💬│general

[neat axle]

Hey guys!

Welcome @wide lark

[wide lark]

nice to be here ser!

[prime wind]

Nice to see you here ma frends

[deep shoal]

hey hey

[analog nebula]

🐻👍

Oh I can't change my name on here

was going to change it to ayu or something but perhaps when I've perms

[neat axle]

You should be able to do it now 🙂

[analog nebula]

thank you ❤️

[neat axle]

@analog nebula are you also in the team? 👀

[analog nebula]

Actually, yes. But perhaps best to talk to mindspheres about it

he has briefed me rather briefly so to speak

[neat axle]

Awesome ,will message him

[analog nebula]

yeah he hasn't really assigned me roles or talked me though the discord things, mainly just on Leonardo itself

😅

[limber forge]

Hey )

[candid horizon]

Hey )

[wary raven]

Hey)

[neat axle]

Hey guys! Welcome @limber forge @candid horizon and @wary raven 🙂

Welcome to the Leonardo family.

It’s a pleasure to have you around!

@analog nebula and @fast tiger will be your guides

[dusty atlas]

🖐️

[neat axle]

Hey @dusty atlas!

[dusty atlas]

when will the test start?

[neat axle]

We're putting the final touches on the app, let me ask the team and get back to you in a bit

We expect to launch the alpha before Christmas 🙂

We'll also be having some contests in the upcoming days to reward our most active community members, as to give them additional perks 😉

[dim hazel]

when will it be possible to read whitepeper?

[modest badge]

gello guys

[dusty atlas]

great, I'll wait for the news

[neat axle]

Not a crypto project sir 🙂 it's a webapp for an AI tool

[dim hazel]

I thought your project in the field of NFT

[neat axle]

It does have that use case, it can be used to generate assets for web3 projects

[modest badge]

Can you tell me if the app will be paid or free?

[neat axle]

Main use cases we've found are:

• creation of assets for games
• assisting artists with the brainstorming of their projects
• keeping consistency among the assets created

[neat axle]

We'll have a generous free tier during the alpha testing

[modest badge]

Oh that's cool!

[neat axle]

Need to fix this bot (the Gm ban) haha, will do so briefly

Curious, how did you guys find Leonardo's discord?

[deep shoal]

There's some great examples on our website and Twitter if you want to see more - https://leonardo.ai/ + https://twitter.com/LeonardoAi_

To be clear, this is AI tooling that can create stylistically consistent assets for any use case - Web2/Web3/entertainment/fashion/marketing - all potential use cases

We're offering model fine tuning at scale, with a full social platform, and with control over outputs that you don't see on Midjourney or standard Stable Diffusion

Oh man that bot, we gotta fix it 😆

[limber forge]

an invite came in the mail after subscribing on the site

[neat axle]

@wide lark there's your answer 🙂

thanks Funky 🙏

[limber forge]

np

[devout gale]

Hi guys. what's going on here?

[neat axle]

Hey Rinari, welcome to the Leonardo family, we're just kickstarting the Discord server

[frail grail]

I think you should definitely touch on the cryptosphere and NFT

[neat axle]

Definitely, it's one of the biggest use cases we've found! In #🖼│image-share I shared some images we generated with our AI by training a model from an NFT project I'm friends with (Bad Kids)

hi @severe summit, excuse our bad behaved bot, we're working on fixing it so it doesn't delete the greetings 🙂

[sleek moat]

hey there:)

[neat axle]

hey Korok! Welcome 🙂

Mind helping us answer some questions in #👑│introductions ? 🙏

[vital flower]

👀

[mystic river]

Hello everyone!

[fast tiger]

Hello, @mystic river. 👋

[knotty kayak]

YOOO

[analog nebula]

Don't forget to make #👑│introductions if you are comfortable with it! We also got a few questions and answering those would greatly help us tailor and improve things towards people's needs and wants 🙂

[vague arch]

hi

[analog nebula]

Welcome!

[soft sparrow]

👋

[analog nebula]

Please do feel free to make #👑│introductions , we also got a few questions and your response will help us improve our service towards your wants and needs! ^-^

[low garnet]

Привет

[midnight cloud]

[wheat grail]

hi

[analog nebula]

Welcome!

[muted sorrel]

Hi folks. Neil here. I’m building a new media brand to cover ai and web3 called W3 Guru.

I’m interested in this tool both as something to cover as a story and to use for our art production.

[analog nebula]

Thank you for your interest! We hope you'll find our service to be suitable for your needs 🙂

If you're comfortable, it would be great to make #👑│introductions , we got some questions and your response will aid us in improving our service in regards to our userbase. ^^

[odd lintel]

Hey!

Saw something poppin up on twitter. Its looking good!

[odd lintel]

Very curious how leonardo ai van transform my pixel art!

[midnight cloud]

Can't wait to get my hands on this! Probably the most ingenious project to pop up in 2022.

[analog nebula]

Thank you for the kind words!

[neat axle]

Hey Berny! If you can you provide me with 15 base assets, I can run some tests and see 🙂

[hybrid hedge]

Looking forward to use this! 👏

[dapper depot]

gm, glad to be here and see what's to come

[neat axle]

Behind the scenes 👀

[surreal merlin]

hi all I just joined via email

thanks 🙏

[neat axle]

hi @surreal merlin! welcome

[quaint dirge]

Just joined so excited about this project 🚀

[sinful holly]

[wanton pelican]

Hi everyone

[wide lark]

does anybody know modern point-and-click games? I remember spending hundreds of hours completing quests

Mr @sour prawn even mentioned he'd love to create this type of a game on his own some day

[supple violet]

thanks for inviting

in mail

now how can i acces

[sinful holly]

For point and clink games I heard of one called Chinatown detective

[wide lark]

Old Broken Sword, Full Throttle, and the modern Deponia were my favorites

[sinful holly]

Isn’t Return to Monkey Island one also

[neat axle]

We're expecting the launch of the tool around Christmas time ( #📌│faq )

[supple violet]

alright

💯

[wide lark]

never played it, though know it's a classic!

[supple violet]

its good , but every amazing ai apps are always waitlist 🥹

[analog nebula]

You'd likely be in the first batch of users 🙂

still fairly early I'd say

[supple violet]

yes

alright is this using stable diffusion for 3d game assets ?

[analog nebula]

I believe they're actually developing their own in-house solution for that

or at least in the process of R&D

[sinful holly]

You could play the original Monkey Island then Return to Monkey Island. The original games are about $10 on Steam and the modem sequel is $25. Probably on GamePass also

[supple violet]

am i allowed to invite people to discord or no

[drifting phoenix]

Hi everyone!

[supple violet]

hello fellow new comer

[drifting phoenix]

Hey @neat axle ! One of your loyal twitter followers here! Feels good to be part of this! 👍🏻
Wish you and the team good luck with this!

[midnight cloud]

Runescape comes to mind lol

[fast tiger]

I love that game. 😄

[midnight cloud]

Lol me too, haven't played it in a while. I still remember stepping foot into the game in 2005, ahhh good memories.

[worn breach]

Hello all, very happy to be a part of this!

[fast tiger]

Of course, feel free to invite your friends. 🙂

[neat axle]

hey ser 🙂 appreciate it

[shadow lantern]

Hey all! Any tutorials you know of?

[analog nebula]

what sort of tutorials specifically?

[dry matrix]

hi

[analog nebula]

Welcome!

[high ledge]

Ola all, Hello from Portugal.

[analog nebula]

yes

welcome all!

and Jo D

😉

Please do feel free to make #👑│introductions everyone , we also got a few questions and your response will greatly help us improve tailoring our service towards your wants and needs! ^-^

[odd lintel]

Gonna prep that!

[odd lintel]

Ill run a few base wizards trough my generator and sent them to you!

[nimble anvil]

Hello!

[analog nebula]

Welcome to our server! ^^

[wind bramble]

howdy

[analog nebula]

Welcome!

[twin idol]

Hello friends

[analog nebula]

Hey there 🙂

[odd lintel]

Send DM 😉

[nimble anvil]

Does the welcome email take a while to arrive?

[analog nebula]

Possibly

I think they're sending them out in batches

[midnight cloud]

Can't wait to get a hold of it!

