💬│general

[vestal patrol]

Lmao no you got it wrong

You wanna see my actual crush

[hard galleon]

duh

[analog nebula]

You just fell for her bait

[vestal patrol]

Idc lol

[hard galleon]

now im gonna make all the ai art of the rock and use it to control your world

im gonna deep fake dwayne into saying your name

[vestal patrol]

How do you know I'm not lying

[hard galleon]

and how much he loves you

[vestal patrol]

¯_(ツ)_/¯

[analog nebula]

this should really be in #💭│off-topic

[vestal patrol]

Nah it's ai

[hard galleon]

I mean we're using leonardo to take over justhavens crush so

🤷‍♀️

stress testing

[vestal patrol]

Lol realistically I don't have a celebrity crush

[hard galleon]

i do

hideo kojima

[vestal patrol]

My heart belongs to my boyfriend 👍

[hard galleon]

but thats obvious

[vestal patrol]

Actually no it wasn't until you told me lmao

[hard galleon]

don't look at my models then

[vestal patrol]

Am I not an early contributor

Or is that donating

Prob is

Idk anyways

[hard galleon]

i gave up two kidneys for this purple

all so ayu could play soccer again

[analog nebula]

I don't play soccer

[hard galleon]

you do now that you have my kidneys

[analog nebula]

😦

[hard galleon]

or.. or did i waste them D:

[vestal patrol]

I donated 15 kidneys

:)

[analog nebula]

For Leonardo invitation

[vestal patrol]

Yes

It was a bribe

[hard galleon]

yeah that was just so you could make the markiplier bot

you gotta donate 15 more

[vestal patrol]

Ayu what are you really doing with these kidneys

[hard galleon]

dress

[analog nebula]

trade secret

[opaque wyvern]

Why is the 'Seed' input box read-only?

[analog nebula]

Coming soon

[opaque wyvern]

ah ok

[analog nebula]

Yeah we got like a few UI elements on there that don't work since they are upcoming things 😅

[opaque wyvern]

would be nice to be able to type numbers into the textboxes too 🙂 (instead of drag sliders)

[quasi nimbus]

I don't know what it is, but each day I feel less and less passionate about A.I generated art... I'm losing something. It has made me realise I have much to learn about my own creative skills though. Anyone else feel this way?

[analog nebula]

Can relate somewhat to be honest. But seeing how certain subject matters can be interpreted in an old artist's style (especially one that has inspired a lot of digital artists I look up too) has been very fascinating

[quasi nimbus]

I think on that note, i'd be nice to challenge onself to create 'something' in somones style by research and so on, making it into a mini dissertation or project could be fun and interesting...

[halcyon ore]

What is Negative Prompt?

[opaque wyvern]

tells it what to exclude from the image

as you'd expect, it rarely works - I often try telling it to exclude shadows, but it nearly always adds them anyway 🙂 I guess it just doesn't understand what a shadow actually is

[halcyon ore]

Ah! Thank you

[analog nebula]

Basically as photostorm has described

[river holly]

So what kind of things do I use a dataset for? Training the AI to create the same character in multiple pictures by uploading a bunch of faces?

[opaque wyvern]

that's one possible use for it, yes

[river holly]

Is there a tutorial somewhere? Do I just upload a bunch of photos of the same person and give that person an identifier, which I can then use in the prompt?

[opaque wyvern]

there's a guide on the site how to use it

[river holly]

In which channel? FAQ? Because I'm looking for it right now.

[opaque wyvern]

on the site

[analog nebula]

https://app.leonardo.ai/faq

[opaque wyvern]

FAQ & Help

[analog nebula]

alternatively: https://sweet-hall-e72.notion.site/A-Guide-to-Getting-the-Most-out-of-Leonardo-AI-aae753df4cf64bc7bbda99790c8e448c

[river holly]

smacks head Thanks!

[analog nebula]

No worries 🙂

[halcyon ore]

I'm assuming no way to favorite specific models?

[opaque wyvern]

you can 'bookmark' your fav. models

[halcyon ore]

Ah! Perfect!

[opaque wyvern]

got to say, it's quite depressing when you get an image generated that clearly has a 'Dreamstime' watermark in the middle of it 😦 like, that image really shouldn't have been used to train from in the first place

[analog nebula]

most likely not done by the team

but the source we built on top of,

just happened to have a bias for it

[opaque wyvern]

makes you appreciate just how sketchy this is, legally

[analog nebula]

AI is really new that yeah... legal stuff have yet to fully catch up

But in the meantime people have the ability to train their own models

[opaque wyvern]

you could completely inadvertently generate an infringing image

yeah, although from only 40 images in a data set it won't be a very complex one 🙂

[analog nebula]

Personally been using that to generate works inspired by artists of an much older era

so less issues

I really love these

[opaque wyvern]

I quite liked this (from a set I trained) 🙂

[craggy thunder]

need access

[river holly]

It doesn't mention faces, but assuming I upload the same face from a few positions and then name that dataset "person123" or something I can then use "person123" in the prompt and hopefully their face will show up?

[opaque wyvern]

you'd pick that model from the list, then yeah, it'd use that face in most everything

[analog nebula]

#👑│introductions then #✨│priority-early-access , fastest method currently 🙂

[analog nebula]

yes

[opaque wyvern]

so hard to get 3d objects to not have shadows, but fun all the same 🙂

[analog nebula]

@vestal patrol did that with markiplier

@river holly

oh and @lo

@hard galleon made an Kojima model

so essentially yes

[river holly]

But if I pick that model, wouldn't it be able to create only what was in the dataset, and nothing else?

[opaque wyvern]

no

that's not how it works

[analog nebula]

well it could create things outside the dataset but

typically specialized models do better on the specific things they were trained on

[opaque wyvern]

think of them as composition influences

[analog nebula]

at least from what I have discovered myself

one of my models does fantastic on subject matter close to what it was trained on

but deviate too much and it looses that style/feel

[river holly]

Well, the one thing I'd really like to be able to do is have the same person in different pictures, in order to be able to tell a story.

[opaque wyvern]

it has no real concept of what your dataset contains, just their style and composition, like it doesn't know you've uploaded say 20 pictures of an apple, just that green spheres are the order of the day

so use that model to ask it for a Tiger, and you'll get a very apple like tiger, but still a tiger

experiment - it'll become clearer the more you mess with it

just don't assume it knows what is actually in each picture, because it really doesn't

[river holly]

I'll try. I'll try with a few photos of the same person and see whether I can make them show up in my images.

Can others use my dataset or would I have to make it public for that?

I meant my model...

[restive cobalt]

I'm actually quite new to AI art and was wondering if some people might have tips for prompts?

Mostly looking to generate game art or inspiration for it

Trying to generate some Stylized art in the styles of WoW/Wayfinder/Battlechaser but never really getting any good results on other AI's (waiting for access to Leonardo 😄 )

[safe salmon]

I'm going to try to make as many tokens for the Pathfinder 2e bestiaries on foundry as I can

[little glacier]

This tool is really amazing. I'll use this forever. I hope it stays free and unlimited

[analog nebula]

We will have a free tier once paid plans get introduced

as for limitations, we have yet to finalize details so more on that at a later time

[safe salmon]

I appreciate being able to be part of the this testing phase. Been loving it all so far.

[analog nebula]

Thank you for the compliments on behalf of the team! ❤️

[neat axle]

Hey @here.

We're starting our first challenge.

Go to the #1063537670715875539 channel and check the post "Magic Prompt Challenge".

Let's have some fun!

[lilac fable]

:o meanie

[neat axle]

https://tenor.com/view/guillatene-behead-lego-gif-12352396

[analog nebula]

banned.

