🙋│image-help

[low bear]

Need help with your prompt? Can't figure out how to get the result you want? Want to figure out how to get that specific detail in your image just right? Ask here!

This channel is meant for any questions regarding help with image generation, from prompt building to help with the content moderation filter. So if you have a question, ask away!

[night raven]

Well I will go first! What kind of prompt or sentence should I use to make a knight/paladin hold a sword and shield properly? 👀

[warped locust]

Hello
/ image, / imagine or /create doesn't work here, Images are generated from the webapp

[tough terrace]

ok

[feral bough]

Just use an XL model

[night raven]

Lol

[frosty crater]

i can't generate a missing body part like ,calf ,and piece of leg and shoe and also i can' fix the hand on my image ,im trying using Canvas but i lost already many tokens ,anyway i can post the image here if needed

[feral bough]

Nice channel @low bear very good idea

[night raven]

are you sure about that? ;p

[feral bough]

Generally I've found the XL models to be more accurate with weapons. But it's no guarantee of course

[night raven]

can always give it a go for it

[stable carbon]

i've heard multiple people say that the pirate element makes people hold swords better. havent tried myself tho 🤷‍♂️

[cold jolt]

image to image no longer behaves like it used to. PLEASE FIX THIS. yes, before you ask, i am aware the tool has moved locations and is under image guidance. But it appears to no behave like it used to, in fact it seems to do nothing at all. if i have an image with strength turned up to 90% or more I should get a result nearly identical to my source image, instead i'm getting complete chaos as if it is ignoring the image entirely.

[low bear]

Hey there! Can you share a screenshot of the settings you're using under Image Guidance?

It could be you have the wrong option selected or the slider is turned off

[cold jolt]

that was my first assumption too, but i don't, not that i can tell

[low bear]

Could you send a screenshot of the settings?

[cold jolt]

nevermind. so i have to turn off prompt magic for it to work?

[low bear]

Ah yeah, prompt magic turns it into Image Prompt, which is a different feature

That takes the concept of the uploaded image and takes it into account in the generation

Helpful if you want to have a certain style for example

[cold jolt]

i'm trying to match styles between book covers, and it is no easy task, but hopefully this is an improvement from before and not a step backwards.

[low bear]

Ahh I see, thats awesome! It is definitely something to get used to but in the end I think it is quite the improvement yeah :)
If you need any more help don't hesitate to ask!

[cold jolt]

have you tried text image yet? I'm wondering how good it would be with typography on a book cover. so far AI has been random enough that i've had to frankenstien the end result together from dozens of attempts.

[low bear]

I haven't tried it yet personally but I've heard very great things! However, it is mostly for big text, small text can get very funky

I have however noticed that Alchemy with Prompt Magic V3 is fairly good at text

[cold jolt]

that isn't surprising.

[low bear]

and I mean freeform text, so prompt only

[cold jolt]

the REAL challenge is publishers won't allow you to use fonts that aren't licensed, meaning that for the moment there is no legal aparatice in place to allow for AI generated fonts.

[low bear]

That is kinda weird, so if you make your own text art it isn't allowed?

[cold jolt]

yep!

[low bear]

Thats crazy

but yeah the above are some images I generated with alchemy + prompt magic v3 without any image input, only a prompt stating what text I wanted

[cold jolt]

so basically i cheat, i have the AI text on a layer, and hide the original font underneath on a different layer, so that the PDF generated has something to reference and check for a license

[low bear]

hahaha smart

💀

what is going on with that face

that is the most cursed loki ive ever seen lmfao

also however not allowed on the discord so I'm going to have to delete that, apologies @broken void

[broken void]

😦

it was tasteful ruby.. but ok

[low bear]

this discord needs to be appropriate for all audiences as we do not have NSFW channels here

[low bear]

absolutely, but still not appropriate for this community haha

sorry!

[cold jolt]

i've messaged Amazon and asked them for them to reconsider their policy but got nothing back from them. I honestly think there is a massive disconnect and the people in charge have no ida how AI tools work.

[low bear]

damn

[broken void]

well i think we should have a nsfw channel but ok, point taken

[cold jolt]

oh i thought image guidance would allow me to submit photos to use as style inspiration but it doesn't appear to be working that way. you can not have more than 1 image to image? 😦

currently Amazon is asking all authors to voluntarily report if they use AI generated art, but their definition is so incredible broad that it includes basically everything done on a computer, so all photoshop tools, filters, resizing or automated image fills, basically everything. it is ridiculous. Their intention i believe is to target artists stealing other people's work, but instead they are just outlawing everything, throwing the baby out with the bathwater.

[low bear]

That is actually what Prompt Magic is for

When you turn that on you can upload multiple images

[sharp remnant]

#1161671205745590322 make a name adeel in style of nike

[agile adder]

hey guys, ive generated a image like a portrait, how can I "zoom" more on the face of the image?

I have this, but I want to focus more on the neck to the head, zooming more on the face like an avatar

I've written a focus on a face of a smiling villager.... and idk how to focus more on the face

[warped locust]

You can try "Close-up"

[tough terrace]

give a some prompt for making Electronics

[pseudo schooner]

morning all

[chrome slate]

Image-to-image in Leo now always produces the same pose. Yet, I've already set it to 0.10. Before, when you set it to this value, it would generate images with different poses depending on the prompt I provided.

[vast shard]

why it dosent add letter A

[vast shard]

this also makes this

all but not what i need

[open echo]

Image generation works on pattern recognition, so it doesn’t know exactly what a letter is, and what’s its purpose, only a bit of what it’s supposed to look like.

It’s also based on randomness and chance.

Try asking for “MA” instead as it’s a more precise pattern.

[vast shard]

it shows this

can you maybe try make one? @open echo

i just need one good logo

but it dosent work

[open echo]

Can’t right now 😅

I need to get may daughter to dance class and she decides NOW is the time to throw a tantrum

[vast shard]

oh sorry

[open echo]

Hey no worries 😅
It’s me who immature and chatting while I should be doing parent stuff haha

Just to show you an old exemple I did, some movie posters came out good, some were ok and others just were garbled mess.

Generating text is 100% a dice roll

[open echo]

Yup…! 😅

[agile adder]

thanks! I will give it a try!

[dim zodiac]

OOOooo my new fave channel

seems like this was made for me 😄

[night vault]

https://tenor.com/view/good-good-evil-smile-evil-grin-racoon-gif-17020508

but what is #professional-chat for again?

[open echo]

It’s “normally” for people who use Leonardo in their work.. professionally😜

[winter spire]

Initially I think it was intended as a sort of... priority channel for paid users? But it depreciated quickly to whatever it is now 😅

[cloud bane]

Hello, I am a very beginner and I cannot make images right?

I try to describe
Corgi and Shiba smiling, corgi has whole brown head with white nose and collar, shiba has long legs"
but it seems not working?

how can I remove the forehead white mark of corgi and make shiba taller

[dim zodiac]

this is a tough thing to get right, so don't feel bad about it! Two separate subjects with specific details often will get blended by the AI

are you using AlchemyV2 with this? We could try some weighting

[cloud bane]

yes I think so

[dim zodiac]

so is that your full prompt there?

[cloud bane]

Yes, and I try agian with
"Corgi and Shiba smiling. Corgi has no white mark at forehead but white nose and collar. Shiba is bigger with long legs and fox face."

Now I got 2 corgi

[dim zodiac]

haha yes this feels like it will be a bit of a token sink... I got two Corg-shibas

[cloud bane]

AI just hybrid them?

[dim zodiac]

When you are looking for specific details on two subjects and describing both, it does become difficult for the AI

there are ways around it, but it takes patience and effort

[cloud bane]

I used a image guide but seems not helping

[night vault]

it is prone to, yes. cant get a dog and a cat either.

[dim zodiac]

this is the prompt I am working with "a (brown headed Corgi: 1.6) and (tall Shiba Inu: 1.6) play together, both smiling, joyful, captured in professional DSLR"

the weights are added to try and separate the two subjects, but it will still fuse details... I'm adding in some image guidance to see if I can get it to separate them more

getting somewhere

still a bit of blending though

tried to get it more realistic annnnd FUSED

[night vault]

just concede to the great canvas @dim zodiac

[dim zodiac]

yeah well... I've already thrown some canvas in here 😄

but yeah, I feel we could generate something close enough to edit in canvas

I feel the AI knows "Corgi" far better than it knows "Shiba Inu"

which is true for a lot of breeds

@cloud bane it might be easiest to generate an image you like of the Shiba, then a Corgi with the details you like, then combine them in the AI Canvas

[night vault]

that’s a feature of ai

[cloud bane]

thanks a lot, it helps

[livid rune]

please sign up here : https://app.leonardo.ai/ in order to generate images

[livid inlet]

https://discord.gg/FmyCmNJZ?event=1168570928058343495

[spiral axle]

ow maybe she is creating my bird collage today 😮

[spiral wyvern]

my bird collage

[tame seal]

Great bird collage!

[hexed kraken]

Where do you guys make the images? I can´t use the /imagine here?

[night raven]

Heya there You can’t use /imagine here you need to sign up at https://app.leonardo.ai/auth/login if you need some extra help you can check out #🔰│getting-started have a lovely day ❤️

[hexed kraken]

Ok, thanks!!

[night raven]

You are most welcome make sure to post the images in #🖼│image-share once you generated some of them!

[buoyant dirge]

How can I see my rank here

Please tell

[karmic pulsar]

Go to #🤖│bot-commands and type !rank

[night vault]

you need some help?

[alpine carbon]

hello fellows! i'm getting difficulty in just remove an object from an image using the "erase " feature in Canvas. is this possible? i mean, when we use the erase button, we need to input a prompt but it will generate a replacement, i just simply want to remove it from the composition and not add a replacement. is this possible to do? 🤔

[low bear]

That is certainly possible, yes! I usually do this by writing a prompt such as "blank space" or describe the scene as if there was no character in it, in your case semething like "Empty fantasy building interior". I do recommend making the selection box bigger, as otherwise the AI cannot draw inspiration from the rest of the scene to fill the space.

[alpine carbon]

i will try this! thank you for the tips! 🙂

[low bear]

You're welcome, don't hesitate to ask again if it still doesn't work out, we're here to help!

[feral bough]

Yes it can be done, you can also try putting just a space in the prompt box. In addition to what Ruby said, I would make the shape of the erased section a bit less specific (i.e. erase a slightly larger area in a smoother shape)- if the erased area is shaped like a character, the AI might fill it up with another character again

[night vault]

one note, your canvas area is not ideal. AI only see what is inside bounding box is in your case almost empty. Make sure to have at least half of canvas cover your original image to let it take into context what it should fill in.

[hardy topaz]

so where exactly do you go to make image?

[livid rune]

https://app.leonardo.ai/ sign up here please

[undone jackal]

Hi everyone, If I want to generate a photorealistic image with a model wearing a specific bag, does Leonardo ai able to do things like that?

