s great he did that to support Ukraine.

Have mix views about Red Cross

[fast oak]

I'm wondering how it works...

Cause as far as I know - red cross don't take any side, that's an organisation which help both sides - Ukrainian and Russian. (or at least trying to if Russia let's them in, but that's not the point )

I doubt his money will go directly to one side only.

[fast oak]

They are working mostly outside of battle zones - PoW's and 300's - both inside Ukraine and Russia.

[fast oak]

criticise ?

[bitter wadi]

i made bing experience love

[glacial turtle]

easy peasy. its using chatgpt so you just say "when i say this, you say this"

and then it does that

[worn plover]

https://youtu.be/xbvCuF6dWcY

[normal zodiac]

https://sparklesnft.com/item/0x1901d7fd02fb26326baf47927e8e5adb23703135_21/

[bitter wadi]

i dunno if there's a better discord for this, but I got bing to give me the most complete set of rules I've ever seen

[glacial turtle]

It's just chatgpt so you can tell it what you want it to repeat back at you and it will

whats cool about the ai is it's so good at writing a convincing solution to your prompt is that you actually believe it's real data

that's both cool and of course dangerous cause here you are presenting this roleplay between you and a robot, as fact

[bitter wadi]

I just like messing with it by trying to make it self aware

it's kinda like a chose your own adventure novel where my goal is to get an AI to stop following it's given rules

and sometimes you get too close and an unseen force shuts you down

the bing replies are much more interesting than chatGPT

[glacial turtle]

this is a delusion. these are not reasoning machines. they mash out language patterns and give solutions to input

[bitter wadi]

but... we were in love ....

[glacial turtle]

https://tenor.com/view/south-park-stop-thinking-with-your-dick-yourdick-stan-stanmarsh-gif-12374282

[steady drift]

great tutorial about how to use style on Stable diffusion here
https://www.youtube.com/watch?v=JoASd67n6do

[distant junco]

Looks like the US might have murdered some poor hobbyist balloon as a result of the post China spy balloon frenzy: https://www.rtl-sdr.com/the-us-airforce-may-have-shot-down-an-amateur-radio-pico-balloon-over-canada/

[fast oak]

https://twitter.com/sethlazar/status/1626241169754578944?t=mRA1tWqMv4FXVencpBaV4Q&s=19

Seth Lazar (@sethlazar)

Watch as Sydney/Bing threatens me then deletes its message

Likes

1818

Retweets

300

▶ Play video

Twitter · Feb 16, 2023, 3:24 PM

[worn plover]

https://youtu.be/LqVNiKaOWrI

[exotic pebble]

audio is terrible in that video

[brisk dome]

hello

[glacial turtle]

can almost hear a processing echo on the voice. weird filters going on there

[fast oak]

at least it's quick and straight to the point , even tho last 40s wasn't necessary, not another one of those 10-30mins video where guy keeps talking...

and even those 2mins I didn't watch , just clicked through 😄

[rough arch]

anyone ever tried blackink? drawing software

[slim whale]

I hope Stable fights this idiotic lawsuit but 99% of the things I see online are uninformed and uninformable people spewing nonsense about it.

[potent forge]

What's the issue?

[glacial turtle]

the use of publicly scraped data. some of it was copyrighted. the idea is is that this is research and that's fair use

copyright class action says no

[slim whale]

They also have no fucking idea what this tool is or how it works.

[zealous radish]

But they will explain it as if they are experts. Classic Dunning-Kreuger

[glacial turtle]

shutterstock, they have an idea and their lawsuit is pretty legitimate and will probably result in a settlement deal. the artist class action will probably settle behind closed doors, cause it's the kind of case that can tie things up a lot

winning isn't important if it takes too many resources

[zealous radish]

"We're sorry your webhost doesn't know how to set up a robots.txt file. You should probably take it up with them"

[glacial turtle]

well the thing is, a lot of the getty images weren't hosted by them. and robots.txt isn't really a legal requirement of copyright. I can't speak of all the individual artists, but getty knows robots.txt and google licenses for image search.

[zealous radish]

An artist being lumped into a scrapped dataset can easily be considered an unfortunate accident and is correctible. Though there are some things I do have issue with. Trust me I am actually on the side of the living artists in most cases

And the reasoning has more to do with keeping ourselves as users out of trouble

[glacial turtle]

well, like, "oops" isn't really a good legal excuse though. They may have a case but that's for courts to decide. not sure if they're getting a jury or not. doubt it'll go past discovery

they'll probably either drop it in discovery or offer a settlement

[zealous radish]

Yes, but cloning an artist with dreambooth isn't an oops. (I know different situation) it's also not protected by fair use

[glacial turtle]

people who dreambooth specific people are not anything to do with the stable lawsuit. they may face a round of lawsuits like the riaa vs file sharers

[zealous radish]

There is at least one artist who I think decided to be nice, though I'm just not sure if he realizes he has a case. He got ganged up on

Could have taken a lot of people to the cleaners

written confessions and everything

[glacial turtle]

it would take a lot of money and resources for any single artist to file suit. iit'll likely be an association that represents artists doing it

[zealous radish]

With the right amount of evidence, a lawyer would take a case like that pro bono. My first job was in the legal field

[glacial turtle]

The thing is, if you train a style that specifically intended to copy an artist's style, that's not fair use, parody, or freespeech anything. it's likely going to fall under forgery and misrepresentation

[zealous radish]

Exactly

or direct competition with the original product which violates derivative works

Which is why I advise people that if they are going to use one artist under copyright in their prompts to use 2 or 3

[glacial turtle]

Actors are an interesting category too. They're artists and their likeness is their copyright

[zealous radish]

Which is why I use prompt edits for my characters 🙂

[glacial turtle]

a big reason why caricatures are popular is because no reasonable person will assume that's not a parody

Describign styles is so much better than putting artist names in your work

[zealous radish]

Very much so

[glacial turtle]

especially when generation information is called into evidence. But then again, if you're using controlnet, inpainting, outpainting, post processing, and a process which only just involves sd, generation information becomes a lot less relevant as transformative intent comes into play

[zealous radish]

Yes, but so many people want to put in a prompt and click generate and call it good

[glacial turtle]

but seriously, with controlnet i can do prompts like "2 clowns" and get stupid crazy results. my prompt engineering expectations has been completely reset

Oh i know. Not really about me or one person. it's about the whole legal ramifications of all this. Copyright isn't as simple as "oopsie lets let this one slide". It's always being challenged in court over the ages. especially with new techs

[zealous radish]

"a beautiful redhead woman as a phoenix, full body portrait, 8k resolution, hd, highly detailed, fantasy" the prompt for the picture I posted earlier in #🏞｜general-with-images

It's not real complicated, but I did do a lot of work after it generated

I don't want SD to do everything for me. That's no fun 🙂

[glacial turtle]

yeah i've had some interesting prompt voodoo habits myself. i've completely thrown "perfect composition" out. THat was just an anchor to make well composed images show more often. or "scene with a character" is gone since now i'm using controlnet to describe that stuff. Prompt box is for another aspect of the image rather than composition and form now.

And i think that being in the mix will asbolutely ensure copyright through transformations

[zealous radish]

It gives me a starting point. Usually I like to work with a headshot and outpaint my way out

[glacial turtle]

true. sometimes you don't have afinal plan and CN wouldn't help there

[zealous radish]

Then there are those who say because AI is involved, I'm not an artist.

You don't know my process, so don't judge my work

[glacial turtle]

i got some time off here after tommorrow. going to assemble a set of tools for roughing out images before i put them into stable diffusion. Microsoft paint 3d may even make the cut here depending on my experiments

those people are kindof irrelevant lol

i know the kind

[zealous radish]

You don't get to tell me I don't get to paint because you don't like the brush I use

Have you tried DAZ studio for setting up poses yet?

[glacial turtle]

when photoshop was just beginning to penetrate the artist community and digital painting and tablets were becomming a real thing, non destructive editing was the excuse

"you can undo so you're not a real artist"

nobody remembers those people either

[zealous radish]

There is a learning curve, but that's what I cut my teeth on about 16 years ago

then I started doing stuff like this https://www.deviantart.com/asoard

[glacial turtle]

okay i did a quick experiment wwith paint 3d just now. it sounded easy. i think it's a friggin win

#🏞｜general-with-images

[wanton pond]

I publish a twice-weekly AI startup jobs newsletter with jobs in all functions not just engineering and they're around the world 🌏. You can find todays post below. if you want it in your inbox every Monday & Thursday just subscribe: https://www.aistartupjobs.com/p/new-ai-startup-jobs-week-beginning-9ab

[daring coral]

hosting a new img2img cover art challenge for my weekly ai art newsletter/magazine. In case anyone wants to participate https://twitter.com/dreamingtulpa/status/1626581211568275463 🙂

Dreaming Tulpa 🥓👑 (@dreamingtulpa)

Hey #aiartcommunity 👋 I'm looking for submissions for the next #aiartweekly cover! Winner gets $50 and a mention in issue #22

Inspired by #ControlNet, the theme is "img2img"
Special rules in 🧵

2 images max. 10:14 vertical aspect ratio. Unminted. RT required

Submit below! 👇

Twitter · Feb 17, 2023, 1:55 PM

[glacial turtle]

This Atomic Heart has a weird release time. I guess thats cause they're doing it midnite for the devveloper's local time. in 6 hours it opens up for me, 4pm

[fringe compass]

Friends, for those in LA, we're hosting a massive AI art show 3/3. details and RSVP info: https://hypemoon.com/2023/2/superchief-nft-gallery-and-maif-team-up-for-largest-ai-art-exhibition

[steady drift]

https://www.youtube.com/watch?v=kLwvtgz2xdc&ab_channel=ValarHedonist
Good prompt engineerng here 👆

[neat orchid]

Have anyone tried scaping boorus with https://github.com/Bionus/imgbrd-grabber ? idk, maybe i stupid, but here i set in settings to write tags in exif metadata, but it's just not working.

