💬│general

[analog nebula]

Because our website isn't really touch compatible at the moment but I will talk to the team internally about it, though it might not be high on the priority list

[hardy peak]

Just image makes it look lime a touchscreen, basically for some reason my pc image treats itself like a tablet.

To ask is this also free and where can I find this?

[proper pier]

If a model keeps cutting off the head of a character, how can I fx this

[proper pier]

Community models

[midnight cloud]

On the website 😄

[hardy peak]

Ah okay

[analog nebula]

Finetuned Models > Community Models

Use the search function 🙂

[midnight cloud]

[analog nebula]

❤️

[hardy peak]

Ah that might work for me

Sorry lol

[midnight cloud]

All good!

[hardy peak]

Forgot that was there

[analog nebula]

Thanks for being around by the way, will slowly change over shifts to you guys

[midnight cloud]

Some really cool community models out there, definitely look around!

[midnight cloud]

haha it's all good! 😄 Feeling any better today?

[hardy peak]

I'm not sure if mine are on there but ya folks can try mine out, I noticed I couldn't create bandages or scarring so I fed a model that and got some of my designs that way

[analog nebula]

A bit but yeah the tiredness is settling in. I need more time away from screens 🙂

[hardy peak]

Just let me know if you want me to be help support for folks needing help on their generations, also got kicked put of a server yesterday for discussing AI

[analog nebula]

Ouch, was it an art server?

[hardy peak]

....that might also explain why

[hardy peak]

Ehhh? Its not appropriate to say WHAT the server was but it was sort of like that more for an animator (I also dabble in animation work and plan to get back into and use blender for renders and animation) and I was explaining using AI generated art to make framed landscape designs to sell for cheap, more just a neat thing to have nothing more (nothing digital, would only be physical copies due to not being able to protect copyright) and I was accused of being a snake oil salesman trying to put artist out of work and then kicked from the server and said owner blocked me before I could elaborate further

[analog nebula]

Yeah sadly you have to tread lightly with AI in art spaces

I'm quite wary of it as well being an artist myself

but most people tend to be more closed minded about it sadly

I try to approach AI with an more balanced artist mindset

[hardy peak]

True, I have said before to a friend that I'm not looking to put anyone out of work and would like to promote it as a bridge between tech and art, I even said one of the things I wanted to use it for was to sell references for people paying for commissions

I can't draw for *** however I can make okay animations and I enjoyed and still do enjoy learning how to be a prompt engineer, as it helps with knowing drawing is not something I am good at.

[analog nebula]

yeah it's going to be difficult especially at the frontier of new tech that will change things for a lot of people

[hardy peak]

Well with what midjourney is being sued for I'm not surprised by the hatred but I'm more.....I have a reaction image for this hold on

Before any replying

Here

That's how I feel bout the "misdirected anger" on AI art

I'm not above telling someone off if they pull some stuff

Like the one guy who used it to finish an artist piece and declare it as theirs.

Not sure how others feel on me with my soapbox but if asked I will stop

[analog nebula]

Well I mean you could? I tend to do the reverse, by having ai start an piece and then paintover it after

[hardy peak]

That works but the guy basically declared that the artist, who was livestreamjng the designing process and hadn't finished it yet, was stealing his work and it fell under copyright

I can understand that and jt also angers me both because making a similar design to something doesn't make you the original placeholder to what you yourself created with the machine, finished or not he made it first, sonic OC's be damned.

And it makes me mad cause all it does is make Prompters look bad, and ropes us into the NFT people and the frigging monkeys (and I mean the artwork and nothing else)

I have to specify the last part cause we don't know eachother behind the screen and there is always a possibility of being accused of any type of ism

I shall stop digging the hole now

[analog nebula]

this is true

[hardy peak]

I will be saving this conversation in screenshots and showing to said friend later as this is the first time I've run into an artists that uses AI with there work too.

[analog nebula]

Not actively but I do experiment with it because I can see where we are headed

It's basically like Edo period Japan and Western science

[hardy peak]

imediate thought process after trauma of world historg teacher using the last samurai for historical accuracy

Sabaton - shiroyama

[analog nebula]

lol

I was actually referring to Sakuma Shozan

[hardy peak]

Also I'm not sure if I can ask you about anime style designs but what would be the correct terminology for the design style of slice of life stuff like my dress up darling, uzuki-chan wants to hang out, and don't toy with me nagatoro

I've been trying to re-create with character designs and coming up with blotchwork

[analog nebula]

do you mean for prompts or

[hardy peak]

For prompting but also just due to the fact I can't find a correct name for the style and I'm trying to make characters with it.

[analog nebula]

sadly don't think japanese prompts work but you can try adding things like normal clothing , everyday clothing

oddly I found luck, though disturbingly so by adding kyoto animation to the prompt while using DreamShaper 3.2 model

[hardy peak]

I was meaning for how the animes look, like that!

Hold kn

[hardy peak]

Sort of like this like you can tell she's drawnish but she looks semi realistic.

[hardy peak]

Aside from the fingers, it looks great

[analog nebula]

I just use "anime style" and also append "kyoto animation" but I think we'll need an dedicated model as we don't have one currently so the style varies since "anime style" covers a whole lot of styles

perhaps "cel shaded" might help

kyoani has too specific a look

[hardy peak]

Alright, thank you for the assistance and for actually talking as I figure your busy

[analog nebula]

Oh I am not

just kind of overstaying though I already am on for 12 hours 😂

its my job but yes I should be resting

[analog nebula]

Hopefully you would be able to figure something out as this is of interest to me as well

But I do have plans in the near future to create an dedicate model for japanese style work so hopefully I'll make some time for it

[hardy peak]

if you need anymore employees i'd gladly help, not sure if it pays but i'd like to make money off of something like this already so i figure i might as well make money now.

[fast tiger]

The way to do this is to contribute to us as you are doing. For example, if you like to be an early contributor you can:

Participate more in #🖼│image-share, create twitter threads about Leonardo, give feedback about Leonardo, create useful models, etc. 🙂

I mean, be someone active here, because that enriches the community and we really appreciate it. ❤️

[hardy peak]

fair enough, i'm currently unemployed and looking for work, i also need to get my 2nd W-2 before filing taxes so i can make sure i get what i need for trying to start a side business, thank you for the response though @fast tiger

[fast tiger]

Well, I've been quick to say it, but you may be interested in doing it too, as you create connections with others.

By posting your projects or what you are creating a force with other artists/studios.

[hardy peak]

fair enough, fair enough,

[empty hollow]

Thank you so much!!!

[verbal sparrow]

I'm in my early 40's and I really feel the kind of excitement that ppl would have in front of the internet in the late 90's... so much promising opportunities are emerging right now, I'd suggest to start engaging actively early on bcz we can feel that it is only the beginning of the journey. This is an advice also to myself first and foremost.

[hardy peak]

thank you for advising me, i do appreciate it, have a pic for these trying times

[frigid wind]

Hello guys, I saw in some community images "Upscale 2" option, but I couldn't see this one in my generated images. Is it better upscale then just "Upscale". What is this functionality could you explain me please?

[fast tiger]

if two people upscale an image at once, there will be two different upscales

[frigid wind]

thank you, got it

[neat axle]

Added a new challenge to #1063537670715875539: Car Challenge

This challenge consists on creating the best looking car using the models of Leonardo Creative, Select or Signature.

[neat axle]

Most liked images in Leonardo this week

[sharp zenith]

So people can like our public umages?

[deep shoal]

Yep by clicking the heart

[normal fjord]

Hey all just wanted to introduce myself as a new member

I’ve been fascinated with the rise of ai tools and stoked to be a early adopter

Been really blown away by the level of detail and extent of possibilities in Leonardo. Happy to be here:)

[analog nebula]

Welcome and thank you for the introductions! Have you managed to get an invitation yet? 🙂

Also thank you for the compliments on behalf of the team! ❤️

[normal fjord]

Hey 👋 thank you. I just signed up on the site a few moments ago.

Only just confirmed my email

[normal fjord]

Well deserved

Courtier?

[analog nebula]

I see, #✨│priority-early-access should be visible to you now 🙂

[normal fjord]

What’s this magical role do

[analog nebula]

Yes it unlocks that channel

you'll be able to get an invitation much faster than though our website, just an thank you for joining us and introducing yourself ^^

[normal fjord]

Much appreciated! 😃

I’ll fill out the form now

Thanks @analog nebula 🙂

[analog nebula]

It's a pleasure! I'll send out the invitation immediately when you're done ^^

[normal fjord]

Phenomenal. Your awesome!

🤙

I’ll confirm in a second

[analog nebula]

Thank you!

[normal fjord]

Okay submitted ✅

[analog nebula]

Invitation sent!