[open echo]

Image generation is based on pattern recognition, so it may look « like it » but the logo may be off a bit (but sometimes it comes out good 😊)

[robust gulch]

Does anyone have any suggestions on creating low waist clothing in photoreal? I have tried everything I can think of adding parenthesis to low waist, describing the hips or hip bones, saying that the clothes sit under the hips or waist or saying that the hips are uncovered. Any ideas at all would be appreciated.

[modern mica]

hello , i am newby , where do i need to write the prompt for generation

please help

[fathom owl]

https://discord.com/channels/989166677390426132/1080534053272436767 here is the link you need.

Maximus isn't going to like your name🤣

[spiral wyvern]

go to leonardo.ai 🙂

[feral bough]

There can be only one!

[half cairn]

How do I generate image is Leonardo

In

[safe sable]

ive been trying to create images with Leonardo and it doesnt seem to create what I am looking for. What am i doing wrong? my prompts are- little girl standing beside bed with a monster under it-ill get a little girl next a bed but no monster.

[spiral axle]

check DM 😄

[robust gulch]

Anyone have suggestions for celebrity faces? They always seem to come out as a either twisted or exaggerated approximation of what they should look like. Or sometimes nothing like them at all. They often have smudges on their eyebrows eyes or lips and depending on the person sometimes the whole face is picassofied. It seems to help if I turn the depth of field to low for some reason though.

[vapid lodge]

porque no puedo hacer imagenes en este servidor?

[pure ivy]

:wave: Welcome to the Leonardo.Ai community!

To use our webapp to generate images, please create an account on https://app.leonardo.ai
This message was requested by ooo3r1cooo. Our apologies if this message did not answer your question.

[little dawn]

I'm working in canvas attempting to get rid of an extra fairy wing on the male but I must have spent over 100 tokens so far and about all i can get is this blotchy color. And other weird stuff but mostly just different looks of this.

My bounding box is as big as I can get it.

Anyone have any suggestions?

[dim zodiac]

What prompt are you using to get rid of it? Also are you inpainting (with mask) or deleting and outpainting the erased area?

[little dawn]

inpainting with erase.

[dim zodiac]

I personally would erase the wing, take some of the background with it, then set my prompt to the "halloween haunted house... etc" leaving out the subject part

if you erase, you should use Outpaint

[little dawn]

ok. thats a bit better but still a bit strange. hmm. does the model matter? i used alchemy and photoreal etc which i can't use in here

better background but a skull hanging out in the air. lol

[night vault]

from your bounding box, if you delete the wing only one side it will try fill the wing back (given context) of one wing left

one way you could do is reduce width of render so it the box only see one wing which can be easily removed

[dim zodiac]

yes I would also suggest reducing the size of the box

as long as it's more than half full with previous image it SHOULD give you results that match

I would use something like Absolute reality to try and match close-ish with Photoreal

[little dawn]

ok. did those couple of things. I think I can live with this vrs.

[dim zodiac]

there we are

[little dawn]

thanks for the tips 🙂

[supple scaffold]

/imagne realistic, hyperdalhistic, depth, medieval

#🙋│image-help message

Alguém que possa me ajudar?

[fathom owl]

Have you tried image to image to hellp it recognize better?

[robust gulch]

Yes. I think it would work if I could use a variety of pictures but with the limit of 1 on photoreal it doesnt really help.

[snow finch]

Hello, anyone here with experience in trying to transform such sketches into images? I have tried several image guidance options and different models, to some result.
My main issues is that it does not recognizes the shapes as characters, and whatever I do, it adds fishes, corals, etc, INSIDE the characters.
Any help would be appreciated.

[maiden jay]

Hey how I can generate perfect image.
Like a Boy standing in night rainfall, sad/crying in gray jacket ,chibi

!rank

[fathom owl]

Perfect is relative. do you mean perfect for you?

[maiden jay]

Like I don't know how to explain

[fathom owl]

So there isn't likely to be a "perfect" image generated. Unlless youprecieve it to be. and someone ellse my not see it as perfect. That said it willl depend on your prompt and model selections as wellll as upscllaing etc toge the image perfect in your eyes. there is no 1 step process. unless you relaay like am image that was generated by the model. I know it isn't definitive answer but you are asking a rellitive question. practice and trial and error arre the things you have to do to make images you think are perfect

[subtle karma]

Just look for what art style u like then just try different things out

[subtle gulch]

Have you tried using the line art option?

[subtle gulch]

It’d also be helpful to include in the prompt exactly what you want, like a cartoonish clown fish and an octopus, which a tropical ocean backdrop.

[maiden jay]

Ty

[maiden jay]

Yeah sure.

[frosty crater]

Guys the prompt you used are long or short? i mean is better write a short prompt or this not so important? because i have a problem with mine

[winter spire]

Feel free to post your prompt/image example and what you're looking to improve.

[frosty crater]

this is my prompts : A professional cinematic shot, sci fi style of a sexy futuristic bounty hunter,half body shot, wearing a highly detailed, cybernetic suit ,black hair, green cybernetic eyes, ruddy skin, detailed skin texture, detailed clothing, cinematic, hyper-realistic, blockbuster movie VFX, set in a futuristic desert city, dystopian, sand crawler beautiful face, dynamic shot angle, (Highest image quality, outstanding details, ultra-high resolution)

negative : nudity, nipples, unrealistic, ugly, 3D model, render, photograph, badly drawn body, poorly drawn face, badly drawn clothes, badly cybernetic suit, washed out image, closeup

Im using Albedo XL model with Prompt Magic ,dimensions 512x768

my issue are that Leonardo generate more like colse up image than what i want

[snow finch]

This looks much better than anything i could accomplish, can you share with me which model, style, prompt, you used?

[winter spire]

You're asking in your prompt for a "half body shot" and a "beautiful face", so the AI is going to try to make those features prominent in your image. If you're looking for a full body image, try adding "full-length shot", "standing", or some other indicator of how much you want to see near the front of your prompt.

[frosty crater]

like this : A professional cinematic shot, sci fi style of a sexy futuristic bounty hunter,"full lenght shot", wearing a highly detailed, cybernetic suit ,black hair, green cybernetic eyes, ruddy skin, detailed skin texture, detailed clothing, cinematic, hyper-realistic, blockbuster movie VFX, standing in a futuristic desert city, dystopian, sand crawler beautiful face, dynamic shot angle, (Highest image quality, outstanding details, ultra-high resolution)

[winter spire]

Seems reasonable.

[frosty crater]

ok ,thx im going to try and post the results here 👍

@winter spire still the same

[winter spire]

Next step would be trying to ask for things like, "distant" or " running" or something that would force the AI to generate the rest of the figure

"armored boots" even

[frosty crater]

@winter spire A professional cinematic shot, sci fi style of a sexy futuristic bounty hunter,"full lenght shot", "CLEAR FACIAL FEATURES" wearing a highly detailed, cybernetic suit ,black hair, green cybernetic eyes, ruddy skin, detailed skin texture, detailed clothing,"armored boots" cinematic, hyper-realistic, blockbuster movie VFX,standing in the mid-ground of a futuristic desert city, dystopian, sand crawler beautiful face, dynamic shot angle, (Highest image quality, outstanding details, ultra-high resolution)

[subtle gulch]

Model: DreamShaper 7

Prompt: Naïve oil painting illustration of a cute cartoonish tropical fish and an octopus, cel shading, flat colour blocks, extremely sharp focus, vibrant ocean backdrop, elaborate consistent background elements
Style: None
line art: 0.88

[winter spire]

In this case, you would probably be well served moving your framing terms to the front of your prompt, to make sure those are read and included

[frosty crater]

@winter spire better but now the face are very bad

[winter spire]

The more distant an object is, the less resolution it's going to have to be detailed. You could try running the refiner/upscale on it, or perhaps generating at a higher resolution

[frosty crater]

as a free user i can't no more use alchemy, photoreal or the refiner, so i need try to generating at a higher resolution

[winter spire]

A downside of Albedo in this case, the upscalers don't apply to the XL models seemingly.

[frosty crater]

much better with a higher resolution @winter spire 🫡

[smoky silo]

How to create bro

[frosty crater]

moving "CLEAR FACIAL FEATURES" at the start of the prompt make her face very very good thx to @thin tinsel for the trick❤️ 👍

[sand spade]

Hello,how do I generate images on Leonardo?

[livid rune]

hello sign up here : https://app.leonardo.ai/ thank you

[robust gulch]

Any tips on seperating clothing colors? I'm prompting for a yellow shirt and blue jeans. Even seperated by commas and with 4 sets of parenthesis it just seems to pick randomly and even throws in white, grey or red sometimes.

[robust gulch]

I'm so glad I came to ask a different question and saw this. It has been driving me crazy! Thank you!

[frosty crater]

believe me i understand you 💯

[spiral wyvern]

playing around with image guidence.. haven't really tried the depth yet,, but I think the guidence is improved a lot. 🙂

[livid inlet]

Daily Challenge Is Here #1171600291754422322 TOKEN PRIZES!!! 🥇 500 🎁 🥈 300 🎁 🥉 200 🎁

[sinful torrent]

gm

[night vault]

put horn and tail in the front, things in the back is easily ignored

[feral bough]

@sullen ravine These kinds of images with very revealing clothing REALLY have to be spoilered.

[knotty kettle]

Hey, everybody, enjoy today!
I have a little question : what about spaces in the prompt? is this really necessary? Is there a difference between a prompt with and without spaces (I’m talking about spaces after commas)? for long prompts, we sometimes lack usable characters... I try to trick, but I doubt the coherence of this approach, someone would have an idea on the subject?

🤨 ❓

[robust gulch]

Anyone know how to keep proportions normal? I was getting pretty normal pictures at 776 pixels tall. I changed it to 1344 and it either stretches the person or doubles up a body part and stacks them on top of each other. Double legs, stomachs or chests.

[subtle gulch]

There is no difference with or without space after commas.

[subtle gulch]

That’s because the AI is trying its best to fill up the space according to your prompt.

One way to avoid the duplicates is to use negative prompts, but a more effective way is to provide more prompts about the backdrop for the AI to fill in the extra pixels outside the dimension it is trained on.

[robust gulch]

I'm on photoreal so no negative prompts unfortunatly. That would make alot of things easier.

More background description I can work on.

Thank you!

[feral bough]

Generally stay away from too high dimensions because you will get duplicates with that.

This is a common problem

You should mostly stick to somewhere in this area

But the 1360x768 is already too high in my opinion

[robust gulch]

I decreased the resolution and it was still stretching and chopping off arms even sometimes. After going down a little bit further and adding perfect anatomy and proportions to the prompt it is not perfect but good enough.

[feral bough]

So what were your final dimensions on the image that had bad limbs? Ive never had that problem on Photoreal

[robust gulch]

Too high

1344 x 2016

At 800 x 1200 with perfect anatomy and perfect proptions almost at the beginning of the prompt it is only an occasional issue.

I didn't even realize I had set it that high if I'm honest.

[feral bough]

At these dimensions you should have little issue. I try to keep the 'average' of the two dimensions (in this case its 1000) below 1000. You can always upscale afterwards for higher resolution

And "perfect anatomy" and "perfect proportions" arent really needed with PhotoReal if you use the right dimensions because it has terms like that built into the pipeline 🙂

[robust gulch]

I'll try it without it and see what happens.

