🙋│image-help

[gray crypt]

hi, where i can find the new element popo surealism?

[gray crypt]

found it:)

[real lynx]

Hi

[real lynx]

I'm asleep

[manic root]

Eager to level up your skills or dive into big projects? 🚀 Get personalized support by visiting #👋│accademia and typing /mentoring. Your learning journey is about to get a boost! ✨

[drowsy plover]

do you know why with a logo controlnet (edge) works and with another one not?

the image is similar (black logo on a white background)

[real lynx]

We make the logos with the Phoenix preset

[drowsy plover]

but you can't use controlnet with phoenix

[real lynx]

There's no need

[real lynx]

Large pink cloud-shaped letters that say "BSK".A charming and enchanting photograph captures the essence of Winnie the Pooh sitting on a tree stump amidst the enchanting Hundred Acre Wood. With a gentle and tender expression, Pooh delicately holds a dandelion, blowing on its fluffy seeds, creating a magical moment. The background is illuminated by a warm golden light filtering through the leaves, with ethereal lines intertwining with the branches, adding to the whimsical atmosphere. The cinematic quality of the scene is enhanced by the artist's attention to detail, capturing the whimsical charm of the forest, with intricate textures and vibrant colors filling the image. Post-production effects subtly emphasize the warmth and coziness of the scene, evoking feelings of playful tranquility and natural beauty. photo, cinematic

Whatever you want in letters you have to write its description and put it in quotation marks

[thin tinsel]

what do you mean?

[spiral axle]

someone can help me create Zeus from One Piece in a photorealistic design ?

[karmic pulsar]

how to create furry characters ?

like this ?

[real lynx]

It's easy xD you just put fox ears on it and that's it :3

[karmic pulsar]

which model should I select for this ?

[real lynx]

I used semi realistic anime

Also use premium character reference tools

[cloud barn]

I'm trying to download an image larger than the edit region in canvas editor

can anyone help me out?

[real lynx]

Shrink your image from the edges to fit it into the editing box

[cloud barn]

But then the image will lose quality

Nvm I figured it out, move the editing region away from the entire image, press download and you should be able to download pictures outside the size limit of the region

[thin tinsel]

Zeus the Thunder Cloud?

[spiral axle]

yep 😄 ... with the funny hat

[thin tinsel]

Yes, you can make it photorealistic.

[spiral axle]

show me ❤️

if I try One Piece in the Prompt I get some other characters

[thin tinsel]

@spiral axle

[spiral axle]

wow

couldnt do it like this

[thin tinsel]

Prompt:
"suspended in the blue sky, A hyper-realistic anthropomorphised cloud, soft and fluffy. Tails off, He has little eyes with white eyelids, a neautral expression, a little cap with vertical red and white stripes with yellow peak. The whole image is ultra highly realistic."

@spiral axle I put legs in the negative prompt, but it still wanted to do legs.
I think you can cut them off

Image 2, I did with Content reference,
The other two are prompt alone, with Phoenix.

[wise drift]

/walks in only find furry
lol

[sly obsidian]

i used up my tokens, is there a way to still generate images even after using up my tokens?

[manic root]

Hey there! Would you like a helping hand with your learning journey? 📚 Our dedicated mentori can offer personalized guidance anytime. Just pop over to #👋│accademia and type /mentoring to get started. Let's ace those goals together! 🚀

[wanton belfry]

no, but it'll be refilled tommorow

[spiral axle]

there is still Realtime Canvas and Realtime Gen .... but he cant save those images

[dim zodiac]

Hello, we try and avoid solicitation in our general channels.

Once you have been in the discord and participated a while, the #💞│self-promotion channel should become available, where you can showcase your project and ask about artists. I will be deleting your ask

[sacred yoke]

Hello again!

has anyone used a preset by Playground called Fantasy Worlds?

Its on PG2.5

I would like to try and achieve the style i'm getting there on Leonardo

anyone care to help me?