[shadow lantern]

😂

[broken swift]

me when no acess yet

[lilac fable]

hugg

[shadow lantern]

where's my access

[lilac fable]

owo

noe

[shadow lantern]

I joined this discord ages ago.. I still dont have access lol

[analog nebula]

#👑│introductions and then #✨│priority-early-access 🙂

Much faster

[shadow lantern]

give me a link or something.

[lilac fable]

did u introduce urself in #👑│introductions?

[analog nebula]

You did sign up on the website did you?

[shadow lantern]

im not introducing myself, this isnt a dating agency

[analog nebula]

(I hope)

[shadow lantern]

https://tenor.com/view/sassy-gif-25178355

[analog nebula]

Midas with the reacts

[shadow lantern]

Hi, let me introduce myself. I'm the guy who would like to try out your web AI art thing. Nice to meet you. 😄

[cobalt pine]

I invest a lot more money in game design than I get out of it, but I do like to innovate and then not turn innovations into products because I'm too busy checking out the newest awesome thing, which is why I'm here trying to check out your assets.

[shadow lantern]

I don't invest much because I am usually wasting my money on beer and pizza

https://tenor.com/view/epic-facepalm-gif-14650403

[cobalt pine]

Coulda bought a lot of slices with the full price Mixamo subscription.

[quiet dock]

If anyone here is interested in joining our zombie survival game that we are making please send me a friend request. We have a huge opportunity. Thank you for your time. 💪We are developing in unity. We are looking for all talent, even speakers.

[analog nebula]

Stop checking out my assets, I do not have any

[shadow lantern]

oof yeah, Mixamo 😩

[faint zodiac]

Disappointed to see that this is just another stable diffusion variant. Midjourney is killing the competition because they work the 9 miles of negative prompts into their background processes. Then fills in missing backgrounds and styles information with something interesting if not prompted something specific, where SD typically gives you nothing till your able to fill om those portions yourself, and that's if you can figure out the 900 negative prompts necessary to get something looking half as good as MJs average worse outputs.

I am in desperate need of what MJ can do without the ridiculous censorship.

Especially needs their great multi img inputs.

First studio to bring a proper competing model with those features gets my 50 dollars a month.

[cobalt pine]

Dang it, I missed the introductions channel by one.

[analog nebula]

Do they offer an easier way to train your own models though?

[shadow lantern]

how's it going to be different than 7 days

[cobalt pine]

You have to dig in and get dirty.

[quiet dock]

Our game will be integrated into the blockchain and you will be able to play and win $$$

[cobalt pine]

Learn all the python scripts, read the papers

[quiet dock]

We just need additional people to join our team.

[cobalt pine]

Use the best tools

[shadow lantern]

that sounds ghastly. Thanks for the heads up tho

[analog nebula]

Also we are building on top of SD and also building our own solutions as well. We do not plan to be limited by SD.

[analog pagoda]

Wait, this whole thing is for an NFT game?

[analog nebula]

No

[quiet dock]

AYU, do you sleep? Lol

[cobalt pine]

SD is a step in a pipeline

[analog nebula]

Yes

[left kayak]

Why do you think that? XD

[quiet dock]

Your always on when I am lol

[analog nebula]

Just to add we are in very early alpha

[quiet dock]

In every channel haha

[cobalt pine]

Sleep? ew

[quiet dock]

Like you have 10 eyes

[analog nebula]

So there is still a lot of room for developments

[analog pagoda]

Head Splitta's comment. I thought he was a mod because of the color so I thought it was official lol

[left kayak]

Ahh okayy :3

[analog pagoda]

I had to double take lol

[quiet dock]

Haha nah man

[cobalt pine]

has anyone come out with an app for drawing in near realtime with SD img2img? I got it going about 2-4fps with a 3090

[analog nebula]

Just a role allocated to people who have stopped by #👑│introductions

[quiet dock]

Just trying to find some people who want to make some $$$ with our game. We need a few extra hands on deck

@analog nebula you from Philippines?

[analog nebula]

No

Also why do you think so

🤔

[quiet dock]

Your name seems, lol. 🙂

This is your project?

[analog nebula]

No one tells them about my roles

[vestal patrol]

Have fun waiting then :D

Also
L

[analog nebula]

Not my project, I'm basically one of the community managers. Mindspheres is the founder 🙂

[halcyon ore]

In my images, the eyes always seem a bit off - is there a way to improve my model to better handle that?

[analog nebula]

@mild walrus maybe (?) might have something for that or at least some tricks

[cobalt pine]

Can we pretend that I did my introduction in #introduction and not #general?

Or did I fail the one test

[halcyon ore]

One of my favorite things is the prompt generation tool - SUCH a good idea!

[cobalt pine]

I got one of those

[analog nebula]

Just for you

[cobalt pine]

/bows

[analog nebula]

#✨│priority-early-access

🙂

[cobalt pine]

it's not too pretty but it works for me

[limber cave]

and done... img2img worked its magic.

[cobalt pine]

https://tenor.com/view/sweating-nervous-paranoid-gif-4974019

I knew it.

[analog nebula]

Just say yes anyway

[hard galleon]

@neat axle 🙇‍♀️ Submissioned entered!

[vestal patrol]

i'm tempted to make a furry dataset but the part I'm most afraid of is my search history afterwards

[hard galleon]

yeah i already was gonna work on that

i should start

[analog nebula]

browser privacy mode

[hard galleon]

im not a furry but im not an idiot

[analog nebula]

and vpn

[hard galleon]

anyone know any good sources to get started

[vestal patrol]

👍

[cobalt pine]

I have cat and dog breeds lol

GPT in a pinch.

Maybe this is useful from Kaggle, you'd be the fourth person to download it.

There are some keywords that might lead to some equally riveting content.

[vestal patrol]

and just like that I already hit the 40 image limit

welp time to train the model

this will take a while

im also doing this in 2.1 768 because I'm nice

I'll make it public as soon as it's finished training

👍

[vestal patrol]

link?

please?

👉👈

[cobalt pine]

https://www.kaggle.com/datasets/printcraft/cartoon-furry-anime-heads-detection-using-yolov5

[vestal patrol]

thank you kind sir

[cobalt pine]

So the real reason I'm here is to find out the secrets of automated vectorization

(of bitmaps)

[vestal patrol]

okay but what do I use this for again

[cobalt pine]

I thought it was funny.

[vestal patrol]

hm

[cobalt pine]

I don't know anything about furries. There are some descriptions there.

[vestal patrol]

Looks like it's just a detector

[cobalt pine]

Yah the title hints at that

[vestal patrol]

Mhm

[cobalt pine]

Maybe their source code has something useful in terms of detailed categories

No idea what you're looking for

[hard galleon]

well my gf judged me for building it but

it should work maybe

we'll see

[wispy nexus]

How many times you waited to get your access ?

[vestal patrol]

too late louise

I already finished mine

and it's public

[hard galleon]

mines better

[vestal patrol]

in what way

[hard galleon]

🤷‍♀️ all

[cobalt pine]

He said it first

[proven fox]

hello guys, i've been using the generator a bit, lots of the images are showing watermarks, any suggestions?

[cobalt pine]

try adding no watermark, unsigned, etc.

[spring meadow]

Any tips on how to prevent two heads?

