💬│general

[analog nebula]

But simply put it yes 🙂

[vale inlet]

Thanks a lot Ayu.

[analog nebula]

A pleasure!

[vale inlet]

The prompt generator is such a nice addition

[analog nebula]

Yeah pretty clever to integrate such a feature

We plan to integrate various solutions in the future, like 3D and audio as well

instead of having disparate services you can get all the things done in a single platform

that is our goal

[vale inlet]

When you say 3d, do you plan to offer the function to download the 3d model after generation?

[analog nebula]

ideally yes

[vale inlet]

Very interesting

[analog nebula]

the tech isn't quite there yet, but it's one of our goals

[young surge]

I'm so excited to try this. Can't wait!! Love the community.

[analog nebula]

Thank you on behalf of the team ❤️

[vale inlet]

Welcome @young surge =]

[young surge]

If you ever need a Graphic/Web Designer, I will be joyfull to contribute. (Maybe my english isn't the best because I'm from Argentina, I speak spanish)

[analog nebula]

Thank you~! @distant tide , for your consideration if we do require another 🙂

[young surge]

I am currently working full time in a Studio (responsible for MrBeast website, just to throw some flowers hahaha), but if I can help with anything I will for sure be available.

I'm still waiting for the invitation, but I can see my craft being improved by this. I've already designed some shirts with AI artwork, but Leonardo's output is another level.

Sorry for the spam, if you wish to talk privately I will be here. Thanks again!

[analog nebula]

No worries on that 🙂

@young surge by the way do try #✨│priority-early-access if you haven't, else kindly ignore this message ^^

[young surge]

Yes! I've done all the required steps haha, I just need to wait.

[analog nebula]

There's a lot happening so I might have lost track if you did it

[analog nebula]

Ah great 🙂

Should be sent out by one of the other team members eventually!

[young surge]

Thank you!

[analog nebula]

A pleasure 🙂

[fluid grotto]

Is the tool for creating 2D character model sheets available yet?

[analog nebula]

not yet I believe but you could try training an model with some examples

[fluid grotto]

That is useful. Thanks

[grizzled raft]

Hello, just a follow up on the "deletion" suggestion! It would be nice to have an option to toggle if we want to save each picture by default or not. Like if i'm playing around maybe just keep track of the last 5 generations or so, so I don't clutter the generation history, if then i find a good one i can "pin" it so it is saved until i delete it.

[analog nebula]

@marble moat , just an fyi, apologies for the ping 🙂

[shadow lantern]

Does anyone have a guide on how to create sprite sheet for my game?

[analog nebula]

Not me personally no, and we haven't done up an official guide yet either 😦

We are working on an solution for it though

[shadow lantern]

Thank you

[wide lark]

question for everybody: what's your favorite game object/ character?

I've just generated a series of isometric houses in The Sims style and it brought me good memories (posted inside #🖼│image-share channel)

[analog nebula]

Won't say game object/character but I have a fondness for Arkane studio's artstyle, though its such a signature artstyle that's quite obviously their own

[wide lark]

what games did they make?

[analog nebula]

deathloop and the dishonored series

[wide lark]

wow, that's some creative art

[analog nebula]

the thing I liked about their games is it felt like exploring paintings in a way

[wide lark]

never played them though

[analog nebula]

install it now

[wide lark]

yes, looks so

[bold crescent]

hello, leonardo.ai was pitched to me as generating 3D models by a friend, but afaik this was a misinterpretation, the service will create 2D images only?

[analog nebula]

currently only 2D images yes, but we do aim to implement solutions for 3D in the future 🙂

[bold crescent]

ah ok cool, would be curious to see any research / tests / work in progress on that front, fascinating stuff.

[analog nebula]

Nothing significant shared publicly yet, aside from this tweet: https://twitter.com/LeonardoAi_/status/1598690332669272065

[bold crescent]

ooh that does look promising!

[analog nebula]

yeah

exciting stuff

[hard galleon]

@analog nebula fun fact: i'm the person who exposed the true plot of deathloop

all because someone sent me the narrative writer's slide from her resume

and it was super small but i could squint and recognize the text

tldr: it's part of dishonored universe

(im futa-snake)

https://bethesda.net/en/article/1P5zGScDTYTqfmoQya8Hlp/deathloop-is-set-in-the-dishonored-world

i just like mysteries 🤷‍♀️

the narrative writer had no idea it was her fault it was leaked and she got upset with me until i linked the source of the file

which was her own blog

i could just squint and read blurry pixels and knew enough "CONTEXT" of the story to fill in the blank

As she had made it SUPER tiny

[analog nebula]

Yeah I suppose people need to be more careful about what they share

[hard galleon]

https://www.svg.com/719013/is-deathloop-set-in-the-same-universe-as-dishonored/

This supposed Arkane presentation states (via Redditor Futa-Snake)

😅

People of all sorts are always watching!

https://imgur.com/a/oPiPBs0 Here's what I deciphered

v smol

Now I could just use AI 😄

But I didn't have gigapixel or similar then

https://www.pcgamer.com/the-latest-theories-from-gta-5s-ufo-hunters/ problem with using so many usernames

its hard to share your clout

https://twitter.com/LOWROAR/status/851837172453748736

Ai is quite powerful 😂

I always consider it the "EQUALIZER"

A zero can be one hundred.

[austere lion]

So to understand, when Leonardo AI is able to generate 3D models, will that give 3D game artists the ability to generate models they need for games and pop them in? Because WOW, that would be a significant time saver on that front

[hard galleon]

@austere lion I would think a modeller would even be able to train it to create similar models to their own style based on their portfolio.

[austere lion]

That is insane, this tool would be used across the whole process from storyboarding to actual modelling

[hard galleon]

Saving a ton of work on the "CREATION" and moving it "EDITING"

[austere lion]

Exactly!

[hard galleon]

Personally I like hacking up code, and I can ask ChatGPT to give me the stuff I could care less to understand

Then edit the nitty gritties.

Anyways my Create A Character Model is now life. 😄

Fun times ahead of us!

[hard galleon]

Have any of you met Lia27?

https://twitter.com/_Lia27

https://twitter.com/_Lia27/status/1601224507330822147

https://twitter.com/_Lia27/status/1614984762980016128

@analog nebula https://twitter.com/_Lia27/status/1609948133903028224

Lia's pretty fun

https://lia27.ai/

ChatGPT x Leonardo

@deep shoal Might give you some ideas. 💡

If Leonardo had a voice too, man, that would surely speak loudly. 🎭

Wouldn't need to go anywhere else!

The idea behind lia is more like: https://www.youtube.com/watch?v=m9QU69958JU

It's cute that she has a phone number too.

Lia in your pocket.

[analog nebula]

Possibly in the future, if there are viable options to optimize mesh outputs

But then again, 3D game artists could just generate and then optimize the outputs themselves

But yeah there's definitely an optimization step between just generating and popping them into a game. Question remains if it can be automated or has to be manually done

[austere lion]

This has lots of potential, incredible...