[night vault]

Haven’t visit playground since last year. Do you have galley? you cam dm me or start new /mentoring thread

[sacred yoke]

If anyone is up to the task I can provide some promts I did there and what was the outcome

Sure I will @night vault thanks

[storm jungle]

Hi there, if you are a paid subscriber on the Artisan plan or higher you get unlimited tokens, meaning when you run out, it will go negative, meaning you are now in 'Relaxed Generations' mode, this rests daily to 150 until your new month resets fully, relaxed mode means slower generations, but I can assure you there is hardly any difference, in relaxed mode things like upscaling won't be available, but everything else will be including 1000 monthly separate tokens for describe with Ai and random prompt generation etc, but auto prompting for Phoenix is part of your unlimited generations.

Keep in mind that any tokens used in relaxed mode will not be taken off your new starting month acc, as it started fresh again.

Hope this helps.

[carmine nymph]

Hello, if I have let's say a game character image standing. Can I use Leonardo to make (the exact character) but in a sitting state for example? is it possible?

[dim zodiac]

Yes, you'll use Character Reference and describe the whole scene and clothing etc carefully

Character reference will handle making sure the face looks the same

[carmine nymph]

Sorry, where can I find that option?

[dim zodiac]

It is an image guidance tool found if you click the little image icon in the prompt box

[real lynx]

If you make him a coffee he sits down :3

[warm skiff]

I'm wanting to make a model so I have a consistent face. When I'm making the images with the set seed what other options should I be looking at setting.

[graceful panther]

what part of this is bad ?( A cyberpunk character wearing a pink rabbit mask with a cybernetic eye, dressed in dark leather jacket, holding a gun and firing it, the gun is in the middle of a shot with muzzle flash, smoke trailing from the barrel, shell casing ejected, the scene is filled with intense action, debris flying, motion blur on the character's arm, gritty urban background with graffiti, dystopian atmosphere, dramatic lighting with neon reflections, cinematic shot, high resolution, realistic textures, dynamic pose, detailed cyberpunk aesthetic. )

im getting this error

Stricter prompt filtering is in place when using Character Reference. Your prompt appears to contain suggestive or inappropriate content and has been blocked. You may modify your prompt and try again.

[wise drift]

Maybe the gun

[graceful panther]

?????????? ok i know now i guess it is that sencerd

[sacred yoke]

hey mate I sent you the info on DM earlier. Dunno if you saw. If u need me to start a mentoring thread lmk

[real lynx]

Hi

I have a solution for that

Check your prompt only if it allows me to create images

I chose the ones I liked the most and in the end I only chose the last one.

I passed that image through this tool that describes the image and gives you a prompt similar to the image

Then I uploaded a photo of myself to see if it worked and it did.

I think it's a good solution

What do you think?

[manic root]

Need a hand boosting your skills or conquering a big project? 🚀 Our mentors have got your back! For one-on-one help, pop over to #👋│accademia and type /mentoring. Always here to support your journey! 🌟

[graceful panther]

Hi

[graceful panther]

I think I will give it a try and see it sounds awesome and productive

[wise drift]

That’s avatar worthy lol

[marble cape]

Hey so this art I did years ago trying to generate, it’s a clock that has one dark dot like an eye as you can see and dark teeth

Can anyone help me with this?

This how it got generated which is nice, it has a nice dark theme to it but Ive been trying to see how it would with that one black dot in the middle and those things around are or suppose to be dark teeth

[real lynx]

Hi

[real lynx]

I don't know where the eye is 👀

The idea is that where the round clock is there are teeth and inside what is the mouth an eye

[wise drift]

He’s saying the 12 is just a black dot, and is the only enumeration on the clock. The tribal teeth design is just a background on the clock face, not a functional part of the clock. MJ keeps overwriting both the dot and the tribal teeth background with Roman numerals for the hours enumerations.

[undone furnace]

If you wanna create this kind of image you can use the Describe with AI features, If you wanna add/change something you use the Edit with AI feature or add a prompt manually.

[wise drift]

Correct terminology for the numbers/tick marks is “indices”, but I don’t know if the AI recognizes such or not

[marble cape]

Just like the image I have but more like reality

[iron ferry]

Question: How can I get this woman to look at her phone? I keep trying to look straight ahead and it is difficult to freely adjust my gaze.

[wise drift]

They’re neither eyes nor teeth lol

[fresh herald]

Trying to use Finetuned model on alchemy image guidance page and everything is "not compatible"; add elements, image guidance.
Is that because model trained with SDXL 2 and not 1.5?