[analog nebula]

I don't think it's for the access to the service however, that is expected to come out around Christmas 🙂

[midnight cloud]

Yeah, either way still excited!

[dusty cedar]

hello hello excited be here

[analog nebula]

Welcome!

[barren dagger]

So what is everyone excited to build?

[midnight cloud]

Hmmm gonna try some emblems, maps, maybe a few other things. Wbu?

[barren dagger]

Kind of excited to create game characters if that is possible, let.

Do we get any sneak peaks behind the scenes of Leonardo? Like is it using Unreal Game engine or Maya, as a backend or is own code base?

[analog nebula]

this community

[neat axle]

@barren dagger here's a sneak peak. We're not doing 3d (yet)

[limber cave]

Looks good, can't wait to give it a spin. What are the subscription / usage fees going to look like?

[neat axle]

hey there! there will be a free tier moving forward, but subscription fees have not yet been set

[limber cave]

what would also be cool is to generate images at different tiers. like you have a basic shield. then you have one that has some metal, and every step up gets more complex/looks cooler

[carmine tide]

Hey there so I was wondering what exactly the leonardo ai service would be and what the plans are going forward. The website says creating "assets" but is that just images, 3d models, or what exactly?

[fast tiger]

Hello, @carmine tide. 🙂

This can help you a lot what Leonardo.Ai is all about.
#💬│general message

[neat axle]

Welcome @noble lark!

[noble lark]

Hi!

[neat axle]

@analog nebula fixed 🙂

[analog nebula]

Welcome Jason! Have always admired your work in the 3dtotal books! 😄

[carmine tide]

Thank you! However I still am confused on the actual output you get from the "model". Is it a png of the generated asset or is there more to it?

[analog nebula]

currently a png yes, but we are looking into 3D capabilities in the future as well 🙂

[noble lark]

It has been years~ and I have not been active online for last 10 years.

[analog nebula]

indeed!

You're definitely a lot more low key these days

[weary jacinth]

Hey all 🙂

[ebon canyon]

Gm Gm

[low arch]

based

wen launch

[fast tiger]

We expect the launch to be around Christmas.

[low arch]

can't wait to try it out 😍

[rose stump]

Hello everyone

[bitter plover]

gn

[merry mason]

Clarity (I hope): The alpha will launch around Christmas. There is a free tier. Details of the premium plan have not been released.

[quaint dirge]

Howdy makers!

[spiral hinge]

gn

[fast tiger]

That's right. (#📌│faq)

[arctic canyon]

hi

[dusty spear]

gm everyone

[supple violet]

hi

[limber cave]

so for those of us who use MJ how many images have you made with mj?

[supple violet]

i use niji . journey

and sd local

[limber cave]

nj/mj use the same image count 🙂

[supple violet]

yh i mostly interested in vector art and photo realism

[limber cave]

i did wombo before mj but then found mj and just used that

well till i got the nj invite then made tons using that

[supple violet]

i paid nj

10$

[limber cave]

unlimited fast was like crack

[supple violet]

got nj and mj both

in 10$ 💯

[limber cave]

if you got in the beta for nj it was unlimited fast for free

yeah, a few weeks ago they turned off the free fast

[supple violet]

imma start selling vector art

on graphic resource sites

[limber cave]

what sites besides adobe stock allow selling ai art?

[supple violet]

freepik

pretty popular

[limber cave]

yeah ive used freepik before

[supple violet]

so will they allow me

to sell nj art

[limber cave]

also with the basic NJ plan, you can't sell the images you make.

[supple violet]

mods said me tho

[supple violet]

u cant sell others but can do mine

[limber cave]

only if you have a standard subscription. You don't get commercial terms with a basic subscription

[supple violet]

bruh i have commercial lol what u on about

[supple violet]

https://media.discordapp.net/attachments/1017943946166550620/1052931113624027186/MJ_new_plans_fin.png

with 10$ yes

commercial

[limber cave]

oh yeah, they changed the plans 2 days ago

[supple violet]

yes

[limber cave]

before with the $10 plan you didnt get to do commercial

[supple violet]

ah i see

but now its worth

and good

[limber cave]

the standard is a much better deal.

i can easily generate 500-1000 images in a day

[supple violet]

i purchased mj back when it was just started

30$

when faces were crap

[limber cave]

i guess you could do the $10 plan and rate images every day to get more fast time.... but the unlimited relaxed is the money

[supple violet]

if fast hours expire 200 mins

i willl be switched to relaxed ?

on 10$

unlimited

[limber cave]

no

you only get unlimited relaxed with standard/pro

[supple violet]

ight ok

[limber cave]

once your 200ish images are done for the month you have to pay by the hour or earn more fast by rating images

[supple violet]

ok got it

[analog nebula]

Morning, and welcome everyone! 🙂

[unborn moon]

gm guys

[red tide]

Hi guys

[analog nebula]

👋

[odd lintel]

Morning!

[odd lintel]

Were are you from? Just curious 🙂

[analog nebula]

in asia currently! But I sleep at really odd hours 😅

Singapore to be specific

[odd lintel]

Haha. Nice!

[analog nebula]

Ah I wish

it's probably not doing my health or appearance any good 🫠

[analog nebula]

Please do feel free to make #👑│introductions everyone, we also got a few questions and your response will greatly help us in tailoring our service towards your wants and needs! 🙂

And welcome to everyone new!

[jolly wing]

Hello guys!!

[neat axle]

Hey! I’ve started running some tests on the dataset you sent me

[neat axle]

Wsup Future!

[spring yarrow]

Hey hey

[analog nebula]

Welcome to the server! ^^

[spring yarrow]

Thanks 🙂

[analog nebula]

If you'd like, you can make an #👑│introductions 🙂

Got some questions in there too, and the responses will help us tailor our service better towards the needs of our userbase

[sinful holly]

Today I was going to start a new project but I actually finished it earlier in the morning instead of sleeping. So, I guess I’ll just chill for awhile

[analog nebula]

Very nice, but do get rest!

[odd lintel]

Curious

[slim wolf]

hello

[fast tiger]

Hi, @slim wolf. Nice to have you around.

If you could, it would be perfect if you could answer the three questions in #👑│introductions, please. 🙂

[glacial sail]

Hello

[fast tiger]

Welcome to our server! 😄

[sinful holly]

I slept this time. Now, I can be up another night crafting

[nimble anvil]

I've tried a few times but still not getting the welcome email

[fast tiger]

Yeah, now you're in turbo mode!

[sinful holly]

Right now I’m organizing. Some folders then some training.

[fast tiger]

Hi, have you checked your spam box? If it's not there you can send me your email to check it, please. 🙂

[coarse igloo]

Hey I'm new can someone point me to where I can figure out how to use this ai?

[nimble anvil]

Yeah I checked everywhere, I'll send you a DM! 👍 :edit: can't send DM

[fast tiger]

Hi, you can't enter yet. Around Christmas is when it will be open (you can check it at #📌│faq)

[fast tiger]

Perfect, ty! 🙂

[nimble anvil]

I've sent a friend request, the DM got rejected

[fast tiger]

I have already accepted you!

[burnt bramble]

Hello everyone. Thanks for the invite

[fast tiger]

Hello, welcome. Nice to have you here.

[burnt bramble]

Just getting acquainted with the server. I will be able to give into later tonight.

[faint vine]

Exco in the building

[hybrid hedge]

Hi 👋

[faint vine]

What’s up @neat axle ?

[blissful glade]

Hello so how do I get started with this?

[hybrid hedge]

It's not available yet.

[supple violet]

wait untill Christmas and new year around

[blissful glade]

Oh ok

And to make this clear. Leonardo helps with any asset?

[neat axle]

hey wsup Exco!

[neat axle]

at the moment, it helps with 2D assets. Our ML team is researching the 3d aspect of it

[distant tide]

Currently 2D, but yes any kind, not just limited to game assets 🙂 If you have any ideas let us know and we can try and get some pre-trained models ready for you at launch!

[sinful holly]

It’s a new day

[lost ember]

is it coming to 3D too?