[normal fjord]

https://tenor.com/view/spongebob-spongebob-squarepants-gif-26050988

[austere field]

Hi! General question, is/will it possible to implement Leonardo via API into a game directly, to generate random items for example? Or is that not a consideration due to traffic amount etc.?

[normal fjord]

Oh snap 🫰 I’m in!!!

Thank you again @analog nebula you made my morning

[analog nebula]

There will be an API available in the close future, so such an implementation should be possible. @marble moat may be able to give an better answer to this inquiry however 🙂

[analog nebula]

You're welcome! 😄

[normal fjord]

This is truly awesome. I’m having a blast. Such a powerful tool

[analog nebula]

Glad you're enjoying it ❤️

Do try the community models as well if you have not!

[normal fjord]

Oh!?

I’m loving the ability to generate better text prompts too

[normal fjord]

Where do I find this feature?

Found it

Wooooah

[void fulcrum]

how can I use my trained model to generate images? thanks

[normal fjord]

😆 love it

[void fulcrum]

never mind found out

[analog nebula]

Great!

Was about to type it for you

[zealous thistle]

How’s it going? My friend recommended I try out this app. I actively use MidJourney and have an account for a few miscellaneous pieces on Instagram @vacantbliss — How’s this tool? He seemed to be pretty stoked about me trying it. Lol

[analog nebula]

Welcome to our community! Please do check out #✨│priority-early-access , you'd be able to get an invitation much faster though that 🙂

[elfin shuttle]

Can someone tell me please where I can report bugs?

[analog nebula]

#989166677910519850 would be ideal 🙂

[topaz siren]

This platform is “groundbreaking” once thru the fog

[topaz siren]

“Fog” is user side 😄

[solid haven]

Hi all, is there any way to get Leonardo to understand backgrounds? I'd just like to add a background to this image but it refuses under all prompts?

a succulent joint of ham on the bone, cel shaded, magical world behind, inspired by borderlands.

Have tried, 'in background' 'surrounded by' and even 'object' in 'foreground' 'object' in background

Is there some colour code to use to help it understand it needs to add to the background?

[analog nebula]

are you using a custom model? Would like to know your settings so I can give it a try

[solid haven]

Yes 'Magic items'

[analog nebula]

oh you have an img2img

[solid haven]

Have tried every combo of settings, if you drop image init strength below 3, it will start making backgrounds but meat icon turns into something else entirely

Yes for consistency on UI, I've found img to img is the only way to go and no 'prompt magic'

Got 15 out great, last 2 refuse to gen

[austere field]

Not exactly what you want, but I'd use the "transparent background" feature and create another image to my liking, then just overlay them

[solid haven]

Yeah this is a good suggestion but doesn't work for nice shadows and lighting, may just have to give up but so close!

[analog nebula]

magical word background

forest background

perhaps

Don't have the exact model so I can't replicate

[solid haven]

Yeah I'm just wondering if the dev's have any suggestions on how the algorithm can identify and understand backgrounds

[analog nebula]

@mild walrus would be the person to ask 🙂

But most likely it'd be something for inpainting feature

which will be coming soon

hopefully within this month

[sly kettle]

Hey everyone, just joined, is there a way for noobs to play with this algo?

[solid haven]

Thank you

[analog nebula]

Yes, fastest way to get an invitation if you have not gotten one is to make an quick intro in #👑│introductions , which unlocks #✨│priority-early-access 🙂

[sly vapor]

yoy

[analog nebula]

But yeah once inpainting tools are released it would be much easier

[sly kettle]

Thanks Ayu!

[analog nebula]

A pleasure!

[analog nebula]

Welcome!

[errant seal]

hello 😇 ... excited to find new adventure in art

[long crest]

where do i find the information again about the differences in signature, select, and creative? i've been looking but couldn't find it

[fast tiger]

Hello and welcome, glad you are experiencing this new adventure. 😄

[fast tiger]

You are going to "Finetuned models":

[long crest]

Thank you very much!

[fast tiger]

My pleasure 🙏

[long crest]

I kept looking at the model list and the "added ones" thinking it would be there, should have just clicked the back arrow haha

[analog nebula]

Can happen 😅

You can press explore other models

In the model list

Would bring you directly to the Finetuned models page

[long crest]

Imagine that I was looking way too hard

[analog nebula]

We definitely need that site tour thing when we've more things in place

[long crest]

honestly it was that easy I was just "stuck myself" on the creating page and all the info I was finding was just on the community models so that was on me more than anything haha

and first thing I did was jump on the FAQ page rather than back out lol

[neat axle]

we should include this in our FAQ anyways haha

thanks for pointing it out

[analog nebula]

We definitely need more FAQs

[neat axle]

video guides are also coming soon

[long crest]

I'm glad I could help, as simple of a mistake as it was I'm sure it'll happen again hah

[analog nebula]

[neat axle]

I'm re recording some of them as they're a bit time consuming to create but worth it

[long crest]

I understand completely even if you're given the footage it can be painstaking to edit (especially if you're a perfectionist like most of us)

[analog nebula]

AI it all

Can't wait to be automated

[long crest]

Oh yeah I forget there is AI now 😉

[analog nebula]

😂

[neat axle]

well, the voice is AI generated

but sometimes it does some funny things, like mispronouncing

[analog nebula]

Oh

[neat axle]

let me drop a funny audio haha

[analog nebula]

Needs an Italian accent

Because Leonardo

[neat axle]

Leonardo... ayeeeee

[analog nebula]

Ayyyyyy

[long crest]

Aiiiiiii

[sand mortar]

hi all

How can I use the app?

[analog nebula]

You'll have to wait for the actual invitation to be sent to your email, it'll contain the application link. Fastest way to get one would be to make an quick intro in #👑│introductions then go to #✨│priority-early-access 🙂

[sand mortar]

Ok thanks! I know you should be tired answering this same question, so I appreciate it!

[analog nebula]

No worries, though it is definitely true! ❤️

[tame zodiac]

Hi

[buoyant wedge]

🎊

[hardy peak]

How does one become an early contributor

[neat axle]

some of the activities:

• participating in the #1063537670715875539 channel,
• sharing images on social media and linking the post in here,
• helping other community members in discord,
• sharing your models
• sharing your generations in #🖼│image-share and the upcoming image forums

[hardy peak]

Hmmm

I have twitter but I fear I shall be crucified, I will do some of then and show concept art of a logo for my company

[neat axle]

sure, those are some of the options, but mainly being part of the community (Discord) and such will be enough 🙂

[hardy peak]

True true but I like the AI and I am gonna be looking for using it for business ventures

[analog nebula]

understandable I do have a bit of difficulties as well since some of my friends are artists 😅

[sour token]

What's the requirements for becoming an early contributor? Should we be posting on social media. We have been working on a game using this.

[vestal patrol]

How do you become early contributor

[analog nebula]

Yes

[vestal patrol]

I already use the app all the time

[analog nebula]

That would be one way (social media sharing) 🙂

[vestal patrol]

Oh we're already talking about it lol

[tired violet]

Yaay

[sour token]

I think we just posted at the same time @vestal patrol

[vestal patrol]

Lol

[tired violet]

I am an @early contributor

Wait

Why I don’t have the role?

[sly forum]

NOT THIS AGAIN, IT'S EVERYWHERE

[fluid grotto]

I've been posting Leo images on Instagram daily

And making them into youtube shorts as well

[neat axle]

😮 can you share your IG? will love to see it ❤️

[fluid grotto]

Does this make me an early contributor?

[fluid grotto]

https://www.instagram.com/p/CnuhZsghXtB/?igshid=YmMyMTA2M2Y=

[sly forum]

I only shared images (with some prompt shown), does this make me early contributtor?

||don't ask why I put this video here||

[fluid grotto]

https://youtube.com/shorts/5PeZlxOTYuI?feature=share

[sour token]

I feel like its a bit more involved then just using this thing but I may be wrong

[fluid grotto]

Yeah. Contributor might involve more responsibilities than using it

[sly forum]

I SWEAR TO GOD

IM GONNA GO INSANE

IF I SEE HER ONE MORE TIME

[sour token]

old gf or something

[sly forum]

uh no

it's something like DÉJÀ VU I SEEN THIS GIRL BEFORE-

[obsidian cloud]

Stupid question. Where are people creating images? Discord bot?

Adding bot to our own Discord?

[analog nebula]

Though the web application

that would be SD's base training data

which would bleed over

Yeah we are built on top of SD with several custom solutions

so expect data from SD to be used as well

[hardy peak]

Question is there a time limit on any challenges as I have an idea that can be done

[analog nebula]

@neat axle would have to answer that

yeah

[neat axle]

at the time, no, we want to encourage participation

[analog nebula]

we don't use watermarked stuff to train our models

[neat axle]

once we see them getting more engagement we'll set up rewards and a panel of people that vote for them

[hardy peak]

Hmmm can the community make challenges or have a community challenges area? I have some ideas I'd like to promote if that's okay

[neat axle]

oh this would be a nice thing to have, @analog nebula can you set it up?