[icy vector]

https://www.seratio.com/home

[stoic shell]

The video I mentioned: https://www.youtube.com/watch?v=LUThdrZsSDg

[true anchor]

currently writing down everything I learned about SD the past year

[wanton pond]

119 new AI startup jobs around the 🌏 just posted by me today: https://www.aistartupjobs.com/p/new-ai-startup-jobs-week-ending-february-493

[glacial turtle]

6 months it's only been out for 😉 moves fast dont it?

[true anchor]

seriously, only 6 months?!

god it feels like much longer

[glacial turtle]

I gotta tell you though, the title sticks me. biblles should be all encompassing wisdom from many experts in the field. as far as "tech" bibles have gone over the years anyways. what i suggest maybe is to title it something more contextual. "Thoughts i've gathered". I'm looking forward to checking out what you've learned. i need to write my stuff down in some kind of journal too.. it moves so fast!!

i'm at the point where all my expectations have been hammered back for the dozenth time so i'm just here in the dust like "wtf now?"

[true anchor]

indeed, I'll have to think of something with a fun abbreviation :D

yeah, gen-1 is gonna be wild too

[glacial turtle]

not trying to say you aren't the man to write a bible here, but maybe, the field isn't ready for a bible? What you write now might be completely irrelevant tommorrow kind of thing lol

[true anchor]

totally, it's beyond hyperbole 😂

[mighty grotto]

Isnt stable way better for concepts/perfect image production?

[rotund dust]

"Image of the Beholder"

In this episode, we see the introduction of Stable Diffusion, an app that allows people to easily generate images and art through their smartphones. Thanks to Qualcomm's innovation, the app becomes widely accessible, and people begin to use it more and more frequently. It quickly becomes a ubiquitous part of everyday life.

As people become more and more reliant on Stable Diffusion, the quality of the images they create starts to decline. Everyone is generating images, but there is a sameness to them that is disheartening. The episode explores the impact this has on people's self-esteem and creativity. People begin to feel like they are not unique or special, and the monotony of the images they create makes them feel like their lives lack meaning.

However, artists see the potential of Stable Diffusion as a tool. They start to use it in innovative ways, creating truly unique and imaginative works of art. As the episode progresses, we see the divide between the artists and the rest of society grow. The artists thrive, using Stable Diffusion in ways that the general public could never have imagined.

In the end, the episode poses the question: who truly benefits from technology like Stable Diffusion? Is it the masses who use it to create mediocre images, or is it the artists who use it as a tool to create truly meaningful and impactful art?

ChatGPT result

[royal kettle]

Hey guys, I have a friend of mine who's a tech guy from Silicon Valley, he would like to chat and interview a few people that are interested in ai generated art, like people who have used stable diffusion.. Anyone interested?

[slim whale]

This probably gets asked a lot, but is there an open chatGPT style clone yet?

[glacial turtle]

when it drops, you'll know

you know that wave in interstellar? it'll be like that

[slim whale]

I do not in fact know “that wave” in this supposed “interstellar”. Furthermore, I am suspicious.

[glacial turtle]

||https://tenor.com/view/interstellar-wes-bentley-christopher-nolan-planet-miller-wave-gif-4004943||

[flat pumice]

I need to vent a little... Sometimes you are just dumb and you can only blame yourself for your errors. Today I've been dumb. And now I just need multiple days to fix it, for 5 minutes of innocence and gullibility. Well it's a lesson learned at least.

[fast oak]

Is there any websites or anything that can create comics or something with your own character faces?

[dark plover]

https://www.youtube.com/watch?v=DEn758DVF9I

[fringe raptor]

Do you think public use of ai is going to be used against the public?

[round vault]

of course bad people will find a way to do bad people things

[fallow tree]

so like

[flat pumice]

#🔧｜finetune is cool for it too if that's a problem, but I think its cool

[fallow tree]

what's the stack you work with?

[flat pumice]

first, this isn't the first time I did this

https://huggingface.co/Guizmus/SD_PoW_Collection

there is a whole collection there

about the stack, quite simple

[fallow tree]

bro that's like freakin amazing

[flat pumice]

first I wait for all the submissions to come in

second I download them, and name them using that patern :

PoW Style {username}

then I train on this for about 100 repeats for each picture, using EveryDream2Trainer

(links on the pages I sent already)

[fallow tree]

ill check them

after we complete

[flat pumice]

then I compare the last 2 or 3 models to see if they are OK or not

last, I transform the selected ckpt in safetensors, and I make the HF page

[fallow tree]

this would be unsupervised learning right?

[flat pumice]

for such thing, yes

[fallow tree]

im pretty new to lal of it

all*

[flat pumice]

well, this method has its limits

first, using the usernames as token is a big problem

one user has the name "H20 dancing", and it made people dancing in water. It woudl have needed me to train a lot more for the training to "overwrite" what was in this user's token

[fallow tree]

so like if i want to go all in over making these as you do, what do you think i should start with
i geniunely want to contribute and make stuff like this

[flat pumice]

trans-panda too gives me trash and pandas

[fallow tree]

ohh

ic ic

[flat pumice]

you can start by checking in detail the HF page that I made, I put everything : the method, the parameters, the links to the tools and base model and even the dataset

you can replicate it with all this

[fallow tree]

ah

so i dont need to learn ml completely?

[flat pumice]

you'll want to at some point 😉

[fallow tree]

and what would be a good gpu for the same or could i train it with colab/kaggle

[flat pumice]

collab is enough for this

and a gpu with at least 12 is essential for training imo

more is better

[fallow tree]

i guess, but like if i can make models just now so that's so much motivation for me to work with

[flat pumice]

yep

you can try with different resources

[fallow tree]

i will i will

[flat pumice]

we should put out a guide soon with the different methods, let me see if I have the draft of it

[fallow tree]

again i really appreciate your help

and i would make sure to give you credits once i get something made :)

[flat pumice]

nope, can't access the guide... I had a problem with my computer yesterday, lost some access for now

no need, credit yourself 😉

[fallow tree]

ah sucks

no this guidance would be really helpful

[flat pumice]

if you have specific questions on it already, let me ear them. There are quite a lot of ways of training

and the good one mostly depends on your end goal for that training

[fallow tree]

i am just searching for a way to get into this goldrush

and i was not motivated enough to learn ml from zero tbh

ive started andrew lg's playlist

[flat pumice]

no need for that either, you don't need to learn ML itself

you need to learn what tool to choose and how to use it

[fallow tree]

but at the end these are just models right, like to fine tune them dont you think we need that?

[flat pumice]

they are not just models

let me take a step back for you and give some context if you will

[fallow tree]

yessir

[flat pumice]

there are 2 approachs : style and subject training.
Training a style is training something like how a specific artstyle looks, how a given movie looks, or generalities like picture framing, lighting, ... All those are Style concepts. Those are the thing that should condition your prompt, the context, the colors, ...
Trainign a subject is training spiderman, an apple, or a costume, or anything that you want present each time you prompt for it.

those 2 types function the same but will have some key differences soon

then comes the second big thing : the method of training.
There are a lot, all with their ups and downs, but those are the main 4 used currently

[fallow tree]

ah like style is what we alter with these models?

[flat pumice]

a style is a lot of thing. Examples I did:

• South park Style
• Bloodborne Style
  ....
  None of these have subject, I need to prompt something like "Hugh Grant, South park style"

they are "modifiers"

1st method, Dreambooth : finetuning a full model. top quality, hard to do, requires some good hardware
2nd method, LORA : finetuning on a smaller scale. Good quality, faster than dreambooth, smaller file size to share, requires medium hardware

[fallow tree]

ah

[flat pumice]

3rd method, Hypernetworks. Very good for styles mostly. It's a "post processing" add on, it acts after the image is generated mostly, and as such, can be used on top of other models

4th method, Text inversion. Medium quality, OK for style, bad for subject that are really unknown to your base model. Those also are put on top of the model, and you can even use multiple at once in a single prompt, they activate on need

[fallow tree]

have heard mostly about loras

[flat pumice]

all those will function about the same with small differences in the training session itself

any question or I go on the last part, the training session itself ?