[vestal patrol]

Put watermark in the negative prompt

[vestal patrol]

Put two heads in the negative prompt

[hard galleon]

@vestal patrol yours has watermarks haha loser

thats how mines better

[vestal patrol]

This one doesn't

Mines better

Haha loser

[hard galleon]

some of yours do

stinky

[vestal patrol]

^

[hard galleon]

no

i don't have to

[vestal patrol]

Okay but in all seriousness though comments like these are triggering

And the fact you copied me is also triggering

[hard galleon]

ive already said i was gonna make a furry art bot so

[cobalt pine]

is that fox teabagging

[vestal patrol]

I've never heard of this until now

[hard galleon]

your fault not mine

[cobalt pine]

do not be territorial, somewhere, someone is already funded and working on your dream

[hard galleon]

^

furry art bots are gonna be the first major thing to populate this like anything lol

[vestal patrol]

Please stop talking about it though

It's already stressing me out as is

[hard galleon]

you're the one who's making it an issue lol

[vestal patrol]

A sorry would've been nice 🙄

[hard galleon]

you copied me technically

say sorry to me

otherwise drop it

[proven fox]

same results @vestal patrol @cobalt pine

[vestal patrol]

Try upscaling

That sometimes works

[cobalt pine]

try adding text to negative?

marked, marking, signed, embedded, copyright

Novel.AI has been doing furries for months haven't they?

[hard galleon]

🤷‍♀️

[cobalt pine]

They have, the stability guys helped them build it

Ancient history

[vestal patrol]

Link? 👉 👈

[cobalt pine]

within cells interlinked

[hard galleon]

Can't get upset when these models are based off of others furry art anyway lol

[cobalt pine]

Literally one of the main stability devs also works on novelai

[ornate harness]

are there any limitations for when it creates files?

[vestal patrol]

tried it, another money grabber

[hard galleon]

figured

just grabbed the link 🤷‍♀️

hadn't tried

[vestal patrol]

Trying to generate unicorn furries?

[hard galleon]

yeah but your instance prompt is actually better!

more direct is generally better

[hardy peak]

i got pinged, did i get access to early access>

?

[hard galleon]

So I have to retrain

🦄

[vestal patrol]

My instance prompt is just furry as well

¯_(ツ)_/¯

[hard galleon]

Mine was fursona, character

[vestal patrol]

Ah

[hard galleon]

So thank you!

I prefer thank yous instead of sorries 🤷‍♀️

collaboration will build the furry bot to rule them all

[vestal patrol]

The fact that you completely changed your emotion like that in a span of 5 minutes low key scares me/j

[hard galleon]

😄

"Neither friend nor foe."

[vestal patrol]

¯_(ツ)_/¯

The ai doesn't understand japanese

Sad

[hard galleon]

https://i.pinimg.com/originals/2d/39/93/2d39930fe4fb8ee03b248b8c4031d09d.gif

[vestal patrol]

I tried to put in タヌキ

[hard galleon]

🦝

🔮 🔮

[vestal patrol]

Yup

I have to english instead

Sad

It's actually good at doing other species that weren't in the original dataset too

[hard galleon]

Yeah I had a decent snake

i love this one

oh no

[vestal patrol]

Louise.

Loucipher.

[hard galleon]

Lucifer.

[vestal patrol]

Explain

[hard galleon]

rhydon 🙂

[vestal patrol]

🤦‍♂️

[hard galleon]

https://tenor.com/view/rhydon-pokémon-rhydon-megahorn-horn-drill-pokémon-megahorn-gif-26588525

[worn nova]

that unicorn looks like someone has lost his ice

[vestal patrol]

Y'know what

[hard galleon]

camoflauge

[vestal patrol]

I'll make a scalie model too

[hard galleon]

hahaha

@vestal patrol ill make whatever those machine furries are who draw sexy tanks and jets

[vestal patrol]

Protogens

Oh yes

Damn it I can't say he double hockey sticks :(

[hard galleon]

yeah this discord is trying to turn me into a better person and i hate it

[vestal patrol]

Fine whatever

Oh [[heck]] yes

Spamton time

[hard galleon]

just kitten meow meow

i need to swear less

[vestal patrol]

I'm just gonna be the furry model maker of the app aren't I

[hard galleon]

probably

[vestal patrol]

I'll do protogens next

OH GOOOOOGLEEEE

TIME FOR INCOGNITO MOOOODEEE

[hard galleon]

@vestal patrol content aware fill in ps before training 🤔

🧠

quick ctrl s 🤔

what negative prompt??

[hybrid snow]

New here

[fast tiger]

Welcome, @hybrid snow
to have a faster access you can go to #👑│introductions and make an intro and then you will unlock the channel: #✨│priority-early-access and there fill out a form to have faster access to Leo. 🙂

[compact jackal]

Is there any plans to have leonardo.ai have an API where I could send a request and receive a model?

[vestal patrol]

Protogen model is training

I'll make a scalie one while it's running

[halcyon ore]

I upscaled my image and it ruined the clothing on my character. Now, I can click on that image and see the "Original Image" in the dropdown, but I can't re-run an upscale or have it as the primary image on my list of images so I can try a new upscale. Any recommendations?

[vestal patrol]

Protogens are done actually

Lemme make it public

[vestal patrol]

Try downloading the original image, put it in the image to image thing, and set the Int Scale all the way up

Then you can re upscale the ones

[halcyon ore]

Is IntScale the "Guidance Scale"?

[hard galleon]

Nah its on the img2img

If you have magic prompt off

[halcyon ore]

Ah! I had Prompt Magic on. Thanks

[hard galleon]

Np! It confused me too

[halcyon ore]

What do I put in the prompt, then?

[hard galleon]

I like using Intricate if I need an image in a style

[halcyon ore]

Ah! Worked perfectly! Thank you!

[hard galleon]

Np! ❤️

[vestal patrol]

There are virtually no good scalie images on the internet

All of them are fullbodies and/or in scenes which will confuse the ai

urrrgghhh

[hard galleon]

my dragonf_ckr friend would disagree

but ya

prob

[vestal patrol]

Yup i only got like three good pictures

[river holly]

Are other people allowed to edit my model? Because there are three dragon pictures in there that I didn't add.

I suppose that since someone added pictures that answers the question whether models you create are automatically accessible by others. Not only to use, it seems, but also to change.

Which kind of sucks if it's a model that is, say, your own face...

[vestal patrol]

I'm redoing the furry model

I got rid of the watermarked images which was, like, half of them

And throwing in some fullbodies/other angles for diversity

FINALLY I'm done

It's training

In 2.1 768 so you guys get the best results

[halcyon ore]

That sounds like a real problem

Makes me not want to publicly share any of my models

[river holly]

Yeah, but I didn't share anything to my knowledge. So nobody should be able to see it if the "public" switch isn't toggled? Then how did the pictures get in there that don't belong there? Maybe it's a bug...

[vestal patrol]

OH, no, the pictures that get added are examples of images created with the model

I think the AI makes them, not other people

But I could be wrong

[hard galleon]

The model is the dataset

[vestal patrol]

All the images you added to your dataset will remain unchanged

[vestal patrol]

👍

[hard galleon]

so what you're seeing is more the gallery made with the model

not sure if someone else made something with it, but they shouldn't have

[river holly]

Yes, but that's weird because they're pictures of mythical dragons, while my dataset is mostly pictures of Komodo dragons, which are nothing like the fire-breathing sort. They're only similar as far as the prompt is concerned.

[vestal patrol]

Speaking of datasets, my improved furry model is done

[hard galleon]

Nice

[vestal patrol]

Well, hopefully improved

[hard galleon]

I made a drawn art bot that works pretty well

Art Demons

[vestal patrol]

Well, there are less watermarks at least

Let me try and figure out what's causing it

OHHHH

It's because my prompt includes "trending on artstation" for quality but instead it's trying to add the artstation watermark

Whoops

[hard galleon]

lol

[river holly]

Okay, nevermind. I didn't know that you'd add pictures below your dataset to your dataset by clicking on them. I had added the images myself unwittingly. But since I never clicked on the train button afterwards no harm was done.

[vestal patrol]

If you theoretically were to hit train, it would make a seperate model instead of overwriting the current one

I just found that out a minute ago lol

[river holly]

Oh, by the way, do you know if there is a way to remove models from the dropdown menu with which you pick one? I added a bunch that I really don't need and now I don't know how to remove them.