[hardy peak]

I have Lotta ideas using community models as well as trying to use the monthly days thing artists actually do

[hot mountain]

Hey I posted my work in insta but does leo have an account that i can mention? Also what are the ownership guidelines for leo generated image's?

[leaden harness]

Can this AI make spritesheets?

[hardy peak]

Lotta new people here, both good and suspicious

[leaden harness]

Suspicious?

[naive sparrow]

yes, it's not perfect but it can

[leaden harness]

Me? Never.

[leaden harness]

Sweet.

[analog nebula]

yeah it's a bit of a hit and miss right now

[vague oriole]

How can use this tool, who can help me, but, i not have acces for now?

[analog nebula]

Yeah you will have to wait for the actual invitation to be sent to your email

[leaden harness]

How long does that typically take? My pc is fried right now anyways.

[analog nebula]

Fastest way to make an quick intro in #👑│introductions then go to #✨│priority-early-access 🙂

[analog nebula]

See the comment I just made ^^

It's a lot faster than though the website since I'm sending them out personally 😅

[flat aurora]

how do I use the software to generate models?
Im on the website but I can only generate photos

[analog nebula]

There's no 3D functionality at the moment, but we will be looking to expand into 3D in the future when the technology is more viable

Currently we are doing R&D on it but we are unable to provide an estimate at this time

[flat aurora]

oh ok

so the model-share channel is for what?

[analog nebula]

Sharing the generative models

basically custom AI models that can generate whatever subject matter/style they have been trained on

[flat aurora]

ok

thanks

[analog nebula]

No worries 🙂

several animation frames yeah

yeah don't generate in odd aspect ratios

you tend to get oddities like that because the ai is trained on 1:1 aspect ratio

yeah the model is trained on 1:1 ratios, SD's like that

you get weird results typically if you go beyond a certain aspect ratio

better to extend the canvas once the tools for that are out 🙂

[hardy peak]

Try putting into the prompt 3d model and fiddle around with jt, also check community models as some of us have actually made some that could help you

[wintry oak]

how to create my art

in leanrdo

[analog nebula]

They are talking about 3D output specifically I believe

which isnt possible at the moment

[hardy peak]

Ah

[analog nebula]

You'd have to get an invitation to the web application by email. Fastest to make #👑│introductions then go to #✨│priority-early-access 🙂

[hardy peak]

I was thinking they'd want a concept of it first

[wintry oak]

im accepted and im ok

[amber sluice]

hi

how are we today?

[wintry oak]

so how to create in this platform

[hardy peak]

On the website there should be a place to get to the app

[hardy peak]

Bit suspicious of new folks and some kf the questions being asked.

[hot mountain]

Hey I posted my work in insta but does leo have an account that i can mention? Also what are the ownership guidelines for leo generated image's?

[hardy peak]

Check the terms of service on website

[analog nebula]

He's not a new person

😅

[hardy peak]

Nah I mean some of the folks that recently been showing up

And some of the questions got me a bit paranoid

[analog nebula]

All rights belong to the person who generated the image but do check the TOS, link is on the sidebar on the left 🙂

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

[spark cosmos]

try using magic prompt with "sprite sheet, frames, animated" on 768x1024, it's not much but it's a start

[snow oak]

Hi, how can I do to generate images

[hardy peak]

Animation is currently not supported, They can't animate, you'd have to find someone to do that or to do it yourself

[leaden harness]

Phenomenal.

[analog nebula]

will do, going to figure it out in team chat

[hardy peak]

Hard to tell who's new a bit as I'm recently using place aswell but I've been here a bit, just been silent

[analog nebula]

go to #✨│priority-early-access and register though the link in there 🙂

Will be sending out invites shortly

[spark cosmos]

you can then reuse the best outputs as input for the prompt magic img2img and expand on it

[analog nebula]

You seem to know what you're doing 😮

That's really neat!

[leaden harness]

It's a good start. Animation is hard.

[spark cosmos]

yeah, I wouldn't personally suggest this necessarily for animation. I think there are better approaches which are not currently supported on leonardoAI

[leaden harness]

Maybe in the future.

[spark cosmos]

thanks 😁

[naive sparrow]

https://cdn.leonardo.ai/users/fea491cd-7a5b-4d13-a5be-c5ad8fb9919c/generations/20179449-0209-43b6-9ae9-0df834db49ce/Isometric_Fantasy_a_monastery_at_the_bottom_of_a_mountain_3d_vray_render_3.jpgfor pixel art it can make reallt good mini enviroment(which can be the base for something more)

[unkempt grove]

Wait, so I should have been telling this channel all about all the people I have gotten on this platform to be a Early Contributor? 😏 Actually even considering using Leonardo in our game dev course for graphics 🥲

[leaden harness]

I have trouble coming up with art from scratch. I always have to base it off of something at the very least.

[naive sparrow]

something like that

[long crest]

great tool for that, I am the same way haha

[leaden harness]

SHOOK

[spark cosmos]

maybe if I talk about depth2image for sprite sheet generation here the team will implement it 👀

[cinder orchid]

Just signed up and tried this, can leonardo generate 3d models or just images?

[hardy peak]

I do 3d animation so I will suggest this for anyone who needs models and props

If you want just sfw use open3dlab

Alright that's there now for the other thing

[fast tiger]

Not for now, but we are working towards that.

[spice pond]

Howdy @neat axle 😎

[unkempt grove]

Tell me more

[hardy peak]

If you want nsfw models (note I have some but mainly due to the fact said models are heavily customizable with outfits, also DONT SHARE IMAGES MADE HERE IF SPICY, this ain't that place as far as I know) use ||smutbase||

[spark cosmos]

basically there are models that respect the "depth" of the image, so you can maintain the shapes while editing the content in a way.

So you could take a generic spreadsheet with a person running for example, and then change the content to be "spiderman" and you would get a running spiderman spreadsheet as you can see in the picture above.

[mellow elm]

Hi

[hardy peak]

For sfm (source filmmaker) models to port to blender use sfm lab

These are all made free and created for folks to easily download and I promote them as I use them

They have weapons, props, items, characters, etc for blender users like me, once I get all my stuff downloaded the plan is to re-learn animation and use some renders for Leonardo to see what I get.

Also need to ask random questions after folks start talking again as I not be trying to dominate conversation, I figured blender users here could use jt

Also there's supposed to be an add on for illustrator in blender

[cinder orchid]

how can i generate nsfw contents? When I tried that it gives "content moderation filter" error

[hardy peak]

On here or the app?

[cinder orchid]

the app

[hardy peak]

Hmm I have only gotten this from a few long prompts, normally I wouldn't ask this but dm me it or ask the higher ups about it as I'm not part of them, just a helper.

[wet falcon]

How to participate in contest?

[hardy peak]

Currently there's no contests, only challenges, please check in the table of contents on your left and scroll down to challenges for them

[neat axle]

briannn wsup

[spice pond]

we chillin'

wen spacedollars moon? 😄 good times

[fast tiger]

You have to go here: #1063537670715875539

[cinder orchid]

I can't add you and you don't accept messages from non-friends, just try something like "naked sexy girl"

[hardy peak]

Alright one sec, lemme get up and crack my neck, gotta finish the laundry and I shall check.

[wet falcon]

How to access modules?

[hardy peak]

You mean models?

On the app scroll down to and select finetuned models and you will see the options for the officials and community models.

Would show via images however I am currently bout to take shower, brb

alright, i have a banana, i have taken a shower, time for me to get back on logo designs and community stuff here.

[fading musk]

Hey! I have a question, not sure if this is the right place but if not I assume I'll be guided in the right direction 😇 Are you guys working on a feature where you as a creator get to decide what you want to publish instead of making all creations available for everyone to see instantly? Only publishing what you're proud of could be a nice way to build a profile and could also create that community feeling on the platform itself. That way you can follow people knowing they only publish their best work 🙂 I have seen it on some other platforms and really like it. Was just wondering if Leonardo.ai was also working on something similar.

[amber sluice]

that why?

[neat axle]

yeah we have this one: @leonardoaiofficial

https://instagram.com/leonardoaiofficial

[hardy peak]

before i respond to blade

there was a recent update today for the models you have create and images you generate on whether you wish for said model to be public or not, i would assume the model generations from it are hidden to the public

[neat axle]

just starting to post there some videos

[spice pond]

I am going to spend so much time with this lmao

[neat axle]

nope, the model is hidden but the generations are not, unless you apply for professional settings

[hardy peak]

reason being with the current lawsuit towards company, not sure the name but midjourney is based off of it, and they're being sued, so me hearing many people asking about legal aspects of the leonardo model makes me nervous that some of these people are looking to start drama, though i'm not looking to accuse anyone, just staying on my toes.