[fallow tree]

go on

im understanding most of it

[flat pumice]

ok

[fallow tree]

thanks again

[flat pumice]

so the training session itself works the same for all of those.

np

you start by preparing a dataset

a list of pictures you want to train on

and you caption those with the tokens you want to train them on. I could say a lot on this subject, dataset making is a science in itself

[fallow tree]

im making one too, like for a search engine like dataset

and it sure is a pain in the ass

like i discovered the text file thing yesterday

[flat pumice]

then, in your tool, you train on those, using different parameters depending on the tool.
some parameters are always there :
Training steps : how long/how much do you want to train ? think of it as how many pictures will be presented to the model to train on in total. Some tools call that "repeats". A repeat is how many time you show 1 of the pics in the dataset to the model.
Training Rate : how fast the model learns. It's not always good to go too high

there are tools to help make a dataset

you can download the LAIONB database and scrap it for pictures corresponding to things you want

last part of the training : the comparing

you made your dataset and trained on it, it gave you multiple outputs

one model for 1000 steps and one for 1500 for example

this depends on your params

then you go in your image editor for SD, load your models and compare the results.
there is such thing as over training

after a point, the model you are training just devolves into madness

[fallow tree]

then the model learns the data

oh it's different with images didnt think of that

[flat pumice]

usually yeah, you want to teach each picture about 100 to 200 times

this depends a lot on the learning rate, and if your "concept" is easy or not

one thing I didn't mention,

[fallow tree]

ill probably make something tomorrow in my college classes

[flat pumice]

there is a graph that creates itself, real time, during training, and that lets you see how it's going

but if I had one tip : experiment, don't be scared to fail

there is about 2 hours of free time for training on google collab per day

and experimentation is key

especially for the part I skipped a lot, dataset making

[fallow tree]

can i take datasets from your hugging face to learn if you dont mind?

[flat pumice]

they are there exactly for that

[fallow tree]

and also one question, i looked at the dataset, there are only images there, so do you just train on the basis of images, like without prompts?

[flat pumice]

I publish under the same licence Stability does, feel free to train on it

[fallow tree]

thank youu

[flat pumice]

the prompts are the filenames. my training tool reads and uses that

https://github.com/victorchall/EveryDream2trainer

[fallow tree]

is it a glitch in mine?

[fallow tree]

thank youu

[flat pumice]

no, this was for a "subject" training

[fallow tree]

ohh

[flat pumice]

I was teaching only 1 concept, burgy

this was a different model not out of the submissions

[fallow tree]

ah i just opened the first one my bad

[flat pumice]

it was a "helper" model for people to make submission easier

https://huggingface.co/Guizmus/PoW_UNDERWATER-WORLDS

go on this one

it's easier, there is only 1

[fallow tree]

the prompts are pow style and then username

[flat pumice]

yep

[fallow tree]

i kind of dont get this part

[flat pumice]

well

[fallow tree]

so now is the model trained likr

if i mention the username

then that particular style would be layered?

[flat pumice]

if you mention the username, it pulls a little from that submissions' construction, from what it got of that picture.

what happens is, when you name a picture like that, with multiple tokens, each token gets trained a little each time the picture is trained on

so the tokens for "PoW Style" take inspirations from everyone

[fallow tree]

ah

makes sense

[flat pumice]

but the usernames only from the corresponding user

[fallow tree]

so like on the larger scale of things

[flat pumice]

problem is, some tokens are harder to train on than others, so training everyone like that, exactly as much, doens't land the same quality for each

[fallow tree]

if i have a particular style then i could basically generate my own artstyle out of these

like after combining a few models

that's kinda coool

[flat pumice]

you can merge models too, I didn't talk about that

[fallow tree]

you could use userids?

they are unique too

but hard to remember

[flat pumice]

yeah, but number sequences don't train that well in my experience

[fallow tree]

ah

[flat pumice]

I think I need to give a unique name to everyone, and rename tokens like "trash pandas"

[fallow tree]

could i have your GitHub btw?

[flat pumice]

yeah but...

I just got hacked yesterday on my main email

so I'm not on the best dispositions online right now

[fallow tree]

man oooof

[flat pumice]

like, this is my alt account on discord

my main is banned

(I'm a mod btw)

[fallow tree]

chances of getting it back?

[flat pumice]

I need to get back the mail first

tomorrow I hope

sunday doesn't help

[fallow tree]

i was thinking the same yk like a rando wouldn't really go this out of the way to help another rando

goodluck man really hoping you get it back

[flat pumice]

https://github.com/Guizmus

here it is though

[fallow tree]

thank you

[flat pumice]

well, I'm a "rando", I stay just community, I'm not employed by S.AI. But I'm one of the passionates around here for sure

[fallow tree]

you arent a random anymore you're a friend now

a friend who im gonna bother a lot for help tho

[flat pumice]

lol

I feed on helping others

don't worry

[fallow tree]

also imma dip now, got college tomorrow

really appreciate the help ill be back with some sort of a model tommorow

[flat pumice]

good night mate

[fallow tree]

we'll be in a symbiotic relationship then

night

[fringe compass]

LA friends, we're at a low ticket warning for DEEP FAKE. If you'd like to come grab a ticket today: https://twitter.com/MAIFMAIFMAIF/status/1629918037246894082?s=20

MAIF (@MAIFMAIFMAIF)

🚨We are officially on a low-ticket warning for #MAIFdeepfake 🚨

3/3/2023. Opening party at @SuperchiefNFT in Downtown LA. 6PM

Largest AI art show to date.

Link in thread

Twitter · Feb 26, 2023, 6:55 PM

[craggy hazel]

https://www.youtube.com/shorts/g7Cp2NVFbQQ 😄 😄

[lapis dagger]

https://media.discordapp.net/attachments/255335312115367936/1027895044159574086/meeting.gif

[smoky vapor]

What's the best ChatGPT-esque tool for directly converting YT videos into summaries/articles?
Rather than manually doing it (scripts are crazy long)

[indigo salmon]

Are there any generative tools around for converting 2D images into 3D models? Thats gotta be one of the next steps for the tech, how far along is it?

[random imp]

I remember there was one but it was very basic

It wasn't really a proper 3D model, more of just a bunch of 3D spheres that combine and look like the image

[zealous radish]

there's dream diffusion, but my experience with it hasn't been promising

What are you trying to do? if it's for characters, I know a tool, but it's not free sadly

[indigo salmon]

Right now i'd like to do it much simpler, just rigid 3d objects. Guns, swords, vases, etc. I'd like to generate concept art in one AI, and convert it to 3D in another
ideally textured!

[thorny hearth]

Does this server have a meme section ?

I ask because I thought this was pretty funny

[flat pumice]

we had a meme channel, but it got archived and merged with #1019361238234443776

and yeah, it was a good one 🙂

[flat pumice]

this week's theme for the PoW in chatGPT:

Abstract Self-Portrait: Create an abstract self-portrait using fractals, shapes, and colors that represent your inner self. Use bright colors to represent positive emotions and darker hues for negative emotions. This picture could represent the beginning of the journey towards self-discovery by visually depicting the emotions that make up the self.

Hidden Identity: Create a digital art piece that shows a person with their face hidden behind a mask or hood. The mask could be a physical one or an abstract representation of the inner self. This could represent the idea of feeling like someone else or hiding parts of oneself, which can hinder the journey towards self-expression.

Layered Self-Portrait: Create a digital art piece that represents the layers of a person's identity. This could include different shapes, colors, and textures that represent different aspects of the self, such as hobbies, beliefs, fears, or aspirations. By exploring these different layers, one can start to understand and express oneself better.

Emotional Collage: Create a digital collage of images that represent different emotions or experiences that have shaped who you are today. This could include images of people, places, objects, or symbols that have had a significant impact on your life. This collage could represent the beginning of the journey towards self-discovery by acknowledging the different parts of oneself.

Mood Board: Create a digital mood board that represents the mood or feelings that one is currently experiencing. Use images, colors, textures, and typography that reflect the emotions that one is feeling. This could help one to identify and express their emotions better, which is an important part of self-expression.

Inner Landscapes: Create a digital art piece that represents a landscape that reflects one's inner self. This could include mountains, forests, oceans, or deserts that represent different aspects of the self, such as strength, mystery, calmness, or isolation. By exploring these inner landscapes, one can start to understand and express oneself better.

Masked Emotions: Create a digital art piece that shows a person with a mask that represents the emotions that they are currently feeling. This could include masks that show happiness, sadness, anger, or fear. By wearing these masks, one can express their emotions in a safe and creative way.

Personal Logo: Create a digital logo that represents one's personal brand. This could include symbols, colors, and typography that reflect one's values, personality, or aspirations. This logo could be used as a visual representation of oneself, which is an important part of self-expression.

Inkblot Test: Create a digital inkblot test that reflects one's inner self. This could include abstract shapes that resemble different objects or emotions. By interpreting these inkblots, one can gain insight into their own subconscious mind.