I do wish Photoreal still had a negative prompt. Not necessarily to prevent weird anatomy but In this specific case to stop colors getting mixed up. I type in blue jeans and yellow shirt and I get that only sometimes. It often switches them or adds in white grey or red for some reason. Given I am pretty new at this and it may just be user error.

[robust gulch]

Yeah without perfect anatomy and proportions I'm still getting funky stuff. But it's not a big deal. I have room to leave them in.

[worldly zealot]

Has anyone tried to created a 2D game sprite sheet? e.g. an idle character or building? or an explosion?
I tried but I don't get what I want, e.g. I once tried to create a ship, but each ship looked completely different, had different sizes etc.
Or are my expectations wrong? 🤔

[robust gulch]

I've seen people use sketch book style that might do something like what you want but I haven't tried it so I can't say for sure.

[slender thistle]

/imagine realistic incense

[robust gulch]

You have to use the website not discord to generate images. At leonardo.ai

[steady plaza]

can i create logo with leonardo ? if i can, which model is good for it ?

[gentle current]

"How do I create an image in Leonardo AI?"

[smoky gull]

It creates a stunning photograph of a coastal sunset, in which the waves gently break on the beach, surfers enjoy the waves, the sky turns warm colors and the sun is just on the horizon. He adds seagulls flying over the sea to give it a more realistic and scenic touch. The image must convey serenity and beauty."

[livid rune]

to all new users please MAKE SURE you have signed up here : https://app.leonardo.ai/

[robust gulch]

This is the first time I've ever used the canvas. What is a good prompt just to fill in the existing space matching what is already around an area I am erasing?

[winter spire]

Sometimes just a space will do. " "

[robust gulch]

I wouldn't have thought of that. Thanks!

[robust gulch]

What is the best way to use the canvas tool to remove something from an image and match the area around it. I have an image I like but it has some random red string? across the chest. I have tried erasing and then in painting but it just fills in the line with beads or stitches even with negative prompts. But this is only the first time I have done this. Anyone mind helping me out?

[crude mango]

Xmen storm realistic photo, make her a nigerian woman, elegant, powerful. show lighting behind her. - @vocal fable (fast)

[dim zodiac]

uncertain if you're trying to prompt in the discord? We only generate online

app.leonardo.ai

[feral bough]

This is Leonardo.ai, not MidJourney.

[noble kettle]

Person in the living room of this house watching his cell phone

Imagina/Person in the living room of this house watching his cell phone

[feral bough]

Come on, just read a few messages above you and you'll see that spamming prompts here does nothing

[subtle gulch]

Of course you can. There are plenty of samples on the community feed and also tutorial videos on YouTube in how you can create those.

You may start with Dreamshaper 7 and the XL models and include “vector logo art” in the prompt.

[subtle gulch]

One prompt I usually use to remove items is “background, consistent style”. If it doesn’t work well, I prompt for exactly what the background should look like.

[robust gulch]

Thank you! I'll try it as soon as I get a chance.

[steady plaza]

Thank you

[worldly zealot]

Thanks for the hint, I will try it out!

[unborn jacinth]

image-help

#🙋│image-help

[livid inlet]

Daily Challenge Is Here #1171962520316162098 TOKEN PRIZES!!! 🥇 500 🎁 🥈 300 🎁 🥉 200 🎁

Weekly Challenge Is Here #1171239119582666853 TOKEN PRIZES!!! 🥇 1500 🎁 🥈 1000 🎁 🥉 500 🎁

[magic quiver]

Hello everybody , i have a question , are leonardo ai canvas generated images private or are they shared to the public just like the image generator

[raw thicket]

Work you do in canvas is stored locally.
It isn't stored online and isn't publicly viewable.

[devout kraken]

I'm having a hard time using the negative prompts. is there an example as to how we should provide them?

e.g I told AI "No texts" in negative prompt (using the API) but it still generated some images that included texts. adding no distorted faces or hands worked wonders though. any thoughts?

an example of said generation

prompt: No distorted hands, faces or any body parts that might be seen as such. No diagrams. No text. No detailed faces. No bright colours. No brands or anything that might cause a copyright infrigement. No pens. No books.

[winter spire]

Both your prompt and negative prompt should be given in positive terms.. Take a look at some of the prompts on the Community Main page for some examples of phrasing and terms.

[devout kraken]

thank you, I'll keep that in mind

[winter spire]

@hasty shore Leonardo images are not generated on Discord. Please sign up at https://app.leonardo.ai/auth/login and have a look into #🔰│getting-started

[knotty kettle]

TY for your answer, we tested, thank you also for your confirmation. enjoy your day! cyah!

[livid inlet]

Daily Challenge Is Here #1172324934941409391 TOKEN PRIZES!!! 🥇 500 🎁 🥈 300 🎁 🥉 200 🎁

[robust gulch]

How do you prevent a piece of a prompt from bleeding over into other parts of it? Particularly colors? If I mention for example green eyes suddenly in half the images the clothes or background or hair are green instead.

[winter spire]

Color bleed has been a recurring issue. Sometimes grouping colors with their objects will work, sometimes it'll be ignored. If you're planning on specifying colors, I'd recommend specifying more than just one color, that way your image doesn't become immediately flooded.

But I'm sure there is a better and more complete answer.

[dim zodiac]

(psst, ask them if they pay)

[winter spire]

Oh hey, are you on a paid plan?

(shoot, forgot about i2i)

[robust gulch]

Yes

I do group them with the object and usually I'm listing 3 or 4 different colors throughout the prompt but getting them to stay with the specified object seems to just be a matter of luck.

[dim zodiac]

ok knowing that you're on a paid plan, you have a few more options

are you using alchemy? If so you can try to weight the different specified colours

multiple colours in the same prompt will become more and more difficult to control

[robust gulch]

I'm using photoreal so no alchemy

Ok I'll just keep making heaps of the same prompt and keeping the ones that match. Thank you!

[dim zodiac]

you have options still!

the best would be Canvas

you can generate the image with the majority of the correct colours, then mask and recolor the eyes, or the coat, etc

[robust gulch]

I've only used canvas once. I guess I'll have to learn how to use it better.

[hushed oar]

how do i make flat printable labels? every model and prompt i use give me a picture of an actual bottle with the label on it.

[winter spire]

I'm sure somebody's made a good guide for using Canvas around here somewhere... maybe start here for a good primer? https://discord.com/channels/989166677390426132/1126679661741609120

[robust gulch]

Ok thanks!

[dim zodiac]

are you trying to get it to generate text? What kind of a label are you looking for?

[hushed oar]

i want to print a label for a bottle of homemade bbq sauce, no text, only the "theme" i need, but no matter how i prompt it, negative and all, i keep on getting pictures of a bottle rather than a flat label

[winter spire]

For reference:

[dim zodiac]

aha good sleuthin

I would just ignore the "bottle" portion

[hushed oar]

yeah, exactly what i get every time

[hushed oar]

but how do i explain what it is for then? just mentioning label for barbecue sauce without mentioning bottle whatsoever?

[dim zodiac]

Barbecue sauce label, masonic elements, dark colour and gold, flat illustration, vector art, UHD

I would start with something simple like that

now... still a long way to go 😄 need to finetune it to not give so much garbage text

but this could be printed and stuck around a bottle

same prompt, does give a bottle, but could still be a useful label

I'd buy your Bour Sauce

😄

This was done using the new AlbedoBase XL model

[hushed oar]

yes, that worked so much better, thank you very much!!! now, since we're talking.....what's the best way to get rid of any letters and words? i'm having a hard time even when i use negative prompts...

[livid inlet]

Put bottle in the neg prompt as well, it works for me not getting t-shirts when asking for t-shirts designs.

[dim zodiac]

negative prompts are the way to go

unfortunately the AI knows that a label usually contains words

so it tries to put them in there

instead of calling it a label, you could call it a sticker only

[livid inlet]

Put text in the negs as well, not always effective but worth a try

[dim zodiac]

and put "text, words, watermarks, typefont" in your negative

[hushed oar]

yeah, i've tried but it keeps on giving me bottles, i think it's because it's for a sauce so it automatically assume i need to see it on a bottle....idk....

[dim zodiac]

yeah, it's both very advanced, and very stupid

[livid inlet]

What theme are you after? I'll give it go

[dim zodiac]

it'll make connections that you don't want 😄

what are the key flavours in the sauce

or what do you want people to think about

[hushed oar]

it's for my Lodge bbq, so i want masonic elements including george washington (which is the name of my lodge)

[dim zodiac]

git'r dun son

ship it

stock it

I'm buying a full VAT now

high definition sticker, masonic elements, dark colour and gold, campfire and George Washington, flat illustration, vector art, UHD

play around with that and see if you like what you get

hell, take these images if you like them enough

[hushed oar]

thank you very much kind stranger!!! i'll tinker with these and try to get a little more BBQ in there 😆

[hushed oar]

looks great, can you share the prompt please?

[livid inlet]

Logo, George Washington, Compass, All Seeing Eye.... LDXL model

[hushed oar]

that's great folks, thank you very much, still a bit of work to do but your suggestions helped a lot! hope i can return the fav....

[dim zodiac]

be sure to tag me when you get something you like please! I'd love to see what you come up wwith

[livid inlet]

Logo, George Washington At A BBQ, All Seeing Eye, Square And Compass

[tall jasper]

robot con una tijera en la mano id:guide #💬│general

Barbecue sauce label, masonic elements, dark colour and gold, flat illustration, vector art,
I would start with something simple like that

[karmic pulsar]

Hey guys, I have an oral presentation this morning. I just have to create a first magazine page for the Sephora brand (French perfume and cosmetics brand) by putting perfumes, cosmetics and skincare products on this page. This Magazine will be to promote the Christmas holidays. Thanks in advance

[tulip widget]

Im' new here and I have no idea what to do. This is different and new for me. How do I make images?

[robust gulch]

You have to go to https://leonardo.ai/

[low bear]

Does anybody have any advice for the settings/prompt to use to generate icons in a simple flat style such as this

[karmic pulsar]

mujer llevando niño al colegio con bebé en sillita y con vaqueros y camisa blanca

[obsidian yoke]

/imagine In a magical forest, there lives a little animal called Giraffe Lele. Lele's neck is longer than other animals. Whenever it stretches its neck on the treetops, it can see the beautiful scenery deep in the forest.

[livid inlet]

Do you know what type art that is, what you'd call that character? I got this, but can't seem to make the character that solid white you are after.

[low bear]

It is a vector SVG icon

[livid inlet]

Getting closer!

[livid inlet]

Its hard to get the toilet character without a toilet lol

And I can't seem to get it simple enough

Very Basic Simple Logo, Flat, 2D, Icon Of Solid White Toilet Stick Figure Working On A Laptop, Vector SVG, Black Background..... using the LDXL and LD model's

[low bear]

Yeah this is my problem too

[low bear]

Thanks tho, you're getting much better results than me!