[neat axle]

let me know if you're a game studio/game dev and will help you get set up

[hardy peak]

may i ask how one does that? i'm looking to use it in a business idea and am wondering how that works?

[neat axle]

any drama will be handled swiftly, first with kindness @analog nebula and @fast tiger are very good at it, but if not, I'll bring the guillotine

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

[hot mountain]

Thank you, I'll mention the account ☺️

[cinder orchid]

I just want my nsfw images cmon bruh

no drama

[neat axle]

let me get back to you in a couple of hours

[cinder orchid]

😎

[hardy peak]

take your time, i'm paranoid for reasons

[neat axle]

we are very flexible with NSFW images, as long as they're not illegal

[cinder orchid]

i can't generate nsfw images tho...

this -- #💬│general message

[neat axle]

please try going to the settings and check if you have the NSFW setting on

[cinder orchid]

This is my setting

[neat axle]

what's the prompt you were trying?

[neat axle]

im not sure "girl" would be accepted

as that can create CP

which is forbidden

[cinder orchid]

oh

[hardy peak]

put "adult" in there

or "adult Woman"

[spice pond]

What is the best way to specify for pixel art? Trying to outline the exact number of pixels in the prompt, some issues

[hardy peak]

TIL that in using leonardo within 3 seperate tabs, the tab will generate from the model while the other tabs will still show the generations.

[cinder orchid]

how are yaw gonna monetize this, like how is this free?

[neat axle]

we'll have subscription tiers in the nearby future

[neat axle]

@mild walrus might be the best to answer this.

[cinder orchid]

any estimates on when 3d models for game dev would be supported?

[neat axle]

still in research phase

[hardy peak]

should there be a Q&A area in General with slowmode on?

like just a thread or something or a seperate channel?

[neat axle]

Yeah, this is a good idea, perhaps a q&a channel

[cinder orchid]

Are you guys accepting funding? Like, can I invest?

Or are yaw hiring?

[hardy peak]

read #📌│faq for further detail and no they're not hiring to my knowledge

[neat axle]

Are we hiring?
Yes, we're looking to fill ML and DevOps roles right now.

[cinder orchid]

Are we hiring?
Yes, we're looking to fill ML and DevOps roles right now.

[neat axle]

Seed round, looking for VC leads

[cinder orchid]

where r u guys located?

[neat axle]

Australia

[hardy peak]

not bad

[cinder orchid]

darn, taking remote positions?

[neat axle]

yeah

[cinder orchid]

can I send yaw my resume lol?

[neat axle]

open a ticket here and we can take it from there: #989166677910519850

[cinder orchid]

wow the weird job application i ever submitted

done

[hardy peak]

what's an ML?

[fluid grotto]

Machine learning?

[hardy peak]

huh, probably.

[midnight cloud]

Yes lol

Machine Learning.

[cinder orchid]

i can't find stuff about support for 3d models... research phase as in like, this year? eventually?

[midnight cloud]

No date set yet.

[cinder orchid]

darn

[hardy peak]

gonna be working on some stuff for promoting and getting other stuff done.

[midnight cloud]

Awesome! Sounds good. 🫡

[unkempt grove]

I have passed my first 2 weeks of introduction to data science and AI, I should really get that job now

[hardy peak]

what's the evolving forest?

[midnight cloud]

It's another project we partnered expedited access with.

[unkempt grove]

3D models are an entirely different breed I believe. The thing is that it "feels" so easy in my mind when thinking about it. Just to stretch out and "guess" volumes based on color relative to surrounding colors, then allow the user to "squeeze", "pinch" and "inflate" and voila!
Buuuut, I know that's not how it works

[elfin shuttle]

can models be trained by generating more photos? is there any way to train a model (Asides training with 40 images) to work better?

[hardy peak]

but WHAT is it? like is it a game, or is it something else, as i have no clue what they are.

[midnight cloud]

It's a sci-fi fantasy game with nft assets within it. It's pretty cool!

[midnight cloud]

Like having more outputs?

[unkempt grove]

I did love this idea though. Would be really sweet to have some sort of base model for sprite sheets. In the end, I would love something like being able to generate "body parts" for character creation etc. Right now, we get some really good looking models, but using them for mostly anything else would mean a lot of work.

Would also be nice to be able to "mark" parts of an image and label them, then use those labels for "re-posing" an image

[elfin shuttle]

No, like having more precise output.

[midnight cloud]

Your best bet is training.

[elfin shuttle]

is it possible to add an option for users to tell the AI a photo has flaws for example or its not what they are looking for to correct itself? something like an Algorithm?

[midnight cloud]

Not yet!

[midnight cloud]

You could try to prompt your way through but I find that to be a bit difficult and it doesn't work all the time.

[elfin shuttle]

Yeah, I'm good with it. I just thought I was missing something cause my new model couldn't depict lips or fingers in some cases correctly

[spark cosmos]

Clip art studio has "mannequins" you can pose, if you run a depth2image model on them you can obtain what you're describing more or less.

[midnight cloud]

Nope, you're good! Great questions also. Yeah fingers will always be a challenge, it's a hit or miss for me haha

[spark cosmos]

Is there a depth2image model planned @midnight cloud?

[midnight cloud]

Hmm good question, at the moment I'm not sure but it doesn't seem like something we can't do! Will take note of this! Have you tried prompting it in and seeing if that works?

Probably not the ideal outcome if you did it that way lol

[spark cosmos]

it's not exactly something that can be prompted, the depth2image model has a slightly different architecture I think, but certainly img2img can do a bit of what I described.

I also know that stable diffusion made an inpainting model with a similar albeit weaker ability. No clue how that works however.

[midnight cloud]

Noted! Will see if there's any plans for this.

[hardy peak]

will be making a google doc for a list to all previous and future reviews, will also use this for typing out the reviews posted.

[unkempt grove]

Know of any depth2image models?

[hardy peak]

@amber sluice

[spark cosmos]

base image, leonardo select, leonardo select, depth2img, depth2img. Here's an example for "a woman doing the v sign" and "batwoman doing the v sign" for non-depth and a depth model. As you can see the original pose is preserved much better in depth2img model, furthermore it loses the "white" of the original mannequin and instead draws the required subject. @unkempt grove @midnight cloud

[midnight cloud]

Got an answer for ya, there is research being done for depth2image but there isn't any ETA on it just yet.

[spark cosmos]

https://colab.research.google.com/github/camenduru/stable-diffusion-webui-colab/blob/main/stable_diffusion_v2_depth_webui_colab.ipynb

[spark cosmos]

okay, thanks 😁

[midnight cloud]

Absolutely!

[unkempt grove]

Nice! So this is basically for running a local SD model I assume?

[spark cosmos]

yeah, it's literally an SD 2.x model, I use colab because my laptop doesn't have a gpu!

[fast tiger]

Already have access to the platform?

[unkempt grove]

Damn. I have a RTX2060, so should be... OK

[midnight cloud]

Nice! I'm not too far off aswell, RTX 2070 super.

[unkempt grove]

Nice! I feel kinda scammed by the 2060, but now that I am a student... Well, affording something new is a lil' tough

• I spend money on raspberry pi's and tools and such instead

[hardy peak]

posted the doc in my forum area for anyone wishing to use it, i'll be using it for a sort of promotion personal image sharing area for me if that's alright with the team and yoss

[spark cosmos]

personally I just want to live my life on the cloud, a computer on the cloud, gpus on the cloud, gaming on the cloud

[midnight cloud]

Love Raspberry Pi's

[midnight cloud]

Lol the future mindset, love it.

[unkempt grove]

Until you get hacked on the cloud and someone steals the brain you uploaded to the cloud 😂

[hardy peak]

alright, that's done, now time to mess with character designing a bit more as well as work on the logos and mascot more.

oh yeah, also gotta work on that one selfie

[unkempt grove]

Same. I just got my hands on a third one, going straight into my little bracket of stacked pi's. May be building a new PSU to power all 3 at the same time

[midnight cloud]

Sheeeesh lol Goodluck!

[unkempt grove]

Haha yeah, I've heard PSU's are a little hard to build.

Dangit, guess I'll go re-install Stable Diffusion on local machine

Just re-installed my computer for the first time in 4 years 🥲

[midnight cloud]

Your computer is probably thanking you right now lol

[unkempt grove]

Haha yeah, kind of what I was thinking as well. As I was installing drivers for my mobo, something went a little wrong, so I backed out of it and restarted the install. Asus then decided to retick the boxes saying I wanted Norton, Daemon Tools and some other bloatware. I was so close to just re-re-install my computer again

[void fog]

its funnny I did that 😆

[spark cosmos]

now upscale it with leonardo 💪

[unkempt grove]

What in the world am I actually looking at

[spark cosmos]

stares in disbelief

[midnight cloud]

Lol for some reason I really like this.