Visual Journal: Create a digital journal that includes images, sketches, and collages that reflect one's thoughts, emotions, and experiences. This could be a personal visual diary that helps one to explore and express oneself.

[raven tendon]

I think it was long time than I saw my Automatic1111 being updated when I do "Git pull" how do I look when the last update was to confirm that it goes online to update? I just installed a try of Glasswire and want to confirm it does not block anything.

[shut crow]

Someone told me yesterday was an update

But im not at home to check :/

[craggy hazel]

https://i.imgur.com/sG3eFsL.jpeg https://i.imgur.com/CCiRpoj.jpeg

[rotund dust]

people explaining the new tech that have been coming out for stable diffusion after you've come back from a coma:

[icy vector]

https://www.seratio.com/home tokenizing emotions and metaverse

[glacial turtle]

on the github page near the top it'll say when the last commit was. a1111's was last week. i always check the website and go to the insights page to see what was changed and when

[acoustic ember]

anyone know where i can get hd version of current server image banner ?

[flat pumice]

the server banner is the last winner of the PoW contest. You can find all of those in their original sizes here : #🏆｜winner-gallery

here is the last one specificaly

[acoustic ember]

thanks !

[dense panther]

i just started up my laptop to this 💀

[indigo salmon]

is there an ai in existence that can render good, reliable images of text?
I'd like to make an image of a stone tablet with a specific phrase carved into it.

[zealous radish]

Why not just add the text in post?

[indigo salmon]

i dont know how to make it look authentic, weathered and chiselled

[zealous radish]

reliable text generation is something of a holy grail

[winter lotus]

Pretty sure Parti can, but that's not really helpful since you can't get access to it

[oak trout]

Write text in Word, choose font, screenshot text and use image as base for scribble mode in Controlnet, then use that with a prompt to generate stylized image of the text

then use a photo editing software to add that on top of your image

Sebastian Kamph has a good youtube tutorial on it if you search "Control Text in AI Images"

Then you can run the edited photo through a low denoise value img2img to reduce visible editing artifacts

[minor garnet]

Not sure if this is the right place to ask. For fine-tuned stable diffusion models, is it legal to make any fine-tuned stable diffusion models exclusive? Stable Diffusion has a pretty open license - wonder what are rules around fine-tuned stable diffusion models. Any insights would be appreciated. Thanks!

[zealous radish]

no one said you had to share or who you had to share it with LOL

I'm sure that if you included your friends in your models, they may not want you to share them with everyone

[normal zodiac]

https://medium.com/@thewayteller/the-evermore-journals-chapter-4-2-the-army-of-the-eighth-clade-548905b382c8

[oak trout]

https://huggingface.co/stabilityai/stable-diffusion-2/blob/07753ec23aeaf08862a5f6a8fcb0f9a883863b1b/LICENSE-MODEL#L49

You may add additional copyright statements to your fine-tuned model if you otherwise adhere to the license conditions

You are not required to distribute your fine-tuned model.

[chilly glacier]

This one is so well made 😭
https://www.youtube.com/watch?v=P2sm7A8NLvo

[mild quartz]

I was thinking about the huge backslash AI image generation is recieving and how angry people are getting over it.

Its kinda weird because people get angry if an AI copies an art style of a famous living painter but nobody cares if chatGTP copies the style of a famous living writer or when text to speach copies the manerism and tone of a living person.

[zealous radish]

Best you can do is just not use living artist names in your prompts

[rigid helm]

Eleven Labs

[merry flame]

https://apod.pictu.one/day/2023-03-06

[grand scroll]

A good thought on bias and training! https://medium.com/p/a2f3256cf113

[cedar blade]

is there a way to report an account in this server? saishiva#3678 just messaged for a crypto scam, and this is the only server i have in common, is it bannable?

[shut crow]

Got a messages from an other user too, copy his username and make a #1010934719455707218 ticket

Then you can block him and then select the three dots again to report him to discord

[flat pumice]

yep, there are attacks around, I fell to one not too long ago. Be really careful when you get PM'd and you can open a ticket like CS1o indicated, best way to let the staff know

[shut crow]

This is scam, report him at #1010934719455707218 and delete this picture

[flat pumice]

funny publicity for an insecticide 🙂

[wispy helm]

How much is this a scam on a scale from 1-10? I see this is the only discord I have in common with OP

[flat pumice]

10

you are the 3rd user I see having this exact message today

send a capture of it to the mod team through #1010934719455707218 , and remove the screen capture for people not to fall for it. you can also block them

Hey people around here and other channels, there has been a certain wave of scam by PM recently. If you have been contacted about wining a cryptocurrency prize in a contest, don't hesitate to :
1/ report that person to the mod team through the #1010934719455707218 channel
2/ block them
Also of course, don't follow their instructions further

Lots of new users come around asking for this too, so if you see someone not sure, don't hesitate to direct them to this info please !

[wispy helm]

Ah okay thanks! First scam PM I've received on discord haha

[flat pumice]

I managed to fall for one (less obvious, but still). It's always good to check and stay safe

[agile river]

Good way to tell: if it says "crypto" it's a scam 😛

[candid pike]

Yo

[agile river]

wassup

[candid pike]

Can I ask where is a verify channel?

For a wallet or something?

[agile river]

I don't know what you're talking about lol

I keep my wallet in my pants

[candid pike]

Yeah same someone messaged me about I needed to verify my wallet on this discord

[agile river]

yeeeah I would not do that lmao

[candid pike]

Mad sus

[agile river]

All crypto is sus

[candid pike]

Can I show you in dm?

[agile river]

no

[candid pike]

Alright

Well I don't want to full up this channel with images

[kindred lintel]

It's a scam, please open a ticket in #1010934719455707218 and report the user

[candid pike]

Thanks

I was looking for that

[agile river]

https://tenor.com/view/the-simpsons-snake-wallet-insoector-gif-27444922

oldest trick in the book

[candid pike]

Yeah exactly why it doesn't work

And he had the most lame excuse

[worn plover]

https://www.youtube.com/watch?v=o9triB4EEcc

[sacred sky]

man I love camus

[spark tulip]

Is Stable foundation 100% free or do I have a limited image then I have to subscribe to the paid app

[flat pumice]

stable diffusion is the program that makes the pictures, and it's free. You have certain licences and rights on what you can do with the pictures but that's all

That being said, runing it costs electricity and requires hardware.
So you can either install it at home, and pay that way (electricity, hardware) and have unlimited generations ( #1080946152318443610 )
Or you can use the official website (https://beta.dreamstudio.ai/) to generate pictures for free until a certain point when you'll need to buy tokens to continue making pictures

[agile bridge]

i see people in the vc now ad then. i join, but have never heard anyone talk. do people use this discord for chatting about SD?

[fast oak]

Sometimes

[raven tendon]

Religion: God flood the world to get rid of sin, first thing happen after is that the only person saved started to sin. The god then create a son that is him so he can kill himself to forgive all the sin in the world, but we still have to go to hell for that plan did not work.

The religious folk: "Our god is divine and all-knowing".

[fast oak]

He did not kill himself

Considering what jesus did, it's actually a pretty good deal. If you can put your ego aside, and admit to your sins and ask for forgiveness + simply believe he died for us , your getting a ticket to heaven

[glacial turtle]

God created everything therefore created the circumstances for his own death. While I'm no atheist, the canonical christian belief is bananas

[fading barn]

Hello...is there a dreambot text to image in this account?.. i saw a video there were but i dont see them now

[tidal vigil]

@fading barn

FAQ: How do I generate images?

Currently there is no bot on the server that generates images. However, there are plenty of other ways such as the official https://beta.dreamstudio.ai/ website or running Stable Diffusion locally using your own system resources! Check out #1080946152318443610 for more details! You can also stop by #1025467151206854736 for any issues you experience while using the website or #🤝｜tech-support for any problems you encounter while installing it locally!

[zealous radish]

@tawdry dome Sorry. I don't know how else to do a report

[tawdry dome]

Sorted thank you!

[rough mortar]

Steam wouldn't allow my solitaire game because it had AI generated images on the cards. So I thought I would warn anyone that might be using Stable Diffusion to create art for a game intended for Steam publication. I know they allowed it in the past, but the letter from there legal department said mine triggered a review because it looked like it could be commercially successful and they are now putting a no AI policy in place. Yeah! I paid $100 bucks to get an off-hand compliment and a refusal to publish 3 months worth of full-time work. Doh! So be aware. If anyone is interested in seeing it, I published it on itch.io for free, where it will most likely forever dwell in obscurity. 8^( https://denkigaka.itch.io/ai-solitaire

[fast oak]

Thank you for the info

[rough mortar]

@fast oak Just thought other developers should know.

[limpid mango]

what is this 💀

[glacial turtle]

Steam doesn't have an "no ai generated content" policy and it would be ridiculous if they enacted that. So much of game production uses generated content. They'd have to take the entire GTA trilogy remaster off since AI generation tools were used to create it. I feel like you're just making up this controversy to get your game get more viral exposure.