[gray burrow]

Hello

Im looking for someone with art skills

Something like this

I really want some arts like this for my daily posting on reddit

[hard galleon]

I made a manga bot @gray burrow

You could probably generate something similar if you use the right prompts

[gray burrow]

Really ?

[hard galleon]

Yep!

[gray burrow]

Nice

[soft owl]

This Ai is amazing really love the help generate prompt feature

[hard galleon]

It's really good at giving new ideas

Even if you don't use the prompt it gives directly

I always like "Circular PFP" at the end, creates art people might use

I don't mind if someone "Steals" it

[small flame]

I am glad there is "isometric sci-fi" model, I was looking for some AI that would not have problem with sci-fi buildings

[gray burrow]

Another doom fan ?

[small flame]

DOOM fan here, I want some art for my hobby project, programning ogame-like clone but with my ideas.

[steep sierra]

should i be getting concerned? ??

nvm let me delete it finally

[obtuse forum]

wohoo got my email

[steep isle]

I've been using midjourney and stable diffusion -- how is Leonardi.ai looking in comparison?

[hard galleon]

@steep isle It's the next step imo

If midjourney is a bunch of seperate "FORUMS" generating stuff, Leonardo is the ISP

Tons of collaboration potential, tons of power behind the image generation

[rancid pier]

so i just got access, i thought i could generate 3d models, not images?

[fast tiger]

For now it is not available, but we are working on it, it is one of our objectives. 🙂

[vocal osprey]

can leonardo refrence other media such as if i said generate a charecter that would fit into the terraria world could it do that?

[hard galleon]

Yep!

It would take some training but

This is an example of stuff similar to the "SIMS"

@marble moat @wide lark (just tagging whos online)

[marble moat]

cheers

[hard galleon]

o7

they hit a few of the channels so

[marble moat]

yea I banned them, should have removed their messages I think

[hard galleon]

oh yeah ban

ye

o7

@vocal osprey same prompt but in borderlands style

⏳

All depends what model you use and what it was trained on

Also depends on the settings used, same prompt but with "PROMPT MAGIC" on

both of which look straight out of borderlands

[vocal osprey]

ok thanks

[hard galleon]

🫡

[hard aspen]

hello! quick question, does anyone know if the models you fine tune on this site are private or public?

[mystic river]

woah this role system is really cool, anyone know what bot that is?

[mild walrus]

I believe there is somewhere you can toggle to be private or public

[hard galleon]

@hard aspen @mild walrus Models created are private until toggled public! 🙇‍♀️

It's under settings when you hover over it!

Images generated with models are public though until pro I believe!

So watch yourselves! 😈

I do need to test the report feature. 🤔

Public by default will be like NSFW

They will show up under here

But if toggled public, they'll go to the community models as well.

I like that there's no updoots personally, those can be gamed, imagesets generated cannot.

Kinda allows everyone to be equal and skirts the need for bots

I can confirm some services like these do work as advertised so..

Search gallery then just allows you to "PROMPT" for models to "PROMPT" further 😄

[sick gale]

how to train it to get spritesheets ?

[hard galleon]

@sick gale put in sprite sheets in 512x512 in the style you want

instance prompt "sprite sheet"

🤷‍♀️

[balmy solstice]

Hi

[analog nebula]

Currently could be done by creating dataset of spritesheets but we are looking into an more reliable solution

Have a look at this post by one of our developers: #💬│general message

tldr: You could but it would be an painful process of generating to get things right. We're looking into solutions on this.

[sick gale]

Another question, once I have a character, how to make different positions of the same character? each generation it makes different alternatives

[analog nebula]

try img2img feature

again not the best but you'll have to work with what you got currently 🙂

also try this : #💬│general message

but omit the stuff behind

[sick gale]

great, thanks for helping

[analog nebula]

A pleasure 🙂

Hopefully you'll manage to figure something that works!

AI is really about trial and error

[sharp zenith]

Hey guys

I just removed my access

Recived*

I have to say the developers did a excellent job, the quality even though stabe diffusion based is phenomenal

[analog nebula]

Thank you on behalf of the team! ❤️

We do have custom solutions built on top of SD yeah 😉

[sharp zenith]

I'm currently training the ai off of my 3000 something midjourney prompts 😈 but yes wonderful job guys i would have to say quality wise based on first impression you are tied with midjourney for #1 and blue willow follows you both with the #2 spot.

But the amazing thing is you guys added the beautiful freedom here

[analog nebula]

Wait till you try the custom models and train your own 🙂

[analog nebula]

Definitely freedom to a larger extent. But we still do moderate for highly questionable usages 😅

[sharp zenith]

Also a perfect hand!

I had to run that in the midjoruney facebook group 🤣🤣🤣

Oh man this has me so excited 😍

[analog nebula]

Happy to hear that 😄

Personally love my team so it's been fun seeing things grow

Can't wait to see how the future turns out for Leonardo

[sharp zenith]

You guys did something different than evwry other ai out there including midjourney, this training your own dataset is going to take ai to new levels

I would love to join a team as revolutional as this

[analog nebula]

Indeed! Not forgetting that we are actually striving to make things as easy to use as possible 🙂

People who do art are already experimenting with incorporating Leonardo into their workflows

[sharp zenith]

I've gotta say I personally enjoy it being web based

[analog nebula]

Oh we do have plans for mobile apps too

😉

[sharp zephyr]

😉

[sharp zenith]

Oh man

Android first please 🤣🤣

Oh man, I literally went to bed and got this email and now I'm wide awake 😅

[analog nebula]

from what I know, iOS version might be the first

Though I myself only have an pixel

[sharp zenith]

I personally dislike apple

[vestal patrol]

I do want an app because the website itself isn't the best on mobile I find and I'm not always on my computer

[sharp zenith]

Beautiful ui on the website btw

[analog nebula]

I used to use apple products, still have an appreciation for the designs. But pixels are a decent android alternative

[vestal patrol]

^^

[analog nebula]

yes

[vestal patrol]

I meant ^^ for this

[analog nebula]

We are working on it

[sharp zephyr]

watch app ?

[vestal patrol]

The UI is really good

[sharp zenith]

Yeah it's gorgeous

[sharp zephyr]

i havent seen it

[analog nebula]

No

iOS mainly but not the watch I believe

[sharp zenith]

So freaking clean 💪💪💪💪

[sharp zephyr]

ok

[analog nebula]

also I don't think a watch app for an AI content generation service is practical at all

considering the amount of content you have to display vs the limited size of the watch

[sharp zephyr]

it might be with holograms

[analog nebula]

Assuming they even come up with holograms

[sharp zenith]

Oh boy my dataset is done 😈

[analog nebula]

But for now, not really practical to develop for watch devices

[sharp zephyr]

yet

[analog nebula]

yet.