[spark cosmos]

emoji time

[unkempt grove]

Yeah I really think it looks good, ngl

[spark cosmos]

your picture is actually Dracula

I didn't manage to fix the arms 🤣

[void fog]

That's what you don't want to see in the image with negative prompt 😅

[wary minnow]

/imagine cool man

[hot mountain]

hey, that doesn't work in this discord. if you want to use this kind of generation i would suggest visiting the midjourney discord.

[midnight cloud]

[wary minnow]

👌, how I can generate some image in this server?

[spark cosmos]

https://leonardo.ai/ you sign up, wait to be whitelisted and then go to https://app.leonardo.ai/

[midnight cloud]

If you want to speed things up. #👑│introductions , share a bit about yourself!

[unkempt grove]

This depth2image is so cool. Can't wait to try it out!

[spark cosmos]

using leonardo to finish the image is what makes the magic work though

[unkempt grove]

Yeah I get that!

[spark cosmos]

same picture before img2img and upscaling 😄

[unkempt grove]

HAhah daaamn. What kind of prompts do you give it to just remake a picture like that?!

[lone mortar]

i think i saw a site where you can input your image and it will generate the prompt

its new. still sick.

[spark cosmos]

https://openart.ai/promptbook I base myself off this promptbook personally, suggests a lot of nice keywords to get nice pictures

Check out the magic words words chapter, that's the juicy part

[unkempt grove]

Thank you!

[red flame]

Hi

[fast tiger]

Hi!

[unkempt grove]

Hello!

[midnight cloud]

Hey there!

[unkempt grove]

Tried a character sheet

[midnight cloud]

😂

[spark cosmos]

What is the original and what is the prompt?

[unkempt grove]

Probably a too varying image

I tried things like "A Knight. But also "Sprite Sheet of a Knight" etc etc

[spark cosmos]

Was that with the depth model?

[unkempt grove]

Yeah

On Collab

[spark cosmos]

I'll give it a shot too

[unkempt grove]

It was pretty damn good here though with my image from leo

[spark cosmos]

Yeah it's solid

You could try making less lines at the time

Or upscaling the picture

It might just be too many small things to change

[unkempt grove]

Yeah. That was my thought as well

Oh well, have to spend some time with irl fiancée, I'll keep experimenting tomorrow 😄

[gaunt pier]

It didn't get the geography update I'm guessing😂

[spark cosmos]

👍

[spark cosmos]

it's partially because you spelt it mediterranean wrong, the model is interpreting part of your prompt as "medit" as in meditation. Still quite a bit darker than you'd normally get on google I guess.

also I suggest using leonardo select to draw real people, unless you want it to be more of an illustration, then creative is fine

[gaunt pier]

Omg I was so oblivious to it😂 Well thank you for the free grammar lesson🤝

[cinder orchid]

Why is the sprite sheet cut off?

@unkempt grove how did you get a full sheet?

[spark cosmos]

@cinder orchid we had this discussion earlier in this chat, you can read through it. The short of it is that for sprite sheets, leonardo currently isn't the best way. It's great for pixel art and lots of other things though, and its upscaling is a beast

I think they're also working on developing some spritesheet models though

[hardy peak]

pardon @neat axle i have a question if you don't mind me asking it in #💭│off-topic since others be busy here

[cinder orchid]

Is it possible to get it to use the input image as the character rather than as an image style? Like, I'm looking for it to generate an image of that same character in the input but "attacking a flying ship", but looks like it's only using it as a style

[spark cosmos]

very very difficult, I'd start by turning off prompt magic so that it actually uses your picture as a start

my suggested workflow would be inputting an image with the flying ship and the creature, just a simple job on paint, and then running it through img2img without magic prompt at high init strength to fuse them well together

[cinder orchid]

don't think it worked... it just created variations of the same image

also, how long are the images saved? Can I expect it to still be here should I log in days later?

[spark cosmos]

they're stored on the website forever

[cinder orchid]

cool thanks

[midnight cloud]

https://tenor.com/view/forever-sandlot-the-sandlot-gif-13526090

[neat axle]

sure, just tag me

[hardy peak]

hope people take to the challenge

[zealous anvil]

Can anyone point me in the direction of instructional docs for generating a complete 3d model with this (if even possible)? I see the basic image generation capabilities but that doesn’t seem much different than Midjourney and similar things out there.

[spark cosmos]

the upscaling feature is much better than what MJ has for example, 3d technology is still in research

model fine-tuning is freely available as well

[dapper shale]

So much better than MJ in upscale oooh nelly

[spark cosmos]

prompt magic, forcing the model to really listen to your prompts, and also prompt magic img2img, which lets the model draw inspiration from an image rather than just modifying it

[finite wave]

😍

[spark cosmos]

also way better for realistic pictures

honestly, I just use both, each has their strong points

plus leonardo is free right now

[fast tiger]

Indeed, as say @spark cosmos, we are investigating 3D, as I told you in the ticket.

[zealous anvil]

The instructions from the email you referenced are TOS and email sign up. They really don’t explain the various settings. I understand if certain things are still in concept or dev but it would be good to further explain the functions available now.

[fast tiger]

That will be done eventually (it's in process), if you want a guide to creating models in Leonardo here's one:
https://app.leonardo.ai/faq

And as for the word prompts here is another one:

https://openart.ai/promptbook (It's for SD but applies to Leonardo)

[zealous anvil]

Okay.

[spark cosmos]

👀

[daring aurora]

💀

[long crest]

it's interpreted our language too much now it has sarcasm 😂

[dapper shale]

Burn on humanity ouch

[zealous anvil]

There is a delete function that doesn’t actually delete.

[fast tiger]

As if it went to the potato bin? For now there is permanent deletion.

[zealous anvil]

Yes, the permanent deletion pops up the message but didn’t seem to delete. Also, on the mobile view, text does not seem to wrap so a good portion of the prompt is cut off. This makes it really hard to mod it.

[fast tiger]

The message implies that everything has been permanently deleted. 🤔

Yes, but they also plan to release a version first for iOS and then for Android.

[zealous anvil]

Want me to send the screenshots showing the message and what fails to delete? The UI is also kind of funky because delete greys and then becomes active again after it is pressed. There is no normal confirmation that delete was successful. This is on iPhone.

[fast tiger]

Yes, of course, you can do it in the ticket you opened a while ago, please. 🙂

[vestal patrol]

aw man guys I was gonna go to waffle house, they're supposed to be always open but they were closed, how sad

[long crest]

dang if a waffle house closed it must have been a good reason

i've seen them run on half a person before lol

[analog nebula]

Waffles?

Do want some

[river holly]

The non-wrapping is the same way on the PC.

[light timber]

Mmmmm waffles

[analog nebula]

Sadly I did not place an order for one for dessert 😅

[unkempt grove]

Dude. How? Could you share your prompt? I mean obviously it's far from perfect, but its a step in the right direction

[violet stirrup]

haha I've been trying to do a sprite sheet myself today and I've not been having much luck.

[unkempt grove]

Nope, it's really hard. To be fair, the above one is not in Leonardo, it's using depth2image with Stable Diffusion running on Google Collab. I did get some pretty sweet results from a possibly usable 2D asset in Leonardo though! However, I would have to animate it myself

[little lintel]

GN

[spark cosmos]

It's actually not depth 2 image, but it's not leonardo either 👀

It's a really hard spreadsheet to work with though, I definitely suggest using a simpler one to start

[sly kettle]

gm🫡

[midnight cloud]

Gm!

[onyx wave]

how can i use a data set?

[fast tiger]

Already have access to the platform?

[onyx wave]

yes, just started a few hours ago

[fast tiger]

Oh very well, this guide can help you a lot:
https://app.leonardo.ai/faq

[gloomy kraken]

Nice onboard!

[fast tiger]

There is also another one:
Prompt Guide: https://openart.ai/promptbook (It's for SD, but applies to Leonardo)

[fast tiger]

Welcome! 🙂

[onyx wave]

I've already read it, so once you have your own dataset the images will be created according to the dataset?

[onyx wave]

Thx!😁

[fast tiger]

That's right!

[compact flicker]

Can I export a model to Maya or Unreal?

[onyx wave]

Okay, gracias!

[fast tiger]

A pleasure! 🙂

[flat aurora]

what does fine tuned model settings do?

[midnight cloud]

Gives you some parameters to use for your image(s), I would mess with them and see what they do!

[flat aurora]

thx

[serene oak]

yo

how can i use this?

am i whitelisted?

[long crest]

from what little I know, you can start here I think https://discord.com/channels/989166677390426132/1064120500781068319

[midnight cloud]

#👑│introductions go introduce yourself first!