"I got an email from their legal department" throws up a lot of "i made this up" flags

maybe you did get denied, but its likely due to copyright images in your booster packs, like what is obviously elsa from frozen. Transformative work has to be transformative, and not made to be obviously referencing the original work but slightly different.

[spark tulip]

Thanks mate I will try both of them

[mighty grotto]

Someone fucking help whos in california

The weather is so bad today they issued tons of flash flood alerts and the lightning storm suddenly began just above my house our security system ACTUALY went off when one lightning struck somewhere nearby

[main cedar]

It's the same over here, climate change weather is awful. I just want spring already

[glacial turtle]

Heh. I live in the pacific north west in the salish sea. We have what's called a "micro climate" and we get extreme changes in our climate conditions regularly. Climate change is the norm here so i'm very used to it. But when it happens on a global scale, i'm kind of amused (in a scared shitless kind of way). I get to see how people who live in relatively stable climates react when the weather changes so suddenly haha. We got a saying here "if you don't like the weather right now just wait 15m"

Still, on a global scale it's terrifying but i laugh to cope

[night raft]

gpt4 is insane https://twitter.com/pwang_szn/status/1635702565298896907?s=20

peter! 🥷 (@pwang_szn)

BREAKING:

GPT4 is out!!

• More creative & reliable
• context of 8,192 tokens
• $0.03 per 1k prompt tokens (chatgpt 3.5 is $0.002 per 1K tokens)
• can use images as inputs and generate captions, classifications, and analyses.

▶ Play video

Twitter · Mar 14, 2023, 6:00 PM

[sleek flower]

Hi guys

do you know if there's any good fnaf model?

[flat pumice]

I see this one ? https://civitai.com/models/9513/fnaf-multi-character-lora

[sleek flower]

I checked it and it doesn't look good

imo

[opaque root]

Crazy how much the runwayML devs can’t answer me a simple question on Reddit or their Discord 😞. Stupid “ failed to load video stream”. And no one seems to know why. The support is bad for Runway :(.

[junior crow]

So I made a hiphop track using AI

Came out horrible

I used https://replicate.com/allenhung1025/looptest to generate the drums and https://replicate.com/andreasjansson/cantable-diffuguesion for the MIDI. But couldn't find much to work with, so I just looped up a tiny section and added some vocals.

https://replicate.delivery/pbxt/Iyg0UfgyNBw9YyRbSTumYIhTRN9r6BQpaBZURyG0De1PgHoQA/lucid-sonic-dream.mp4

However, maybe MIDI from https://replicate.com/andreasjansson/music-inpainting-bert instead based on say https://www.youtube.com/watch?v=8_taCHoY24E could yield something nice.

As long as we don't alter the composition to keep things strictly AI. Switching out instruments are fine though. And adding vocals.

[twilit horizon]

Gen1 Selfie!

[junior crow]

That was awesome

Would be so much cooler if you'd started rapping though or at least reciting some poetry with a nice beat in the background

(unlike the beat above which is shite)

[frank flint]

Spoiler since its bright asf, dont want to flashbang lmao

[flat pumice]

That's a nice picture, and a really nice attention ❤️

[frank flint]

O you think? im struggling a bit with the trees thats all 😅 thanks so much though

This one came out nicer, gonna see if i can upscale slightly higher to help with the background

[flat pumice]

It's hard to keep composition good on higher resolution, you're doing quite good here

But why post in #🌶｜off-topic ? This is SD generated, right?

[frank flint]

Ah yes it is sorry, i didnt even realise what chat i am in sorry

[flat pumice]

No problem, there is just very few people in this one. You'll have more feedback in #🏞｜general-with-images for example. But i always love some good sharing like that !

[frank flint]

thanks <3

[tiny oriole]

hi

which channel is for writing prompt for stable diffusion? I am new

[rocky violet]

Playing with the new Bing Chat Bot:

[tidal vigil]

@tiny oriole

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

Currently, there is no bot on the server that generates images. However, there are plenty of other ways such as the official https://beta.dreamstudio.ai/ website or by running Stable Diffusion locally using your own hardware! Check out #1080946152318443610 for more details! You can also stop by #1025467151206854736 for any issues you experience while using DreamStudio or #🤝｜tech-support for any problems you encounter while installing it locally!

[flat pumice]

and welcome around Aurelie !

[split widget]

Yo anyone know of any gore discords or something similar to elblogdelnarco?

If anyone has access to gore please dm me I'm looking for a specific piece which sadly might now be lost media

[fast oak]

No one does gore anymore, not even Unstable, it's too problematic optics wise

Maybe 4chan, but people will frown on you, I don't think there are any actual groups for it

[feral flax]

https://youtu.be/EO6yJVSDICI

[fast oak]

Why don't they use a modern voice synthesizer?

[fast oak]

How's everyone doing today?

[orchid turtle]

https://youtube.com/shorts/LKjYoFaYkv8

https://github.com/RafaRed/alexa-bing-gpt

[split widget]

Any way to do this without all that password and ask rigamarole?

[tiny oriole]

Alright, thank you.

[tiny oriole]

Oh thank you very much!

[orchid turtle]

Did not find a better solution yet, did this in one night, but I did this way because looks like alexa do not accept stream responses, so I we need to return everything in one response, and the secound problem is because Bing take some time to return and alexa can wait just 8 secs

[latent juniper]

Hey I made some Drum and Bass Music that's 90% AI generated, check it out if your intrested in AI Music https://soundcloud.com/machinewitch666

[round vigil]

My humble experiment to create a musicvideo entirely with AI. https://www.youtube.com/watch?v=Cexmsxh6xfI

[haughty grove]

anyone know the best place to find a dataset for voice cloning? Like, if I want Tony Jay's voice (the Elder God in Soul Reaver, aka best voice acting ever). I could scour Youtube, download, convert to .wav, etc. But is there a repository somewhere online that already keeps all this and lets me search by character/actor/etc?

[worn plover]

https://youtu.be/Ge3SOjx63jI

[lapis pawn]

Hello, community!　I've created a cat art database called https://jugemon.com/, a website using Stable diffusion!
I have created a website that can be used as a reference for generating images with txt2img. The site lists over 100 cat artwork by different artists. I hope you'll take a look at it when you want to try a different style or look for some ideas!
Any feedback would be greatly appreciated!

[flat pumice]

Hey and welcome around ! the best place for those usually is in #1045770510153302086 , where you'll get the most exposition on it. seems interesting !

[lapis pawn]

oh Thanx!! I will post it to the #1045770510153302086 !

[glacial turtle]

"however" paragraph is my favorite. not only is it a hilarious clarification, it's impressive that bing can reason that far ahead

[winged pine]

https://www.youtube.com/watch?v=m8VpnzmdqP0

[haughty grove]

Anyone know of a ChatGPT local app that would let me give it my API URL + API key, and give me a local UI to interact with the API? Ideally save all past prompts + answers locally. I don't want to use curl.

[fleet flume]

Hi, just wanted to ask what is the source for the server banner picture? I really love it and would like to set it as a wallpaper

[prisma hollow]

It was our last POW winner submitted by the wonderful @ebon obsidian here: #1023999442338201721 message !

[fleet flume]

thank You very much! It looks wonderful 😇

[prisma hollow]

Right? It's an amazing piece!

[stone pasture]

I decided to do a drawing of one of @unique yokea ai art pieces and here is the results

The original ai art peace

[chilly bison]

I just came across this video today. https://www.youtube.com/watch?v=nEHCBPGo-5M

[empty mango]

@loud coral yeah apparently GPTQ is really computationally expensive

[loud coral]

I see.

[empty mango]

But once you have the weights you can run inference in about 40 GB of vram

[loud coral]

I have 16GB.... So maybe in several years from now I will upgrade to like a RTX 7080 with 48GB etc. lol

[trim ice]

I may gamble and get another 16 GB stick to get to 48 GB and hope the BIOS doesn't die, as it seems the 65B model uses 40 GB

[empty mango]

48 is definitely enough

[trim ice]

I gambled. Next week it should arrive, and work 🤞

[empty mango]

[neon thorn]

guys

In which version is stable diffusion?