Better to focus on mobile devices first

the ones that people can actually properly use our service on

We could build an app for the watch, but it would be such a gimmick for now unless there's a way to display things properly

[vestal patrol]

Not everyone has a smart watch either

[analog nebula]

this too

But yeah main point is that it's not really practical to run something like our service on current smart watches anyway

[sharp zenith]

....leo...yep my new favorite

😈😈😈😈

[sharp zephyr]

[sharp zenith]

Lmfao how did you read my mind..m

A little annoyed midjourney rejected my comparison post

[vestal patrol]

2023 :D

[sharp zephyr]

why did they reject it?

[analog nebula]

it's an competitive world out there

[analog nebula]

business

also

lol

discord rejected my images on hands

its apparently nsfw

[sharp zephyr]

🖐️

[vestal patrol]

they're just scared imo

Leonardo is better

[analog nebula]

its knows the artist trick to hide hands when you can't paint em

😬

[hard galleon]

it learns

[sharp zenith]

Spamming and advertising

[analog nebula]

it's only fair they would

we would have done the same

[sharp zephyr]

😮

[analog nebula]

cheers anyway 🙂

[sharp zenith]

They're just mad cause Leo made perfect hands 🤣

[analog nebula]

we cant hold all these glasses guys

[sharp zenith]

Yeah kinda annoyed me I made a post addressing that admin and so maybe he will ban me oh well,

[analog nebula]

Oh please don't 😅

People will find out about us anyway, though various means

[sharp zenith]

I get not wanting competition on your page, but at the same time you should be willing to embrace it just kinda made me lose respect for them

[sharp zephyr]

🫂

[analog nebula]

We do allow for mentions

just no direct advertising

comparisons and the like sure, just not blatant advertising of competitive services ^^

[sharp zenith]

Yeah my post was a comparison of midjourney blue willow and Leo, and in the post I added that Leo was mjs #1 compettior and was curious of their response I guess it was blocking my post 🤣🤣

[analog nebula]

also yeah it needs to learn that wine glasses do not bend when people hold them 😅

(this is an private custom model of mine btw)

[sharp zenith]

I'm guessing you are an artist ?

[analog nebula]

I am, but the models I make are based on older eras, of artists from the past that I like for their style etc.

so around 1920s-1980s or so

[sharp zenith]

Oh my

Thats brilliant

[analog nebula]

photorealism is never the goal, something more textural, more organic

but that's just my personal preference

[sharp zenith]

The crazy part is the potential with Leo is you can literally train it on life...m

[analog nebula]

basically yes, you can tailor solutions to your needs and preferences

So for my case it would be that more textural painted look

[sharp zephyr]

cool

[sharp zenith]

I enjoy your images for the fact that it just reminds me of the older times when thi gs were simple

Very vintage

[analog nebula]

sometimes its good to know what you want, makes it easier to come up with models and also select what images generated are actually what you're after

having an eye for the visuals 🙂

[sharp zenith]

Yeah that honestly is what excites me that level of versatility

Like I would consider Leo it's own platform of ai separate from SD DALLE and MJ for thet reason

[analog nebula]

not perfect

but as artists and hollywood people would say: we'll fix it in post

[analog nebula]

One of our goals is to eventually bring together all the disparate solutions into one suite of tools that you can use to generate a variety of content, not just 2D images

[rancid harness]

Leo Ai is so much better than Midjourney! I’m impressed guys! Please don’t make it too expensive tho! 😅

[analog nebula]

We do plan to retain a free tier as we shift to introducing paid plans 🙂

[rancid harness]

That’s awesome

I really like the interface and the general feel to it. Also making your custom learning models 🤩

[analog nebula]

Easily creating custom models is definitely one of our key features 🙂

[rancid harness]

Where are you based?

[analog nebula]

Most of the team is based in Australia, some in Europe and some across the Atlantic, I myself am over in Asia

[rancid harness]

Coool!

Is there a mobile app in dev?

[analog nebula]

Yes, from what I know, it would be developed for iOS first

Android later

Cannot give an approximate at this time but it is in the works 🙂

[rancid harness]

Awesome!

[hard galleon]

chatgpt develop me an ios app for leonardo.ai ty

ok i do sleep gn

[analog nebula]

🤔

Perhaps that's what our dev is doing

[hard galleon]

that's what everyone on linkedin is doing 😇

[analog nebula]

(I'm joking)

[rancid harness]

I'm an older dude here 35 yo and this moment of time has been my dream since a little kid! AI's working for us while we enjoy life and create for the sake of creation (and food of course)

[hard galleon]

hey im 31 we're not that old we're just grown in ok

🌲

[rancid harness]

I still feel like 20

😅

[hard galleon]

Kids in a sandbox, my friend!

🏰

[rancid harness]

Amazing

I have two major feelings tho

Either soon the world enters into a catastrophic war and all servers and communication satellites are destroyed and this is the most we got to enjoy life with this technology OR Life will be bliss and AI nano and bio technology will make us live our best years, plus a mega improvement as cosmic human beings

[hard galleon]

yeah basically

which is why i'm kinda all in on this from an altruistic perspective

i want this project to succeed

[rancid harness]

Exactly!

[hard galleon]

i had to wait for an invite too!! 😤

[rancid harness]

😄

[hard galleon]

But I'm certainly trying to give others to tools and ideas to build further

"BRANCH"

[rancid harness]

Ha! I made my first ones and will make them public

Based on 80s and 90s airbrush paintings

Super fun!

[hard galleon]

With tech, the more numbers you add, the faster it can multiply, at least from my perspective so

[rancid harness]

True!

[hard galleon]

If I can load enough models for everyone to really sink their teeth into leonardo 🤔

[rancid harness]

You will become LORD

[hard galleon]

I see a tower of babel of sorts 🤷‍♀️

Art speaks all languages so

[rancid harness]

That’s the ultimate truth

[hard galleon]

I do need sleep though, but it's been a pleasure! 🙇‍♀️

Speaking of.. it seems like a lot more people have made theirs public.

There's a lot of models in gaps that didn't exist before 😄

[rancid harness]

Awesome

Good night

[analog nebula]

❤️

There's a reason why we have a community like this and features like the community models too. Collaboration is something we want to see happening 🙂

[grim spire]

I use caffeine

[hard galleon]

@grim spire i use dexedrine

@grim spire and caffeine

[summer pilot]

hola

[restive adder]

Hello, is the image generator already available publicly?

[analog nebula]

Yes but though an invite only early access program

You can either sign up on our website (https://leonardo.ai) or make an introduction on #👑│introductions and then taking at look at #✨│priority-early-access. The 2nd method is much faster 🙂

[burnt aurora]

Just want to say, loving the new additional features! Just noticed the Delete All and the Tiling options now available.

And the Prompt Magic has been really beneficial! Thanks devs for rolling out the features like this!

[analog nebula]

Yeah we still got plenty to come in the next few weeks! ❤️

[frail flume]

i have one question. there is option for NSFW training model. does leonardo allow NSFW creations?