[long crest]

Sorry, there ya go ^^

[serene oak]

ty sir

i'm already in

kinda early here lol

[long crest]

good deal and good luck

[serene oak]

can i make money on freelance using it?😁

[long crest]

I saw this on the faq
Will I own the images I create?
Yes, they are yours to do with what you wish.

[serene oak]

how can i create logo?

[midnight cloud]

Logo? Hmm you'll have to try to mess around with some of the models. Maybe a fine tuned model might help with this.

[proud canopy]

Hello guy, nice to be here

I am Whale Validator and want to have an early access

[fast tiger]

Hi, if you want to have an earlier access enter #👑│introductions and you must make a presentation to unlock the #✨│priority-early-access channel, you fill out a form in a few hours you have your invitation in your email.

[proud canopy]

Done. Thanks

Now waiting for the whitelist Email

[calm comet]

About the active community, will it also "honored" or "seen" as being active, if we post 3d models we made from the images we created?

or just the images itself?

[fast tiger]

What we think is that everything is welcome in the community (everything that enriches and benefits everyone), although for now we don't have 3D assets.

Precisely in the #📢│announcements it was put that being an active member of the community here on Discord is a key factor to become an early contributor.

[amber sluice]

I hope you had a spectacular day

I was a little absent because I was giving love to the family and pets

[fast tiger]

Thank you very much, likewise 🙂
I also spent today with my family. It is the best treasure one can have. ❤️

[amber sluice]

We must take advantage of them while we can.

[analog nebula]

post them actually, it would be nice to see how people have turned the outputs from Leonardo into 3D models in their own ways!

[serene oak]

ok ty

[cursive estuary]

I was curious, how does the magic prompt work? Is it an a.i. that changes the prompt or something like this?

[fast tiger]

In this tweet he explains what prompt magic is all about:
#🐤│twitter-feed message

[wooden glacier]

I would like to ask if the project has any plans to issue token

[hardy peak]

Finished the logo

[coral folio]

@young surge

#💡│suggestions message

I saw your suggestion but actually there is already a dropdown menu where you can choose from original, upscaled, without bg etc

[real wraith]

I want to ask, why i can't create my own models on leornado ai. I follow all the instruction and have upload some image i want to train it, and when I want to create the models it say "An error occurred submitting your request:
internal error" like this, can anyone help me with that. Because my first own models run as it and don't get that error

[analog nebula]

We have no plans to issue an token. Although we are an layer between web2 and web3 we do not have any plans to be an web3 project ourselves currently. 🙂

[analog nebula]

It may be an server side issue. You could try again now perhaps and it should work.

If the issue persists do let us know. 🙇‍♀️

[thick isle]

Morning Iam invested in evolving forest i have followed all steps for early access what is the next step to gain access? Thanks

[analog nebula]

Would be something for @rancid grail to answer

[real wraith]

okay thanks

[analog nebula]

No worries. Apologies again for the late reply

[real wraith]

it's okay everyone has another business

[analog nebula]

It is a weekend but I do check in from time to time regardless ^^

[real wraith]

I have to try it again but it still has the same problem

[analog nebula]

Understood. @marble moat fyi

Actually

[analog nebula]

Your instance prompt may be too long in this case

Apologies for the ping peter

Keep it to 1-3 words if possible

[real wraith]

ahh i see i will try to short it

okay

[analog nebula]

It's just an short string of text to trigger the model to use it's training data basically 🙂

[real wraith]

thanks, it works know

[analog nebula]

Glad to hear! 👍

[gray vigil]

so... has anyone really used ai generated images for a game ?

[viral gate]

Hey

[dusk moat]

i think its missing something

[thick isle]

Thank you I’ll wait for response

[analog nebula]

🤝

A pleasure fellow rainian

[wintry oak]

hi

how to create desgin in platform ?

I need answer right now please

guys ??

[shadow lantern]

this is great!

[shadow lantern]

what do you mean by "creating design in platform"?

[summer flame]

What is the technology behind Leonardo? Is it stable diffusion?

[analog nebula]

You'd need to get an invitation by email that has the web application link. It's not the one with the discord link

[wintry oak]

already accepted

but now I don't know using this platform

[narrow sandal]

Loving this AI !!!!

It’s super dope

[analog nebula]

Have you managed to sign in?

[narrow sandal]

It puts out 3D quality it’s next level 4K an it’s makes other AI look terrible the way it spits it out at a quick rate on point every time

And the magic button oh lord that’s from the heavens

[analog nebula]

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

[narrow sandal]

It’s very good for building concepts for ideas many different things

[shadow lantern]

Is there a way to have 7:4 or 16:9 images instead of square ones?

[analog nebula]

There are width and height sliders on the left sidebar

[shadow lantern]

cool I'll search for it

[analog nebula]

Do note that as SD based models are trained on 1:1 ratio images, you will experience strange results

Going beyond 1:1

Outpainting tools will be available soon that will allow one to expand the canvas with additional content

[shadow lantern]

having this on your own gallery is sooooooooooo cool!!!!!! I love it!!

[midnight cloud]

Glad you like it! Looks good!

[hot pine]

Leonardo staff, please make Leonardo Signature great again! The past few weeks it's gone from make beautiful images to creating the same vaporwave wasteland no matter the prompt

[long crest]

just.. a little 😂

[midnight cloud]

I don't think anything's changed to the point where you can't create beautiful images anymore lol 😄

[hot pine]

Definitely changed. I even used the same prompts for a four weeks ago. Nowhere the same images.

Also Prompt assist on Signature returns badly spelled text documents. Which would be cool every now and then but that's literally most of the generations... vaporwave wasteland or text images spelled badly

[loud mauve]

https://media.discordapp.net/attachments/980667360614678570/1019101541573066772/meme.gif

me watching this chat waiting for my access like

gonna be a long couple days 😂

[midnight cloud]

Won't be long! Just enjoy the views for now lol

[midnight cloud]

Could you share an example with me so I have a better understanding where the errors happening as far as the text goes?

[hot pine]

Sure. Where do I send the images?

[loud mauve]

Im writing a list of prompts to have ready haha

[midnight cloud]

Hahaha, I did the same thing. You'll love the prompt generator.

[midnight cloud]

#989166677910519850

[hot pine]

Ok. Will send them in about 30 minutes

[shadow lantern]

why are some images with watermark on it?

I mean this

[loud mauve]

its iterative yeah? can you request the same image with no watermark

[shadow lantern]

is this possible? I don't think so

[loud mauve]

I dont know. Ill stop until I get more familiar

lol

[midnight cloud]

I would give it a shot. Did you happen to use an image that had a watermark previously? Img2img?

[shadow lantern]

no... just text to image

[midnight cloud]

Try adding watermark or watermarks to the negative prompt.

[shadow lantern]

using this imagen as image to image?

it still gives me the image with watermark... I'm afraid I'm using some not legal images due to authors rights

well, not me, the app I mean

[midnight cloud]

There's a section where you can toggle negative prompts, it should be right next to the prompt magic button. Click that and input watermarks to that while using image to image.

[wintry oak]

guys help me

i can't

please help me

[shadow lantern]

I've just tried it.. but keep giving me the image with watermark

[shadow lantern]

is it nourrished by images with copyrights? just to know if I can have problems when using it on my feeds

[midnight cloud]

Ahh dang. I guess SD picked up some stuff with watermarks, weird. I don't think there's a way around it if negative prompt isn't working.

[midnight cloud]

What's wrong.

[shadow lantern]

that's something to fix... there can be problems with that I think...

[midnight cloud]

For sure. Noted

[sly kettle]

Sup y’all, can I get some advice on how to talk to the algo when using an uploaded image? For instance, can you somehow give it an image of a head, and have it use that head on a body, or does it only do variations of the image?

[hot pine]

Oh yeah forgot to mention that. It has started give a camera icon watermark now for the past week

[analog nebula]

Probably SD's underlaying data

[analog nebula]

What are the settings specifically, do you have prompt magic turned on, also which model are you using?

[hot pine]

Leanardo Signature was great for removing watermarks on Img to Img

[analog nebula]

Inpainting and selective features would likely be easiest to achieve such a goal then toughing it out with prompting, though currently they're still being working on. We should be implementing them hopefully within this quarter of the year

[hot pine]

I have getting them on all the Leonardo's and with and without prompt magic

[analog nebula]

I'll bring it up during the team's meeting tomorrow

🙂👍

[analog nebula]

Could you please kindly send a screenshot of what exactly you are having trouble with? It'd be much easier that way for both parties 🙂

[analog nebula]

Image generation AI typically have issues spelling because they aren't trained on language but merely what visually looks like language (pattern recognition etc)

So you get an semblance of text but it makes 0 sense typically or is not exactly correct

[hot pine]

I understand that...but what I'm saying is prompt assist in Signature only yields text images

[shadow lantern]

this

[analog nebula]

Ah. I have misunderstood

[hot pine]

Totally understandable

[analog nebula]

Will mention this in tomorrow's meeting. If possible could you send your prompt and the settings?