[analog nebula]

our team is doing R&D on that basically 🙂

but for the initial launch it will be 2D only

[hybrid hedge]

Sounds like a good start 👌

[analog nebula]

Thanks! It'll be coming out very soon, though invites will be sent in batches

[blissful glade]

I mean it will do for now

[brazen marlin]

2d assets are prettier imo, very glad leonardo is starting with 2d ❤️ maybe it's personal preference, but 3d assets tend to look... funky, even with thousands of dollars poured into them. very exciting that alpha is soon !!!! you guys are doing fantastic work ❤️

[soft sparrow]

seems like the workflow is just going to end up being generation of 2d images from various angles, then pipe it into 3d generation

[distant tide]

Maybe initially as a sort've psuedo-3D if the results are solid. However, we'll definitely be allocating R&D resources towards solving that problem. One thing that is for sure though, we definitely don't want to be releasing anything mediocre, so when it does come out, best believe it'll be 🔥

[olive crow]

hello everybody!

[modern mural]

yo!

Leonardo will be my cheat code for my DnD campaign.

[uncut panther]

when do you think leonardo could be truly efficient in making ai-generated 3d assets?

[limber cave]

So how do we sign up for the alpha?

[paper yew]

Through the website.

[undone gust]

Is it built on top of 1.5 or 2.0?

[analog nebula]

@distant tide or @neat axle are probably more capable of answering that, I'm mainly on the community side of things 🙂

[frosty heath]

Morning all

[analog nebula]

Morning! 😄

[distant tide]

Hey! Yep initially we’re building on top of SD with our own custom training and render pipelines which our ML guys have been hard at work perfecting and will continue to iterate on. The platform we’ve built is super modular and scalable which will allow us to integrate new models very rapidly as they come out 🙂

[undone gust]

Thank you for responding.
What version of SD are the models built on?

[lavish jacinth]

Morning

[analog nebula]

morning 🙂

[lavish jacinth]

I am a motion designer from brazil and I would like to use "Leonardo" for creative concepts, but I would like to do it using the client's logos, is that possible?

[analog nebula]

possibly, if a model can be trained on logos

but I'll leave it to the experts to give an better answer 🙂

morning Jo!

[sinful holly]

Slept early. Woke up and made some snow globes to fill time

[neat axle]

Yes, you can train it on your clients’ logos or anything you’ve designed

For example, I trained a model on my mom’s handmade sweaters

So it also works for other verticals other than game development, such as fashion, logo design, etc

[cold fox]

Hello

[deep shoal]

Hi @cold fox

Welcome!

[cold fox]

So i was thinking

I was gonna use leonardo for a garden warfard fangame to create concept art

İs it adapted to pvz style games ?

[deep shoal]

There will be a number of pre-trained models on Leonardo, some of which could be used in that context. Alternatively you could train your own, or try out models trained by others on the platform.

It really comes down to the style you want

Early access will be free, and once we bring in paid plans we will continue to have a free tier

[cold fox]

Me omw to insert 2837839282 pvz images to Leonardo:

[deep shoal]

😆

iOS to start, Android after

Yes, based on Dreambooth + we're using our own custom render pipeline and there's a big feature set being bulit out for the front-end.

We'll also have a load of pre-trained models, and there will be community models too.

ETA = before Christmas

Have not finalised the intakes yet

We'll likely start slow, make sure everything is working, then ramp up.

Their loss? 🙂

First in

Although we may also give some users priority early access as a thank you if they are particularly helpful with respect to the community / raising awareness and that sort of thing.

[sinful holly]

Ohh an iOS app. That’s great. Lately I’ve been feeling a lot of developers forgot about us for website or android

[hybrid hedge]

Will there be a limit of how big the resolution of an image can be? Will it be 500x500 or larger?

[paper yew]

IOS deserves to be forgotten about tbh.
I hate how they treat developers.

You are forbidden to emulate Mac on a non-apple machine.
So you're forced to buy a mac to build apps for mac / IOS.

You're forced to use their store, and they also take a 30% cut from any revenue you make.

[distant tide]

You'll have the choice between training a model at 512x512, 768x768 and potentially 1024x1024. We've also integrated some really cool smart upscaling features where you'll be able to get generations up to 3000x3000

For sure!! Honestly the possibilities are endless, we're so keen to see what the community does with this, we're sure there are a bunch of use cases which we're not even aware of 🙂

Yeah very well may be possible, we haven't extensively tested this yet. But with the right training data and parameters it should be doable

[azure igloo]

Hey what's up, I just joined

[fiery nacelle]

How do I use this? I haven’t found where to actually try the platform

[fiery nacelle]

Saw the faq after I posted question lol smh

[slim snow]

When does the test start?

[unborn moon]

hello guys

[stuck garnet]

hey guys

i’m a developer this new ai is going to help so much

is the ai capable to generate 3d models at a low poly count

[deep shoal]

Not yet re 3D unfortunately. 2D to begin! We are working in this area though, as are all the big ML brains around the world.

One of the main challenges is training a 3D model, which means obtaining a lot of high quality 3D models for the dataset.

[formal wedge]

Hi im not a programmer but i like design so im super curious about all of this!

[charred kraken]

I have been using midjourney for a while. And LeonardoAI's ads looked promising and so here I am. Can't wait to test 🙂 also I am a graphic designer, the kind who is not afraid of AI. Infact I welcome it. It's gonna be very useful. But I have already started to dab in other feilds for my future, just in case 🤣😉

[hybrid charm]

Whassup

[ivory kayak]

I have very little experience with AI and modeling in general, but this seemed interesting, so I thought I’d check it out.

[analog nebula]

planned to be out this week, invites will be sent in batches, to whoever signed up first

but they will make exceptions for active community members 🙂

Also welcome everyone who joined recently! Nice to have you all on-board 😄

[hybrid hedge]

Can't wait to try it out, been waiting for months! 😄

[ornate ore]

Hello! Insterested of new stuff like this right here! ✅

[analog nebula]

Welcome! 🙂

[ornate ore]

Need to learn what this is capable atm.. but fast question first; avatars?

[analog nebula]

a model can be trained on that yes

so just feed it a bunch of samples of what you want

we'll have an easy to use interface just for that

[ornate ore]

Thanks.. i'll check it when i get back to my cpu

[analog nebula]

currently we're dealing with 2D image generation. 3D is in R&D 🙂

Oh we aren't out just yet

but the alpha invites should be going out soon, we plan to launch it for testing this week.

If you've signed up pretty early you will most likely get an invite, else they'd probably be sent another time as we ramp up the number of users

[ornate ore]

Early stages then.. 🙂 I'll just hang around and wait what you come out with

[analog nebula]

yeah, stick around, watch how things evolve with us 😄

[ornate ore]

So, Wen??!! 😆

Nah, i learned patience with start-ups

[analog nebula]

haha

just startup things? 😛

[ornate ore]

What's the difference between start-ups and things that are starting up?
Not my first language 🤫

[analog nebula]

🤔

well I mean you already kind of have to have something in order to start it

like an engine for example

[shadow lantern]

Super existed for this.

[analog nebula]

Same here 😄

I can't wait to see how the platform turns out. I've had a bit of input in the site design but I haven't seen the built version of it

[slim snow]

OK，

Looking forward to your products

[slim snow]

Have you guys considered getting into the blockchain?

How to qualify for early testing

[analog nebula]

Not part of the plans as far as I know

People are free to use it for blockchain projects, but we have no intentions of getting into blockchains 🙂

[analog nebula]

If you've signed up on the website (https://leonardo.ai/) you should be all set. Invites will be sent out on a first come first served basis

We will be sending them out in batches as we expand the userbase

However; we can make an exception for active community members

[distant tide]

Very easily, get keen!

[distant tide]

So excited to see the built version aye, not long now 😀

[hybrid hedge]

Glad to see a community that isnt spamming for access 🙂 Good job everyone.

[analog nebula]

❤️

[analog nebula]

Definitely lovely to see 😄

[sinful holly]

I fixed my sleep schedule (maybe) it’s still early

[shadow lantern]

this looks very cool

[analog nebula]

Thank you! And welcome! 🙂

[meager summit]

Any other Members here who are using Flutter in some sort?

[frank wedge]

Good evening all

[formal karma]

What are the types of files you’ll be able to generate from this?