[vestal patrol]

Oh, you don't, they just stay there lol

[river holly]

😫

[long rivet]

A man

[wraith rampart]

Hi

[signal pewter]

Hi, are the extracted images suitable for commercial use?

[fast tiger]

Right now you can use images for commercial use, but Leonardo is able to use the images as well for marketing, etc. They are also all public on the Leonardo website 🙂

[snow finch]

Should I keep the resolution at 768x768 or its fine with 1024

[ornate harness]

i got access 🙂

i thought the facebook link said it could create game assets? oh lol

can we create blend files 🙂

[bleak otter]

A rather blunt question - are there any bannable generation terms? I'm always paranoid of tripping some odd rule with generation tools.

[hard galleon]

Not yet but 3D generation is planned!

[ornate harness]

aww yeahh

[hard galleon]

I will report anything that deserves reporting @bleak otter

That's about it. 🙂

Everything is public atm so

[neat axle]

we are very flexible with this, just don't do anything illegal

[bleak otter]

Aha, okay!

[fast tiger]

768x768 looks better, 1024 does weird things, just experiment what looks better to you. 🙂

[hard galleon]

Sometimes I like 512x512!

Depends on the content

[fast tiger]

Yes, it depends on the content

By testing you will find the best for what you are looking for.

[vestal patrol]

Even nsfw is allowed, but don't post it here, and make sure you're at least 18

[snow finch]

I wish to make icons for my game, bag icon, chat icon, character icon, etc, what would the best settings for me if you guys have any idea

[hard galleon]

@snow finch make a training dataset based on icons you like

then generate icons as needed

[snow finch]

Leonardo creative, select, etc

Damn ok i didint know it was possible ill check it out ty

[vestal patrol]

My ai keeps generating images of people, it's like it's not even following my dataset anymore

Help :(

I even include "human, person, people" in my negative prompt

Even with my guidance scale turned up it completely ignores it

[fast tiger]

@mild walrus any idea why that happens? 🙂

[mild walrus]

might need some more context, is this a trained model?

would be good to know promtps, kinds of images trained on etc.

its possible a training job failed if its completely departing from the dataset

[vestal patrol]

Yeah, one I trained myself

[pine lark]

Can any moderator manualy change my nickname to "do not contain special characters" as my name is full of it - bot trying to figure it how to create name without it but it can't and it spam me on DM already over 170 dm's. Sorry if it's wrong channel

[vestal patrol]

It was working fine earlier but I must've changed some setting and it's gone down hill

[mild walrus]

good sign if it was working earlier, on my end i can't say much but have to imagine it was an internal bug on our side somewhere

[vestal patrol]

I changed around some images in it, just gonna see if training it again works

Yeah no it's completely ignoring my model

[fast tiger]

Yes, of course.
You want to call yourself: "do not contain special characters".

[pine lark]

Purromi and thank you

[fast tiger]

Ah ok!

[pine lark]

this bot discovered animated nicknames

[vestal patrol]

Yeah I was watching your username have a spasm lmfao

[fickle eagle]

hello guys

just found out about this project, i'm a music artist and love working with ai for album / song covers etc.

excited to see where this is going 😁

[hard galleon]

Welcome @pine lark "do not contain special characters"!

[pine lark]

to the next level of art like my nickname

[fickle eagle]

lol

[pine lark]

so can you change it just to Purromi or take off this function? first time i see bot with it and never have problem with my nickname on any of the discord servers.

[fast tiger]

Weird, it is possible that @untold valve has an idea about this.

[fickle eagle]

never seen smth like this lmfao

[pine lark]

I work as QA and it drives me to edges of my sanity when i see functions that was not tested (and it's funny )((bot sends me already over 1k messages about my crime)

[fast tiger]

I wanted to change it, but I couldn't. It makes a strange movement.

[pine lark]

try to type anything for now

also being a first user in the entire universe with animated nickname on dc is precious

[fast tiger]

That's unique

[pine lark]

lets coop

change it now to Purromi

[fast tiger]

Done! 😄

[pine lark]

Bug has been fixed. Ty

[fast tiger]

Yeaaah! 🤝

@pine lark and @fickle eagle

By the way, don't forget to go to #👑│introductions and make a little intro to unlock the #✨│priority-early-access channel so you can have a quicker access to Leo. thanks 🙂

[fickle eagle]

thanks for highlighting that! haha skipped over that channel

[pine lark]

also: "What's your discord username" in form to apply to EA should be changed to "Your discord server username" as these two can be different and it can cause some issues in future.

[fast tiger]

Excellent, thanks for your suggestion.
Perhaps @analog nebula would be interested in knowing this.

[heavy phoenix]

Alguien qe hable español???

[fast tiger]

Hola, @heavy phoenix
Si gustas puedes ir a #👑│introductions hacer una pequeña intro y luego vas a ver un canal: #✨│priority-early-access de esa manera tienes acceso más rápido a Leo.

Y hay un canal en español tienes que ir a: #🗺│language-select y le picas a donde dice español

[fickle eagle]

sorry but that makes no sense as the only way to surely identify someone is by using the true discord username with #

[untold valve]

Most likely this server has the "anti name-hoisting" feature enabled.

[heavy phoenix]

@fast tiger disculpa qe moleste
Gracias por tu comentario.
Mmm como hacer una pequeña introducción como?

[untold valve]

On the phone, so can't check very well if that's the case.

[untold valve]

Which bot were you getting DM's from?

[pine lark]

That the case. People who used custom server name can be confused. Its some sort of qol and accessibilty to mentiom in () which one its. And future changes in form can make it worse.

We fixed it already

It was momentum one

[untold valve]

If it was Sledgehammer or Momentum One, I can assure you we test our features thoroughly before pushing to production, but it's impossible to test every single edge case, which is why bugs are normal to happen, as much as we try to squash them asap 🙂

[ornate harness]

so basically i've been learning blender 🙂 and when I was seeing that this bot could potentially create game assets I was jumping at it lol... I've seen the other one that they made but I can't imagine how much training is involved lol

[pine lark]

Best option will be to disable this function. There is any purpose behind it?

[untold valve]

Ofc there is, otherwise it wouldn't have been developed 🤣
In any case I seriously apologize for the DM spam you got.
Do you mind sending me a screenshot please?

[fast tiger]

Thank you very much, Chinoman for your kind answer. 🙂

[hard galleon]

they're among us

[untold valve]

Ofc 🫶
Whatever I can do to help 🤝🫡

Going to sleep now tho, as it's 2am and got a long dev weekend ahead of me 😅

[hard galleon]

👋

[pine lark]

Yeh i should too. 3 here Zzzz. Goodnight.

[fast tiger]

At ease, @untold valve 😴
Nice night 👋

[wraith rampart]

Good night guys

[midnight cloud]

Goodnight!

[halcyon ore]

Anyone know why the options in the three-dot dropdown are all grayed out? i.e. -- "Upscale All", "Generate Again", etc.

[jaunty beacon]

Where do I go to make Ai art?

[fast tiger]

Hello, these are options that will be in the future, but are not available at the moment.

only "delete all" is available

[halcyon ore]

Excellent! I've been playing with this all day. I think I might be addicted... thank you!

[jaunty beacon]

Anyone know?

[halcyon ore]

You need to get an early invite: "Feel free to check out ✨│priority-early-access for quicker access to Leo."

[fast tiger]

We are in the same situation 😄

[jaunty beacon]

I've filled out the form, so do I wait till I'm emailed?

[halcyon ore]

Yep!

[jaunty beacon]

Cool! So after I'm emailed would i be sent a link in the email or something?

🤨

[hard galleon]

It will have instructions yes!