[analog nebula]

Got it. I'll discuss with the team about this as well 🙂

[stoic abyss]

I have a question. I want to commission an image via image-to-image feature. The original is that of a female humanoid, without any suggestive content at all. However, the image Leonardo gives me is sexualized, to say the least. Can you suggest any anti-NSFW and anti-nudity prompts for the Negative Prompts space that aren't over-the-top? Sorry.

[hot pine]

Prompt: Beautiful Sunny Day Negative off Prompt Magic: On

Also the Superman logo shows up quite often now as well

Prompt: Beautiful Sunny Day Negative: off Prompt Magic On

The camera watermark that started showing up in the past 3 days. Never got this one before in a month of using Leonardo

Gave you guys some of the bad...now the good....I will never use Midjourney as long as Leonardo is around!

[analog nebula]

Oh yeah we did notice the superman thing

[analog nebula]

Much love from the team!

❤️

Apologies for the late reply was away for a while

[analog nebula]

Try adding things like naked, nude etc into the negative prompt

[zealous anvil]

I’ve tried various models but they all seem to ignore prompts for perspective on portraits, It zooms in and crops close instead. Prompts to vary face worked best when ethnicity was specified. In the absence of that it would ignore some things like “big nose.” I also still wish skin was more realistic and less perfect/plastic.

[hardy peak]

As evil as this will make me sound, I plan to use both for making money, hope that don't make me a bad guy to anyone here.

[zealous anvil]

Not at all. Do it. I am sure many others will.

[analog nebula]

Perspective?

[hardy peak]

To make clear though I have no intention of using others artwork from Twitter to create art, I mainly grab the stuff I feed the models data for on pinterest then convert to jpeg format, as well as using multiple models

[zealous anvil]

For example perspective prompts like medium view, long view, show head and torso, etc.

[analog nebula]

Ah

[hardy peak]

What's in your data?

[analog nebula]

Perhaps adding an focal length might help too

[zealous anvil]

Can you give an example on how to word it?

[analog nebula]

Also try moving them higher up the prompt

[analog nebula]

'18mm lens'

'fisheye lens'

'telephoto lens'

'35mm lens'

Don't use camera for 35mm or it'd give the analog film look instead

35mm camera typically gives the film back look

[zealous anvil]

Will do. Thanks. Also noticed it is putting odd creases in necks (like extra wrinkles) and makes older women in photos look much older. A 40 year old woman looks 50 for example.

[analog nebula]

I suppose ai being ai..

Hopefully inpainting features will be helpful for that

Once they're out

[zealous anvil]

Overall realism is better than some ai I have tried for portraits so hope it will keep getting better. Will post some stuff on the other channel to show.

[analog nebula]

Glad to hear! And yeah hopefully the models will be finetuned further

[zealous anvil]

What is the correct channel for show and tell here?

[analog nebula]

#🖼│image-share or #1070290451560665158 or #1070592065366413362 (for personal galleries)

Last one is a little special

You get to make an post all about your own stuff

[crisp lotus]

hi how copyright work with the model

[stoic abyss]

Thank you very much for your wonderful answer. I am sorry but I was busy so I couldn't thank you right away. I will try it out tomorrow.

[analog nebula]

All rights belong to the person who has generated the images

[warped sleet]

how soon will i be accepted to leonardo.ai?
because I've been waiting since December.

[analog nebula]

To add though, you should read section 3 of the terms of service regardless: https://leonardo.ai/terms-of-service/

[warped sleet]

ok

[analog nebula]

You should have gotten an invitation by now, perhaps it's in your spam inbox?

If it isn't, you can make an quick introduction in #👑│introductions and then take a look at #✨│priority-early-access

Also the above response regarding the terms of service was not meant for your inquiry

[warped sleet]

I already did it on 👑

[analog nebula]

Ah

[warped sleet]

and i don't have spam

[analog nebula]

It was before the new role and channel got implemented

Give me a moment

[analog nebula]

Should be able to see #✨│priority-early-access now 🙂

[warped sleet]

ok

[analog nebula]

Invitations will be sent out shortly

[warped sleet]

okay, i see

[sharp sigil]

I am eager awaiting my invitation as well 🙂

[warped sleet]

if i made a poll when will i be added?

[sharp sigil]

Cool just got in. So I am interested in text to 3d. Is that going to be available soon?

[analog nebula]

Poll?

[analog nebula]

Not that soon. Tech is still in infancy so it might not be something available that readily

We are doing research on it however

[warped sleet]

ok

[sharp sigil]

Right but is that something that Leonardo ai is going to offer?

[warped sleet]

form*

[fast tiger]

As Ayu commented, it is being investigated and it is something that Leonardo will offer, but not so soon.

[fast tiger]

The process is very fast, in 24 hours or less you will have the invitation in your email.

[velvet bronze]

Good Morning!

[fast tiger]

Good morning!

[velvet bronze]

GM! have an awesome day!

[midnight cloud]

Gm!

[fast tiger]

Thank you, beautiful day!

[zealous anvil]

Some images with multiple people seem to just replicate the same pose or person over and over. Also saw some replication with city scenes.Strange shadows too.

[summer flame]

What is the technology behind Leonardo? Is it stable diffusion?

[analog nebula]

custom solutions built on top of SD 🙂

[summer flame]

any further info? any whitepaper, docs, anything?

[crisp lotus]

is there a way to generate images as private

[river holly]

It's clearly a scene from the original Body Snatchers movie...

[zealous anvil]

@analog nebula fish eye lens prompt did not give expected results.

[shadow lantern]

Hey does anyone know how to use the website

I can’t do anything after singing up

[midnight cloud]

LOL very cool actually.

[midnight cloud]

Make a short #👑│introductions and then you'll be guided to the next step to get access to Leo sooner.

[shadow lantern]

Ok

[zealous anvil]

Yes but I really, really want to be able to do camera photography angles especially mid range (showing half the body) and distance (showing full body) as well as different camera type effects. If anyone gets good results with that please say how you did it.

[zealous anvil]

Looks good. What was exact prompt?

[spark cosmos]

"fisheye" should be enough

I used leonardo select and prompt magic as well

CFG 7, 60 steps, although I doubt those matter too much

[zealous anvil]

Ahh maybe because I put a space in “fish eye”and maybe anime style confused it.

[spark cosmos]

try both with and without prompt magic, both work really

[fast tiger]

As this is not a crypto project, there is no whitepaper. We have a guide to create models:
https://app.leonardo.ai/faq

[spark cosmos]

yeah, fish eye is a completely different thing according to the model

[fast tiger]

Yes of course, just give us your Leonardo app username and when Ayu is here, she will take care of it.

[shadow lantern]

I don't know why in portraits upscaling eliminates good details on faces... Not upscale gives better results

[spark cosmos]

upscale uses your original prompt, if the details are mentioned in it you have better chances of preserving them

[shadow lantern]

but I mean... skin and eyes... are better in non upscale... for example: this is non upscale and the upscale

[spark cosmos]

you could reupload the non-upscaled version to img2img using 0.9 init strength, and give another shot to upscaling it for example, eyes however are notoriously difficult for AI to do

this will still not preserve your picture as much as you'd like, but it might give you an upscale that you actually like

[dire oyster]

So ive played with leonardo, and ive got to say this is the best online user interface for ai art, thank you DEVS!

[marble aspen]

What can I do to get a full body version of this?

[spark cosmos]

unzoom, but it's still a very basic version of outpainting so don't expect miracles

or prepare to try a bunch at least 😁

[marble aspen]

How do I go about that, like A dragon knight in full battle regalia, full body character, her sword drawn and ready to face her enemies, clad in black, silver, and gold colored armor; unzoom or something else?

[spark cosmos]

press the three dots

[marble aspen]

oooh, thank you!

[devout bridge]

You could also try image to image,. (the icon with a left and right arrow)
And try adding a head / zooming out with a new prompt.

[fast tiger]

From the team, thank you very much! It is nice to read this kind of comments. 😃

[spark cosmos]

here's another possible approach with img2img

[marble aspen]

Oh that's cool, so just multiple iterations adding more details

[spark cosmos]

they're distinct generations actually, they start from the same base image I did with paint though, didn't try upscaling them as I'm doing other stuff atm

if you get the colors right, the AI will fill in the rest for you though

and you can also just crop some good boots off of your generations, apply it to your perfect one, and start over from there with higher init strength on the img2img to avoid losing details

[dire oyster]

you can see the effort and love the team put into it, every bit of positivity is well deserved,

[shadow lantern]

Is there somthing in Leonard like the "seed" of midjourney?