[livid inlet]

I think it will take either making a model, image to image or maybe image prompt to get any better tbh

[subtle gulch]

🤔

Perhaps you can try the keyword “isotype” and see if you can achieve better results?

[robust gulch]

Any tips on keeping the background in focus? I have the depth of field set to high and it doesn't seem to make a difference.

[low bear]

You could put "blur" in negative or "no blur" in positive

[robust gulch]

I'll try no blur. I'm using Photoreal so no negative. 😥

Still blurry. If anyone has an idea I would be very grateful.

[feral bough]

PhotoReal does have a Real (no pun intended) preference for depth of field (i.e. blurry background)

[robust gulch]

Ok just something I'll have to live with then. Thanks!

[winter spire]

@near dune Leonardo images aren't generated via Discord. Head on over to the website and sign up to begin creating images -> https://app.leonardo.ai/

[winged crypt]

Where do you even begin to make the images? I can't figure where to start...

[winter spire]

Leonardo images aren't generated via Discord. Head on over to the website and sign up to begin creating images -> https://app.leonardo.ai/

[silk holly]

Dragonite Pokemon for coloring page, clean outline, winter in the background

[fathom owl]

https://discord.com/channels/989166677390426132/1080534053272436767 you can't generate images here or in any channel in this discord. follow this link and folllow directions!

[livid inlet]

Daily Challenge Is Here #1172687229039153162 TOKEN PRIZES!!! 🥇 500 🎁 🥈 300 🎁 🥉 200 🎁

[dim zodiac]

@low bear @livid inlet using "vector art" and "clipart" will help you achieve that look

[robust gulch]

Would someone mind taking a moment to explain to me what Raw mode and Contrast boost do? The little question mark doesn't really make sense to me.

[subtle gulch]

RAW mode: toggling it on can help generate images closer to your description instead of following the presets of your choice of model or alchemy.

Contrast boost: the higher it is the more contrast the image will have. In other words, dark gets darker, bright gets brighter.

[robust gulch]

Darker or more saturated?

Nevermind some googling answered that.

Thank you for the help!

[subtle gulch]

The left one has it set to 0.0
The right one has it set to 1.0

[winter spire]

Contrast Boost: 0 vs 1

You're quick on the draw, Siloa 😅

[ebon palm]

create

[subtle gulch]

Leonardo doesn’t have a discord bot.
Please go to https://app.leonardo.ai to register and generate images there.

This discord channel is for image sharing and discussion only.

[robust gulch]

There is less of a difference than I expected with the contrast boost. Thanks for the help guys!

[ebon palm]

Thanks

[subtle gulch]

Feel free to come and ask questions, join challenges for extra tokens or simply share your masterpieces.

[cerulean spire]

Hi all, I’m trying to create an image that represents myself. I’m originally a construction/civil engineer, but work in import/export and supply chain. At the same time, I’m a clever software developer, and have an interest in finance/stock market.
I played around with several prompts and models, but can’t seem to find myself in any output. I’ll share some, and please help me out with either prompts to try or your own trials to recreate them.

[night vault]

nice side by side comparison, the effect is too subtle tbh

[fathom owl]

so are you trying to get al of those aspects into one image? that many different things is a big ask for the ai

I fee llike it did a pretty good job based on the description

did you prompt that specific art style or was it random?

[subtle gulch]

It is difficult for the AI to generate many different aspect separately in one image. A more effective approach would be generating different components separately with similar style and use the canvas function to combine, merge and modify them.

[cerulean spire]

Aaah, thank you.. I didn’t know about the canvas function. Thanks sooo much for the idea 💡

[cerulean spire]

I’m very new at this, so I wasn’t following a certain standard, just changing the prompt after every generation. I also went through the public images and tried to use the art styles of the ones I liked. Not sure if there’s a resource that shows the different art styles I can use in my prompt.

[fathom owl]

https://discord.com/channels/989166677390426132/1125488677095739514 https://discord.com/channels/989166677390426132/1091125166064533504 Here are some good resources to learn art styles and appllications. It's a good place to start.

[cerulean spire]

Amazing resource, thanks so much. Exactly what I needed

[fathom owl]

no worries . I found it very useful! should be right up your alley!

[tall jasper]

dog black

The left one has it set to 0.0
The right one has it set to 1.0

[livid rune]

https://tenor.com/view/infp-yeah-about-that-gif-25010945

[compact trellis]

Hello, any advice about how I could generate or improve a prompt to make realistic environments and rooms ? There's so many keywords available, what would you use ?

For now I have: wide shot, subject, photography, photorealism, 8k, hyperdetailed, cinematic lighting, colors I want emphasized.

[dim zodiac]

That's a good start! I would add some camera details as well personally

Minimally "captured in professional DSLR"

[compact trellis]

Ah yes, that's a good word.

I need to deep dive into the large pool of guides fo Stable Diffusion. I roughly know the order of keywords one should include but damn There's so much by categories like what medium, styles ect.

I also need to figure out a prompt to make cute recognizable emoticons of characters. 🤔

[quasi dove]

i need to create an art like this but i struggle with the prompt

its an abstract art

[fathom owl]

Be sure to use models that best handle reaistic outputs. The Leonardo visionxl, diffusionxl and albedobasexl do well with it and in more basic absollute reality 1.6 and dreamshaper do well for me. if you are paying go with photereal and you should really have no issues as long as propmping is good.

[fathom owl]

Are you using abstract in the prompt? ie abstract art, abstract style etc? also try adding specific abstract artists to the prompt like "in the style of "xyz.. you can also fine tune the look with things like ink wash, oil paint, whatever.

[quasi dove]

ok ill try

i use abstract style

[fathom owl]

okay if you are using abstract style just try giving it support with other keywords. also you may want to try other models. if you go to the community collections you can find models trained specifically for abstract art style and even specific artists.

[warped locust]

Hello, Does anyone have a prompt for E-sports logos?

[wanton belfry]

like this?

[warped locust]

More on this style

mascot sports logo

[wanton belfry]

any specific animal in mind?

[wanton belfry]

something like this?

[warped locust]

A griffin

[warped locust]

But I wanted him in full body, I'll see if I can do it

[wanton belfry]

it can happen might need some additional push

[warped locust]

yeah

[wanton belfry]

there you go

[warped locust]

Wow, that's kind of what I wanted to do

[wanton belfry]

this is through image prompt

even better

[warped locust]

I liked this, I'll try to reproduce it here, thanks for the help

[warped locust]

Image Guidance?

[wanton belfry]

yes, use this image...you can maybe even img2img this

[warped locust]

thanks Kentox

[wanton belfry]

yea img2img is better, just use folklore element

[warped locust]

can you send me the initial prompt?

[wanton belfry]

Stunning Logo Design with Griffin Mascot, Fullbody, E sports, flat colors, simple background, traditional design elements

[warped locust]

thanks

[wanton belfry]

no problem

I like to help wherever I can, however I have no clue about training models, Canvas etc

[karmic pulsar]

hi i need help . how to work with the peoples face, as you can see the face of the peoples are ugly and deform. how to make it perfect .

[livid rune]

using the refiner (8 tokens) can help but in a crowd that big surely the faces will be blurry (mostly)

[karmic pulsar]

thank you , yes i try refiner and still not perfect . i will sleep it out 🤗

[livid rune]

you are welcome, usually i use for 5 or 6 characters

or for that picture :

[robust gulch]

I've found putting clear facial features in the prompt helps sometimes.

[frosty crater]

Hi everybody im starting a new project but very often Leo ignored my word Full body ,im using Only Absolute Reality with Prompt Magic

[karmic pulsar]

oke i can see that , so i will tr to work out with mentione a group of 4 or 5 people .it could work . thank you so much

[karmic pulsar]

absolute reality ?

[frosty crater]

yes

[wanton belfry]

depends on your prompt, it might not prioritizing it

generally you have generate a few more to see if its still ignoring

then drag it up in prompt

if it still ignores, try prompt magic

[frosty crater]

but this is possible only under 200 words as im using Leo for free

[wanton belfry]

you can still generate stuff, you just need to see that it prioritizes it.

[frosty crater]

this is my prompt im trying to make this short :(((full body ))), ((( clear facial features))),portrait a seductive female Tarzan,looking at camera,standing in the middle of the jungle ,cinematic shot, adorned in revealing leopard skin attire, amidst a lush jungle setting ,photographic masterpiece ,intricate details: piercing gaze, confidence and allure, flawless composition, vivid colors, and stunning clarity, enchantment display (best quality:1.3), (extremely detailed:1.2), (photorealistic:1.2)

[wanton belfry]

it works

what models are you even using?

[fathom owl]

ran it once in vision xl no prompt magic...surely you can get better with a few generations did not change prompt at all

2 tkns

[wanton belfry]

might need to spoiler that before mod warns

[fathom owl]

no worries 👍🏽

@frosty crater

[wanton belfry]

problem is that he's adding "Clear facial features" so the prompt is focusing on face

[fathom owl]

diffusion?

[wanton belfry]

albedo

[fathom owl]

exactly

[wanton belfry]

however I have different ways to get output

[fathom owl]

lol

as do i

just don

t need them

[wanton belfry]

even with original prompt its not bad at all

[fathom owl]

agreed

[wanton belfry]

Last resort, do a img2img with a character standing in middle and decent comp with low strength

this does take 3, +8 if using alchemy refine

[wanton belfry]

I've checked it, it works....remove clear facial features

there isn't too much wrong in prompt, just overemphasized needlessly

if facial features mess up, try refiner

[fathom owl]

1 off image absolute reality, prompt magic .55 without going through too much but a few gens likely would get ful to feet.

[frosty crater]

Absolute Reality

[frosty crater]

i can't use refiner ,Alchemy or Photoreal

[fathom owl]

😬

[frosty crater]

using Albedo is good but the face is not

[fathom owl]

face looks good. hands...not so much

[clever fiber]

Hi, the sunset colors are too saturated, I need to wear sunglasses to view this. Not sure if this is right model to use and if there are any prompts to make the colors appear a bit more natural. SDXL 0.9, Leo style, used improve prompt for sunset. Thank you.

[robust gulch]

I've been using (muted colors) on my prompts in photoreal but it only helps sometimes but it's better than nothing.

[clever fiber]

Thanks, I'll give it try. fingers crossed.

[fathom owl]

Try using descriptors that tone it down or dim the lighting effect

[clever fiber]

🥹 last 3 tokens

[clever fiber]

Thanks, I'll try that

[low bear]

@bleak forum Hey there! Feel free to send the images you've been getting here and explain what you want instead so we can help you :)

[bleak forum]

wron image

[low bear]

I see, what kind of image are you looking for instead?