Is it still all PNG/JPG

[hybrid hedge]

My guess is JPG/PNG from experience with other models

[fast tiger]

That's right! 🙂

[round oracle]

Hey guys!

Good to be here

[wheat silo]

Howdy

[fast tiger]

Welcome, guys!

[meager summit]

👋 👋 👋

[wheat silo]

@fast tiger Has your team explored the use of AI derived heatmaps for procedural terrain/biome generation/management? (moved this question from introductions... didn't seem to belong there)

[fast tiger]

Interesting question, no, we haven’t, but the team is now aware of this idea. 🙂

[wheat silo]

https://tenor.com/view/youre-welcome-gif-21319408

[jade hornet]

so how can I check asset generation atm

[fast tiger]

We’re not open yet, we’ll be sending out the invitations in the upcoming days. It will be done in stages.

[sinful holly]

I scrapped up a concept idea. Now just waiting for soft launch to see how it renders out

[wheat silo]

@fast tiger are you guys just doing image generation, or are you developing for other types of AI usage?

[distant tide]

Initially PNGs, however we're exploring vector formats as we speak, looking to do SVG next

[unkempt dawn]

curious about this too - messed around with riffusion the other week, which basically just is img2img or text2img of audio spectrographs, and then converting the spectrograph into audio files. but the model was trained on rhythmic sound iirc, so i had to go by adding 'music' to the negatives to get anywhere to usable territory. i could see this work out very well for creating foley/sound fx work, if learning set contains these spectrographs rather than music and beats 🤔

[distant tide]

Definitely going beyond just image generation, we've built up a solid team now and have great development velocity. We'll be including generative text at launch and will continue to iterate on it. And as we go we plan to introduce generative audio, video and 3D as well all in a cohesive manner.

[distant tide]

Riffusion is on the radar! And the application of spectrograms for game fx is something we're looking into 🙂

[wheat silo]

If you get into NLP, can you make a data set that has anachronisms removed 😛 Hard to get an ai to work with dialogue when it wants to talk about things that are period inappropriate 😛

[wheat silo]

From a dev perspective, being able to integrate AI easily into non-art related parts of the dev pipeline and grunt work (animation rigging) would be as useful, if not more so, than more art generators.

[polar maple]

meowwww

hewwwooooo

what is this thinggggggggg??????

[midnight cloud]

New tweet is fire.

[neat axle]

haha beat me to post it here 🙂

[analog nebula]

Sweet

[midnight cloud]

Haha I always retweet anything tweeted from Leonardo.

[analog nebula]

Can you scale the bit size

I want more sliders

[deep shoal]

Scale the bit size?

[analog nebula]

Oh like say

The pixel grid

So you can make it even more pixelated

Or more defined

[deep shoal]

Mm, you could do that in post. Or you can train a model on your own dataset!

[polar maple]

meowwwwwwww

i will paw youuuu

explain me thissss thingggggg

[sinful holly]

Oh those are nice

I been playing with realism so I wonder how close to unreal I could get

[polar maple]

meow

[deep shoal]

Cats or cat people, either way

[vocal osprey]

Can it do other kinds of pixel art eg: houses or items or is it limited to just characters?

[deep shoal]

We're working on training more models, but you can also train your own

[vocal osprey]

That’s awesome

[deep shoal]

If you have a dataset you're good to go. If you want to DM me one I can run a test training job too.

Only need ~15 images

This model was trained on 15 fantasy characters, but you can deviate quite far from the training data

e.g. aliens, no problem

[vocal osprey]

Nah can’t rn not at my pc but I’ll try train my own model when it releases

[analog nebula]

closing in on 1000 members

😮

[polar maple]

are you a bear?

[analog nebula]

yes

very much so

[polar maple]

you are scaryyyyyyyy

[analog nebula]

unfortunately yes

but helps with my role here

[polar maple]

i will paw you

and hissssssss

[analog nebula]

please don't

[polar maple]

i want color role

why it is green?

[analog nebula]

because the founders decided it should be green I suppose

I don't set those things around here 🙂

[polar maple]

gimme a different color

[analog nebula]

also no we barely have color roles honestly

[polar maple]

or i will hiss

[analog nebula]

Well you'd have to make do with the green

we only have 3 roles that have colors right now

[polar maple]

😭

[analog nebula]

2 of them being staff roles

The team probably wants to keep things simple for now so we don't really have vanity roles that have different colors

[polar maple]

you can use icons instead

for that

what is this AI for?

[analog nebula]

generating art assets mainly but it's pretty versatile so can be trained to make a lot of various things

[polar maple]

can it make 3D things?

[analog nebula]

Not for now, but it's undergoing R&D 🙂

[polar maple]

😭

the ad i saw was showing 3D things

[analog nebula]

it'll be only 2D initially

well 3D looking things

actual 3D is still a while away

[polar maple]

ummmmmmmmmm

[analog nebula]

but we're doing R&D for sure

[north arrow]

tbh same, but im still excited to watch the progress 😄

[deep shoal]

Some of the images we're showing on the website are 3D-looking, but they are 2D isometric images to be clear 🙂

As @analog nebula says - 3D is a work in progress. The first thing we'll probably be able to achieve is generative textures for existing 3D models. Would that be of interest to people?

[analog nebula]

I believe we have 1-3 people in introductions who have expressed interest in that yeah

and I made an suggestion as well which has votes

so definitely something of interest

Personally I think texture generation would also be of interest to concept artists who can throw them onto 3d models/blockouts for a base layer to paint on top of

bonus points if you guys can figure out texture generation that outputs the various maps for shaders

would make material creation so much more procedural in ways

than just relying on quixel maps

[unkempt dawn]

hmmm there is the depth2img model. something like that could perhaps help with adding bump mapping shaders or whatever's the thing these days

not sure how well it'd do when resolving image depth of a flat texture, but maybs that aspect can be trained too

[analog nebula]

normal mapping too

[tough cedar]

I've been following tweets and reading in here and I haven't seen it exactly answer (tbh, probably something worth putting in the FAQ) and thus I have to ask:

A lot of people are anti-AI art due to the less than ethical sourcing of training data, how is the data for this model been sourced? Has it been trained on on artwork with artists permission/open sourced assets?

Sometimes this question makes me feel like a Karen "Um excuse me, it this AI Ethically sourced I only use Ethically sourced and free range AI"

[mild walrus]

The upcoming stable diffusion 3 model will be trained on a dataset that allows artists to opt-out of. That said, it is ultimately up to the users to decide what they would like to finetune on, but we are working on database of royalty free images to select from when building your training set

[meager summit]

Will der also be a way to generate Sprites? Like Charachters for such isometric game?

[analog nebula]

You can basically train Leonardo on an dataset based on what you want, and it will be able to generate sprites too 🙂

It's just really about having the right dataset of what you want. The UI is fairly simple to use ^^

Hopefully the sprites will be consistent enough, we'll find out as we improve on things!

[analog nebula]

Extremely

[meager summit]

Sounds very promising. Is there some sort of Whitepaper to understand a little bit about how your Models learn? - Maybe if its more understandable its easy to get a consistent style out of it.

[primal fiber]

Hey

So how exactly do I use the AI?

Is it launched yet?

[vocal osprey]

It’s not currently released yet

[primal fiber]

Alright

[vocal osprey]

I believe it will launch early alpha some time this week. Correct me if I’m wrong

[primal fiber]

Possibly, they have mentioned it's gonna be released during Christmas

[analog nebula]

expected sometime this week yes, before christmas

[vocal osprey]

Oh this server is growing super fast

5 new members in like 10ish minutes

Is there currently a way to generate models that we can download with Leonardo or are we limited to 2d isometric images?

[analog nebula]

2D for now, 3D is still undergoing R&D

[vocal osprey]

Alr thanks

[bold halo]

Hey ho

[low shell]

I'm in

[jolly raven]

Hello, can it create some 2D tiles set like a blob of grass ?

[low shell]

We'll find out soon enough. They are saying first intake is this Friday for alpha.

[analog nebula]

you can train it to 🙂

[analog nebula]

hey Draz!

[low shell]

Hey Ayu 🙂

[analog nebula]

Welcome!