[analog nebula]

yes

Though it is likely to be part of an paid only feature in the future

but for now it's available for usage to anyone

[frail flume]

ok thanks ❤️

[analog nebula]

A pleasure 😄

[grim spire]

Redline💀

[hard galleon]

my lungs will kill me before my heart does 😛

but its also only 5mg a day and its prescription for adhd soo

[grim spire]

Awww = (

[hard galleon]

no

thats a good thing

[grim spire]

Redline the cooldrink the energy one they banned

[hard galleon]

ohh neat

lol

[grim spire]

Cause ppl died drinking it

[hard galleon]

i just know 4loco for similar so

[grim spire]

They got fast af

[hard galleon]

im fast but unfocused without

physically im not but mentally im also zoomy but can't play games cuz im all over

so the dex helps keep me productive

[analog nebula]

when you think about it, it's pretty clever though

instead of banning nsfw outright

[hard galleon]

It also keeps everyone safe of dirty perverts like me from using the model without paying tax

[grim spire]

Tru cause now ppl need to pay to make a fake only fans

[grim spire]

❤️

[grim spire]

I'm on another discord server that generates only waifus for free

and u can talk to the ones you made

They have nsfw rooms hehe

For gening

[hard galleon]

idk i can generate real life black mail so waifus seems pretty small time in comparison

thus

$

[grim spire]

Oh no

[hard galleon]

mhm.

I don't, but it's certainly within the capabilities of "NSFW"

Money helps keep track of those abusing the system

"CHECKPOINT"

[grim spire]

Yeah

Also If theirs a free tier it stops minors

From doing "STUFF"

[hard galleon]

It's also easier to ban a credit card than it is to ban an ip

even if it's only $1 a month

[grim spire]

Did an ai make your pfp

[hard galleon]

People hear paid and think $100 a month but I feel, personally, the more people know about leonardo, the cheaper they can keep subscriptions for everyone 🤔

several ai + human hand

[grim spire]

Which ai

Holy

[hard galleon]

ive been doing this for awhile 😅

[grim spire]

Dk what any of that is u speak Chinese or smth

[hard galleon]

yeah

ni hao ma

[grim spire]

Indeed

[midnight cloud]

[sick pawn]

Is there a current limit for image generation?

[toxic arrow]

while the platform works great, the original advertised purpose gave the impression that the tool was abble to creat real "game assets" meaning you could download actual 3D geometry. I was worried because when that tool becomes a reality it will disrupt the whole industry.

[analog nebula]

No limits, only on how many simultaneous tasks you can run at one time

[sick pawn]

Thank you!

[analog nebula]

You're welcome 😄

[spark cosmos]

I'm very impressed with your work

[analog nebula]

Thank you on behalf of the team ❤️

[wraith rampart]

I already filled out the form

[analog nebula]

The google form or the one on our website?

[wraith rampart]

The one on the website

[analog nebula]

That would take a bit more. If you would like to get in faster, simply make an self-introduction on #👑│introductions and then go to #✨│priority-early-access which will unlock once that is done. 🙂

Registering though the website essentially puts you on a waiting list for an invite, but though the method I mentioned above, you should be able to get one within 24 hours

[wraith rampart]

Done

Filled out the wl from

[teal sable]

hi all, I've got my access today, I was trying to generate a Train Model, but the site is returning "An error occured submitting your request: internal error"

am I doing something wrong, or is there any server error?

[chilly breach]

I finally received my invitation! I'm so glad!

[analog nebula]

Might be an error server side

@marble moat @mild walrus (?)

[analog nebula]

Could try refreshing the page too or waiting a while before retrying

[wraith rampart]

Waiting on my email

[analog nebula]

Should be processed within 24 hours

[wraith rampart]

Ok

[teal sable]

I've refreshed and tried a few moments later, and the error persists

[analog nebula]

Will have to likely wait for Peter to come back online if the issue still persists

Let me try training a model just to check

[upper lava]

I'm sory if you already ansered this cuestión but, can you make a 3D model with leonardo?

[analog nebula]

No worries! And no Leonardo doesn't current have 3D capabilities. We are looking to implement 3D in the future however!

[analog nebula]

Seems to be working on my end. Try training with another dataset perhaps and see if the error still occurs

[teal sable]

OK thanks, I'll try again

[analog nebula]

I'll see if the training finishes successfully

[upper lava]

Ok, is there going to be an anauncement or anything?

[analog nebula]

In the future yes

[upper lava]

Ok thanks

[analog nebula]

But currently we are still improving on 2D functionality

Model training completed on my end

[teal sable]

I understood my mistake, I've put too many words in the prompt, separated by a comma

[analog nebula]

Ah

For training the model?

[teal sable]

Yes

[analog nebula]

You just need a really short keyword 🙂

https://app.leonardo.ai/faq

Do read the faq as well, it's very helpful for model training

not bad at all

[keen grail]

how can i give suggestions?

[analog nebula]

You can do that though #989166677910519850 if writing a suggestion though the bot is too complicated

Typically it is /suggestion suggest <your suggestion here>

if you want to submit an suggestion but a ticket can be easier for some people

[keen grail]

ah ok

thanks

[analog nebula]

👍

[neat axle]

We're about to begin a weekly challenge, any suggestions?

Some of the ideas we've been toying around with are:
Strange characters
Character sheets
UI Elements
Icons
Fine tuned models

[analog nebula]

Could even do a vote if needed once we narrow the selection down

[vestal patrol]

How do you keep the ai from making people have seven fingers lol

I already put "deformed hands" in the negative prompt but it ignores that completely

[plain cave]

I think UI elements would be fun and interesting!

[analog nebula]

Just hope that AI grows to not mess up hands 😅

but yeah hopefully in the future outpainting works on hands

[left kayak]

This is such a nice and intresting project

How is this (being) made? With an API or so? Litterly have zero clue xd

[analog nebula]

A bunch of custom solutions built on top of SD

[left kayak]

SD? :3

[analog nebula]

We got quite a number of secret sauces in the mix

[analog nebula]

stable diffusion

[neat axle]

we've got some OGs from the Disco Diffusion times which have been building custom solutions on top of Stable Diffusion

[left kayak]

Ohh okay, looks awesome

If I introduce my plans in #👑│introductions, is there any chance I can try this out in like a few days?

[analog nebula]

within 24 hours roughly

It'll unlock #✨│priority-early-access

just by making an intro in #👑│introductions

though that particular channel, invitations typically get sent out within 24 hours instead of the usual wait

Just something to reward people who stop by our discord and made introductions 🙂

[vestal patrol]

I was thinking, in the distant future

Once you figure out 3d generation

You could add support for doing gifs and stuff

Emphasis on "distant"

[analog nebula]

exponential growth in tech is a thing 😉

Moore's law

^

[vestal patrol]

👍

[upper lava]

How opened a call?

[analog nebula]

you mean an voice chat?

[upper lava]

Yes

[analog nebula]

Right below the language channels

clicking on either of the 2 above ones opens up a room with the respective amount of slots (10/50 users)

Though people aren't really using the voice channels at the moment

[tribal cloak]

UI Elements could be interesting

[nimble anvil]

I'm not sure if I'm doing it right, but I created a dataset of 36 selfies/head shots and then trained the model, but I just got normal image results, is there something else I should be doing? Thanks!