[bleak forum]

i want a market with mythical creatures

[low bear]

I see! What model are you using? You can see it here

I will try generating some images of a market with mythical creatures :)

[bleak forum]

Leonardo Diffusion xl

[low bear]

@bleak forum I will try to improve your prompt but please stop sending your prompt in random channels as you cannot generate here in the Discord server :)

[bleak forum]

ok

[low bear]

I noticed that Leonardo Diffusion XL is just very bad at mythical creatures

That is with dreamshaper v7 (first try non-upscaled so it doesn't look very good but it does have mythical creatures haha)

[bleak forum]

ok thankyou

[low bear]

I also recommend prompts such as "photo of a fantasy market with a lot of mythical creatures walking around", especially if you're using Alchemy and/or Prompt Magic

And other than that just play with the wording a bit until you get what you want :)

[bleak forum]

i lost my 90 points due to wrong image

[low bear]

I'm sorry to hear that! However, if you're on a free plan you will get more tomorrow

Free users get 150 tokens every day!

[bleak forum]

ok

so what model should i use for mythical creatures

i have doubt i want to make a market consisting of mythical creature to so which model i should use

[frosty crater]

Albedo is one of my favourite models, in using this to add realism from normal stuff to sci fi 💪🏻👍

[low bear]

I would recommend DreamShaper v7

RPG v5 could work too

[bleak forum]

ruby can you tell what model should we use for generating mythical monster

[burnt sable]

so ive got this pic, but i wanna have it with different colors?

anyone know how to get around that

it keeps giving me different generations

[fathom owl]

try image to image with a different color in the prompt

[burnt sable]

do i enter "change color to orange" or how would i have to word the prompt?

[fathom owl]

that could help added to prompt or ask specifically for the color in the prompt as a descriptor

[burnt sable]

ok

thanks

[fathom owl]

sure thing ...hope it heps

[burnt sable]

how strong does the guidance have to be?

[fathom owl]

the higher it is in the image to image slider the more it will want to duplicate the actual image. I would start in the middle and adjust as needed. you are going to have to try a few times I would imagine

[burnt sable]

ok, can i ask you more if ill start having issues?

[fathom owl]

I'm going afk in a few min but if you are stilll having issues I wil be on later. plenty of helpfull folks here so they may jump in in the meantime

[burnt sable]

alright thanks

[burnt sable]

if anyone has a different solution, i gladly appreciate it

[night vault]

just recolor? try image guidance with edge or depth one. asking for “teenage boy in orange suit” etc

[burnt sable]

that actually helped quite a lot, more than i anticipated since i used it before.

thank you so much

[night vault]

glad that helped out

[robust gulch]

So when I generate 2 people side by side they tend to end up with identical faces, anyone have tips for preventing that?

[wise parrot]

When do the coins regen?

[night vault]

12am utc

[wise parrot]

I’m British

Idk what the thing in here

[night vault]

that’s midnight uk time now

[wise parrot]

K thanks

[night vault]

are you using alchemy? if yes, try disable them to test

[robust gulch]

Nope just Photoreal

[wanton belfry]

another way to sort it is multiple descriptions

but might still merge into one

[robust gulch]

I gave a rough description of each but not their facial features I suppose.

[wanton belfry]

you dont need to provide too much detail if its not focus, because you might get zoomed in

[robust gulch]

Well that kinda helps some, but the majority of the images still generate identical faces. Man I wish I had a negative prompt for Photoreal.

[night vault]

it is known behavior of some alchemy settings. maybe under the hood of photoreal is the same setup. I dont have enough experience with multiple face let alone PhotoReal that might give come clues here

[robust gulch]

I appreciate the help regardless. Thanks!

[livid inlet]

https://discord.gg/SdHurdvz?event=1173662671988609115

[wise parrot]

How do I get my images on my camera roll instead of being a file?

[thin tinsel]

No one understands what you mean. Try rephrasing.

[livid inlet]

Weekly Challenge Is Here #1173774312252117012 TOKEN PRIZES!!! 🥇 1500 🎁 🥈 1000 🎁 🥉 500 🎁 24 hours to get your votes in for last weeks challenge here #1171239119582666853 🎉

[robust gulch]

I'm getting tired of not having a negative prompt in photoreal, which model could I use to get a similar effect and how?

[dim zodiac]

The new Vision XL model is great

or Absolute Reality, Deliberate, RPG or Leo Diffusion all handle realism well

[robust gulch]

Great I'll try them out. Any particular words I should prompt or camera types?

[dim zodiac]

It really depends on the effect you are looking for... Whether you want polaroid style, or film grain, or something more like a modern DSLR

[robust gulch]

Modern

[night vault]

prompt with camera name, like Hasselblad, torso shot of xyz

[robust gulch]

Yeah I know nothing about photography. Would someone mind telling me a few cameras to try?

[night vault]

Hasselblad or canon 5d, anything will work. you just need to ask for any of them.

[robust gulch]

Ok thanks!

I apologize for what is probably a stupid question but I tried googling and looking at the faq and I still don't understand. What exactly does alchemy do?

[winter spire]

It's a proprietary pipeline, an additional layer of processing that improves your image output.

What that is specifically? Can't say for sure.

However, using 1.5 models you can use Alchemy to configure a number of things including Contrast (how deep the dark colors are) and Resonance (additional details and clutter)

[robust gulch]

Ok thanks!

[winter spire]

If you're using Alchemy, make sure to try out the various pipeline options. They are tailored to different styles, anywhere from anime to photorealism

[robust gulch]

Yeah I've only been using photoreal. Time for me to branch out a bit and figure things out. Have to try different things out and see what I actually like. Lots of slightly different prompts and models with the same seed.

[winter spire]

Don't sleep on the new XL models, they are remarkably powerful especially in combination with Alchemy

[robust gulch]

Will do thanks.

[dim zodiac]

#🖼│image-share message this is an example of an image I prompted in Vision XL using photography Alchemy

[robust gulch]

Wow that is impressive!

[dim zodiac]

and you can certainly use Negative Prompts with the XL models

[robust gulch]

Awesome

[frosty crater]

Alchemy can trasform things in Real ,unfortunately after one week if you are a free user this stop.You need subscribe a payed plan if you would again use this powerful function 👍

[burnt sable]

anyone know how to have a character with heterochromia

[wise parrot]

Posted this a few days ago and I’m trying to rephrase this

I’m am having trouble getting my images as regular pictures on to my iPhone 11 but I can only save them as files

[night vault]

just prompt for it, I have success with anime one

Sticker Design, aerial lens shot, (konosuba style), medium shot of 1girl chibi, onmyoji, Centered, (bakemonogatari:1.3), 🫣😽, Stained glass, witch, whimsical, Crystals, Flat shading, Cheerful, lay of light, heterochromia red blue, (colorful splashes), soft tetrad pastel color, vector art, fantasy art, Illuminated, bioluminescence, ((luminism, rayonism))

[burnt sable]

I've tried it but it gave me the same one for the two eyes

[night vault]

maybe try with prompt magic?

[burnt sable]

I'll try it tomorrow or the day after. I've used up all of my tokens.

150 token limit feels like dust :(

[livid inlet]

Daily Challenge Is Here #1174136761203568640 TOKEN PRIZES!!! 🥇 500 🎁 🥈 300 🎁 🥉 200 🎁

[young locust]

Hey guys! Just a quick newbie questions - I'm trying to generate something like the first image (lonely girl next to the wall, winter, very simple sketch-like pencil art)

What model should I use? The second image is Diffusion XL image to image (looks too realistic and detailed)

The third image is Ilustration V2 image to image - I kind of like it but maybe you know how to make it look even less detailed and more sketch-like

[fathom owl]

if that is the prompt you can try adding negative prompt to let it know you don't want it realistic. ie. put realistic, detailed , etc in the negitives. or make pencile sketch the first thing in your prompt so it prioritizes it

[wanton belfry]

there are multiple ways...give me the prompt you are using

I'll try to do it after I am done eating

but yes you can to a extent handle quality

[wanton belfry]

it wont drop further

[stuck nova]

START

#START

#🔰│getting-started

[feral bough]

Please click on the link you sent and it will have instructions on how to generate images on the webapp, NOT ON THIS SERVER

[robust gulch]

Ok I've got a weird question. If I'm generating a woman that is wearing makeup they often have very heavy thick square eyebrows that are colored dark brown and don't match their hair. Anyone have any suggestions? I've tried putting thick, heavy, colored in, filled in eyebrows and eyebrow makeup or thin and natural eyebrows in the prompt with only minor changes.

[fathom owl]

what kind of eyebrows are you trying to get? some thing specific or just something that looks normal?

[robust gulch]

Just normal

Not like this

[fathom owl]

is that in photorea?

[robust gulch]

Vision

[fathom owl]

ok

I'm gonna check something real quick

[robust gulch]

Perfect thanks

[fathom owl]

would you say that is any better?@robust gulch

[robust gulch]

Yes the color is still off but its definetly an improvement.

[fathom owl]

I would assume they didn't get dyed with the hair lol. that is not specificallly prompted for eyebrows just off the basic prompt. it is in albedo base xl. I run a lot in vision xl and it tends to be that way with brows most times. have you run it in albedo?

[robust gulch]

Nope

[fathom owl]

I would try it. and then if you can refine with prompts specifically for what you want it may give you more what you want

[robust gulch]

So what did you do differently than me to get them shaped like that?

Is it just albedo vs vision?

[fathom owl]

yes likely. my prompt has no real specific descriptors other than pretty or attractive woman. I looked through my gens and compared my vision stuff to albedo that was a difference i noticed so it may be as easy as model swap

if not you may get a better result with prompting in that model

[robust gulch]

Ok that's an easy answer. Thank you so much for the help.

[fathom owl]

no problem, youare very welcome. I hope it helps

[manic mist]

hello friends, I have a problem. When I try to generate an image for example "a little girl in garden" it creates 2 girl in picture. How can I solve this? Which prompt should I use?

[wanton belfry]

you can also try adding more weighting

[manic mist]

i did it but unfortunately it's happen again

Can I upload my exapmle picture here with my prompt?

[wanton belfry]

why not

[manic mist]

here is my promt

generate a little boy who has black wavy hair the air seemed to shimmer with a magical energy. The trees whispered secrets, and the flowers bowed as the duo passed by, in the heart of the enchanted forest

[wanton belfry]

if you have large space, it'll find something to fill it with

[manic mist]

yes I use 16:9

Is there anything to solve this because I need to use 16:9

[wanton belfry]

no, but I'd reccomend to add more focus to prompt, like closeup

so it'll focus on one character

[manic mist]

Okey. I will try it. Thank you so much

Understood 👍

[autumn wing]

Hey,
I trained a model on « mountain outline » so basically very very few strokes (the first image is an example of the style of images I am trying to make).
I’m trying to use image outline to image to load a picture of a mountain and get it’s outline with the model I created unfortunately 99% of the results are way too detailed compared to the training data (and what I’m trying to get) and the one’s that are close enough to the style don’t look like the mountain… :/
Anyone knows how I could solve this?
Maybe I haven’t trained on enough images?
Otherwise does anyone know what wording I could use to not have « shadows » like the model is currently doing? (I tried shadow in negative prompt but it didn’t work)

[dim zodiac]

try "simple illustration, single line drawing, monochromatic"

include detail and shading in the negatives like you were doing

@autumn wing (sorry didn't "reply" on my first response there)

[autumn wing]

Thank you very much! Will try!!

[uncut marten]

I wish to generate image on the topic "Knights & Jousting". Can any one guide me which model is best?