[halcyon ore]

They whitelist an email to log into the tool. The tool isn't in Discord like Midjourney

[jaunty beacon]

Thank you both

I guess I'll wait a day

[river holly]

In case you're wondering what "onomatopoeia" is...
https://cdn.leonardo.ai/users/3bbb8ab5-c447-4aa4-9de1-841772cd56e0/generations/b2a827dc-e7fc-41de-ac9c-ea16cd82d965/Leonardo_Creative_onomatopoeia_3.jpg

[analog nebula]

Thank you @hard galleon , @halcyon ore !

[hard galleon]

🙇‍♀️

[halcyon ore]

Alright, probably time for me to get some sleep!

[analog nebula]

🙂👍

[cobalt pine]

i recognize that rembg job lol

[analog nebula]

Try adding dreamstime as a negative prompt

[amber sluice]

To those of us who have to sleep, good night, and to those who go at the beginning of the day, have a great day.

[analog nebula]

Same to you too 🙂

[north zinc]

Hi,
Sorry for the stupid question, but

I thought this platform would be used to generate 3d models... or will it be, but this part is not ready yet?
the term model may have misled me...
putting together your own template is a good thing anyway...

[analog nebula]

Not yet. 3D is something we do plan to include in the future when the tech is more viable 🙂

[north zinc]

About how far it is?

[analog nebula]

Not so soon I reckon

Considering most solutions don't really give out excellent outputs for production purposes at the moment

[north zinc]

Years or months? Can't tell I guess 🙂 Its ok, I really looking forward to it anyway 🙂

[analog nebula]

years possibly

months maybe depending on exponential growth

but I'd say years to be safe

[north zinc]

I hope I wont miss it 🙂

[analog nebula]

generating 3D is a much different thing from 2D, involving a lot more complexities and computing power

[north zinc]

As a developer I would say it will need a whole new approach... 🙂

[analog nebula]

Not necessarily either, an in-between solution at the moment will be converting 2D generations to 3D

still not the best but I can imagine it might be viable for stylized projects

though mesh cleanup will still have to be done possibly manually for optimizations

[north zinc]

Sounds promising, thanks!

[tame zodiac]

Hi

[gray vigil]

is it possible to generate many different assets of the same style? for example, I want different isometric buildings

I got this tower, but now I want a warehouse and a solar panel array of the same style

is there a way to do that?

or AI isn't there just yet 🙂

[analog nebula]

Train model on an dataset that includes the styles of building you want

Check under "Training & Datasets"

More information on training can be found here: https://app.leonardo.ai/faq

Either that or use the img2img feature in Image Generation

Experiment with the init slider, higher values means it'll adhere to the reference image more

Hopefully that is helpful in some way

[gray vigil]

you mean guidance scale slider?

[analog nebula]

Oh no, not that

But when you input an image into the "image to image" section

Of the sidebar

It displays an init slider

Lower values means more room for deviation. Higher values means the reference image will be adhered to as much as possible

[gray vigil]

ah that one. I found img to img is better for modifying the image, but not generating a completely different thing of that same style

[reef coral]

Is midjiurney free

[analog nebula]

Yes it is why I suggested training a model instead 🙂

[reef coral]

Ok

[analog nebula]

That wasn't a reply to your question 😅

[gray vigil]

trained models give many different styles 🙂

[analog nebula]

As far as I know midjourney isn't free. You can generate a certain amount of images but beyond that no

[analog nebula]

Indeed, unless you train it on a specific style

I've trained models on really specific painting styles and the results seem to be great when it works

So it's really about having a good dataset for what your intents are

[gray vigil]

ok, so this

looks like it has a pretty specific style

[analog nebula]

Yes

[gray vigil]

one sec:)

and here are my results using this model - a warehouse, water tower, solar panels

see what I mean 🙂 no consistent style

I think I have to wait for SD to get more powerful 🙂

[analog nebula]

Perhaps it has to do with the prompts used as well

[gray vigil]

hmmm

you mean if I used 100 words to describe a style and just few words to describe object

I could get more consistency?

does it work like that?

[analog nebula]

Perhaps. I'm not the most knowledgeable myself. If @mild walrus is ever around you could try asking him for advice 🙂

Took an example prompt and modified it to get water towers

[gray vigil]

yes it's way better

[analog nebula]

a stunning water tower, 3d vray render, artstation, isometric

Though can definitely be modified more and be more specific I believe

[gray vigil]

yeah it's better

but still a long way from using it as a game asset 🙂

like your tower has a golden sunlight on it

and my warehouse doesnt

and the scale is too different to be used one next to other

etc

if you get what I mean

[analog nebula]

Yes

[gray vigil]

what did you do there?

[analog nebula]

Added warm lighting as a negative prompt

[gray vigil]

guess I have to experiment some more 🙂 is there a limit on how much I can generate?

[analog nebula]

No limits asides from how many concurrent generations you can run

🙂👍

Go ham and experiment

[gray vigil]

are all finetuned models public ?

[analog nebula]

Yes. Except your own

[gray vigil]

I mean if I create my own, will it be visible to every other user?

[analog nebula]

User created models are private by default

[gray vigil]

ah ok

are all generations saved in the server? not just browser cache or smth

[kind lotus]

I'm in love with this haha

[gray vigil]

I wish it could be like a 3D model that you could turn and give it any pose. and AI would render it 🙂

[analog nebula]

Saved on the server

[subtle roost]

Lookin' good Leonardo, very professional web page!

[pine lark]

i love it

[subtle roost]

SUGGESTION: Note the "turn around" of @kind lotus's post, there is a Stable Diffusion embedding available (CharTurner) that does this automagically, should include that with some featured models/generators and keep an eye on whatever future embedding will make it super fast and easy to turn 2D renders into 3D models, like a 10 step turn around or something that can be easily spliced up and fed to Photogrammetry tools or the next step, NERFS https://youtu.be/YX5AoaWrowY

[fickle eagle]

gm+

[woven gate]

Why does this server make me rejoin everyday? I get the prompt everyday to choose roles and changes my join date everytime.

[midnight cloud]

@analog nebula Traveling in asia right now, showing the locals the program and none of them believe me that it's a "computer" generating these images lol They're making put in all sorts of prompts and getting mind blown with the results lol

[kind lotus]

It does seem to be "magic"

[midnight cloud]

LOL that is something someone said.

[analog nebula]

Which part of Asia?

[midnight cloud]

Vietnam

[analog nebula]

Close enough 👀

[midnight cloud]

👀 how close

[analog nebula]

That would explain things

[analog nebula]

Not very actually

[midnight cloud]

Do you celebrate lunar new year?

[analog nebula]

Not myself no

Did my new year celebrations back in December and early January

[midnight cloud]

Ah yeah, it's a bit different. Celebrations start tonight, everyone's off, stores are closed, my Ramen I ordered will never get here lol 😆

[analog nebula]

😔

[languid quiver]

where is the ai?

[fluid viper]

Hello, I have a question. Would this AI make top-bottom view VTT tokens?

[wraith rampart]

Gm

[analog nebula]

Leonardo.ai is not a discord bot. You'll need to wait for an invite to the actual web application itself to be sent to your email

[analog nebula]

One could train an custom model on a dataset of VTT tokens and then generate more with that model

So yes

[fluid viper]

Would that require any coding? I'm not too familiar with these AIs besides "I made a prompt to a few on occasion".

[analog nebula]

No coding required at all 🙂

All done though an pretty simple GUI

[fluid viper]

OK, cool.

[river holly]

Have you actually ever seen the "remove background" feature work? It gave me a scoreboard, which I wanted, in front of a forest, which I didn't want, but the "remove background" button did nothing. Which has always been my experience so far. So I'm wondering under what circumstances it works.