[vestal patrol]

Do you have the model selected?

[analog nebula]

What exactly would you be looking to achieve? You could try inputting one of the selfies into img2img and set the init strength, see if that helps 🙂

Also what JustHaven has mentioned

[nimble anvil]

Yeah I have the model selected, I was hoping to do something like this app...

[vestal patrol]

Apparently Leonardo doesn't know who Markiplier is 😔

help

[analog nebula]

hmm @mild walrus any ideas (when you're around)

[real pivot]

Has anyone figured out how to get a consistent solid background color? Such as white for example.

[wraith rampart]

Hmm

[neat axle]

what prompts are you using?

[merry stream]

How to get early access?

[hard galleon]

@real pivot "WHITE-BACKGROUND" works well imo but you might find better

i feel it pulls from urls then

or img file names?

🤷‍♀️

it is the heart of god and i just poke buttons

i don't understand it but i learn what the buttons do

[fast tiger]

Hello and welcome, @merry stream. 🙂

Please, first of all you have to go here: https://leonardo.ai/ and enter your email, (we do this because we keep track of the people who are registered).

Then make an intro here in #👑│introductions to unlock the #✨│priority-early-access channel and fill in the form, so you will have a faster access to Leo (this is an advantage for those of you who are here).

Thanks 🙏

[errant hinge]

This is really nice so far I haven't been able to do but one prompt and I love the ui. It's very organized. Now time to see what I can generate.

[fast tiger]

Thank you very much, @errant hinge!
The team really appreciates this kind of comments. 🙂

[rancid harness]

hi guys!

[fast tiger]

hello, @rancid harness!

You can take a look at this, it will be very interesting for you:
#💬│general message

[rancid harness]

Yes I'm already in!

Just wanted to share with you the AIRBRUSH PAINTINGS MODELS I made public for the community

super cool 80s and 90s inspired

start with "An Airbursh Painting""
of etc...

[fast tiger]

Perfect, oh yes of course...
With pleasure we want to appreciate your models, you can put it in: #🖼│image-share

[rancid harness]

Great

[feral ice]

Hello Leonardos

Where do I start? Are there any sneak peeks?

Already introduced myself and added my email to the whitelist and filled the form

Kinda afraid of this

[amber wyvern]

check out #🖼│image-share to see what others made

[feral ice]

Might speed up a lot of my creative process

But not sure how what this will lead to

[feral ice]

Output is always an image or could be a 3d model?

[amber wyvern]

3D is still in development last I heard

So still only images for now

[feral ice]

Ok , thanks for keeping me updated @amber wyvern

[errant hinge]

Of course. I just trained a model and have been finetuning the prompts so I'm excited to see what happens. 😀

[errant hinge]

Gonna try it out Sir 😉

[fast tiger]

https://tenor.com/view/fantastic-gif-21010169

[wraith rampart]

Can I make planets

[noble plank]

Yo! I assume that it's not available for testing just yet? 🙂

[amber wyvern]

You can make planets. I suggest training a model on images of planets for best results

[wraith rampart]

Awesome I will try it out when I have access

[fast tiger]

Hello and welcome, Baron Von Scrub 🙂

Please, first of all you have to go here: https://leonardo.ai/ and enter your email, (we do this because we keep track of the people who are registered).

Then make an intro here in #👑│introductions to unlock the #✨│priority-early-access channel and fill in the form, so you will have a faster access to Leo (this is an advantage for those of you who are here).

Thanks 🙏

[wraith rampart]

I have already done that

[noble plank]

Cheers, will do! 🙂

[fast tiger]

Excellent! 😉

[wraith rampart]

Done it already @fast tiger

[fast tiger]

Great, now you just have to wait for your invitation. 🙂

[noble plank]

Then you just gotta be patient, homie! That's why below the form it says to give it some time. :p

[wraith rampart]

I know

[hard galleon]

Worth the wait ❤️

D/w no ones idle behind the curtain!

It'll simply be better when you do see it

It's also a very human project so, so ya know, one person to another, patience. 🙇‍♀️

I'm personally putting a lot in and it's just simply because I wanna see icarus fly!

I haven't even started using any of them to create art for my web store or anything

[tawdry adder]

May I ask how long it should take after following all the steps before it will allow me to set up my account?

[rancid harness]

Hey so the Prompt Magic. What does it specifically do? I am trying with and without applying it but the changes aren’t that big of a deal to be honest. Is there a particular trick im missing? Also the dropdown menu from leonardo style and none, what should be the difference from none and or leonardo style be?

[fast tiger]

It is different from what it does to the normal image. It seems to me that it makes variations, a mix between images and text. Either I'm wrong, @marble moat or @mild walrus?

[marble moat]

Prompt Magic improves prompt accuracy. It also enables the ability to use images as a prompt (which is different from standard image to image)

@rancid harness

See how this reflects the prompt quite accurately

It works differently and is quite experimental at this stage!

[latent basin]

OK!

[rancid harness]

Perfect!

[safe salmon]

I am so looking forward to trying this for my tabletop assets. I just wanna show off assets so bad

I'm really curious to how close it can get to what I need for certain dnd characters, or specific, homebrew items

[crimson pecan]

Hey All

[real pivot]

Any talk about adding a landscape dimension or two?

[midnight cloud]

[crimson pecan]

How are things around here

just learned about this today

[midnight cloud]

They're great! One of my favorite channels to be in would be #🖼│image-share , definitely check it out! 😄

[raven smelt]

Hi everyone

Just got introduced via some sick AI art, thought I'd come in and try

Figuring out how to get access

[fast tiger]

Hello and welcome, @raven smelt 🙂

Please, first of all you have to go here: https://leonardo.ai/ and enter your email, (we do this because we keep track of the people who are registered).

Then make an intro here in #👑│introductions to unlock the #✨│priority-early-access channel and fill in the form, so you will have a faster access to Leo (this is an advantage for those of you who are here).

Thanks 🙏

[light surge]

This would be cool, and Twitter banner dimensions.

[analog nebula]

Currently you can set it manually but definitely would be a nice quality of life feature

[timid shore]

Hey Scoob it's me hoppy haha I'm on Dwayne's name cause he was the only one with hours left on another AI thing but nice to see you here!! I'm excited!!

[crimson pecan]

hey

[proud gate]

Hi folks, new to the server, was wondering if the ai is limited to 2d assets or can produce 3d models as well.
Thanks in advance

[analog nebula]

Currently 2D only but we plan to add 3D features as the tech becomes more viable in the future 🙂

Welcome to our community by the way!

[proud gate]

Wonderful thanks for the response, cant wait to try it out! When is EA available?

Thanks 😄

[analog nebula]

It's already available and we have been sending invites out in batches. If you would like an invite faster, do take the time to make #👑│introductions , that will unlock #✨│priority-early-access , which you can then apply for 🙂

[proud gate]

Perfect! very excited! thanks for your help Ayu, Ill sign up now 😄 !!