[real lynx]

Well I liked it anyway xD

[real lynx]

Candid, close-up photograph of a young woman with medium skin tone and straight, shoulder-length dark brown hair. She is smiling and looking at a smartphone she is holding with both hands. She is wearing a white sleeveless top and a silver bracelet on her right wrist. The background is softly lit, with a light-colored wall and a window, creating a bright and airy atmosphere. The overall composition is relaxed and casual, focusing on the woman's content expression as she engages with her phone.

[wise drift]

I meant he used really weird descriptions for English lol

There are watch faces (faces are what you call the circle part that is inside the frame) that just have a dot for the 12 hour and nothing else. Faces can also have backgrounds. His background is basically what is called “a tribal design on the edges”

[real lynx]

he never asked for his watch to be a monster xD

[gray crypt]

what means when in the prompt you put words in the (), or even double (()) ? thx

[feral bough]

Adding more brackets increases the 'weight' of the word(s) within the bracket, which means that that element will get more emphasis in the image. For example, if you want an image that almost completely red, you'd put (((red))) or (red:2.0) in your prompt. The advice with weighting is to never go above a total weight of ~2.5 because otherwise you'll get weird results.

[pallid haven]

Hi, I need to create a similar image. But I do not know how to write a promt for that. Is there anyone who can help me with that

[marble cape]

Or just those dark teeth would be better

Perhaps in 3D mode

[open echo]

For a precise center piece like that, you can cheat and generate the clock body and the round clock itself separately and then bring them together in a photo editing app like Photopea or other

[feral mica]

hello Sam! we have a tool called describe with AI, you need to upload an image and we get you the description

I got this image with the description it gave me

[velvet tusk]

hi, what negative prompt should I put for a standard 16:9 image such as below, I'm not sure what's causing it and try to make sure every generation pass a minimal standard of not leaving blank in edge

[thin tinsel]

You're not sure what's causing what?

[velvet tusk]

not sure and wondering what key phrase should I put in negative prompt

[thin tinsel]

You said; "I'm not sure what's causing it"
What is "it"?

[manic root]

Need a bit of one-on-one help with your skills or projects? Our mentori have got your back! Swing by #👋│accademia and hit /mentoring to get started! 🌟

[velvet tusk]

"not leaving blank in edge"

[thin tinsel]

In a batch of four how many images have a white edge?

[velvet tusk]

1 out of 4, and it's not that frequent i run into this

[thin tinsel]

I suggest that negative prompt is not going to resolve this, but adding something into the main prompt could make it fill to the edge.
Which model and preset are you using?

Is your prompt a long one?

[velvet tusk]

a high-quality, animated cartoon art by Pixar, a big, friendly Great Dane with a shiny black coat is stretching his legs in a warm grassy park. The camera captures the scene in a wide shot, showing the park full of kids playing with kites. His joyful expression and sparkling eyes shine in the sunlight, surrounded by fluffy white clouds and blue skies.. using Anime XL + Anime preset

[thin tinsel]

Ok thanks. I will think for a moment.

@velvet tusk A big, friendly Great Dane with a shiny black coat is stretching his legs in a warm grassy park. His joyful expression and sparkling eyes shine in the sunlight. A Wide shot, showing the park with children flying kites.. Fluffy white clouds and blue skies. Fill to edge.

I have done 12 images , and not seen the white edge. so not sure if that resolves it or not

[velvet tusk]

ok thanks, after regenerating using the same prompt, I did not get a bad one, so maybe it's a one off thing

[thin tinsel]

I thnk it just happens occassionally, due to the images the AI was trained on

You can always crop it off, or fill it in with Canvas Editor.

[velvet tusk]

Another thing is sometimes I ran into animal have multiple tails, like the image I attached earlier, what exactly should I put in negative prompt? “No redundant tail”?

[feral bough]

You should put in the negative prompt what you DON'T want to see. If you don't want extra tails, you can put "multiple tails" in the negative prompt

[marble cape]

Can you show me an example

[open echo]

Of course!

So it's simply to have the body of the clock, and the face as two different images and then placing them one above the other in a photo editing software (Here, I am using Pixelmator... but Photopea or GIMP are other free alternatives)

Also, ignore the style and colors.. it's simply to show the exemple 😄
When you create your own, you'll be able to prompt exactly for the style you are looking for

[wise drift]

FWIW, if it’s a one-off piece, you’re probably better off just commissioning it.