[low shell]

Thank you 🙂

[analog nebula]

👍

[low shell]

Good to see you in here. Not busy enough in your other job...haha

[analog nebula]

did you sign up early?

[low shell]

I think so. I DM'd as early as I could

[analog nebula]

Oh yeah, well you know mindspheres 🙂

[low shell]

The output from this is fantastic

[distant gull]

how do i use the ai?

[analog nebula]

well you can always just drop us an message I suppose

[distant gull]

i sent an email and it still isnt responding

[analog nebula]

it's not out yet, but will be an web platform based service

[distant gull]

👍

[analog nebula]

invites have not been sent out but the alpha of Leonardo should be out within this week 🙂

First come first served basis though so hopefully you're early enough!

We can't scale up the userbase on launch unfortunately since it's still in early phases

But we will definitely be expanding gradually

I think artkin isn't deeply involved in this

[sinful holly]

Now the sun is up

[analog nebula]

did you manage to fix your sleep schedule?

[sinful holly]

No

Woke up too early

[thick cove]

Anyone can expand me about this ai cause i am new here

[analog nebula]

website is pretty self-explanatory (https://leonardo.ai/) but basically what binger has said. 🙂

In addition you can create models off datasets you upload to it, and basically generate whatever you'd like.

[limber cave]

So the term floating around for people making AI images is synthographer / synthography. What do you think of that term?

[neat axle]

I like it tbh

[hybrid hedge]

Meh, whats wrong with AIrtist 🤪

[slim snow]

How to get the test qualification?

[sullen stump]

Will this ai have an ability to generate 3D assets?

[analog nebula]

There's none. It's a first come first served basis. Exceptions can be made for active community members however.

[analog nebula]

We're starting off with 2D only features first. 3D capabilities is something the team is looking into.

[sullen stump]

Oh awesome! I was also thinking about it, and how would be approaching it... lmao. I would imagine the dataset for it should be huge

[analog nebula]

Perhaps. We'll have to find out 🤔

[sullen stump]

Yeah...

[crimson marlin]

How do I actually use the AI?

[analog nebula]

It'll be web platform based, but it's not out quite yet. We are hoping to launch it within this week and invites will be sent out on a first come first served basis. 🙂

[crimson marlin]

Oh, ok

Thanks a lot! Looking forward to it

[analog nebula]

and thank you for your interest 😄

[crimson marlin]

I can't wait to up my world to the next level of immersion

:)

[analog nebula]

Team's doing some magic here, so can't wait to see what people get out of Leonardo ^^

[slim snow]

When will it be officially launched?

[analog nebula]

Not yet known. But the alpha is expected to launch within this week.

[limber cave]

remember back in the day (earlier this week) when there was only a few hundred people... gosh i miss the good old days.

[analog nebula]

yeah it was a lot more cozy in a way

[limber cave]

is there going to be a private/public like mj or what do things default as?

[analog nebula]

there will be a free tier

and paid plans of course

if that was your question 😅

[limber cave]

nope, it was more like the privacy settings of what we create using Leo

[limber cave]

also, is there any talk/ability about making transparent pngs?

[deep shoal]

During the early access phase, there will be a couple of tiers. Similar to MJ, the base tier will have all generations and models that you fine tune public.

The idea is to promote collective learning as to what works well and what doesn't with model training and generations against those models.

There will be higher tiers which we will give free access to for specific users who apply (and meet certain requirements). In those higher tiers, there will be the ability to keep generations and models private.

[deep shoal]

It's already built in!

[limber cave]

awesome, thats good to hear

[minor stag]

Will leonardo output images or 3d objects?

Sorry if it's written somewhere and. I missed 😦

[analog nebula]

2D images for now, our team is still looking into 3D capabilities

[minor stag]

Can they be rotated?

top left right bottom views?

[analog nebula]

the 2D images?

[minor stag]

yea the generated images

like a turntable

[analog nebula]

most likely not

you'd have to state it in the prompts, but consistency won't be the same

[minor stag]

I see i see

[analog nebula]

But I'll leave it to the others to reply

[minor stag]

Well a couple of more papers down the line.. 😄

[analog nebula]

Perhaps! We'll have to see if the developers have something up their sleeves. I'm mainly on the community side of things 🙂

[analog nebula]

Thank you for replying ms 🙂

[minor stag]

❤️

[sage nimbus]

how many free renders?

[analog nebula]

Will ask the rest of the team on this!

[minor stag]

None! 😄

Nah kidding

[shy atlas]

gm gm

Does this software do 3D models ?

[distant tide]

Still to be finalised, but safe to say we're going to be quite generous 👀

[minor stag]

As I understood it won't

But maybe a couple of months and it will 😄 As fast as things are going with AI these days

[analog nebula]

Not at the moment. It'll be 2D only initially but we are doing R&D on 3D capabilities. 🙂

[neat axle]

We've been training it with pixel art and it has thrown great results

[hybrid hedge]

How limited is it when it comes to which words i can use? Like death, corpse etc?

[ornate rivet]

🔥

[languid sonnet]

qq all

[hard galleon]

I like sharing knowledge as it's worth more than money, so I figured I'd post this here for people as well.

though --ar 16:9 should be changed now for m4

least it'll tell you that tho

Share it to your friends, i don't care, I wanna see a beautiful world.

And putting things into simple terms for people helps that.

The real secret is in the photoshop tips, as that's all you need to keep baking.

[limber cave]

they will ban you for making multiple MJ accounts if they catch you and you can't use any of the free ones for commercial use

[hard galleon]

None of what I do is for monetary purposes 😗

And my accounts paid for!

I'm just a hacker with an unusual MO.

World sucks, I want it to be better.

So, I try to use art.

I have philosophies that go into it that I can't fully explain otherwise I give away the wrong kind of power to people undeserving~

But there's more to it than just art.

Art is culture.

https://www.youtube.com/watch?v=pC3IrqUpm9U

I've been compared to Beggar Vance 😛

[fiery cape]

How do I create images?

[limber cave]

you have to wait till alpha opens up, see #📌│faq

[timber raven]

Love the look of the pixel art!

[fast tiger]

Once the platform opens you’ll be able to generate images through a text prompt.

You can also create images by training a model with your own images. 🙂

[ornate rivet]

Thanks mate!!

[hard galleon]

Never a problem! ❤️

[limber cave]

@timber raven zomg using AI images in games?!

[hard galleon]

heck you should see what kojima uses AI for

hes my fave

certainly gave me a spoop

[timber raven]

@limber cave I've already done it for a couple of my board game prototypes.

[limber cave]

Awesome 🙂 I've just seen a lot of backlash on the boardgame design fb groups about using ai images

[timber raven]

Yep. No comment.

[limber cave]

and none needed. Everyone who uses it, uses how they see fit.

[hard galleon]

If ya'll need a way to defeat any artist whining about AI

Just tell them

You view art as a commodity, I view it as a message.

[minor stag]

And 30$ is not much for almost infinite GPU time

[strong sentinel]

Hey there, quick suggestion: An important part of production ready assets is transparent background. MJ and the likes cannot handle such transparent background. Maybe you want to try and add such functionality to this model

[minor stag]

MJ v4 can do it I think. But I may be wrong

Will Leonardo use MJ underneath?

Asking cause mayube a discount could be offered if you have MJ played, and use the users MJ account to generate. But it's maybe a stupid idea :V

[hybrid hedge]

MJ isnt public

[sinful holly]

I use Stable Diffusion, MD is almost fail proof but there is not a fine tuning

[vocal osprey]

Does anyone else here use chatgpt for things?

[deep shoal]

This is already in 🙂

As above, we've already implemented background removal in the platform 🙂

[deep shoal]

We use Stable Diffusion as a base for the fine tunes.

[deep shoal]

GPT-3 will be built into the generation interface so you can ideate rapidly without leaving the platform 🤓

[hybrid hedge]

Sure! Ive started using it in my daily work for different things like correct code, do better sendouts and so on

[deep shoal]

Like this

along with auto-batching of prompts if you cbf and wanna go make yourself a coffee while it generates

[ornate rivet]

can't wait!!!

[vocal osprey]

That sounds awesome

I also recently realised chatgpt writes code for you so I’ve been messing around making small Unity projects and blender scripts with it.