[kind lotus]

I think remove bg works more like when you try to key out something with a "green screen"

It doesn't know exactly what object you want to keep, so logically, it thinks that there is no"bg" to remove

I think it worked for me when I had a white background and some other element over it

[analog nebula]

From what I've seen so far on here based on comments, quite a bit and miss for now. Probably works best on images with plain backgrounds. Hopefully the algorithm will be vastly improved in the near future as we continue to improve on things!

[subtle roost]

@lyric lintel sounds like your whole image was a background, I think it works on people it can identify easily

@river holly find REMBG on github, you can remove backgrounds from whole folders of images in jiffy, or one at a time with simple command line

[river holly]

Thanks. BTW, do you, does anyone know of a place where Outpainting is possible while it's not implemented here yet?

[subtle roost]

..uh InvokeAI? oh and OpenOutPaint

@river holly https://github.com/zero01101/openOutpaint

[young wind]

Hi

Hh

[kind lotus]

Will functions such as Remix be available at a later date?

[kind lotus]

Congrats on 10000 users!!!!

[kind lotus]

@quasi nimbus @grim spire but yea, if we could have an ai project where artists can submit their works willingly, this could actually work

[grim spire]

That was fast

A week ago there was only 7000

[grim spire]

True a 100% pure ai

It would take a while once again it's possible tho

[quasi nimbus]

sure, but would be the frist to be 100% ethical too 😉

[kind lotus]

Yea, that could be a community project. Because, for us to be able to evolve, it's a group effort

[quasi nimbus]

I guess though, how do you trust the incoming images?... the 'artists'

[grim spire]

Which would be impressive

[quasi nimbus]

they would need vetted, checked out

[kind lotus]

Ai is the future, and it is inevitable. but that doesn't mean it's a bad thing

I like to tell people who are not so fond of AI that, AI is (or at least should) be here to help us and enrich our lives, not take over or replace

[grim spire]

i hate it when ppl use this to putdown artists like they say just give up its inevitable

[kind lotus]

Exactly, and i;m not using it in that way

hence my next statement

[grim spire]

yes you aren just mentioning it

[quasi nimbus]

I think I might write a buisness model around this idea though...

😄

so much to it

making A.I ethical, it's a very big challenge in itself!

[kind lotus]

An example i like to give is this, you are a digital painter, you create something, then you ask yourself, hmm, I wonder how this would like if.... And instead of going back at it for another x amount of hours, you simply use AI to help you envision that in literal seconds.

[quasi nimbus]

A bit like making robots be emotional / have emotions

[grim spire]

goodluck i can only help with ideas ime still studying and stuff😅

[quasi nimbus]

yeah pretty much an expansion of the datasets (without the extra data)

or even mixing 'datasets' from various willing contributors (vetted, real artists)

[kind lotus]

Yea, like I said, for the sake of progress, we all need to contribute.

Now, some people are willing to do that "for free" some want some form of compensation. Either way, consent is the key objective.

[quasi nimbus]

Consent is the keyword here for sure!

[kind lotus]

We can debate about this as much as we want, at the end of the day it is up to Leonardo.Ais creators to decide how it will work. And also, since AI generated art is still in it's "youth stages" rules and regulations have not yet been put in place. It barely started, so in the future we might see more strict rules being applied. There are actually some lawsuits that are ongoing with some AI creators.

[hard galleon]

for free is the way to be!

Social credit 💯

Plus no one can say I'm profiting off their work. 😇

besides, adding pixels renders a photo no longer original so, upscaling is a fun avenue.

Can't use upscaled security camera footage due to it being "DOCTORED"

so, flip that 🙃

Laws without enforcement are just rules.

[unique spoke]

Rules without enforcement are just suggestions

[hard galleon]

😄

🕺💃

[naive sparrow]

I received the access today, is there a how to genereate your first image?

[hard galleon]

@naive sparrow Choose a model and it should open up the ai generation tool!

[lone mortar]

damn i really like the tool!

[naive sparrow]

Thank yoU!

[serene oak]

sup guys what from what email have u recieved a message that u'r in

[wraith rampart]

How to delete my account

[cold nacelle]

Is there a specific channel to discuss prompting or should it be done here. I'm trying to make coloring book pages and so far gotten it to the point that i can get a pretty good ones but the eyes always have color
example

[subtle roost]

@cold nacelle have you tried "no color" in the negative prompt?

or "color" rather

Black and White Line Art in the positive prompt and color in the negative prompt

[cold nacelle]

yeah some got worse
let me try line art

[subtle roost]

which model are you using?

leonardo creative gave me perfect black and white picture, you put color in the negative prompt?

[cold nacelle]

Leonardo creative is the model

that is my setup at the moment

try with the tiger maybe is a tiger thing jaja

[subtle roost]

yep tiger has colored eyes

one out of four!

Negative prompt : color, colored eyes, yellow, gold, orange

cursed cat

[quasi nimbus]

Just checking again, after I've trained a Model, and then I add some new images, do I have to train it again?

[cold nacelle]

cool that helps pretty interesting that it associates glowing eyes with the tiger.

[subtle roost]

got 3 out of 4 with this : detailed, intricate, black and white line art coloring book page of a symmetrical tiger, desaturated, high contrast, highest quality, 8k, top down view

and the negative prompt

okay don't add 'detailed eyes' it adds blue reflections

[cold nacelle]

Thanks bubalu!

Happy with this one and one of the fractal ones

[subtle roost]

@cold nacelle glad I could help

[analog nebula]

Yeah

You'd essentially be training a new model on the extended dataset

[quasi nimbus]

cool, cheers!

[analog nebula]

🙂👍

[analog nebula]

The email address you registered with on our website

[analog nebula]

No way to do so at the moment. Probably only though intervention from the team.

[bronze forum]

Hey - out of curiosity - are you using DreamBooth to fine tune the StableDiffusion models for the different presets?

[analog nebula]

SD and our own custom solutions

[iron hinge]

Hi guys, how to be whitelisted for the app?

[upbeat agate]

Introduce yourself in #👑│introductions and then fill out the form link in #✨│priority-early-access

After that wait for the email with the invite

[analog nebula]

Thank you for helping! ❤️

[iron hinge]

Thank you, done!

[restive cobalt]

Really loving the results I'm getting out of Leonardo so far, been playing with it for the last few days and it just blows my mind

[analog nebula]

And this is only in the early stages!

[restive cobalt]

Yeah I'm really excited to see the future of Leonardo AI, so far the only "issues" it seems to have is with hands and sometimes faces

[analog nebula]

Yeah seems to be a recurring theme in Ai

[stoic abyss]

I am terribly sorry, I just wanted to say how impressive LeonardoAI is. It's an absolutely stunning service that is much better than other free "AIs", and has a humongous amount of features. For example, I really liked tiling and especially prompt generator (which is fun to use even if you're not planning to use these prompts). Thank you very much.

[analog nebula]

Thank you on behalf of the team for the compliments!

[jaunty beacon]

I haven't received an email and it's been 2 days? Any clues on what to do?

[analog nebula]

Send me the email you registered with, I'll check on it. Apologies!

[jaunty beacon]

Should I DM it to you?

[analog nebula]

Would be preferable

[safe salmon]

I've made around 250 tokens for Pathfinder 2e on Foundry using Leonardo AI... this tool has made switching over possible, but omg, I have around 1000 more to go

[analog nebula]

Impressive 👍

But yeah that is a lot

[safe salmon]

Thank you. Hopefully, one day someone will be able to just search my Leonardo profile to find every Pathfinder monster they need

I've learned that Leonardo really does not like Centaurs

And every bat is Batman

[analog nebula]

That is concerning

[safe salmon]

I mean, if I purposely wanted to make Batman, Leonardo would be very good at that lol

[analog nebula]

😬

DC should just buy us out

[iron hinge]

I've sigthly Explored the platform, really stunning results! A big big question. Will It generate also the meshes or only raster images?