[buoyant ridge]

Can someone help me correct my prompt? It's for my thesis!

I need this prompt to generate images of the character from the front, to use as concept art, or is there a way to get character views? Side view, front view, blueprints.

im using this

and generate this

[real lynx]

Hi

Try this prompt and add your Nordic tokens to it

Style and type: The image is a 3D digital rendering. Layout: The image Woman Blonde Hair features six identical animated characters arranged in a horizontal line, each in a different pose, showcasing a 360-degree view. Subject: The character is a young person with light skin, short brown hair, and a casual outfit. They are wearing a purple cap, a pink short-sleeved shirt, purple shorts, and white shoes. The character also has a purple backpack. The background is plain white, emphasizing the character's design and attire.

@chrome kiln

help me, it won't render this prompt, seems fine to me: the end of the world, judgement day, epic battle between demons and angels on earth, destroyed landscapes and cities, photorealistic, very detailed, jean ansell and justin gerard,

[steady rapids]

Hi everyone, I need help creating a logo for my scan group, I’ve been trying with Leonardo AI, but I’m not getting the results I want. Could someone please create one for me using Leonardo AI or any other AI? I’d really appreciate any help!

[real lynx]

Hi, I can create it for you. Tell me more or less the idea.

[real lynx]

Tell me what your scanning group is about so I can try to create something.

[steady rapids]

ok,
we translate comics, manga, and manhwas into Bangla and release them on our website. However, we've encountered an issue: our current logo is too similar to another scan (scan = translation group), so we need to create a new one. The challenge is that I'm struggling to come up with a concept that truly represents our scan. Many scan have logos that perfectly align with their names, and with our name being 'Bangla Comik,' I want something that resonates with the name or suits our concept[translating comic/manga in bangla (Bangla is the primary language of Bangladesh)] or our website. Please help me if you can.

@real lynx

[real lynx]

Some of the ones I liked I don't know if it's the style you're looking for.

[storm jungle]

Very Pokemon-ish

[manic root]

On a quest to enhance your talents, conquer grand projects, or hit those learning milestones? Our mentors are on standby to guide you. Swing by #👋│accademia and drop a /mentoring to summon personalized support anytime! 🚀✨

[soft locust]

Is there an easy way to replace a cat in an image with another cat?

[wise drift]

Could make your own cartoon lmao

[karmic pulsar]

Probably basic question...why when using image guidance will it sometimes say image dimensions cannot be matched even though I haven't changed them? What effect does this have when I just ignore and generate anyway?

[thin tinsel]

Just make sure it's as close as you can get it.
Bits will get cut off otherwise.

[thin tinsel]

Yes, you can do this in Canvas Editor, but make sure the cat to be added isn't a completely different style from the image you're adding it to.

[karmic pulsar]

ty 🙂

[pure geyser]

Achieve this as real as possible other than anime style but lady face should be attractive face

[real lynx]

You could try giving priority to the face with an aerial view and with the fisheye effect of a camera

It would give more details to the face to have it closer

[dusty rain]

Hey i need a logo for my cyber security club i recently started , so i somewhere i need a logo that represents a combination of tech and cybersecurity , as the club works on security in diffrent fileds like ml , app dev , web dev ect
i would like to have a tree as my symbol intigrating some cybersecurity features in it as logo

@real lynx need your help with it

[stable carbon]

if i remember, I may try some things this weekend. but i suspect some others may hop in to help before that 😉
do you have any specific colour/s you want featured prominently..? what have you tried so far..? prompt/examples..?

[real lynx]

Perfect, I like the idea

Something minimalist and serious I imagine

[dusty rain]

how about a tree that is connected digitally and it has a cyberpunk theme

[stable carbon]

Here, literally the first things i tried. 🤷‍♂️
any of them to your liking?

Legacy Mode, Anime XL, Alchemy Dynamic and/or Illustration
"Minimalist simple Illustrated Round Cyber Logo, Digital holographic Tree with many glowing cables as roots, black background, extremely saturated, detailed, best quality"

[real lynx]

A minimalist logo with a tree as the central symbol. The tree is made up of various cybersecurity elements, such as a lock, a padlock, a shield, and a computer mouse. The leaves of the tree are made up of code lines. The background is a gradient of blue and white. The text "Cyber Security Club" is placed below the tree.