[dim zodiac]

I have had good experience with RPG for fantasy art - as well as AlbedoBaseXL

Dreamshaper with the right sort of prompt will work too

[uncut marten]

@dim zodiac Ok i will try and let you know, if it is generating correct images. Thanks for info

Which option should I select like cinematic, anime etc

[dim zodiac]

what are you trying to get? A hyper realistic image? Or something like drawn art?

[uncut marten]

hyper realistic image

[dim zodiac]

then go with Photography or Raytraced

[uncut marten]

ok @dim zodiac

Thanks for help @dim zodiac Worthful image generated

For the above image I wish to see the background image faces in clarity!

[dim zodiac]

you will need to spend more tokens and give it more pixels

try upscaling it with HD Crisp

[uncut marten]

Where I found the option of HD Crisp

[dim zodiac]

looks like

[uncut marten]

ok, I will try

I tried the option you suggested, but image looks same, Output remains same - no changement

[dim zodiac]

yeah background faces are hard to improve

[uncut marten]

@dim zodiac For sports images which model is best in our Platform

[dim zodiac]

again going for realism?

probably Deliberate, Absolute Reality, VisionXL or DiffusionXL

[uncut marten]

@dim zodiac Thanks for info

[autumn wing]

Thank you, it worked!!! Although for some reasons many of the generations have those weird lines in the sky (second was image to image but not the first one).

Any idea how to solve this?
Either way thanks a lot at least I get 1/4 good images now!

My bad this was due to the scheduler!

[dim zodiac]

NICE! Glad you got this sorted and your trained dataset can be put to good use

yeah the lines in the sky seem to be the AI trying to figure out how to put in a sky or something haha, I'm sure you'll sort that with a bit of prompt editing

[violet haven]

Can someone help with the image to image feature? It's creating the likeness of the dog very well, but I want a more interesting bakcground but can't seem to achieve it

I'm using Anime Pastel dream and the prompt: senior dog laying in a grassy field, colorful flowers surrounding him everywhere, mystical bubbles fill the air, background looks like a scene out of a pokemon game

[dim zodiac]

try using Edge to Image controlnet instead of just Image to Image

your dog here is well defined against the background, so the AI should be able to pick up the edges well, and then will fill in the rest with your prompt

Image to Image will try and conserve the colours and composition similar to the input image, so it will mostly look like the dog in the chair, the higher the init weight,

[violet haven]

Ah okay, thank you. I will give that a shot

[still portal]

Weekly themes are free to participate in, no?

[limber niche]

is there a way to get text only, i have looked at other users prompts and tried to copy them but it doesnt work for me for some reason

would just like Remnats stylized, but everytime i try i get some random stuff

like Design the Word 'Remants' using hundred dollar bills, gives me dollar bills with remnants on them

[winter spire]

It's likely the case that people who are generating text are also using Image to Image with source text. Take a look at https://discord.com/channels/989166677390426132/1101980163073396867 for some examples

[limber niche]

mmmm that makes sense, thank you @winter spire

[still portal]

@winter spire

[winter spire]

Happy to help where I can

[copper laurel]

I couldn't go past here. Kindly assist.

[feral bough]

Click "Generation History" to view your generated images

[karmic pulsar]

i am getting wrong image formation i want an snowy mountain where 4 people reciver 4 colur balls,they are fire ,water,grass,air.They have different colur dresses releated to balls

may i please get they way to come out of this

help please

[livid rune]

you are asking for 4 characters and 4 colors..... quite hard to get man

[night vault]

let alone two without mixing 🤯

[livid rune]

i got four characters quite different physically it was a miracle!

[violet yacht]

I am seriously struggling with camera angles. What is the angle , and the object position, terminology for this truck? I have blown almost 1k tokens to get 8 vehicles at this angle and position. almost all come out at a 45

they all generate at a 45 degree like this. I dont want that but idk the terminology

thanks

[winter spire]

Could be isometric, or bird's eye view. Are you using 'icon' in your prompt? It might be trying to generate something that's framed appropriately if so.

You could also try negative prompting for something like 'diagonal, skewed'

Another option would be using Image to Image with a truck that is roughly in the position/framing you're looking for

[wanton belfry]

you can use img2img to trick it in

[karmic pulsar]

ok

[robust gulch]

What style would you call these and how would you recommend imitating it?

[violet yacht]

Image to image wasn't doing well, I need to learn how to use it better. But I was using the word 'gaming asset' which might be like the 'icon' issue you suggested. I put in your prompt suggestions and still comming out at 45 degrees. Is there a term for like, a profile? straight side to side? those terms don't work.

still every one lol

[winter spire]

I think it might help to see what your prompt looks like

[violet yacht]

they're all over the place trying to fix it but I'll post some here...

prompt: bird's eye view of top of f150, posed in profile, camera high angle, board game, realistic

Neg: inaccurate design, incomplete, diagonal, skewed, cartoon, fake, disproportionate, weird, warped, behind,

prompt: isometric view of top of f150, camera high angle bird's eye view, board game, realistic

neg: inaccurate design, incomplete, diagonal, skewed, cartoon, fake, disproportionate, weird, warped, behind,

prompt: top down angle of top of f150, board game asset, realistic

neg: inaccurate design, incomplete, force perspective, cartoon, fake, disproportionate, weird, warped, behind,

...that kinda thing. just floundering about

[winter spire]

Hmm... one other thing you could try is using a wider aspect ratio. Might force the AI to generate in the orientation you're looking for.

There's probably a better solution, but it's a little late and I can't think of it at the moment 😅

[uncut marten]

For "Macro Images of Textures" which is best model. It should be realistic and natural. Please suggest!

[tired coyote]

I want to design a cute 3D avatar that has a hoodie on and is is nonchalant and black and cool with a chain and grills

[feral bough]

Okay, you can do that on the website 🙂

[tired coyote]

I'm asking for help with the prompts

[abstract basalt]

Do you have a specific question regarding the prompt, or an area of your creation that isn’t quite right? Perhaps if you post your prompt here and let us know what you’d like to be different about the creation we may be able to help you or guide you better

[fathom owl]

too many chains 🤣

you description works pretty well...just add some more specifics and I think you can get what you are looking for

[karmic pulsar]

I absolutely love PhotoReal/Alchemy. So I try to morph nature, plants or animal feature with female characters. I noticed if I use plant or sea creature features it always chooses to human face but if I want to create a female human red panda mix for example most pictures are massively on the animal side. I want some feathers, tail, animal ears or other parts but not an entirely animal face. For example I want something like in picture 1 but not like in version 2. Also what can I do to get more dynamic poses and not only the portrait one? I manged to get some but also here it is rare they come out like that like in picture 3 and 4. This is an example for the prompt used with the Red Panda theme (Any suggestions for different prompts or negative prompts to get better results):

"photo RAW, (dark orange, rust, vermilion, white and black: Portrait of 1 ghostly woman with red panda features, human face:1:3, shiny aura, highly detailed, red filigree, intricate motifs, organic tracery, Januz Miralles, Audrey Kawasaki, glowing stardust by W. Zelmer, perfect composition, smooth, sharp focus, sparkling particles, feathers background Realistic, realism, hd, 35mm photograph, 8k), masterpiece, award winning photography, natural light, perfect composition, high detail, hyper realistic,more detail XL" - with AlbedoBaseXL / Dynamic / Alchemy V2

Thank you ^__^

[wanton belfry]

I dont recommend photoreal for morph stuff, it really matters what model you are using. Try it with other models first, which seems to understand your prompt the best, go to alchemy with it

Alchemy will strengthen the coherence of the prompt, if it can understand it that is

generally its best policy to simplify your subject, your prompt is looking like it was meant for Base SD. Newer models are capable of understanding things better

especially XL

coming back to prompt, try "Panda Woman" and turn on Prompt magic

put the subject up high, dont add unneccesary artists, cut out useless keywords.

dont go into alchemy just yet, first test if it works

otherwise its a massive waste

[karmic pulsar]

First and foremost thank you. I am still on trial because I tested a lot of AI generators to find the one suits best for me before I buy a monthly sub. Could it be the case that "prompt magic" isn't visible in trial mode? Because I can't see it where it should be. Just to be curious how would you write and shorten it down considering my prompt above? In this case as I listed I haven't used photoreal (I just mentioned it because I really like it especially if you wanna recreate scenery of cities)

[robust gulch]

Anyone know how I could get a spray painted graffiti look? Like with the paint dripping down a bit.

[livid inlet]

Graffiti Street Art On A Barn, ϕ, Ultra Detailed Artistic, Midnight Aura, Black Ink, Liquid Melting, Dreamy, Glowing, Glamour, Glimmer, Shadows, Smooth, Ultra High Definition, 8k, Ultra Sharp Focus, Intricate Artwork Masterpiece, Highly Detailed, Vibrant, Ultra High Quality Model

[robust gulch]

Perfect! Thank you!

[livid inlet]

Works best with the ABXL model, but you should get some good results using Leo Diff model if you are on the free plan 🫂

[robust gulch]

I'm not.

[livid inlet]

If you DM me a screen shot of your subscription page, I can get you the Patron role on here. Or you can give me a link to your profile and I can get a team member to check but that takes longer.

[night vault]

prompt magic is accessible to all users including post trial. it just does not work with photoreal (or I speculated prompt magic may have baked into photoreal and couldnt be turned off)

[wanton belfry]

Prompt magic is available to everyone, it just doesn't work with Photoreal as Zorgi said. Same thing if you apply alchemy with XL models, it'll dissappear as a option.

shortening a prompt is simple, think of what your base subject is that you are making, like in this case it is Panda woman. then its pose like standing...thinking...eating etc, then finally you add all the modifiers and artists that enhance the scene

Do not start with artists generally, you are caging the prompt because it'll do only as much as the artist is capable of

If I for example add in "Stardew valley" first thing in prompt, I do get bright colors, vibrant look and style however I've now caged it to a certain extent. It might not show other things properly and the stardew style will be dominant. Artists also generally work like that

Usually test your prompt in DS or albedo, they are generalists and can handle almost anything

[karmic pulsar]

Thank you for your help. I also found out the the double dice button really creates great prompts and the outcome too ^^

What's the best way to get the fish out of the picture? Should I try it in Image Guidance (but here I can't use prompts to steer the path in a new direction without Premium) or should I go to the editor? But as you can see, I can't change the square. Why is it always the same size? Because it only covers the head of the fish, but not the rest. Thank you

Btw why when I enter "Woman with Axolotl features" does a fish come up? Nowhere in the prompt do I have the word "fish mentioned". If I would enter "No Fish or No Animals" in the negative prompts, would the Axolotl features also disappear?

[wanton belfry]

Just put in Fish, Animals in the negative...no need for no

[keen ore]

Hello, hopefully someone can help me, I have been trying to get their hands out of their pocket but each time I generate the image - hands are always in the pocket. Anybody know how to fix? I've tried using keywords like "A-pose" or "hands on the side" but unfortunately it doesn't take it.