[analog nebula]

No worries!

🙂👍

[shadow lantern]

Has anyone published anything from Leonardo yet? Would be cool to have a look, like a short story with ai illustration or something

[grim spire]

Check #🖼│image-share

Just scroll through

[analog nebula]

They did 🙂

[grim spire]

K

[signal pewter]

hello

[shadow lantern]

Hi!

[worn cedar]

Hello 😁

[amber wyvern]

👋

[analog nebula]

And hello 🙂

[grizzled raft]

Hello, another suggestion! Please make the deletion (in particular the "delete all") button asynchronous (i.e. you don't have to wait for the images to actually be deleted to close the popup). A little qol improvement 🙂

[analog nebula]

@marble moat 🙂

[proper pier]

anyone got guidance on turning 2D assets created in leonardo into actual 2D assets that can be used

[rose prism]

try the finetuned model cute cartoon to generate like that ad

[analog nebula]

They've not in the server anymore

😅

[rose prism]

😆

[wraith rampart]

Still waiting on my email

[analog nebula]

Currently a bit of a delay. But will be sent out soon. Apologies!

[wraith rampart]

It's ok

Can't wait to create some stuff

[steel breach]

Theres a lot of new people joining :D

[analog nebula]

Indeed!

[steel breach]

Really happy to get the chance to try this tool

Generating 3d models is a feature im really looking forward to

Is there a list of planned features?

[analog nebula]

Not officially... Am unsure if @distant tide can provide any

[steel breach]

Well i'll take what is here first haha

[proper pier]

Legends for adding the delete all feature

[rose prism]

https://tenor.com/view/yay-sponge-bob-happy-celebration-job-gif-19821202

[proper pier]

what difference does this option make

[analog nebula]

Affects the style and rendering though modifying the prompt and other things I believe

Typically I turn it off for my custom models because it deviates from the style I want

[proper pier]

Thanks for the help 🙂

[analog nebula]

You're welcome 😄

[shut dirge]

That's a cool bot

[untold aurora]

Hey friends, great application. So i've trained the Ai with a dataset using 40 images. I'm trying my best to create a sprite sheet. I seem to be making zero progress, I'm genuinely trying to use this to create game assets. Anyone have advice?

[little glacier]

I wanna try to change this terrain with prompts.. how would i go about that?

[analog nebula]

try the img2img feature (below the sidebar on the left when in the image generation section)

[little glacier]

Ah yea i tried that, maybe i should try a smaller image cuz its on 2000 seconds loading now lol

[analog nebula]

😅

yeah might be worth a go

it's really about experimentation

seeing what works

push pull sliders, toggle stuff on/off

[little glacier]

Gotcha!

Ah and u can use ur own trained models as instance prompt coool

[little glacier]

I think i might have found a bug if its not user error: i keep generating the exact same image i uploaded

Whatever i change, the generated image is always the same as what i uploaded

[analog nebula]

is there an image in the img2img section?

[little glacier]

Yes

[analog nebula]

that may be it

adjust the init too

[little glacier]

Ahhh kkk

[analog nebula]

too high and you basically don't get any deviations from the source image

[little glacier]

That was it lol ty

This is amazing

Alrite i got the hang of it now. Took me about 90 minuts lol

[little glacier]

Ok one more question...

The less initial strength i put, the more variety

But i loose texture, they all turn cartoony

High initial strength:

Low initial strength:

So my question is, how can i make it not so cartoony

Oohh it looks like its because of the trained model

[crimson pecan]

Good Morning 🙏

[little glacier]

Got it going!

Im using my own 3d models to generate more lol

[analog nebula]

🙌

Looking great!

[neat axle]

try playing with the step count and the guidance scale too

[wraith rampart]

Will the email be in spam

[real pivot]

Mine wasn't

[wraith rampart]

Ok

[shut raven]

hey there guys i just got early access confirmation email but i cant log in ..

[analog nebula]

Replied in the support ticket

[wraith rampart]

I'm getting the same thing

[analog nebula]

It's not the actual invite, just an confirmation that you're on the waiting list

[wraith rampart]

I still haven't gotten the email

[upper lava]

I can't see the result of the training

[analog nebula]

You have to use the model to generate things with

[upper lava]

I know

I puted like 6 or 7

[upper lava]

I know

[analog nebula]

Did you enter a prompt?

[upper lava]

Oooooooo

Ok I have to generate it

[analog nebula]

Yes, training basically creates an model, you use the model to generate images with 🙂

[upper lava]

Ok thanks 👍👍👍

[analog nebula]

No worries! 🙂 👍

[crimson pecan]

cant wait to use this

[wraith rampart]

Waiting patiently for my invite

[safe salmon]

Lol. Same

One day

[worn crown]

After I create a model and train it, how do I continue to train it? I presume that's how you improve the model.

[analog nebula]

@mild walrus might be able to better advise on that

[analog nebula]

Will be eventually sent out. Currently we only got one person who is able to access all form submissions for privacy reasons and lots of applications. Apologies for the delays 😅

[mild walrus]

continuing to train a model does not necessarily improve it, its possible to overtrain a model. traininng is a trade-off between being able to get a wide range of outputs and being able to specialize in the images you feed it. here's a guide for some pointers on it.https://sweet-hall-e72.notion.site/A-Guide-to-Getting-the-Most-out-of-Leonardo-AI-aae753df4cf64bc7bbda99790c8e448c

[analog nebula]

Thank you very much Ethan ❤️

Will be helpful for me as well

[worn crown]

Yes, thank you! Look forward to reading this!

[wraith rampart]

Oh well

[light surge]

Can you increase the dimensions of an image that has already been created?

[analog nebula]

Not currently but once we implement outpainting/canvas extension you should be able to

It'll fill in the extended canvas essentially

[vestal patrol]

Question for all: What's your favourite model in the app and why?

I personally like the Signature model because I find everything looks really good in that model

[fast tiger]

A very good question: I love "Leonardo Select", because you can achieve a high level of realism.

Super cool. 😎

[wraith rampart]

Why is my invite taking so long

[fast tiger]

Thank you for waiting, but only one person has access to the forms for privacy reasons, and he is currently unavailable (due to external circumstances).

But rest assured that you will receive your invitation. 🙂

[wraith rampart]

Ok thanks for the update

[full gyro]

Is this tech trained on GAN?

Also, what are the competitive advantages over something like say, Dall-E, or Midjourney for creation of game assets or ui

[fast tiger]

My pleasure 🙂

[shut raven]

😂 dw boi , we'll get it for sure!

[dusk moat]

generate speeder's next idea

[sharp zenith]

@analog nebula hey the 40 photo cap on training any idea if that will be expanded?

[fast tiger]

The team plans to increase the 40 limit in the future. 🙂

[light surge]

Looking at other peoples images, when it says; Base Model ~ SD v1.5. Finetuned Model - Leonardo Signature. Does that mean they used stable diffusion originally, then upscaled it?