[stable carbon]

ah, i like Cami's more than mine, tbh
i forgot to add relevant security icons..

[real lynx]

I love creating logos xD

I hope it helps you

Any idea we can try to improve it

[dusty rain]

it is more of a creative perspective in digital word of security how about a lil less dynamic and more sharped edges like a form of connected nodes in network or neurons of mind

@real lynx can u also try this

[abstract stag]

Don't know what I'm doing wrong. Two days ago I rendered this first turtle. Now I want to make four versions of it running together, like the second image. I've tried using the fixed seed from the first render, copying the prompt, remixing, but nothing results with the turtles looking like the one in the first image. Can anybody help? I'm attaching a screen shot of the info from the original turtle render.

[wise drift]

“Circuitry” as a keyword maybe?

[dusty rain]

yup

[wise drift]

This is using a modified @stable carbon prompt but with MJ (cuz I’m on my phone and can’t use Leo rn)

“minimalist simple Illustrated Round Cyber Logo, Digital holographic Tree composed of circuitry lines with many glowing cables as roots, black background, extremely saturated, detailed, best quality

That closer?

[karmic pulsar]

any suggestions on how to make her shoulders symmetrical? I tried fussing with canvas but couldn't get it.

[silent wedge]

Hey guys, I am struggling with creating a 2D character for a mobile game on which i could change equipments. Anyone knows how i could do that ? trying to get equipment pieces that fit on the character

[clever pulsar]

Looking for some general advice on face swapping. I would like the pose and outfit to stay the same as one generated image and am using a high setting on character reference, and medium setting on a close in head shot of the face I am trying to swap. I get a good 80% result. The pose is rock solid, but the outfit is altered more than I'd like. Any work flow advice?

[real lynx]

Character referencia Put it on low

From my experience, one of the two has to be in low so that the prompt can also be displayed in the image.

[clever pulsar]

Thanks Cami, so one either can have fidelity in the outfit or the face, not both?

[real lynx]

My advice is to use albedo and be descriptive with each item of clothing, for example the color of the boots.

Even if you want to give more strength to the prompt you will have to sacrifice strength of the pose.

[clever pulsar]

Thanks, appreciate the advice

[manic root]

Hey there! 🚀 Want to level up your expertise, take on some exciting challenges, or hit those learning targets? Our mentors can help guide you! Just pop over to #👋│accademia and fire away with /mentoring to get personalized support.💡

[pure geyser]

Did you try I am not getting

Or any other can help

[stable carbon]

The 2nd and especially 4th are closer to what I had in mind when I was trying. But please try to avoid recommending other services. This is Leonardo's discord 😅

P.s. Leo has an IOS app already, and should be working on an android one. So you'll be able to use it on mobile soon too, I hope 👍

[real lynx]

I wrote what I put is a camera effect in your prompt

[pure geyser]

Let me know if any helper can do.

Not getting good image on Leonardo thought of switching

[wise canyon]

Hi StriKy, there are a few options and ways to achieve this.
A quick and effective approach is to "Cross Model" the existing image you have.
Using Legacy mode, simply choose a photorealistic model, then load your image into Image guidance and select "Image to Image" from the "Type" drop-down menu, put the strength up very high, and generate a new version of your image.

For variations of the image, you can adjust the "Strength", and also try the other options like "Content Reference" and "Depth to Image", etc
You can really take an image in any direction you want using various combinations of the Leonardo tools, there's no end to the possible outcomes 🙂

Here I used Kino model with PhotoReal on your image using "Image to Image" full strength 0.90, then upscaled it to improve it further.

[storm jungle]

Hi Ellie, you could try use your image as a guidance for Image to Image (Leonardo Vision/Dynamic), set the strength to around 0.50, don't go much lower as the flowers on the dress will start to disappear to a plain black dress, keeping it at around 0.50 will keep similar flower patterns, if you regenerate a few times, you will get a change in shoulder placements, and even at times a slight different hair placement which blocks the shoulder, either outcome will happen for you to decide and go with, let me know how you go.

This was a starting prompt that I used, not sure if it was the best one to use, but got some decent results from it.