[crisp mortar]

Ok so I have a question regarding creating an image. I know if I put something like Jack Black dressed as Deadpool it would pop up, but like what if I wanted to do the same for my friend who is not a celebrity. How would I go about generating an image of my friend or myself or anyone I know who is not a celebrity and that the ai would not know.

[fathom owl]

you would have to use pictures in image to image or image guidance. or create a model made from images of them, but would have to have persons consent etc so probably want to think about that before doing anything along those lines.

[misty elbow]

Hi. I am writng a story. I created character images. But, many times when I want a particular persons image in a new prompt, I can't get the original image back. How can I lock in individual images so they can be resused whenever I need them. Whenever, I need them, I want to be able to add a new prompt, but with the original character in it. Does that make any sense? Thank you.

[open echo]

Hello!

Since image generations is based on randomness, getting exactly the same character again is tricky and will require a little work.

One technique is to keep the description of the character EXACTLY the same between images, and generate a couple pictures until you get a good match.

If you only need a small variation (like a frown instead of a smile, per exemple) you can use the same prompt as the original image (with the added variation) and copy the seed. Essentially, you take out the random aspect of the generation to get the exact same image back, but with the changed detail.

The more involved technique would be to train a model for each specific characters so you can generate pictures with the same facial features every time.

And of course, you can always use photo editing software to manually correct Colors, small clothing details or even take a character from one picture and place it in another.

[misty elbow]

Thank you Tuxman. Interesting, but disappointing, but understandable at the same time. I guess there is no library of images that can be referenced, that can be fed back into the image generator. I will play around and see what I can do. Unfortunately, this app may not be usuable for this particular use.

[open echo]

I've done it 😄
(Your mileage may vary)

https://discord.com/channels/989166677390426132/1087904495381512202

[misty elbow]

Lol... Then I will too. Thanks again.

[karmic pulsar]

I can't create this image in Leonard realistic version

[livid inlet]

Are you having trouble generating the image or is it the result is not what you want?

[autumn wing]

Heyy,
I’ve generated an image of a dog using a custom model, and then I wanted to put it in the canvas and add another dog to the image using a different model, however for some reason, it only generates grass…
I’ve tried different wording, I’ve tried Outpaint on and off, nothing seems to work… anyone knows what I’m doing wrong?
I got grass like this generated like 6 or 8 times already (and with outpainting turned off, it wasn’t even really grass just stretched green)

I really would appreciate if someone has any idea! Thank youuuu

[open echo]

What have you tried so far?

Cause I got this:

[karmic pulsar]

It's not what I want

[karmic pulsar]

I wanted to make his version realistic in the 80's or 70's style.

[open echo]

What have you tried so far?

[karmic pulsar]

Like this

[karmic pulsar]

I did not understand your question

[open echo]

What have you attempted to generate the picture you want, but didn’t work?

[karmic pulsar]

Yes

prompt: Springbonnie It is a yellow rabbit with an appearance Bonnie. Your tender and lightly golden with some green reflection, Inside, there are several springs all over his body, and if they were activated, he would die a very slow death.

This costume has two modes, the endo-skeleton mode, and the springlock mode that employees can use to entertain children

@open echo

[open echo]

oooook.... 😄 Let's start from the begining

Here is an intro to Leonardo.Ai: https://youtu.be/Rukln4nr_Z0?si=TWCQNSWICkZhe8y0

Here are tips to build strong prompts: https://youtu.be/KKQDtBLla08?si=2M6Yw0Hq-lSPVL6n

And then, you'll be able to use more advanced tools like the "Image Guidance" to use your reference image as a strating point, and build on it with a good prompt 😄

[karmic pulsar]

Okay

[open echo]

You can do this!

[livid inlet]

Daily Challenge Is Here #1178122981612990595 TOKEN PRIZES!!! 🥇 500 🎁 🥈 300 🎁 🥉 200 🎁

Weekly Challenge Is Here #1176311480984023140 TOKEN PRIZES!!! 🥇 1500 🎁 🥈 1000 🎁 🥉 500 🎁

Official Challenge Is Here #1175098244410392670 TOKEN PRIZES!!! 🥇 4000 🎁 🥈 2800 🎁 🥉 1400 🎁

[open echo]

Depth to Image will be your friend 😁
(70’s vintage horror movie style)

[bronze ferry]

Any suggestions?

[wary drift]

hi all, I need help with Image Guidance. I'm trying to use a dog photo as a reference for an illustration, but whatever settings I've tried never give me an illustration that resemble the original dog in the result. Basically what I want is to get the original photo subject and transform it into an illustration, I can try to tweak the style later. What's the best way to do this?
I've tried a few different models, best results with AbsoluteReality. Image Input: Strength around 1 gives me ok fidelity (pose, general shape) but it's not the same dog. I've tried Depth to Image and Pose to Image with mixed results, none close enough...

[wary drift]

is this the right place to ask this?

[open echo]

You can try to mix Depth to Image and Edge to Image. You can use more powerful models like AlbedoBase or Leonardo Diffusion XL
You could even mix it with the Coloring Book Element

And of course make sure to specify the style as like "handdrawn illustration"

[mental venture]

hello!!
Dear community, I need your help.
I'm trying to generate ios app icons, but I always get different height to width frame, but not 1:1. It's not about dimensions, which is set to 1024x1024, it is about an icon for iOS app, that usually has borders. I tried different options and prompts as well as "Icon dimensions: 1:1" but had no success, the height of the icon is always longer than width. Did someone already solved such issue or has an answer? Thnx in advance.

[mental venture]

like this one.

[mental venture]

ios app icon frame is not 1:1, but like 1.2:1.

[open echo]

I've done a few, I found that using "Prompt Magic" really helped in getting the right border and the right ratio
https://discord.com/channels/989166677390426132/1102627899644846200

If you get a picture that would be perfect but is just skewed a bit, you can use online tools like Photopea to resize the image and make sure the sizes are 1:1

Just a VERY rough exemple of how you can also fix picture in photo editing softwares

[mental venture]

yes, it's like the only solution to use batch in 3rd party services to resize and then crop images )) because it seems that leonardo continuously generates the same dimension like 108%to100% for such type of results. Thank that you proved my guesses!

[open echo]

Didn't have issues with ratio that much... so make sure Prompt Magic is on, you might want to use a model like Dreamshaper

[mental venture]

oh. I'll try this model! thanks for advice!
i used alchemy v2, so prompt magic was disabled..

[open echo]

Ah! I never had good results with Alchemy doing iOS icons

[mental venture]

oh, ok.)) then searching for countless number of the right promts + "promt magic" + "dreamshaper" is my way then. Thank you very much!

[open echo]

You can see those with the VERY unsual borders are made with Alchemy (Like the steampunk ones in the bottom left corner, and the robots at the top right corners), the rest I did some tests with AlbedoBase XL. Not bad also!

[mental venture]

awesome!! It looks like that in most cases there is a stable (fixed) border radius. That is what I want to achieve ))

[mental venture]

could you please give an advice of Prompt Magic Strength coefficient you used?

unlike "leonardo diffusion xl" + "alchemy" the quality of image is a lil bit worse.. but length and height are equal!! thanks!!!

[open echo]

I pretty much always use it at default value

[robust gulch]

Could someone answer some questions about datasets for me please?

[night vault]

Don't ask to ask, just ask

[robust gulch]

I read the little faq thing on the website and it said only to use a few images but when I did that I ended up getting really wierd results. But I have also never done anything like this before so I could be doing something wrong or just stupid. How many images should I be putting in a dataset?

I put in 8? I think photographs of a woman's face just to see if I could get it to learn general facial features. I ended up with stuff like this.

[night vault]

what is your prompt (both positive and negative) and setting did you use?

I have guess about that noise cloud image but would like to confirm with you first

[robust gulch]

Just the name of the person in the photo. No negative.

[robust gulch]

I added more photos to the dataset and it got better but now it has very angular artifacts everywhere.

[night vault]

you added name of person as trigger word, right?

[robust gulch]

In the dataset? Yes.

[robust gulch]

Oh as the instance prompt you mean? Yes.

[shell stump]

I'm asking the community to help me figure this out. Has anyone succeeded to create patterns or whole compositions from geometrically correct lines or figures of symmetric shapes and sizes using one of the available models? One example is a mandala with complex internal patterns. All my attempts to get precise structures containing straight lines with precise symmetry have failed.

[night vault]

try add more image, i guess 15-20 should do it. and i recommend to prompt for something else too. for example a photo of <person name>, realistic sunset, etc

[robust gulch]

Perfect I'll give that a try. Thank you!

[dusk heart]

I want to create an image with a stark contrast, divided by a wall. Here's a prompt I've been using with Leonardo Vision XL / Environment.
A mossy stone wall dividing the scene. On the left side of the wall, a barren land, overcast, colorless, trees without leaves. On the right side of the wall, a colorful pastel, lush, sunny, grassy field with kids playing, swings, growing roses on the side.
It seems to have chosen only the colorful part and missing the gloomy left side of the scene. Do you have tips for dividing up the scene into starkly different settings?

While the image I envision isn't really like this one I can reference to, it does have a little resemblence in how it divides the scene in winter and summer-ish.

Maybe I can do something with canvas or image to image or something but it's a little hard for me to understand. This is a rough sketch of what I would like the picture to generally be structured. Left side gloomy, right side cheerful.

[dim zodiac]

I will caution that TOS does warn against deepfakes. Something to keep in mind as you train your model

[feral bough]

To elaborate:

[dim zodiac]

You have to prompt well for it and it’s still a bit of a roll of the dice, but yes you can create split images

[dusk heart]

Even oneself?

[dim zodiac]

You won’t get in trouble for making a model of yourself, no

[dusk heart]

Oh wow. I am testing some stuff out with img2img, reusing the output to build two pictures then doing some editing to layer them on top of each other and see if I can get a stone wall covering it at the end

[dim zodiac]

You most likely won’t get in trouble for making images of other real people. What that term is protecting against is people making really gross images, or realistic but fake images of people that could be improperly used

[feral bough]

You won't get in trouble for making a model of a real-life person, as long as the images you make with that model remain 'family friendly' and do not depict the person in a defamatory way. That's what the TOS says at least

[dim zodiac]

If you’re comfortable with photo editing, that is going to be your best bet for having complete control over a split image, generate the same image twice with a change in “theme” from sunny to gloomy or whatever, then splice it in Canvas or your fave traditional photo editing software

[dusk heart]

Still, could you spill some secrets on how to create split images like you did there?

[dim zodiac]

The example I posted is all prompted, but I have a ton of images from the same prompt that totally didn’t work haha

[dusk heart]

😄

Aha

[dim zodiac]

But for sure, that’s using the term “Split Image” at the very start of the prompt, then describing the subject, then describing the type of split I was looking for

“Split Image, high definition, cityscape divided equally in four vertically, each quarter as one season”

I used prompt magic as well

Alchemy always helps with prompt adherence too, if you have access to that

Have also used “split image” on face prompts:

[dusk heart]

I got a subscription but I am a utter novice. I dont know what prompt magic is I should read more tutorials or go on some class. I REALLY dig this and want to get invested.