[fast tiger]

No, they used the "Leonardo Signature" model and, of course, Stable Diffusion is also included in the models. 🙂

[quasi nimbus]

got some issues using the app right now.

[marble moat]

there are a few issues right now we are aware of and working on resolving

[quasi nimbus]

Seems fine for me now 🙂 .

[hard galleon]

https://tenor.com/view/caos-bob-esponja-crisis-patricio-gif-23199341

[latent basin]

I enjoy ai-art and making it.

Can wait to try this system of AI Art; would like to try my style with this one.

[marble moat]

should be completely back now

[hard galleon]

@latent basin "influence"

[native blaze]

Nice got whitelisted access excited to see what I can do!

[latent basin]

haha you know it

[hard galleon]

@latent basin

[latent basin]

Nice

[hard galleon]

Thanks for the idea 😄

[latent basin]

welcome let me get you the prompt.

[latent basin]

tshirt vector, [VDay] graphic, synthwave, vivid colors, detailed, appealing

[hard galleon]

testing v2.1

results of that prompt with various settings

[neat axle]

https://twitter.com/LeonardoAi_/status/1616217973257285636?s=20

a bit about Prompt Magic

[left kayak]

Damn, this AI is amazing

Any idea about how much this AI will cost to use?

[spring bone]

at first glance it looked like this tool generated 3d content as well as 2d. Is this on the roadmap?

[hard galleon]

@left kayak There will be a free plan

[left kayak]

Nice!

[crimson pecan]

This thing is pretty good i got my early access today

gonna keep playing but loving it so far LFG

[river holly]

I've only just arrived. So when you train a dataset, is that supposed to be, say, pictures of a specific person from many different angles, which you can then give a name and use in the prompt? How does that work?

[hard galleon]

Fill a data set with style you want, train it with an instance prompt of WHAT you're looking for from the image set really

[river holly]

Also can you enter the URL of a picture in the prompt, like in Midjourney?

[hard galleon]

It has img2img but it's done via upload not url

donno if url works

[river holly]

I see. I didn't know there was something at the bottom of the controls you could scroll to. I see it now.

[fluid kestrel]

you can drag and drop the image from one browser tab into the app

[fast tiger]

It seems to me that you can only do it when you download the image, then you can drag it to the app.

[fluid kestrel]

strange, i've been using it all day like that, dragging from google images as the source

[fast tiger]

and you can't do it anymore?

[safe salmon]

No worries! I just appreciate being here! It's cool to see more niche AI generators pop up.

[fluid kestrel]

oh, no it still works fine for me!

[shadow lantern]

I got into the beta 🙂 thanks guys! Been doing some testing and will continue to later ❤️

[river holly]

How do I upscale an uploaded image?

[fast tiger]

Once you have loaded an image you go to where the circle I marked in the image is:

[river holly]

How do I load an image unchanged?

[analog nebula]

We do plan to include 3D in the future when the tech is more viable 🙂

[river holly]

I mean, all I see is doing the img 2 img thing, but when I do that the AI changes it (and messes it up in my case).

[analog nebula]

Increase the init

right below the img2img box when you drop an source image into it

[river holly]

That worked! Thank you very much!

[analog nebula]

also current upscaling method doesn't preserve things as much sometimes

just fyi

We will be adding other solutions as well in the future

[safe salmon]

What are the username requirements? I've tried to enter a few and all I'm getting is that the username doesn't meet requirements

[analog nebula]

@marble moat , if you can assist 🙂

[analog nebula]

probably no spaces and special characters

But we should update the doesn't meet requirements message to be more specific yeah

[marble moat]

Username requirements are pretty close to Twitter username requirements

You can always change your username later from the settings menu

[safe salmon]

It works now! Not quite sure what was happening, but it suddenly accepted my username

[pallid birch]

A lot of people have been asking for an invite, what does that mean? how to check if I have already been invited

[jaunty sphinx]

hey trying to do something similar with Leonardo as I did in this dalle vs midjourney but I'm not getting great results at all. any tips?

https://saipien.org/dalle2-vs-midjourney-art-prompts/

Also want to add examples and tips to the dedicated article for Leonardo but struggling with the prompts. https://saipien.org/leonardoai/

best I'm getting

lmao

[lone mortar]

i got acces since today cant wait to use the app this weekend!

[sudden rampart]

Its great!! I'm loving it

[hearty estuary]

how do you do a full image scale? i only got half body when generating an image

[sudden rampart]

Im not sure but trying to find out myself. I'll have an idea in a moment

after I fine tooned what I wanted I hadto get specific " Standing 5. foot f inches tall with long hair and wearing her full suit of earth armor, We are looking at her head from toe. she is wearing her priest robe over it even though she is in female form. "

[wispy nexus]

Yo Do I have to just wait to get into the app or do I have to do specific thing to get early access earlier ?

[hearty estuary]

wow this looks good

[analog nebula]

You can either wait if you've signed up on our website or make an self-introduction in #👑│introductions , which will unlock #✨│priority-early-access . Though that you'll be able to get an invite much faster than the first option. 🙂

[wispy nexus]

oh really just by making a presentation ? cool

thanks for the tips

[analog nebula]

Yes, it's much faster than though the website. Not instantaneous but much faster for sure!

[wispy nexus]

We will know if we are whitelisted when the mail will say so ?

[analog nebula]

There will be an email sent to your inbox basically 🙂

[wispy nexus]

👌

[grim spire]

really creative

[opaque wyvern]

when you 'remove background' is there a way to get it to remove the bg from the upscaled image? at the moment it appears to always do it on the original

[analog nebula]

Still being worked on

Not available for upscaled images at the moment.

[opaque wyvern]

is there a way to share a Personal Feed (to non-members) ?

or indeed even to members

[river holly]

What does "remove background" mean? I did it on a picture and I saw no difference. Is it supposed to delete what's behind the characters and have them stand in front of a blank canvas, or something?

[opaque wyvern]

yeah, when it works that is what it does - but often it'll create really weird results

it's easier to just take the images to https://www.remove.bg/ and let that do it (or do it by hand in Photoshop)

[analog nebula]

Images in the personal feed are actually visible to members in the public feed at the moment

As for sharing externally, simply download the image

Or right click and copy image link for that matter

I've been using the 2nd option to share images on discord

[opaque wyvern]

sure, just gets a little impractical when you're generating hundreds of images - but that's fine, easy enough to just select the best ones

[analog nebula]

https://cdn.leonardo.ai/users/97efe82d-50e4-455a-97db-292f68bb80d6/generations/b6bbfcbb-07ca-4e2f-9f89-3a0408cd79d4/John_Singer_Sargent_v2_A_painting_of_teddy_Roosevelt_2.jpg

[analog nebula]

Yeah there's an share option in the UI itself but I think it is an work in progress

Because you can't use it at the moment

[analog nebula]

Agreed on being able to share your personal feed to non members though

I think it would be nice if at least user profiles arent locked behind the login page

[opaque wyvern]

it's fair enough while still in dev - would likely add a hefty bandwidth cost

although probably nothing compared to the GPU Cloud costs 😄

[analog nebula]

Still I think it would be attractive to potential users

Especially if content creators and the likes use the service

[opaque wyvern]

without a doubt

[analog nebula]

https://app.leonardo.ai/profile/Ayumi

I am pretty sure it is locked 😅

Yeah it is

[opaque wyvern]

yeah, I can see it, but only because I'm logged in

[analog nebula]

Yeah

Hopefully in the future!