[spark cosmos]

under show advanced, bottom left part of the generation page

[shadow lantern]

don't forget we were here from the begining when it will become a paying app!!😉 ❤️

[marble aspen]

Will have to do a better job of cleanup when I can get to my tablet station lol, but that works beautifully, thank you again 🥳

[shadow lantern]

hahaha oh my! how didn't I saw this!!!! and how do you obtain the seed from image?

[spark cosmos]

you can go on your own generations page and click on the image, in the details it will be present

[spark cosmos]

great job, inpainting will help a lot with cleaning up when it's implemented 😁

[shadow lantern]

that's soooooooo perfect!!!! I'm in love with Leonardo!!!!!

[fast tiger]

Yes, the team puts a lot of heart into it!

[fast tiger]

You guys give life to this place... A very beautiful community with great minds.

[soft bolt]

Are there any plans to generate 3d models in Leonardo?

[devout bridge]

From their FAQ channel: "Initially, the launch features are focused on 2D asset and image generation, providing model finetuning and hence stylistic consistency and control. This will rapidly expand to include a whole swathe of features, including 3D when a production-ready solution comes available (either via our own efforts or an open source model)."

[soft bolt]

The dungeons and dragons map generator is rad, but it only does grassy knolls, is there a way to expand its repertoire?

[analog nebula]

Thank you for helping out Vecna! ❤️

[analog nebula]

could try adding more specific prompts like sandy terrain etc

it may be biased towards grassy knolls since the training data might likely consist of mostly that

but it should be possible for it to generate other kinds of environments just by adding more specific words

[shy beacon]

This is the best of best AI tool

[analog nebula]

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

[hot mountain]

Hey, i wanted to make a suggestion but i cant figure out how 😅

My suggestion: i noticed that the button for upscaling isn't the same in the ai creation tab as in your personal feed. I personally think that the out Point arrow's are the better choice but I'm the creation tab this is the icon for outpainting.

[devout bridge]

There's a #suggestions channel

[ebon ore]

@Ayu, I wanted to to check if Peter was able to grant me access to keep all assets private. As I would like to test the AI with some real game assets, which are not meant for the public.

[proud canopy]

When are we launching

[analog nebula]

Apologies for the delay, it had been enabled but I was unable to notify you about it

[ebon ore]

Thank you, I checked a few minutes ago and did not notice any such feature. Where can I enable it?

[analog nebula]

@ebon ore Disable this option before generating to ensure they are private. Do note that this option does not affect existing generated images.

[analog nebula]

Do you mean the full product or?

[ebon ore]

And just to be certain, enabling this will ensure all my uploaded images (dataset) and all generated images will remain private and are not used to generate images for other users?

[analog nebula]

type / then followed by suggestion

Select the one that says suggest

then enter your suggestion and hit send/enter

[analog nebula]

Will however forward this to the team internally 🙂

[hot mountain]

Haha figured it out 😅

[hot mountain]

Thanks 😁

[analog nebula]

No worries! 🙂

[ebon ore]

@Ayu, can you confirm the images I generate, even the uploads (datasets) are all private and are not used to generate content for other users?

[analog nebula]

Datasets are private. As for images, ensure the option I've indicated in the screenshot is turned off before generating.

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

Again do check our TOS, but if your generations are private it will not be used to generate content for others

Nor will they show up in the community feed

[serene quarry]

Hii, i have a question

this ia makes u 3d objects?

or are only images?

[analog nebula]

Only images for now, but we do have plans to expand into 3D in the future

Currently we are doing r&d on that but are unable to provide an estimate on when it can be released to the public

[serene quarry]

Okay, Thanks !!

[analog nebula]

A pleasure!

[coral folio]

Aloha

What's the changelog of the last update?

[analog nebula]

Typing it 🙂

Seeing if I need an approval to share it but midas isn't around at the moment 😅

Will confirm with the rest of the team beforehand

[coral folio]

Gotcha, thanks! ✌️

[analog nebula]

👍

[shadow lantern]

hahahahahha sometimes Leonard is tired of working for us...

[tender torrent]

#💡│suggestions

[real wraith]

Yeah that mistake is common on ai

[vagrant harness]

A futuristic women face art line

[spark cosmos]

signature is a fine-tune of 1.5 which was originally trained on 512x512 images, I think this makes it more likely to have cloning instances when working at 768x768. Should almost never happen with 2.1 finetunes

[analog nebula]

Welcome to our discord! Leonardo.Ai currently isn't available as an discord bot so prompts do not work in here.

If you would like to get an invitation to the web application much faster, do make an quick intro in #👑│introductions , then take a look at #✨│priority-early-access 🙂

[keen rock]

It would be nice to have access to API to create my own bot for Telegram or Discord powered by Leonardo.AI. Is it planned?

[analog nebula]

API is in the works yes 🙂

Can't give an estimate but there will be an API

[keen rock]

Got it! Thank you!

[analog nebula]

You're welcome ^^

[summer flame]

lol white paper isn'y a term only for the crypto industry....I went through the FAQ, there's nothing informative there about the underlying technology and solutions.

[analog nebula]

Yeah we aren't really looking to disclose a lot at the moment because quite a number of our solutions are built in-house

[summer flame]

Solutions is a big word
I don't understand even 1 of them.

Weird.

Tech company that doesn't inform their users on their product? If your afraid someone else will do it before you, then be sure that someone else will.
SD is open source.
But i'm not saying anything about Leonardo.ai, because i have no idea what it's about!!!

[analog nebula]

Apple is a tech company

They have hardly made most of their things opensourced the last time I checked

We have our reasons for doing so, being that some of the features we are developing is proprietary

If you would like to discuss further about an whitepaper, @deep shoal would be the appropriate person to talk to.

[summer flame]

lmao. Apple make iphones, computers

No it's fine thanks

good luck.

[analog nebula]

And software for their hardware

[summer flame]

You seriously compare yourself to apple

you should consider a dif approach

[analog nebula]

I am not

[summer flame]

or just share some info lol

But i understandf

[analog nebula]

I am just saying that generalizing tech companies like that makes no sense

[summer flame]

i'm not looking to argue

[analog nebula]

I am not either

I'm just putting forth our reasons for doing so

[summer flame]

Cool

[analog nebula]

My apologies if it sounded blunt

[spark cosmos]

Midjourney uses literally the same technology and also doesn't disclose 95% of how they do what they do 🤷‍♂️

[summer flame]

It's not confirmed they use SD. Only speculations. Do you have other information regarding that?

[analog nebula]

We do use SD

[spark cosmos]

literally says which ones of their models are built off 1.5 and 2.1

[spark cosmos]

there's very little information that I am not aware of as a user that I would need to actually generate better content with this platform

[summer flame]

@spark cosmos stated that MJ uses the same technology

I was talking about MJ not leonardo

[analog nebula]

Same base technology as SD basically. Diffusion methods.

Just different way of implementation and they have their own inbuilt solutions to generating images

[shadow lantern]

Is the website built to be used on a pc or can it be used on phone also

[analog nebula]

PC. We plan to release native applications on both iOS and Android instead.

Might be difficult to use the website on phone since it was never intended to be. 😅

[spark cosmos]

it really is

[analog nebula]

Will be a lot easier once the mobile apps are out 🙂

[spark cosmos]

looking forward to that

[summer flame]

Yeah well obviously the underlying technology is similar. But it's not the same.

plus, training models with millions of dollars has it's own value.

[spark cosmos]

rather than asking for a white paper, why don't you share with us specifically what you are looking for? Perhaps it's something that even I can explain as an experienced user

[hearty mural]

Just wanted to say thank you to the team and devs ❤️ I got my invite and so far I'm loving it, I haven't experimented much yet but so far it's the best AI platform I've used for ease of use and good results right from the get go. I can't wait to have a go at using the datasets

[analog nebula]

On behalf of the team, thank you for the compliments 🙂

[summer flame]

So try it first lmao

[summer flame]

I don't have specific questions. I want to understand what's behind the scenes.

[analog nebula]

Certainly. We do have our own team of Machine Learning developers working on things as well (they have to be paid and such). I think the decision to not disclose what is behind the scenes is quite fair.

[summer flame]

Maybe ill develop my own UI for SD

[analog nebula]

You are free to do as you wish 🙂

SD is open source after all

But we aren't just going to be about SD.

[summer flame]

exactly

[shadow lantern]

@analog nebula do you know how many time we'll be able to use it for free?

[analog nebula]

Unlimited for now 🙂

Just be careful not to queue up too many generation jobs at once or you'd have to manually delete them 😅

[shadow lantern]

hahahahaha yes... my gallery starts to look like a real mess!!!!! hahahahah that's addictive

[analog nebula]

It's happening to mine too in ways 😅

Used to be mostly oil painting style portraits

But now there are plenty of mecha

[shadow lantern]

that makes me feel better, thanks! hahaha