[vestal moss]

can't wait to see what the ai is capable, I will make great game with your assets

will the ai will be able to generate animation too ?

[vocal osprey]

Oh that would be awesome if it did

Although I’m not sure yet since it doesn’t make 3d models and only does 2d isometric images. I believe 3d is in R&D so maybe we could get animations along with that.

[vestal moss]

well you can have 2d animation juste mutliple assets, like a sword swing that can countains 4 or more images to have a good quality

[vocal osprey]

That’s true

[ivory tree]

this ai looks mega

what type of machine learning does this ai use ?

[deep shoal]

We're testing out sprite sheet workflows, but nothing conclusive yet on animations. It's a process!

[deep shoal]

Pretty permissive tbh. Nothing illegal obviously, but we're not going to go full MJ and nanny you all. There will be a NSFW filter turned on by default that you can disable.

The only proviso is that we may limit NSFW generation to higher tier plans (free during early access for those who have Pro) due to moderation overheads.

[hard galleon]

@deep shoal NSFW filter is all I needed to hear. ❤️

[analog nebula]

As a blender user, I am loving this

[hard galleon]

Thought ya'll might! ❤️

[analog nebula]

I hate texturing as a 2D artist lol

[hard galleon]

I hate texturing as a non artist!

😄

[analog nebula]

Looks to be really useful for concept art and illustration

[hard galleon]

Someone in comments mentioned LOD

[analog nebula]

Which is what I do

[analog nebula]

Lod and uv mapping

[hard galleon]

Yeah

[analog nebula]

Also retoplology

3D artist Nightmares

Those are the things that should be automated

😅

Oh and rigging

I don't think I've met an 3D artist that have enjoyed any of those

Unfortunately not within the realm of what Leonardo will do. But perhaps in the future something will come along. 🤔

[hard galleon]

I mean, we've made it this far already.

I could imagine Gait sensing tech getting mixed in

[limber cave]

I’d be interested in hearing more about the pro tier when it gets discussed

[fast nest]

@analog nebula maybe this will help? https://www.youtube.com/watch?v=1kV-rZZw50Q

it basically learned all the movement by itself

[analog nebula]

Yeah the team will talk more about it when they get to the stage on figuring out the details for paid plans 🙂

[analog nebula]

Ah two minute papers

I watch that channel from time to time. Thanks for sharing! 🙂

[fast nest]

you're welcome! have a great evening.

[analog nebula]

Kind of forget it exists sometimes

😅

[analog nebula]

You too!

(it's morning for me actually)

[rose prism]

😯 Didn't think of using AI to create game assets that would be quite handy

[analog nebula]

definitely really useful, especially if you just need 2D assets. The team's researching 3D generation but we'll be 2D only initially 🙂

[rose prism]

Sounds cool 😎

[analog nebula]

👍

[thin void]

Hi, Just dropped in!

[vocal osprey]

welcome, hope you have a great time here!

[pallid birch]

Hi everyone Im new

Is this Ai able to procedurally generate worlds based on your inputs? Also can it take reference images?

[vocal osprey]

Hi, enjoy your stay. not sure about the first one. But yes you can train your own models to make assets simmilar to what you want by inputing refrence images

[pallid birch]

cool cool how do I start?

are there any youtube tutorials available?

[vocal osprey]

i belive closed alpha starts some time this week and access will be given out in waves

check #📌│faq

[pallid birch]

ty

[analog nebula]

Yes, you'll be able to train the AI with datasets to create models that you can use to generate your ideal outputs

Welcome by the way 🙂

But yeah the process will be fairly simple with the clean UI that we've come up with

[pallid birch]

Thank you, btw do you have rewards for people if they review your product on social media?

[analog nebula]

Perhaps! @deep shoal might be better suited for this question however 🙂

[analog nebula]

What leemo has said basically (thank you leemo!) We're expecting to launch an by invite only alpha this week

It's on a first come first served basis but we will try to make exceptions to active community members

[pallid birch]

awesome I will be waiting eagerly to use this with unity game engine for a horror game I am making with a team of people. Our artist went no contact for some reason so this might make or break us lol

[analog nebula]

Oh that's really unfortunate, hopefully you'll be able to get what you want out of Leonardo, though for now we are 2D only. Hopefully that's suited for your needs!

[pallid birch]

Btw if you guys need funding I recommend having Premium Pass ie pay 500 or something and get like extra features or premium priority on issues you have. Maybe its available to founders ie first 1000 people to purchase that pass etc

[analog nebula]

Oh we are definitely looking into paid plans 🙂

but not during the alpha

[pallid birch]

sweet, all the best 🙂

[analog nebula]

Thank you! Same to you and your team!

[hybrid hedge]

Alright sounds smart, then I want the highest tier 😌😌

[vocal osprey]

Will the paid plans be a once off or subscription?

[analog nebula]

Most likely subscription based. But we'll have to see what Midas or mindspheres say.

[vocal osprey]

Fair enough

[analog nebula]

👍

[deep shoal]

Yes, 100% - product reviews, viral threads, etc - during the early access phase we'll give you freebies and/or access to pro tools depending on the level of engagement you achieve when spreading the word.

[deep shoal]

Subscription. We've got a team of 18 people and growing, and infrastructure ain't cheap 😆

[pallid birch]

Sweet thanks 🙂

[deep shoal]

It's important to restate again what we're building here - this is a platform not just for 2d asset or even 3d asset generation.

We see it as an entire content production suite that incorporates all sorts of generative AI on one platform, and links them together in a logical and cohesive way.

You could ideate on narrative and game design, generate character descriptions and push them to a character image generator. Generate dialogue then matching natural audio to go with it.

As tech progresses, we want to be at the forefront of integrating all the fun and interesting things that are coming out. This is why we have such a big team, and a core ML/AI unit of absolute gurus.

We're also not just thinking about gaming but multiple other areas including marketing, fashion, product design, and so on.

The paltform is very flexible and can accomodate all sorts of uses. It just so happens that we come from a gaming background and that makes the most sense in terms of our principal initial focus 🙂

[vocal osprey]

You should put this in the faq or some sort of Infochannel so people avoid asking the same questions again and again and gives a clear understanding of its many use cases.

[deep shoal]

good idea

[vocal osprey]

It’s cool that your bundling all the different ai tools together in one product because it gets really complicated when you have to work with multiple different softwares with different levels of quality

[pallid birch]

It would be nice to have an Ai that can do teasers, trailers and possible even auto upload to youtube/tiktok/reddit/instagram and then do analysis on what works etc

[deep shoal]

It's also that a lot of the stuff out there is in the realms of enthusiasts only, and not really accessible in a production context.

[pallid birch]

also a customer support ai maybe, also something that finds bugs in games

[deep shoal]

We're not looking at AI video at the moment as there are a few players in that space, but it's something to think about 🙂

[hybrid hedge]

ChatGDP will replace Customer Support real soon

[analog nebula]

One more for the FAQ :>

I'll add it. Thank you for the responses MS!

[analog nebula]

bots have already been doing that tbh

Soon I shall be replaced too

[hybrid hedge]

(Chat)bots havent had the real ability to have discussions, never felt "human" like ChatGDP does! Basically a basic ML model.

But now that ChatGDP can save conversations and be able to continue them when ever is a real game changer

[analog nebula]

Yes

Though empathy would be nice

Some human based things probably still difficult to replace to be honest

[hybrid hedge]

Ofcourse!

[amber wyvern]

I find ChatGPT useful as a rubber duck while ideating, as it can give "feedback"

It doesn't have to be true or accurate, just agreeing or disagreeing with the bot is enough to make up your own mind (imo)

[shadow lantern]

I'm here because of the ad I saw on Reddit, I'm the owner of Unstable Diffusion, I love your website and what you guys are doing.

And I hope I'm just the first of a wave you're about to receive

I think a service like this could go a long way towards spearheading AI adoption in graphics, I look forward to seeing what you do

[steady vine]

hey

[wild sleet]

Hello all. Thought I would give this a shot.

[umbral jungle]

waves a hello.