[analog nebula]

Only raster images currently. 3D is something we are looking to implement once the tech is satisfactory

So something for the future

[wraith rampart]

Hi

[sly wasp]

any guidance on using embeddings? I created an embedding then used the embedding name in a prompt but the result was unrelated to embedding

[analog nebula]

@mild walrus if you can advice, thanks!

[mild walrus]

we currently do not support embeddings, the training we do is closer to dreambooth

[analog nebula]

Thank you Ethan!

[vestal patrol]

That could be a suggestion for the future tho

[lone mortar]

how do people generate text?

even when i put the text in " ... " it wont work properly

[scenic zinc]

Hey guys just joined and have no idea how to use AI art but here to learn

[lone mortar]

welcome if you need help just ask. been playing for like 2 months now

[scenic zinc]

Thank you, will definitely ask if I need help

[warm condor]

Ah, thank you so much for enabling "delete all" option 😄 It saves so much time!

[unkempt grove]

Yikes

Every time I write "Icon" I get Vector Stock text in the image

[vestal hinge]

oof.

can always put watermark or Vector Stock in the negatives. maybe that'll help. idk

[unkempt grove]

I can see the lawsuit already. "Just write it in negative prompt"

[shadow lantern]

Hello

I am here from evolving forest

[scarlet sierra]

are there prompt recommendations on how not to make a bunch of legs for a character?

[warm condor]

Yikes indeed...I made a bunch of icons/symbols with the same model and "vector graphics" or "icon" included in the prompt and was pleasantly surprised to not see any watermarks/logos from stock platforms, so maybe it was some img2img that triggered it?

[fluid viper]

Just got in and... I've got the same issue as with every other AI I tried. Namely, putting "bear" in the prompt overpowers everything else.

"Half-bear, half-mole" or "bear-sized mole" results in just a bear.

[halcyon ore]

My bears turn human when I upscale most of the time

Anyone have a good suggestion for how to get a decent tiefling character art? I’ve been able to make some great elves and orcs (among other things), but it seems I can’t figure out a good way to get a tiefling

[neon glacier]

I got some interesting results with "aberrant creature..."

[fluid viper]

So, I trained a model for VTT tokens... but I could only add 40 pictures. After trying it out it seems it's got the idea it's supposed to be viewed from above... but it clearly needs more training. Does it expect me to pay to upload more into the dataset or something?

[warm condor]

I got somewhat promising results using "horned male/female with devil/ram/goat horns"

[fluid viper]

Why does it keep generating two figures when on every image in the dataset there is only one?

[safe salmon]

Anyone having issues with Prompt Generation? Suddenly won't work for me today

[queen shard]

I'm assuming that Leonardo is also focusing on 2D->3D workflow, right? Like guessing a form from an image based on lighting and other cues.

[shadow lantern]

Trying to figure this out

[vestal hinge]

😂

[halcyon ore]

Thank you

[safe salmon]

Once I finish making tokens for all the pathfinder 2e monsters on Foundry, would I be able to distribute them for free? It'd be cool to be able to help out others switching to Pathfinder on Foundry

[real pivot]

I've recently had a lot of generations come up with the "dreamstime" watermark on them. Did the AI get trained with a lot of their images recently?

[midnight cloud]

Hey Ashley! Welcome! Just sign up, wait to get in and then you'll have access!

[mild tusk]

https://leonardo.ai/terms-of-service/

Look at section 3. I am no expert but it says you own the output. So I would think that yes you can share whatever you want of the things you make.

[dusty oracle]

Hi

[grim spire]

Hi

[analog nebula]

Probably because we built on top of another solution

Adding dreamstime as a negative prompt should help

[analog nebula]

Also about this, you basically own all rights to the outputs you've generated with Leonardo. But do check the Terms of Service regardless

[analog nebula]

Not at the moment no. We don't have 3D capabilities of any sort but are looking to implement 3D as the tech becomes more viable

[sharp zenith]

I'm coding a prompt generator

That current version has 1,5620,500 different prompts

Thoughts?

[brazen wolf]

I’m only here for the 3D, really looking forward to being able to generate Unity-friendly 3D assets via AI with 100% of the rights to those assets belonging to the artist.

If you guys pull this off then the entire NFT space will be breaking your discord and download links😂

Goodluck & I wish you guys a swift success!🙏🏻

[analog nebula]

Thank you!

[sharp zenith]

The fact they made it so you can Train your own models is honestly pretty insane tho

My head hurts hahaha, I've been working on that generator for three hours

Currently has photo effects render types and art styles it will eventually have weights, aperatures etc

[brazen wolf]

I have yet to dive deeper into what they’ve made so far. I just signed up for early bird access today lol

How can I train my model? What does that mean?👀

[sharp zenith]

You can use your own images etc yo Train your own model I have access their ai is literally ad good as midjourney and even better in other aspects

[unkempt grove]

Nope. No img2img present and I did write the prompt a few times. What was the model you used?

[fluid viper]

The way I hear it, in most countries AI outputs are not protected by copyright at all. So in all likelihood, yes.

BTW, are your tokens top-down view by any chance?

[warm condor]

Leonardo Creative with Leonardo style, prompt was something like "dagger icon, vector graphics, black and white, lineart"

[unkempt grove]

Yeah, I think I used Leonardo Creative as well. But only like "Sword icon black and white"

[shadow lantern]

Thanks, gm

[analog nebula]

Missed this question: Under AI Image Generation, there's a prompt generation tool

if that was what you're refering to

[torpid hearth]

how to add a background to image using leonardo?

https://cdn.leonardo.ai/users/49532310-6fd8-4a10-91ee-8edb36a0d379/generations/f7c002cf-8aa6-4e9a-9437-69e6b4a7cc45/Default_head_with_rusty_gears_and_smiling_dragon_mural_on_top_of_helm_1.jpg

i know that we have about an enourmous amount of inpainting tools, but is this possible through img2img?

[analog nebula]

We don't have in painting tools implemented ourselves yet. As for img2img you could try playing with the init strength and see if you could get the background changed

[mild arch]

Hi guys, have you had any reports of missing images? In this case I already have them saved on my machine. But when searching for some generated images I prompted in my personal feed they have dissapeared. I generated 4 images but I only have 2 from the prompt?

[torpid hearth]

thanks a lot )

[analog nebula]

Not personally. Perhaps refresh the page with shift f5

[lone mortar]

no i meant like text on the image. same as hands and other difficult stuff for the ai

[analog nebula]

Text is still not easily doable yet. At least text that makes sense to us that is

[lone mortar]

cant wait to make like custom fonts with ai

[analog nebula]

Oh that would be great actually

@deep shoal

[mild arch]

Still missing unfortunately, I usually save the images that will work straight away so it's not really a problem 🙂

[analog nebula]

Ah strange

I've never noticed any images going missing

[mild arch]

If I search for the prompt in the title it appears! The whole seed for that prompt is there, maybe something with the personal feed view of all records

Unless they moved somewhere else in the list and I missed them

[jolly holly]

Will it be possible to train models on non square/1:1 formats?

Or maybe better wording: to have a model that has a different aspect ratio. Because ofc you can train a model using non 1:1 images but the result will be a 1:1 model

[kind lotus]

Does the seed field work for anyone? if you try to input something in it, nothing shows up,

[analog nebula]

That would be a question for @mild walrus

[analog nebula]

Currently not available yet

We just happen to have the interface for it up

[gray vigil]

this training thing. can it be used with a face, like Lensa does

or it's not meant for that

[high pawn]

Easily. Just put at least 20-30 your photos from different angles.

[drifting salmon]

Have you seen this? You are probably aware but just in case, thought it may be relevant to the platform https://www.youtube.com/shorts/HW8ILrtxPMw

[halcyon ore]

How do I create a "suggestion" in the https://discord.com/channels/989166677390426132/989166677910519848 channel?