[dim zodiac]

Fire and Ice was very popular for a bit

Prompt Magic is a button you can select on the left hand side, and it forces the AI to stick more strictly to the exact wording of your prompt

Default weight on it is typically good enough to start with

[dusk heart]

Do I need a certain model to get Prompt Magic on the left side? I just see output resolution under Alchemy V2.

[dim zodiac]

Oh using Alchemy V2 overwrites it

Alchemy V2 should give you a similar prompt adherence

And in Alchemy you can use () weighting

[dusk heart]

Aha I guess I'll try it later when I am done with this.

[dim zodiac]

😄 you’re finding all the fun stuff

Oh boy all the things I could just unload here haha - but I don’t want to overwhelm… there might be some good resources in #1068246518412361808 or in #1091125166064533504 to read over

[dusk heart]

Thank you, yes I'll make use of what's already been contributed 🙂

I am so happy to see this community got such active members

[dim zodiac]

We’ve certainly got a few!

[dusk heart]

Eh, it's getting there. Feels a little out of place hahaha

[dim zodiac]

I think it looks pretty good!

You have 2 different levels of scale thoough

A full tree on one side makes that swing look MASSIVE

[dusk heart]

Thank you 🙂 I keep experimenting, with different strengths to see if I can find variations. And yes, it looks massive.

Hmm, I wonder how I can shrink it. I probably need to redraw the sketch. But I am learning! Woho!

[dim zodiac]

Very refreshing that you have this attitude. I personally love “failing” at my image gens because I always find I learn something new in those moments

[robust gulch]

Yes I did see that. That is not my intention. I play a game where you hire a bunch of pilots but I've never liked their portraits so I want to run them through leonardo and redo them as celebrities.

Oh and maybe it makes a difference that I have no intention of sharing them in any way. Just for personal use and to pass time to keep myself sane.

[dim zodiac]

All good, I just let people know that whenever I hear of a model trained with any celeb

[robust gulch]

Anyone know the secret to keeping eyes from going crazy when they are not up close to the "camera"?

[dim zodiac]

MOAR PIXELS! or just upscale after

It's just the AI struggling with adding enough detail with the pixels it has

[robust gulch]

I can give that a try. Thank you!

[queen mountain]

i can't seem to be able to upscale my images, why is that

[dim zodiac]

My guess would be... You're using XL models without alchemy V2

And have free tier

[dim zodiac]

You pay for a sub?

[queen mountain]

i'm using a custom model

[queen mountain]

no

[dim zodiac]

A custom model should be able to upscale

They appear as icons below the image when viewing in generation view

[queen mountain]

ohhhhh

Would have been good for it to say that somewhere lol

alchemy only then

[dim zodiac]

Custom models should be able to upscale using the basic upscalers, creative, smooth and crisp I think they’re called?

Huh… ok maybe that’s a change then

Just tested in one of my trained models and yeah, only the alchemy ones I’m seeing now

[burnt sable]

how can i make my character have a black and orange-stripped eye visor?

it doesnt work for me somehow

[robust gulch]

Does the order of a negative prompt make a difference?

[burnt sable]

wdym?

[robust gulch]

Does the beginning of a prompt have more weight than the end? Or is everything equal?

[dim zodiac]

Beginning is most weight

[burnt sable]

idk how

[dim zodiac]

In alchemy you can use () weighting just like in positive prompt

[burnt sable]

oh really?

oof

i am a free user thoo

[robust gulch]

Guess its time to dissect and re order my prompt.

[robust gulch]

Thank you!

[dim zodiac]

For free users, the start of the prompt is the highest weight. Repetition of terms will also add weight

The XL models understand () weighting terms even outside of alchemy

[burnt sable]

like albedo?

[dim zodiac]

Yes, albedo, diffusion and vision XL

[burnt sable]

ooh

thanks

[dim zodiac]

👍

[burnt sable]

ill use it tmrw

ive got no tokens rn

sad

[dim zodiac]

Yeah 150 can go fast, but at least they come back every day!

[uncut marten]

For sports related topic which model best suits to generate mind blowing images #Nostalgic Sportswear

[cold jolt]

so the image to image feature ocassionally gets stuck. i can crank it all the way to 90 which should draw exactly what it sees but it doesn't usually when this happens closing the program and restarting fixes the issue but this time nothing seems to be fixing the issue. any suggestions?

i rebooted, and reset the defaults and it seems to work for some trained models but not others. strange.

it works for leonardo XL but not the other ones

seriously this bug needs to get fixed. i can't continue my work under these conditions.

turning off alchemy has fixed it.

[thin tinsel]

Since the most recent updates to image-to-image I have not been able to create an identical image, even on the highest strength, wheras previously any alteration was negligable.

[winter spire]

If I had to guess, it's because since the Image Guidance update, the max i2i strength was capped at 0.9 instead of the previous 1.0.

[feral bough]

It was never at 1.0 though

However, @thin tinsel, you might be suffering from a bug that I guess is still going on, where the init strength gets stuck, producing the same results every time even if you change the slider.

But getting a 100% identical picture has never really been possible because it has always been capped at 0.9

With Alchemy enabled you will probably get closer to an identical picture.

[charred plover]

is there anyway to correct hands

[open echo]

You can bring the image to the regular Canvas, mask the hands and prompt for ..well... "hand" 😄
It will be likea second chance to generate them

[spark brook]

I can’t log into my account through Google, I created a new account and it doesn’t log in either, what’s the problem?

[thin tinsel]

thanks. I don't think the slider is stuck, it's just my experience at 0.9 is significantly different since the update. It's not a massive issue for me but it can be a bit disappointed when I've fixed something in canvas, and when I feed it back in with img2img my fixes are gone.
I'm happy to give examples to the development team if they want any, but as I say it's not fundamentally ruining my enjoyment. 😄

[thin tinsel]

^^I've tried it again today, and it's massively improved. I cleared cookies and cache the other day, so that might've helped.
On some more simple images, at 0.9 it was an exact match and with a more complex image there were only minor changes. 👍

[eager gorge]

Hi, I can not make a character to appear backwards, I tried "looking backwards" "from back view", "facing backwards the camera" and also tried to put looking to the camera on the negative prompts but all give the same result, the character still looks to me. Any ideas what else can I try on the prompt ?

[night vault]

that is the problem with dataset in general

try SDXL model may help and sometime give desire result but I dont know any words/tips to trigger it

[eager gorge]

thank you will try that

[night vault]

prompt is

masterpiece, best quality, strong stroke, sharp focus, anime screencap of cleric, self-reflection of dark side, copy space, (chobit, granblue style:1.2), (rune factory:1.3), stardew valley, (ni no kuni:1.4), delusional, (Rayonism:1.2), soft pastel color, vector art, studio ghibli, makoto shinkai, fantasy art, glowing neon smoke, bioluminescence, william morris, art deco, (strong stroke), soft lighting, jewel tone, retro aesthetics, focusing on people, unreal engine

[past harbor]

help me prompt?

[valid sundial]

has anyone had any success on generating product photo shoot? I feel like it's rather difficult to keep the product identical to reference while the background dynamic.

[robust gulch]

Has anyone come up with a good negative prompt term to avoid repeated faces in group images? I have tried everything I can think of to no effect.

[still portal]

Still on a free plan, not sure if that helps. Thoughts on IMAGE QUALITY?

@everyone

[feral bough]

Are you aware that this server has 1.7 million members?

[still portal]

@feral bough

[feral bough]

Well, are you?

[still portal]

@feral bough sure am.

[feral bough]

And you thought it would be a good idea to ping 1.7 million people?

[still portal]

Actually, I read someone be reprimanded for that recently.

But noone corrected @feral bough

[feral bough]

Then why did you do it?😅

Lucky for you it doesn't work.

[still portal]

@feral bough can you help with negative prompts?

[feral bough]

I must say I am not really inclined to do that because of the stunt you just tried to pull.

[still portal]

@feral bough

[feral bough]

Maybe if you show some better judgement in the future and don't try to claim the attention of 1.7 million people all at once👍

[wanton belfry]

backview is sufficient, next best thing is to place it further up, next best thing to that is Magic prompt, lastly put weights...if its DS you might need alchemy but with SDXL you can do it just normally

also do not have anything that clashes with the keyword

[eager gorge]

thank you will try those

[arctic moon]

Idk how to create a god like pic of gautam buddha can someone help me ?

[livid inlet]

Keep it simple to start with, you can add things later. I'd start with just "God like Gautam Buddha" if it something the A.I knows you'll get a decent image. Then you can add art styles and things to your taste.

[arctic moon]

Can you create it for me

[livid inlet]

I could try but I prefer to teach you how to fish. I made you a pfp, now its time for you to grow ❤️‍🔥

[arctic moon]

ok I will try my bes

Is it good ?

[livid inlet]

You'll have to show me one you made using Leo

[arctic moon]

ok

this is ai generated

I didn't add anything

[charred plover]

Which model u use

[arctic moon]

idk I just search it and image pops um in the app

[livid inlet]

You shouldn't be taking other peoples images really. Its easy to make your own and there is a free plan.

[arctic moon]

oh

I will try

[livid inlet]

@drowsy garnet Hi again. Looking at your question in the welcome room. I'd say you can sort this quite easily. Chatgpt is not great for prompts but if you give me a few details or even post a prompt you have then I can help.

[slender bone]

@drowsy garnet dee is PRO ... your lucky to have his help hehe

[drowsy garnet]

OK then, in the meantime thank you. I'm doing a project for a desk calendar, it's 13 months long, each month has to represent a specific scene that relates to the specific activity (business) I'm making the calendar for, each scene however I would like it to be reinterpreted by a famous or historical painter/artist. Clearly the idea I want to give is not that of a photo or vector illustration, I would like the artistic and twisting side of something to emerge, either the "old or classic art" side but at the same time the innovative idea of innovative elements and above all contextualised with the client's business. I was trying to figure out which model or approach would be best for this type of work.

[drowsy garnet]

thank you very much! Honoured 🙂

[drowsy garnet]

ah ok! I thought ChatGpt could help me... ok then I'll listen to your advice so I can work better

[livid inlet]

For this I'd use the older model's like Deliberate 1.1. The new Leo Vision XL could also work well. You want to think of just a few key words that you want to see in the image. Don't write long sentences for the prompts like chatgpt gives you.

If you send me 1 prompt you have used so far, I will tweak it and show you how to get what you want

[drowsy garnet]

OK, now I give you examples.

my client is a logistics and transport company, so I am trying to give a concept for each month, I didn't like this realisation because it was fragmented, too grey compared to my request for a 'nice image to show on a desk calendar', I would like to create something artistic that is pleasing and not just a 'conversion' that is then not nice to look at.

This is just one example, but I will have made at least thirty, and none so far has satisfied my taste, even with another artist. Doing some tests, I preferred the Leonardo Diffusion Xl - Leonardo style. I tried some others which were either too 'illustrative' or sometimes too banal or lost a lot of content... I did not activate Alchemy otherwise I would have lost the more porous artistic concept or the 'handmade'. But Deliberate for example I did not use it.