[analog nebula]

Welcome Eray, Vortyss, Paul. Nice to have you here. Feel free to ask if you've any questions 🙂

[umbral jungle]

Saw an advert on Reddit - and was amazed at the asset creation (isometric buildings) and can't wait to see more 🙂

[analog nebula]

Thank you for your interest and kind words!

[umbral jungle]

Also feel like the style is going to fit in very well for my side-hustle project of the moment 😄

[analog nebula]

You'd be quite pleased with the default models that Leonardo will have then 🙂

[umbral jungle]

I'm most interested in whether creation of the assets can be built to have the same style - my experience with AI based generation (not for games) so far has been somewhat hit-and-miss 🙂

[umbral jungle]

Loving the team interaction here as well 😄

In a few weeks you'll be overwhelmed with too many people here to talk to everyone! 😉 🙂

[analog nebula]

I think by feeding Leonardo with the assets of the style you like you can basically train a model to output things gradually closer to your liking

Easily done with the UI

[analog nebula]

Perhaps! We're still preparing for the alpha launch. Hopefully within this week.

[umbral jungle]

Can't wait - but don't rush 🙂

[analog nebula]

Probably might get hectic for a bit

Agreed ^^

[umbral jungle]

Question: For the isometric assets.. can you also generate normal maps? 🙂

[analog nebula]

We definitely have the ai decently ready it's probably mostly just front end and some logistics

[analog nebula]

At the moment no, but we do get it in suggestions and I do hope the team would add it!

I dabble a little with 3D so it would definitely be nice

[umbral jungle]

I mean, it's not neeeeddded but is a nice-to-have.

[analog nebula]

Nice to haves are something we can definitely try to include

#💬│general message

This actually is a pretty good run down of where we aim to take Leonardo

[umbral jungle]

Looks good, can't wait to play 😄

Do you have pricing laid out anywhere yet?

[analog nebula]

Not yet, alpha will be free. The price plans are still being decided, but there will be a free tier. 🙂

[umbral jungle]

nods

Awesome to hear all round - can't wait to see what we can all created.

[nocturne temple]

can i send memes in general?

[south edge]

👍

[analog nebula]

#💭│off-topic might be better. And keep it within reason please, nothing offensive 🙂

[shadow lantern]

@analog nebula I don't know if this is the place for questions, but how would you feel if a AAA developer approached you? Do you think your service would be up to the challenge yet?

[analog nebula]

Could be better answered by the other team members, I'm mainly handling the community side of things 🙂

[shadow lantern]

The potential for cost-savings makes services which do this worth potentially millions

As we all know, AAA game development has margins in common with blockbuster films and launching rockets, in other words, very risky.

But yea, I undertand ~

[analog nebula]

Yeah AAA projects are quite a sink or swim matter. Typically larger studios are able to deal with the occasional bad release.

[analog nebula]

It definitely would. But I'll leave it to the other team members more familiarised with the AI itself and our direction to answer your question ^^

[shadow lantern]

No of course, I understand, it would be awesome tho and I see the tech headed that way.

That's why it's so exciting to see something like this, people have been talking about using SD and similar tech to save money on game dev for a few months

[analog nebula]

Yeah we're definitely aiming to be more of a suite of tools for generative output than merely just game assets. 😉

#💬│general message

Best explained with this post by mindspheres

[shadow lantern]

Wait what do you mean? Oh my god.

Damn, I hope that's more than just dreaming, yea, that's awesome

Yea, integrated game engines and tool kits would change a lot.

[analog nebula]

We'll find out where the founders intend to take Leonardo, but it's definitely a good vision

and makes it different from the various projects out there

[shadow lantern]

How so? I mean what makes it different in your opinion?

I can't argue with the idea they seem ambitious about the right things

[analog nebula]

most AI platforms seem tailored more towards one specific purpose than an suite for creative output

[strong sentinel]

are you actually launching alpha this week?

[analog nebula]

planned to be launched within this week, yes

[shadow lantern]

Hehe, yes, that's kind of true.

[analog nebula]

Invites will be sent out when it is launched 🙂

[shadow lantern]

Very glad to be among the first in the Discord for an alpha as cool seeming as this.

[analog nebula]

Good luck 😄

[strong sentinel]

ok and another question, idk if this has been asked before but: Will we be able to use the images generated in our games? Legally speaking?

[analog nebula]

Yes as stated in the #📌│faq you own all rights to the images you have created with Leonardo

[shadow lantern]

Do you have a game @strong sentinel?

ChatGPT or 3.5 to help with some of the code, custom models for game assets...AI-assisted gamedev almost seems viable.

[nocturne temple]

does anyone know where i can use the leonardo ai?

[shadow lantern]

I'm 99% sure it's not available just yet

And we're the early arrivals

[analog nebula]

It's not out quite yet, it'll be web platform based, and an by invite only alpha stage initially. First come first served basis.

[strong sentinel]

yeah we are making Roblox games

[analog nebula]

Active (and helpful) community members may get exceptions however

[nocturne temple]

ah awesome, thanks for the quick reply 😄

[shadow lantern]

Sounds amazing ~

[analog nebula]

You're welcome 🙂

[shadow lantern]

Yea, if a few developers make significant parts of their game with AI-made assets, that's a ton of free advertising by itself.

I imagine that's a feedback loop which easily feeds into itself once started

[strong sentinel]

I have already tried to use MJ to draw simple stuff for me. The biggest problem, and I suspect this will be true for leonardo as well, is that it is not consistent.

All the assets are beautiful but they do not match my assets

[shadow lantern]

For that reason, it may be economical for Leonardo to especially give deals to certain notable devs just for the publicity

[shadow lantern]

What about img2img? And I believe I saw one special project which allowed for greater prompt editing.

Like, more specificity.

I don't know if they would've incorporated those techniques

[thorn nymph]

I am Interested in the Pixel Art Style that was advertized

[strong sentinel]

I have not heard about this, googling...

[thorn nymph]

This should be more conaistent

[shadow lantern]

I'll see if I can find it for you. The fact that you can order custom models should give some increase control, but I understand your concern there

[analog nebula]

We're definitely keeping an open mind about that

[analog nebula]

the twitter thread does give a pretty good summary of how it was done https://twitter.com/LeonardoAi_/status/1605052359826341894

[analog nebula]

by training a model on the specific style output you want I think you'll be able to improve on the consistency of the output

[shadow lantern]

This is what's possible

[analog nebula]

just by getting outputs and selecting the ones you like and then feeding it back

[shadow lantern]

It's also worth noting that OpenAI just open sourced a primitive text-to-3D , that kind of stuff is more down the road still tho

[strong sentinel]

code is closed right now tho 😦

[thorn nymph]

Crazy, i am still New this whole AI image Generation. I played around a Bit with stable defusion but could Not geht Something good looking Out of it, so I stopped trying

[shadow lantern]

Sometimes, you've gotta prompt engineer a little

Look up the prompts for good images, a lot of them have lots of keywords which some services, like Midjourney, use as "embeddings"

Which is one of the main secrets behind such services

[hazy gorge]

is there any exampel result of UI game, Pixel art weapon , Tile Set ?

[shadow lantern]

Keywords like "high quality, 4k, Masterpiece" are automatically applied to their outputs, which is one reason why images from those services look so great

[thorn nymph]

Makes sense, thanks i will be looking forward to trying some stuff Out

[analog nebula]

nope but all possible given the right training dataset

[shadow lantern]

If you go to Lexica.io you can find prompts, but the Stability Discord might be a safer bet if you're looking for good tips

A friend of mine also wrote a prompting guide, although it's outdated

[thorn nymph]

Thanks 🙂

[shadow lantern]

Hehe, you are so welcome, it's interesting that some people are just hearing about this stuff

[umbral jungle]

Is there an FAQ on how to make the right training sets?

[analog nebula]

Not at the moment. But the UI should be fairly easy to understand. Also it's really just about gathering the right images of what you'd like in terms of style/subject matter etc

There will likely be guides released as we go along

[umbral jungle]

Understood, thank you.

[lavish adder]

Enjoyed scrolling through the thread, interesting to see the process here. Seems like the UI being crafted is really designed towards user experience, which is great.

[analog nebula]

Yes, we've got people working on the UI (I've had some input into it too), and it is looking really clean and fairly easy to use 🙂

So it's not just about the technical side of things but making it an proper experience

[lavish adder]

Really looking forward to diving in and experiencing it.