Anime Girl with symmetrical proportions in a black dress in front of a large moon

[fresh herald]

In inpainting canvas mode, whatever you put in negative prompt will appear in your image 🤦‍♂️

[storm jungle]

I rarely use the negative prompts

@karmic pulsar Here is a quick result that i got, it is not 100% level, but you get the idea, it is more level than your original image, this was done with the setting at 0.46... again remember not to go too low, otherwise the flowers on the dress will go

[karmic pulsar]

Yea, that's pretty much what's happening with me....I'm getting some bangin' dresses though 😄

[storm jungle]

You could also use a closer generated image as a guidance and repeat, depends on what changes in the look that it gives, then you can just upscale it to improve the quality as well.

[karmic pulsar]

Thanks, it does look better, I'm just being finicky

[storm jungle]

You're welcome, nothing wrong with being finicky, I been there many times, I am the same, so I get it beleive me : )

[karmic pulsar]

mwua hah hah...very close to victory is mine! apparently, she just wanted a bit of sparklecore

[storm jungle]

it was a sign : P

[steady rapids]

@real lynx Brother, I talked with my team and came up with an idea. Now I need your help to create a good logo for my scan. What do you think about having 'BC' in big letters as the main part of the logo, with 'Bangla Comik' written below it? I don’t know much about editing, but I think the 'BC' should have a good font and design to make the whole logo look appealing, even if it’s simple. By the way, our website theme is red and black, so maybe using those two colors would be best for the logo. Please try to help me with this.

[real lynx]

I'm not available yet in a few minutes

[steady rapids]

just plz help me whenever ur free

[wise drift]

Are you wanting to learn Leonardo? Or just get the logo?

[steady rapids]

just the logo

[wise drift]

Not a recommendation. Just what I used. Can plug same prompt into Leo with the same settings other guy used.

App doesn’t work that well. Options are all different from the website.

[steady rapids]

brother if u can plz help me

[thin tinsel]

Have you read the guide about making a logo? Has someone sent it to you already?

[storm jungle]

Some logos to work with, maybe @real lynx could get some ideas from.

[steady rapids]

no

[real lynx]

The idea is that you learn to create images, we will always be there to help you.

Posters and logos created with the Phoenix preset

Now I'm available

I'm going to read the instructions you left to see what I can create 🫰

Creating, I'll be back 👋

[wise drift]

Can someone make me a colored comic series with me as the protagonist? I only need like 150-300 issues

[thin tinsel]

Here is the quick guide to get you started. I hope you enjoy having a go at creating some logos.
https://leonardo.ai/wiki/creating-a-quick-logo-with-leonardo-ai/

[thin tinsel]

Hi Silent Sagas, we don't do that here. We are here to help people who want to make their own images, themselves.
It's not for advertising a job, and I think the Mods will have something to say to you about that.

[storm jungle]

Everyone needs chocolate to help brainstorm some logo ideas : P

[wise drift]

Was sarcasm lol

[west pivot]

@steady rapids here it has no has no background

[thin tinsel]

Since written language doesn't have a tone of voice, then I recommend putting an emoji to show that it is a joke.
😆 🤣 😜 😉 🙃 <----- these are suitable examples.

[wise drift]

Deamon have you considered naming your service Bangla Bangla Comics? I think the double Bangla Bangla would really make you stand out. Especially for your logo.

[storm jungle]

We all encourage people to give logos a try themselves, to see what they can come up with. you may be surprised by your own results : )

You can also see some of my logo creations here for inspiration, or that push you need to give it a go.
#1271315316072517714 message

[west pivot]

@steady rapids wrong spelling 🤣🤣 here this is correct

[wise drift]

Phoenix, prompt: “SilentSagas” logo, vector art, mysterious theme, ninja theme

[real lynx]

Some of the ones I made I'm going to have for breakfast 👋

[real lynx]

A logo with the text "Bangla Comik" in black, with the "BC" in big red letters at the top. The Giant letters that say "BC"Inside the words I want them to have comic drawings is designed with a modern, bold font, In small print "Bangla Comik", sans-serif font. The overall design is simple yet appealing, with a balance of red and black colors..Exotic allure, A palette of Glistening Opal White, (Fiery Orange, Galactic Blue, Starry Sky Blue, Oceanic Turquoise), scattering, specular highlights, ambient occlusion, dynamic range, volumetric light, microdetails, physically-based rendering (PBR), ray tracing, realistic depth of field, displacement and normal maps, dynamic lighting, Shot on 70mm, Depth of view, F/22