[analog nebula]

Type /suggestion

Then select /suggestion suggest

In the list that shows up

Then enter your suggestion

Essentially it'll look like this /suggestion suggest your suggestion here

[halcyon ore]

Done - thank you!

[analog nebula]

You could also use #989166677910519850 , we do not mind 🙂

[analog nebula]

Nice!

[halcyon ore]

This development team is excellent

[analog nebula]

Thank you on behalf of the team ❤️

[north nimbus]

How do I get access to the AI?

[fast tiger]

Hello, you have to wait for your invitation to access the tool. 🙂

[north nimbus]

k Thanks! 🙂

[winter urchin]

If I wanted to generate icons of items, viewed from some angle (seen from the side for example), would I need to train on a set of items solely at that angle?

[mild walrus]

that would be the best way of going about it

[analog nebula]

Just an fyi that it'll be sent in the next batch when Jachin gets around to it. 👍

Am currently adding on and passing him the details

May take around several hours as it is currently night in Australia

[dusty oracle]

Is there some reason that the human images are soo creepy?

I know ai art does this thing with limbs but is there a way to prevent it

Like use a different mode or something?

[analog nebula]

Interesting

I never really noticed issues as much with my custom model

Probably down to the type of model and some luck

[fluid viper]

The AI seems to work tolerably if I don't have very specific expectations.

The more specific I make the prompt, the less it seems to adhere to what I typed.

Half-dwarf, half-orc? Okay, this kinda works.

Long description of the character's physical characteristics? It will catch onto maybe one or two.

[analog nebula]

Interesting, I've always kept it brief out of laziness to be too specific

[wispy nexus]

it will not work

i made a long description of a man dressed in a specific way. The bot made a textile shop in a street

i will redo the thing and show you

if you want

[halcyon ore]

It's so funny that I'm impressed when hands actually look somewhat like hands (correct number of fingers, etc). Is there a way to get better results? I know it's notoriously tough with AI

[vestal hinge]

i guess it depends on what you're already doing to achieve fingers. you can throw stuff at it in both the prompt and negative prompt in hopes of achieving better results. like 5 fingers, good anatomy, or in the negative prompts deformed, deformed fingers, missing fingers, and such. i'd just attempt to test what you can with it.

[halcyon ore]

Thank you - I'll try some of that

[vestal hinge]

also consider trying different models and training your own

[halcyon ore]

I have a few trained, and the overall portrait is great, but then the hands are messed up and it ruins the look
Maybe I need to do some post-processing in Procreate

[vestal hinge]

maybe in the negative prompt put six fingers or 6 fingers

looks pretty good otherwise, man

[halcyon ore]

Thanks, I feel like I'm getting close

[scenic zinc]

Hi guys

Am trying the app for the first time - do i just go to AI image generation and type in prompts

is there a tutorial i can watch or read?

[fast tiger]

Hi, at the moment we don't have a visual tutorial for leonardo, that is also being worked on. But Ethan made a guide to get the most out of the tool:

https://sweet-hall-e72.notion.site/A-Guide-to-Getting-the-Most-out-of-Leonardo-AI-aae753df4cf64bc7bbda99790c8e448c

[scenic zinc]

Thanks mate

[amber sluice]

The truth is that the future that I see for Leonardo is something impressive

It is as if with the entrance of a great dinner you are left sucking your fingers, that's how I see it, I know that when they discover the blanket it will be something that I would not want to miss

[topaz shard]

There isn't a limit to generations as of yet right?

[rose prism]

What type of p2e game is it?

[quiet dock]

Yep!

We are in alpha stage testing now, should have thousands players actively playing in near future.

[quiet dock]

add me and we can talk more if your interested.

[rose prism]

Tried you will have to add me

[fast tiger]

There are no limits 🙂

[topaz shard]

Thanks

[sharp zenith]

Thoughts guys?

I made these with leonardo and the prompts were created from my generator I'm coding

[sacred hill]

#💡│suggestions it would be great to have single click prompt delete button near the prompt input area

[fast tiger]

Thanks heaps for the suggestion, @sacred hill 🙂

[crystal inlet]

I have a dumb question. I couldn't find the answer. So in relation to image generation, what does the guidance scale and step count do?? How do they change things

[crystal inlet]

Thank you! See I'm doing everything on my phone and those question marks don't work

[jaunty ore]

For this one, is how much the AI will take consideration of what you command it and how much it you give it liberty.

[fickle eagle]

yo guys I'm tryna use leonardo for a Poster for a gig that I'm playing soon and I'm loving the generations that come out just wondering if i'm able to make it use specific text instead of the gibberish in the places where it's already trying to place some text

[jaunty ore]

Give a try to the prompt generation tool. Doing good for me.

[fickle eagle]

good shout i haven't tried that out yet thank you

[jaunty ore]

And upscaling your images can really help to obtain better result.

[fickle eagle]

definitely, I've already seen that in action haha

https://cdn.leonardo.ai/users/2c8f0489-a0ec-4ac9-96d5-3ded72199e6c/generations/c219b1d4-b1e1-45af-9052-3cd7a7302dde/variations/Default_Amazing_Underground_HipHop_Concert_Poster_2_c219b1d4-b1e1-45af-9052-3cd7a7302dde_1.jpg

just wish i could customize the text in generations like this one

but i know that's prolly asking a little bit too much for now, the generations are of insane quality

[jaunty ore]

You can try work it in photoshop or something similar.

@fickle eagle I worked on this for an hour, what do you think?

[drifting sail]

hello all, excited to explore everything. Came over from evolving forest and just signed up on the website with my email

[fickle eagle]

definitely! that's the plan

[fickle eagle]

looks cool, just those lines in the forest background throw me off a little, but only on closer inspection

[fickle eagle]

ayoo

[jaunty ore]

got any clue on how I could prompt that?

[fickle eagle]

what's the prompt you've used so far?

[jaunty ore]

A graceful young woman with long red curly hair stands in a forest, her eyes full of determination

[shadow lantern]

hi, and gn for now

[fickle eagle]

hm hard to say tbh, I'd probably just try regenerating the background by itself if the background removal works good for this generation

[jaunty ore]

How you can only regen the background?

[fickle eagle]

use the remove background button, generate a new forest background and add both together in photoshop or similiar

[crystal inlet]

I'm trying to generate ONE woman and it keeps giving me two or three, with fused bodies. How the heck do I get it to stop? I've got all kinds of negative prompts about only one body, one subject, 2 legs, etc etc

[fickle eagle]

that happens for me sometimes as well, at least the multiple extremities part but i just generate more instead of trying to fix it inside the prompt lol

[crystal inlet]

Does Leonardo Style vs none make a difference for you?

[fickle eagle]

none looks more whacky to me

[vocal notch]

How do I start

With Leo style

[queen plaza]

Wdym?

[fast tiger]

He meant how to get access to Leonardo, but he already did 😉

[queen plaza]

Ah

That's everyone's question

[quiet dock]

Any1 want to beta test our game? Send me a friend request. Potential $ for beta testers.

[jaunty ore]

Follow emails they sent you

[analog nebula]

This is what I do. Generally easier to just generate in bulk and pick the best options

[analog nebula]

Under User Tools > AI Image Generation

[vocal notch]

Where is that?

[analog nebula]

If you have access to the platform. If not, just wait for an invite to be sent to your inbox

[queen plaza]

I may be interested but you can't advertise

[analog nebula]

It's fine

They aren't advertising an rival service of any kind 🙂

Just please do let us know in advance the next time

Also just an fyi that this in no way means we endorse such advertising

So it's an "at your own risk" thing

[analog nebula]

Could try image to image feature with an high init strength but specify an background in the prompt along with the original prompt