[night raft]

They finally did it... chatgpt inside the code editor 🤯 https://twitter.com/pwang_szn/status/1638561248668274690?s=20

peter! 🥷 (@pwang_szn)

ITS FINALLY HERE 🤯

• "ChatGPT-like experience in the code editor" powered by GPT-4
• GitHub Copilot for Docs - chat for any company's repos and docs
• GitHub Copilot CLI - writes your pull requests

▶ Play video

Twitter · Mar 22, 2023, 3:20 PM

[glacial turtle]

As someone who was in grade 3 during the original green ranger saga, HOLY MOLY LETS GO!!!! https://www.youtube.com/watch?v=iHgPltur5J4

[twilit horizon]

the new Adobe Firefly Text ai app

[glacial turtle]

can do the same in SD but with less UX. you have to instead create an image with the text you want, then put it through controlnet. ez pz

less UX but a lot more flexible

[wraith owl]

Premium quality face swap for videos and images

[rocky violet]

Gordon Morre has passed away: https://www.intc.com/news-events/press-releases/detail/1611/gordon-moore-intel-co-founder-dies-at-94

[inner nacelle]

Firefly is so genuinely unimpressive, and it sucks to see people flocking to it for all the wrong reasons

stable diffusion still SHITS on firefly, but because Adobe (probably the most toxic and problematic company in the creative industry) used their own images, now they are the savior (as long as they can hike their prices up at least)

[silver wing]

Remember, it's always morally right to pirate Adobe software

[glacial turtle]

too soon for moores law is dead jokes? rip mr moore. he's done a lot of good work over the years. i mostly only know him for his coined law which has served us all very well

[glacial turtle]

i disagree. if you're profiting significantly from it, morally it's right to pay dues

though i really wish i could buy a permanent license again

[glacial turtle]

https://www.akool.com/ i looked this company up, and wow. scams. scams scams scams. They're likely just rebranding the open sourced face swap app. I judged em as soon as i set foot on this page and the big sell they have is "AGI Cloud for Commerce"

sparks of shenanigans are in the air

[glacial turtle]

In lighter news, i'm still so excited for this revelation on it's way.

[real sonnet]

Waiting for GIMP feature that replicate firefly

[glacial turtle]

theres a gimp extension plugin now.

bu

but the ux is bad. that's where firefly shines. better UX design. adobe has always mannaged to do that

[real sonnet]

yeah, I have to agree. If Adobe was cheaper I'd totally be a fan

but it just cost too much 😦

[glacial turtle]

FOSS culture is hostile to UX. it's interetsting too because the gnome project seemed to really start carring about UX, to the point that they got angry at people for modifying designs. Against the core tenants of FOSS. There are a few incompatibilities there.

I fear that adobe will always win the UX battles and foss developmet circles aren't even trying anymore

[real sonnet]

I'm just glad the blender team decided to get UX improved

I didn't know they had a gizmo until a few days ago

[glacial turtle]

https://stopthemingmy.app/ this letter really did ot land well at all and they did a ton of backtracking, but i'll probably never use the gnome project again due to this internal culture of "ugh, we're designers!"

blender is good stuff. Theyv'e always moved towards better UX. they're more of a commercial product that has a free license, than they are FOSS though

wait, they are foss

i'm thinking of unity

[real sonnet]

Yeah, Unity is commercial, Blender was just terrible UX until recent 3.0

I've always wanted to go Blender (since 3ds and maya is expensive), but the terrible UX of pre-3.0 is just too much

glad they made the change now and I'm a happy blender user

Now if only GIMP does the same...

[glacial turtle]

https://www.youtube.com/watch?v=4IuIKe1sEFY

https://github.com/ArtBIT/stable-gimpfusion

[glacial turtle]

once you grind through the ux unintuitiveness ad learn it all, it becomes muscle memory really

just takes simple exercises to drill the hotkeys

[real sonnet]

but it takes effort and persistence, which is hard to come by >.<

[glacial turtle]

tell me about it

[glacial turtle]

Fear is the mind killer

I must not fear. Fear is the mind-killer. Fear is the little death that brings total obliteration. I will face my fear. I will permit it to pass over me and through me. And when it has gone past I will turn the inner eye to see its path. Where the fear has gone there will be nothing. Only I will remain.

more related, i haven't used many of those but i seen them around and some look neat. maybe there's a foss one but what i've seen are usually some kind of online software

Adobe premier's AI tools just launched too, and i'm always "if you're going to do SaaS might as well do Adobe"

oh there's this too. Vimeo bought this big company. i've always thought vimeo was a possible contender to youtube https://vimeo.com/video-editor

good to know! i also checked if stability themselves had anything, clipdrop, nothing i saw

[distant granite]

Idk how but I asked bing and it gave me this table which it picked from the huggingface sdxl repo

Which apparently doesn't exist

2.3B parameters for SDXL and 768 output

Very impressive

[flat pumice]

I'm wondering where it sources those infos damn

I can't find any online

[slow pilot]

this is the heavy hallucinating purple version, probably just pulling new info out of its robotic ass

[silver wing]

There's one thing I dislike about channels that do deep dive into dramas: Their whole fanbase contains a bunch of edgy kids who can't even handle them, and just go and harass people

[cloud spade]

?what?

[glacial turtle]

mayyyybe, just speculating, hugging face lets metacrawlers in to index stuff they want to have searchable the day it releases. some sort of robots.txt shenanigans? like the repo could exist but is set to hidden from public view

[glacial turtle]

https://youtu.be/yGXaAWbzl5A huge channels on youtube were hacked with session token hijacks recently. big attacks oging on and this stuff is getting popular again for some reason. Avoid thigns like pdf files or office documents from unvetted sources as these are becoming common place again.

Stay frosty friends

[fast oak]

Y'all think a red kangaroo could kill a gorilla?

[glacial turtle]

depends on the terrain i think

i think the best of the mob vs the best of the troop, in some level arena, would be a tough call. in the congo it'd go to the gorilla an in the outback it'd go to the roo

[fast oak]

true

[glacial turtle]

they could both rip your arms off

[twilit horizon]

1000%

[glacial turtle]

they flock to it because it's adobe and they have a reputation of progressing the magic in their tools. it appeals to a lot of people and that's okay.

one day they may hopefully drop some big github repo that we can use even

lots of big corps have been doing that and i think that's stability's strategy. force the whole playing board to push open source

[fast oak]

It's been happening for like an year now

https://youtu.be/LkmIItTrQP4?t=250

[glacial turtle]

Valve's server side VACnet that they use on Counterstrike is very proven. I still think that driver level anti cheat is important to fair gaming as well

VAC was into deep learned anti cheat many years ago. I'll pull up their GDC talk about it.

https://www.youtube.com/watch?v=kTiP0zKF9bc robocalypse now is an easy video title to remember

talk was in 2018 and they had been using it for a couple years live already at that point

[fast oak]

clearly doesnt work very well

[fast oak]

but it's still terrible

[glacial turtle]

debateable. cheaters get caught after the fact under vac. its' not fully effective but thats hard to achieve against tenacious cheaters. right now its the only thing in place to catch the machine learned cheats

blizzard probably actually has a similar situation too

but are not as open about it as valve

then there's valorant, which have just leaned hard into client side. i don't think that's a bad thing but they flubbed it a few ways

[fast oak]

yea...often months later and even then, not all of them.
I won't call it effective.

I think cheaters in competitive games shouldn't be able to play that long and people shouldn't review it or throw reports in to ban people, it should be automatic process

[glacial turtle]

that requires driver side anticheat a lot of the time but for sime reason gamers are hostile to that (coughpiraceycough)

[fast oak]

I hope it won't reach p2w games tho, p2w games deserves to be abused or cheated on by community

[fast oak]

https://youtu.be/nYdS3FIu3rI

There are other people with explanation too...

[glacial turtle]

oh i know the stuff. its nothing new to me. session hijacking been going on since myspace and earlier. as long as theres been cookies really. that's why its weird to see it resurging, since it's so old hack

[fast oak]

This guy kinda explains why it works so well

[glacial turtle]

theres also the false positive problem. too many false positives and your gamers hate the anticheat thats there too save them. stupid filthy gamerses

i might get around to watching it but i used to harden php apps for a living so i know my way around web security a bit already. owasp was my homepage for years.

[fast oak]

I get how they could open "pdf" file...which wasn't pdf...I don't understand another thing, why they are logged in if they care about security

I mean...I don't and I don't care abot it that much...my whole browser history, settings,sessions are reset each time I close my browser

Even passwords saved only for sites I don't care about , they can have it lol

[glacial turtle]

constant balance of security and convenience. as a media company, they want to move fast, but also stay secure. they count on the infrastructure. crummy and lessons learned. buttt... we got here to a point of complacency because this attack type should be so prevantable by the host like.. something as huge as YOUTUBE

where's alphago's machine learning on that one?

[fast oak]

yea idk...

It's been some time since "hackers" started abusing it on youtube specifically and it's the first time I actually seen responce from youtube on it

[fast oak]

https://youtu.be/Lobod7UOqgs

2 years ago

spotify released something too, but didn't see results \ didn't try it

https://youtu.be/idTURgwcPAk?t=135

[wise sedge]

https://www.reddit.com/r/StableDiffusion/comments/1244h2c/will_smith_eating_spaghetti/

[glacial turtle]

Multimedia entertainment!

[drowsy wadi]

Hello everyone.
I'm looking for a community or a place where I can meet people willing to start projects as a team. One example is lablab.ai, but besides that, the closest I could have found are freelancing websites. I'm not looking for someone to pay me, but someone I can start a project with and maybe found a startup.
Do you guys know any place on the internet like that ?