Any plans for additional threads walking through the process?

[analog nebula]

We'll probably eventually create a guide of sorts!

[low shell]

I need more hours in the day to keep up with everything going on 🙂

This is all great stuff and very interesting to read about as mentioned above

[analog nebula]

Very much so

Mindspheres and co as usual

[low shell]

Exactly. Some very impressive stuff coming out of some very clever people. You're included in the clever people group too Ayu 🔥

[analog nebula]

Am I?

😅

[low shell]

For sure 🙂

Your skills are especially beneficial I would think on a project like this too

[analog nebula]

I do hope so! I am rather uncertain if I'm bringing much to Leonardo sometimes

[low shell]

You always had good graphical suggestions for Raini so nothing more graphical it seems than this. And you as an artist and having created these type of things in a more manual way, I would think your input would be invaluable

[analog nebula]

I suppose, but I'll have to wait for the alpha like everyone too for feedback on the ai itself

[low shell]

Cool. Looking forward to seeing it to. Past midnight here so off to bed. Take care 🙂

[analog nebula]

Goodnight!

[shadow lantern]

is the ai availible yet? If so where do i find the access link

[fast tiger]

Not yet, it will be available around Christmas (you can check it here: #📌│faq).

[analog nebula]

What Yoss said. Possibly within this week if we are on schedule 🙂

[hybrid hedge]

No rush, take your time, make it amazing 👌

[shadow lantern]

Alright, excited to see it 🙂

[analog nebula]

Absolutely. 😉

[distant pollen]

hello

[analog nebula]

Welcome to the discord!

[distant pollen]

hey thx ,,, what is this plz

[analog nebula]

AI service, currently 2D image generation focused but we have plans to offer more than just that

[distant pollen]

thx and it is free ?

[analog nebula]

There will be a free tier, but for now it is a free service during the early access phase (alpha testing), though it is by invite only. (https://leonardo.ai/) 🙂

^ on the website you can sign up for early access if you wish. Though invites will be sent on a first come first served basis.

[distant pollen]

haw can i start withe this , what is the command to enter prompts

[analog nebula]

Leonardo is not a discord bot, it will be accessible via an web platform, so though your usual web browser

Also it isn't out yet but is planned to be launched for early access within this week

[sinful holly]

Soon

[serene pike]

Hype!

[analog nebula]

Very 😄

[umbral jungle]

🤞

[quaint dirge]

I’m getting excited and not for Santa

[fast tiger]

This is the reason why:

[quaint dirge]

Oh yeah!!!

[summer sedge]

Hey thanks for this opportunity!! 😄

[fast tiger]

Welcome, @summer sedge! 😄

[summer sedge]

Thanks! the future is here, the future is now!

[fast tiger]

Oh yeah!
What's coming up is awesome...

[ocean jackal]

hello everybody

[fast tiger]

Hello, @tacit cairn and @ocean jackal. Nice to have you around. If you could, it would be great if you could answer the three questions in #👑│introductions. Please. 🙏

[craggy plank]

Hey everyone 🥳

[amber wyvern]

Would it be possible to generate images incorperating previously generated or drawn images.
Like making concept art of a scene with a specific asset which you already have?

Just thought of it, and realised that I don't know any tool that does so. (besides drawing it by hand)

[fast tiger]

If possible. 🙂

[amber wyvern]

If not, I guess I could photobash the layout and generate an image with that as a reference...

Anyways, something I shall try out 😄

[deep shoal]

Yes, you can use img2img with the models - this approach works pretty well actually 🙂

[sacred pecan]

Not sure if this goes here, but this is a recent ai collab experiment. I put in the original concept sketch as a reference (along with text), a few iterations gave me the second image that I then took and finished using Procreate. Would love to make this workflow a little smoother!

[deep shoal]

can you drop the initial image? I'll run a test now 🙂

[sacred pecan]

Sure thing!

[jade vessel]

I know in the FAQ it says we will own the rights to the images we make… does this mean Leonardo was trained only on images that permission was received for this?
How are you guys overcoming the whole “AI plagiarism” thing, otherwise?

[deep shoal]

Do you remember the text you used to prompt?

[sacred pecan]

"shadow archer wearing a hooded cloak and a quiver, 3D render, game asset, cartoon"

[deep shoal]

We aren't claiming ownership rights of generated images, we're passing on that claim to you. The base model is Stable Diffusion, which uses the LAION dataset - at this stage there isn't a permissions system for training data, that's coming with Stable Diffusion 3.0 is my understanding.

However, from a strict legal perspective, datasets fall under "fair use" - and this has yet to be challenged in court. Remember that the model learns from the dataset, it doesn't store the images it trains on, it learns kind of like we do. One good data point for reference is that the source data is something like 100TB while the model is only 5GB.

[deep shoal]

a few variations 🙂

@sacred pecan

Can likely do better when our canvas implementation is live

You'd be able to selectively adjust areas that it gets wrong

[hard galleon]

Hey just curious if anyone here has played around with Topaz Gigapixel yet, or if they needed to be introduced to it?

It's an upscale AI

[sacred pecan]

Very cool! I like how it caught and translated some of the details of the crossbow and the ground! It captured the floating hood as well!

[deep shoal]

We'll have built in upscaling on the platform 🙂

[hard galleon]

Dope! I still find it useful for upscaling things that didn't upscale how you wanted :3

Like if one of the prompts in midge looks cool but then durps out during upscale

I find topaz gigapixel just gives you what you saw

I also use it to upscale other things, like I upscaled the footage of the scranton strangler in The Office

and confirmed that is infact Toby, down to the license plate of his car

It's some unreal tech with lots of uses

[deep shoal]

Actually we have a few upscaling options, including a super clean upscale that doesn't change the image at all.

[hard galleon]

Aweeesome, love to hear it.

As it's my biggest peeve with midge sometimes

I just figured I'd suggest topaz gigapixel though for the toolkit, as it's got uses outside of art if you're creative

something something crack

yadda yadda haha

👼

[vocal osprey]

Gm

[hard galleon]

It's also really good for upscaling old game textures you rip I find too!

Like, I upscaled "THE ORDER" painting from SH

orig

Trying to see missable details that otherwise would just pixellate if viewed another way

[amber wyvern]

Nice, sound promising

[hard galleon]

I actually mostly use AI to solve mysteries heh

A-Eyes are quite powerful.

ChatGPT answered a long running question I had.

https://www.youtube.com/watch?v=hspUCKOZihU but i also do map dev and such

Long story!

[deep shoal]

Hey shot you a DM 🙂

[wheat silo]

Hrmmmm Could an AI map out game mechanics of a TTRPG based on the written rules?

[deep shoal]

Probably!

[wheat silo]

It would be an interesting test A, followed by test b which would be to optimize/balance them

[feral jackal]

Can't wait to try this app 🙂

[candid vine]

Hej guys

Incredible results produced by Leonardo ❤️ How to join alpha? Dropped my on the website

[fast tiger]

Hello and welcome.

It has not been released yet. We are still preparing to launch the alpha phase. We expect this week. 🙂

[midnight cloud]

Loving these outputs on Twitter.

[deep shoal]

https://twitter.com/LeonardoAi_/status/1605695041242271744

Thanks @sacred pecan for the input images 🙂

[vocal osprey]

Too anyone in the dev team what made you guys wanna start creating Leonardo?

[sweet moth]

Gm

[vocal osprey]

Good morning.

[lean carbon]

Has anyone even thought of an ai that makes a list of plans and goals to strive for like achievable goals I would love a program like that I'm only a novice coder tho soooo 🤷‍♂️

[vocal osprey]

You could try to get chatGPT to maybe write you some goals and plans for goals but im not sure how will it could work

I can test when I get back to my computer if you want

[quaint dirge]

What resolution can it output?

[digital lodge]

Hi

[fast tiger]

You can choose to train a model at 512x512, 768x768 and potentially 1024x1024. Intelligent scaling functions have also been integrated, allowing you to get generations up to 3000x3000. 🙂

[quaint dirge]

Nice

[vocal osprey]

Is something like this what you meant?