[safe salmon]

Alright... time to make a bunch of NPC portraits

Will it pass the tiefling test?

[cyan walrus]

What would you add to the prompt to make sure the entire object is visible, I often get objects that are cut-off like this

[halcyon ore]

How would I take a picture of me and have it regenerate it using a specific dataset?

[hard galleon]

@halcyon ore camera

[halcyon ore]

Sure, but is that the "Image to Image" section?

[hard galleon]

sorry i just woke up i had to be a smartass

But yes you'd use img2img

[safe salmon]

Wow, it does pass the tiefling test

[halcyon ore]

Sweet - thank you. Still just getting my feet wet

[hard galleon]

Been at this for literally years so any tips lemme know!

I don't mind a ping

Img2img then drop the initscale between .2-.6 imo

[halcyon ore]

Awesome - thank you! So far, I'm liking this better than midjourney... super impresssed!

[hard galleon]

Can also play with magic prompt which removes the initscale but does wizardry itself

Try to keep the other bar, the generation w/e, to around 10, so it gives an even distribution of prompt/model

But if you want it to look more like the models style, increase it to around 15 or so

Less, 5

[halcyon ore]

This is super helpful - thank you

[hard galleon]

o7

@halcyon ore The opposite thing you can do is take 40 photos of yourself in different angles

Create a dataset

then anything you typed would generate you in that style

"Noir"

[halcyon ore]

Ah! That also makes sense. Interesting. Totally the opposite way of what I was thinking

[hard galleon]

I made a Hideo Kojima bot with that method

that'll generate hideo kojima as ____

(we're friends and hes stolen my face using AI so)

[halcyon ore]

That pretty fantastic

[vestal patrol]

GUYS OMG

THE MARKIPLIER DATASET WORKED

IT GENERATES MARKIPLIER INSTEAD OF A RANDOM DUDE NOW

[hard galleon]

lmao

[river holly]

Can anyone tell me what model is best for turning a line drawing, i.e. a penciled and inked comic book character, into something resembling a painting or a photo, while preserving the facial features and the posture?

[analog nebula]

lol

[hard galleon]

@river holly img2img high init scale

[analog nebula]

^

Try that

high init and mess around with the prompt

probably something like highly detailed, realistic painting/photo of _______

[river holly]

Okay, I'll try. And which model would you use? Creative, or what are the other two again...? Nevermind, I'll just try them all out.

[hard galleon]

Whichever style you want ool

[analog nebula]

Try them all yeah

see what gives the best result 🙂

Toggle things on and off too

[analog nebula]

try increasing the width before generating

[hard galleon]

v2.1, drop height to 512

[halcyon ore]

So, I'm still confused with img2img:

1. Go into tool
2. Upload an image
3. Make sure prompt is empty
4. Select a finetuned model
   "Generate" button is not clickable. I'm sure I'm doing something fundamentally wrong

[hard galleon]

@halcyon ore Needs a prompt

Try "Detailed, Photograph, Portrait, Intricate"

whichever

[analog nebula]

Also outpainting will be a feature we are going to add in the very near future, which will help with such issues too by allowing users to expand on their image (by adding content outside the current image frame)

[halcyon ore]

Got it! Thank you!

[hard galleon]

🙇‍♀️

[cyan walrus]

Oh okay, sounds great 👍

[river holly]

I've never done Outpainting. (Well, I've hardly done anything A.I. illustration related before yesterday.) Is that available here?

[hard galleon]

Not yet but will be

[analog nebula]

Planned feature 🙂

[river holly]

I see. Thx!

[analog nebula]

Will be out within this quarter of the year

Most likely

[hard galleon]

Maybe

We hope

:v

[analog nebula]

Well we have developers not magicians

[hard galleon]

Are you sure they're not? As I'm not convinced

[analog nebula]

I think they're doing quite a bit considering we are technically a startup lol

[hard galleon]

Feels like hogwarts in here

[analog nebula]

hogwarts generator

[hard galleon]

OH BOY OH BOY

grabs all the harry potter stills

YOU'RE A WIZARD, ROWLING

[river holly]

I've always wanted to see what's to the left and to the right of and below the old Paramount Pictures intro from the 1950s...

[hard galleon]

trust me you do not

the shadows don't need a known face

speaking of

[analog nebula]

👀

What does it look like

I need to google it

[hard galleon]

https://www.nvidia.com/en-us/geforce/news/jan-2023-nvidia-broadcast-update/?=&linkId=100000177829126

[analog nebula]

👀

eye contact off

[hard galleon]

NVIDIA Broadcast 1.4 Adds Eye Contact and Vignette Effects With Vir...
Plus new options to mirror your camera and take a selfie.

the last 2 make me laugh so hard

[river holly]

This one.

[hard galleon]

the most basic features

"hey we made it so we can force your eye contact"

"also we made it so you can just save an image finally"

[analog nebula]

we'll never find out

[river holly]

I NEED TO SEE WHATS OUTSIDE THAT FRAME!

[vestal patrol]

looks like 096 is having a feast tonight

[analog nebula]

if it's done by AI it's completely made up though 😬

not to the artist's intentions

[river holly]

So is the matte painting itself. Some Dutch guy painted it. It's not a real life mountain, although the landscape looks like the scenery around Yosemite.

[analog nebula]

Yes, but I meant if AI extended it, it wouldn't be true to the artist's own vision

though I prefer not to go down this slippery slope 😅

[halcyon ore]

Is there a way to search Community Models or are they all on the one tab?

[analog nebula]

Not at the moment

Should be an option though yeah

I'll push an suggestion to the devs

[halcyon ore]

Excellent!

[analog nebula]

An direct link to the model would be helpful too

[vestal patrol]

wee

[hard galleon]

@analog nebula Yes please for that one

It'd be nice to link models

[analog nebula]

yeah I figured

definitely would be helpful

[vestal patrol]

I made my markiplier model public just now

It's not too far down the community models page if anyone wants to try it out

[vestal patrol]

You found it :D

[analog nebula]

amazing

[hard galleon]

bikini, bra, lingerie

[hard galleon]

@markiplier hiii mark

smoking a pipe full of weed down a back path in the forest

[vestal patrol]

The ai finaly makes good pictures lol

[analog nebula]

desert themed stuff

[hard galleon]

markiplier kissing mr beast

[vestal patrol]

lmao not markiplier fat

that second one's like beer belly hillbilly markiplier

[hard galleon]

we're ruining justhavens crush

quick do more