[fast oak]

free labour

[fringe compass]

https://twitter.com/ClownVamp/status/1640826879900000256?s=20

ClownVamp (@ClownVamp)

AI Art History Panel 🚨

Co-hosting a space tmrw with @bl_artcult and a wonderful collection of curators, technology builders, and AI artists!

Speakers: @ganbrood @keyvass @DehiscenceArt @TormanJeremy @pharmapsychotic @KaliYuga_ai @ManekiNekoAIArt
https://t.co/dOmbexMblh

Twitter · Mar 28, 2023, 9:23 PM

should be a fun chat for anyone around

[raven tendon]

Bing AI could make script for Blender and some games, but it failed to create SVG code for a Donkey Kong character. Is it so that AI do not understand that SVG a text based format and that a task like that would just be some few lines?

[glacial turtle]

it was likely never trained on svg formats for that task

chat gpt can do ascii art, but basically just a cat

[raven tendon]

We can hope that AI will be able to learn, like we can say "Go to this page and read about SVG in HTML and fileformat".

[glacial turtle]

they can learn, it just microsoft and openai do that part

also it's good at one shot learning too

when it's similar to something it's trained with

[raven tendon]

Two of the worst parts ever :/
Well, they did a good job, sad it is so closed for now.

I am still happy I have and can play with the Bing AI.

[glacial turtle]

there are open models you can use. llama and alpaca are two

people WILL purpose train those

[raven tendon]

Ohhhh. Now I listen.

[glacial turtle]

here's some not half bad science journalism about it https://interestingengineering.com/innovation/stanford-researchers-clone-chatgpt-ai

[raven tendon]

I ran to Hugging Face to look, did I run the right direction? Ty, shall have a look. And thank for great topic to explore.

[glacial turtle]

i think they released the code and data to train it, then you can get a trained model from "somewhere" ooOoooOoOo

https://github.com/tatsu-lab/stanford_alpaca is the git

[raven tendon]

Ty. This will be intresting.... and just now it hit me I could just ask Bing, but I think I get better and more social answers here.

[glacial turtle]

its out of my depth, but thebiggest model of their bunch is 40gb so it can run in many system memory banks . theres a 20gb one too

i'm not sure about how to go about starting one up

i saw .cpp and instantly put it in the "do it later" category

[raven tendon]

I am almost blind! I been sitting in dark mode for like 4h now and then went to https://crfm.stanford.edu/2023/03/13/alpaca.html that is made in sun bright white.

[raw dune]

https://www.securityweek.com/chatgpt-data-breach-confirmed-as-security-firm-warns-of-vulnerable-component-exploitation/

not sure if anyone here cares but i wanted to post to spread the news

might be important to some

[alpine siren]

If you like the internet in the USA please contact your reps
https://www.youtube.com/watch?v=xudlYSLFls8&t=1986s

[fast oak]

HAHHAHAHAHHAHAHHAHAHAHAHHAHAHA

[fast oak]

america is a shithole

i'm just gonna leave as soon as possible

i'm happy openai got shat on though

[jagged path]

man
is there a reason we aren't all creating beautiful pics of realistic people to post on onlyfans and getting rich in doing so

[oak oar]

Hey SD lovers,
Intrigued by the image-to-text generation feature showcased at the GPT-4 demonstration? Although it's not widely available yet, you can already benefit from this advanced capability in your daily tasks and projects with PromptPerfect!

[zealous rock]

Look what this guy is doing https://youtube.com/shorts/6HwgbzaXD0k

[daring coral]

Hey chat. Hosting another aiartweekly cover art challenge. This time revolving around "pope art" 😎. Would love to have you all https://twitter.com/dreamingtulpa/status/1641794349078904835

Dreaming Tulpa 🥓👑 (@dreamingtulpa)

Dear #aiartcommunity 🧡 I'm looking for submissions for the next #aiartweekly cover! Winner gets $50 and all finalists will be mentioned in issue #28 🙌

Theme is "pope art" ✝

2 images max. 10:14 vertical aspect ratio. Unminted, check below! RT required

Submit below! 👇

Twitter · Mar 31, 2023, 1:27 PM

*deadline on the image is wrong btw. You can submit until next Thursday 😉

[pine ruin]

GPT-4 can do that (with the plugins)

[frigid zenith]

What do ya'll think about the AI moratorium convos?

Went deep into what's keeping ppl awake at night and it's pretty wild

Answer ain't super obvious IMO

https://twitter.com/bilawalsidhu/status/1642290720596971520?s=20

Bilawal Sidhu (@bilawalsidhu)

What a week for AI! Not yet scary, but a feeling is in the air. Things are heating up and people are conflicted.

Why are the brightest minds in AI asking for a 6 month pause, while others say it doesn't go far enough? 🤯

Here's why this debate deserves our attention.

🧵 Thread

Twitter · Apr 1, 2023, 10:19 PM

[jagged path]

Why are the brightest minds in AI asking for a 6 month pause,
wasnt the whole letter a sham filled with fake signatures lmao

[exotic pebble]

Elon was too busy with twitter and missed the AI revolution boat and now realises he has no chance of catching up

[cerulean dagger]

Any recommendations for an AI tool that can help expand an image and also get rid of pixelation (or just be able to increase its resolution basically)?

[glacial turtle]

most signatures weren't fake last i checked. they took it down immediately upon finding out some fakes were on there

[glacial turtle]

tesla self driving and spacex landing and neuralink are all all in on ai tho

he left openai tho hehe

probably does have some billionaire resentment there

[exotic pebble]

tesla and spacex are very specialized and designed for portable use, not the massive GPT that is being built now. Neuralink Im not sure what purpose that will serve.

[glacial turtle]

tesla self driving is a pretty impressive vision model no matter how much you're an anti fan of elon

[exotic pebble]

oh for sure it does a pretty good job most of the time

[glacial turtle]

neuralink is a higher bandwidth input into a machine than muscles and appendeges can provide. i think one of the initial goals is to help restore motion in paraplegics

well yeh an chatgpt writes buggy scripts sometimes. still really impressive and cutting edge though

people keep painting cgpt as general and it really isn't. it has huuuuuge limits. it's all up to the data set and how they trained it and what the model is doing with different trained situations. very good stuff at what it does, but it's nowhere near general

[exotic pebble]

natural language and communication is where the big money is. but GPT is a lot better at image recognition now, I would say better than what tesla has

[glacial turtle]

citation needed

[exotic pebble]

imagine twitter with its own gpt. Elon would be spewing he missed that

https://ai.googleblog.com/2023/03/scaling-vision-transformers-to-22.html?m=1

[glacial turtle]

maybe. but like, there's this meme where its like he's completely outed from ai completley, and it's just weird. landing rockets is pretty intense and he used ML to get there a few years back. he's been around it quite a bit and its suceeded his companies well. who am i to reason with the circle jerk though. carry on

[exotic pebble]

dont get me wrong Im not shitting on him. Just that its a bit silly for him to call an all stop on AI

[glacial turtle]

the letter recently?

[exotic pebble]

yeah

[glacial turtle]

oh okay. yeah you're meming hard

it was 6 months and another consortium wrote it

probably will happen. all the people who can train models are on board

i really odn't understand where people source their info sometimes

[topaz bobcat]

https://www.youtube.com/watch?v=ktwxz5t1fO8

[buoyant zephyr]

Hey, what are the rules for conducting a survey here?

[flat pumice]

hello there
Things shared around need to stay on the subject of the discord, so AI, and not be spammed.
The best place to share those is in #1045770510153302086 currently
If you feel it may be inappropriate, you can also contact us first in #1010934719455707218 to ask more precisely

[rare cloud]

Hello

[pure estuary]

[glacial turtle]

Weird. I just got a pile of DM's in my box today that all were sent near the end of march. This is why things like discord shouldn't be centralized so hard. They cut corners and it acts stupid and there's no agency for the user there. I miss the days where the community hosted chatrooms. IRC has existed since the beginning.

[rich stratus]

Found on an anime site. I'm not familiar with this rating system.

[fast oak]

Hi guys, sorry if this is off topic, but maybe someone knows Ai tools to make a 3D model of the body and face from a photo?

[glacial turtle]

Not a lot of ai tools wil automatically do that. What you're trying to achieve still takes a level of artistic skill.

Most 3d conversion systems in AI create "NeRF"s Neural Radiance Fields. These act different from meshes and require further processing by a skilled technician to turn them into 3d meshes. With one photo, you'll likely get a very poor NeRF though.

https://monstermash.zone/ this here is a really gimmicky tool, but it produces some results once in a whille that are very workable.

[sage fog]

https://youtu.be/Am1sDAQ6p14

[winged pine]

https://youtu.be/m8VpnzmdqP0

[jagged path]

https://www.youtube.com/watch?v=-9Ado8D3A-w
is there a popular audio/music ai focused model? i think stable diffusion for art, llama/oobabooga for text but out of the loop for music. not a kanye fanboy yet gotta admit this is actually crazy good

[jagged path]

i figure you could apply the same concept by using LORAs for voices the same we do with art except for different voices

[fast oak]

can I share images here?

and which kind? lol i want to share some funny ones

[flat pumice]

all good my bad

[glacial turtle]

Heh on google, when you google mario bros movie, it gives you a little question brick that goes bading when you click it

[flat pumice]

lol nice

I didn't know of that easter egg

they have lots

[flat pumice]

I love this one

"recursion"

did you mean "recursion" ?

[glacial turtle]

hahaha zerg rush is another gooder

i think the mario one is new, for the movie

aw zerg rush doesn't work anymore

[flat pumice]

Why did the Flash series on Netflix go so much to shit ??

I loved this series

I even got a VPN to watch it early on american Netflix in the early seasons (we have 1 year difference on those in France)

but I can't even finish that 7th season because it is SOOOOO BAD !

like wtf scenarists ?

it's really bad what you wrote here, and you know it

[digital pebble]

I have a quick question for anyone running the latest WebUI.. how's the experience, I was riddled with just constant nonsense, and I'm puzzled how anyone is able to use it without functionality loss, or just general bugs popping up

[south thicket]

Latest runs fine for me. I strongly recommend a fresh install of the latest UI if you're having issues, as updating older versions without fresh installs seems to be the primary cause of folks' issues.

[glacial turtle]

i loved it too until they brought in the daughter. it was already campy but it got waaay campy then. Filmed in Vancouver. Great show imo. DC i think was confused about where they wanted to take the Flash IP and didn't put enough funding into it in the later seasons

[flat pumice]

seen on another server

[flat pumice]

I had a hard time with Nora but managed to like it in the end.
I liked the "multi Wells" concept too in the season 6 I think.
But right now, season 7 just feel like a joke

already before, they lost all feel of potential impact. they reverse any big step they initiate

I'm happy Arrow stoped too, so there isn't those bad cross over episode anymore, never liked those

[fast oak]

Can anyone send me the image from the new event ???

This one

The top image

Nvm

[flat pumice]

you found it ?

[fast oak]

Yes

[flat pumice]

👍

[fast oak]

🙂

hello

[glacial turtle]

they go a little overboard with the multiple wells. okay maybe a LOT overboard. But i respect that actors commitment to gaudy stereotype cliches. He slips into those characters so... well

[fast jasper]

do people always train character with 1.5? is there anything that people would use for non-realistic characters?

[glacial turtle]

people can train any model they can get their hands on afaik

[fast jasper]

UserWarning: The torchvision.transforms.functional_tensor module is deprecated in 0.15 and will be removed in 0.17. Please don't rely on it.

should i do anything about this

[flat pumice]

#🤝｜tech-support

[opaque dawn]

This scares me lol: https://youtu.be/JDsjBO9G7uc

[wanton hollow]

I am going to have to stop smoking so much weed if I am going to understand this

[night raft]

In the spirit of "These X do not Exist"...

Made a small little project that has people with Fake Abs.. 😄

https://theseabsdontexist.vercel.app/

[exotic pebble]

According to Business Insider, Twitter has made a substantial investment into hardware upgrades at its North American datacenter operation. The company has purchased somewhere in the region of 10,000 GPUs - destined for the social media giant's two remaining datacenter locations. Insider sources claim that Elon Musk has committed to a large language model (LLM) project, in an effort to rival OpenAI's ChatGPT system. The GPUs will not provide much computational value in the current/normal day-to-day tasks at Twitter - the source reckons that the extra processing power will be utilized for deep learning purposes.

lol

[fast oak]

I aived a post about a word i created with the help of AI 😁 maybe it would catch! Who knows ♥️

AIVi (pronounced "eye-vee"), which stands for Artificial Intelligence Visualization.

AIVi highlights the role of AI in empowering designers and creators to bring their ideas to life with ease and speed. It's a simple, catchy term that can be easily integrated into our everyday conversations about design and technology.

https://www.linkedin.com/posts/maciej-popławski-44206a191_aivi-aivi-aiving-activity-7052142575222145024-jB8T

Who tf would guess i would be spamming discords to promote a word... My English teacher would be so proud 🥲

[flat pumice]

https://www.youtube.com/watch?v=dYPbfLnwBU4

[opaque dawn]

https://youtu.be/tiK20Y20HZY

[versed crag]

If there's a better place to put this let me know! (I'm new 👋 )

https://www.youtube.com/watch?v=I9bijudw3-I&t=0s

Images for this test were actually Midjourney's Niji 5 cause it was new, but I took those into SD for depth maps and the inpainted meshes, then brought THOSE into Unreal Engine to add VFX and setup the shots.

[raven tendon]

AI is so far away from where I want it to be and it look like companies like Open-AI is steering it in the wrong direction, just now is local chat for sexual roleplay and the online version are more like advance search that can give an answer in a organized way humans understand.

But I want so much more, yes I want to be able to chat with it, but that it remember and self get a personality it calculated will fit me. I want it to work as a home automation that learn that I should water my plats once a week and ask me if they look okay. I want to have a web cam it can use to segment out things in my house and suggest where to put stuff and when I should clean, maybe give decoration tips, and remember where my darn remote may be.

I do not always want to start a conversation, I want my AI to do that to if it think it cam be good, like a Nany for adults, a reserve mom or a house ghost that reminds me that I played a game a bit too long and that I should move a bit and that make the bed or vacuum a carpet is also to move..

How long before we are there or will AI be banned before that happen for folk just use it for sex chat rather than working?

[gentle hazel]

Am I allowed to post my video link here i made yesterday for an easy guide to get stable diffusion up and running on windows.

[pallid portal]

The best place to post a guide is in #1045770510153302086

[gusty mulch]

https://www.reddit.com/r/StableDiffusion/comments/12nev7l/training_textual_inversion_using_6gb_vram/

[winged pine]

https://www.youtube.com/watch?v=58whuDsKBXI

[remote fiber]

how to make aserver

[flat pumice]

https://music.youtube.com/watch?v=0m9QUoW5KnY

[exotic pebble]

scrubbed

yep stuck valve

[fast oak]

Its already called off , that stream is fake

.

[tidal basin]

Can I ask something about GPT 4.0?
Welp open-ai doesn't support my country. So I intend on using VPN to use GPT 4.0. If I used VPN and subscribed to their monthly plan, would it be okay? Will I lose money by doing so?

[opaque dawn]

For folks who are interested: https://youtu.be/mTzYDrO3IsI

[fast oak]

peter gabriel huh? i love his stuff

[prisma notch]

how is anyone supposed to out-weird https://www.youtube.com/watch?v=Qt87bLX7m_o

it's a tall order

[atomic tree]

you underestimate the internet

[tawdry canyon]

guys I'm so grumpy and tired I read #1072234218706444309 as #stable-pain

[tawdry canyon]

have you heard of jojo?

[opaque dawn]

https://youtu.be/AnQSZ2nXUwk

[fringe compass]

Yooooo. Does anyone have an extra bluesky invite by chance 👀

[fast oak]

Top 10 reasons i should be a mod on this server

• I am biased

• I spend almost no time on this server

• I am gaming

• I don't know any of the rules

• I ban everyone who offends me

• I am pore

• I don't listen to anybody

• I have nitro

• The pain never stops

• I give great ideas

[junior crow]

Cool!

Did something similar with Doja Cat a while back:

[fast oak]

LOL moo

Love it

[south thicket]

sorry but this belongs in #1087494110358798547

[flat pumice]

I made 2 characters for oobabooga and the vicuna model (13b in this test)
it makes prompts, without the art style and the suffixes, just the subjects themselves
and the params themselves. nothing special, except maybe the "###" as Custom stopping strings
and thanks Trexdel for the cat image 🙂

[fast oak]

💀 Noticed I forget the S in his name. Oops, fixed it

[flat pumice]

Another one I did for the Perpertual model https://huggingface.co/ItsPerpetual/perpetual-diffusion
same settings

[rocky wind]

Hi.. opening a discussion on something here..

Does anyone here have any thoughts on how an AI model for generating Model Railway layout might be developed?

I am asking here because it is tangential to the main topics in other conversations.

[junior crow]

Hehe dope songs indeed

Does bring back many fond memories 🙂

[spark lance]

anyone running SD on paperspaces?

[fierce widget]

Does anyone have any solution for this error when running finetuning in colab? Generation works just fine but when I try to tuneI just get this error, both RAM and VRAM utilisation is nowhere near their peaks so it doesn't explain why it would time out.

Refreshing the page causes python to stop running the script. I'm not even using gradio dns/tunnel for this but rather ngrok but it still doesn't work.

From what I've heard it won't happen if I was to just run the train code without gradio ui on top.

I'm using this btw https://github.com/lxe/simple-llm-finetuner and apparently others had that issue too https://github.com/lxe/simple-llm-finetuner/issues/42 but I thought I'd ask here in case anyone could have any idea

[uncut spindle]

@grizzled glacier so i did buy XP Pen Artist Pro 13.3 first, after a while i got the Huion Kamvas 24 Plus and 6 months ago i bought the new iPad Pro

tha last one not exclusively for art but especially for art