[steady rapids]

thanks everyone for the helps, i thing i got some good logos, thanks a lot for this🩷

[real lynx]

I would like to see you creating logos too 👋

[wise drift]

Phoenix uses quotes to determine text to write? Few of the ones that came up when I was messing around would show the text correctly, but then would also put my whole prompt in small text somewhere random on image

[storm jungle]

That is because although quotes are recommended, but it is also not always needed, i sometimes don't use quotes. Leonardo doesn't always know which words to use as text, since you're creating a logo, it looks for words in the prompt, whether in quotes or not,

[thin tinsel]

Yes, I've had that too, where it puts the style of the image in small text at the bottom.

[wise drift]

I gave up on a better looking one after 4 rerolls lol

[thin tinsel]

This is really nice. What are you not happy with?

[wise drift]

Doesn’t really look like an actual magic 8 ball lol. Was messing with it as a reply to Cami saying she hopes that dude makes some logos of his own or whatever lol

[wise drift]

Separate sentences if you do the no quotes? Prompt enhance? Seems kinda weird that it’s not strict on quotes

[thin tinsel]

You can just be really specific with your prompt anyway. I'll show you a good example

[wise drift]

Also, no one caught my Bangla Bangla Comik joke lol

[thin tinsel]

I saw it. but I thought it was a serious suggestion. 😃

[storm jungle]

Well something like (create a high-definition logo using the text "Word")', there is a chance Leonardo will use high definition as a style or apart of the logo since it is part of the same request, so it may or may not get used as part of the text somewhere, Leonardo may not just see logos as a logo, it could think it might be some form of advertisement.

[wise drift]

Don’t Google “BBC” lol. You’ll either get British broadcast channel or a meme lol

[wise drift]

That one looks pretty dope lol

[thin tinsel]

Someone is watching you

[wise drift]

Web app keeps giving me this instead of letting me log in

[thin tinsel]

I think post that query to #❓│platform-questions and see if anyone can help you.

[wise drift]

Lmao. I’m going to take a wild guess there aren’t a lot of Indian people in Argentina lol

[storm jungle]

@pure ivy Can you be a lamb and look into this for me please.

[pure ivy]

Unfortunately I cannot answer that question.
-# Our apologies if this message did not answer your question.

[storm jungle]

pfft

[storm jungle]

@pure ivy can you forwrad this to one of the cool MODS

[pure ivy]

I've sent a message to the moderators regarding the issue reported by SilentSagas 📏.
-# Our apologies if this message did not answer your question.

[storm jungle]

Thata boy : )

[manic root]

🌱 Want to boost your learning curve? Our mentori can help with that! For personalized guidance 📚, just pop over to #👋│accademia and type /mentoring to get started. Let's grow together! 🚀

[wise drift]

Is it generally better to directly use a photo as a character reference? Or to use it as a character reference and reroll until you get it super close to original and then using the resulting image as the future character reference? (So you’re using an image already in the models)

[storm jungle]

Character reference mainly looks at the face, where as a photo image to image will be more closest overall.

[wise drift]

Would I want to use the original face picture as the reference for multiple pictures , though? Or should I try and replicate the face in Leonardo as close as possible to the original and THEN use that image as the reference?

[storm jungle]

Depends on what you're trying to create, you could use character reference for the face, and content references for poses and different styles, using imag2img won't change that much as it will replicate the original.

[wise drift]

There a prompt word for Phoenix for photorealism? Last little thing I did turned into a cartoon

[thin tinsel]

what preset did you choose?

[storm jungle]

Depends on the rest of the prompt and the preset you used, you can use Photorealistic watercolor image, it will still be watercolor

[wise drift]

Style? On my phone so options aren’t the same as the web app

[thin tinsel]

Yes, where is says preset style on the menu under prompt enhance

[wise drift]

Cinematic. Gonna use a different model. Phoenix isn’t recognizing the character in the prompt anyways, so have to switch so I can use character references

[warped locust]

Does anyone have a good prompt for generating a tattoo?

[subtle gulch]

It says 4-15 characters only.

[wise drift]

It’s trying to make me make a new account, and not letting me get to the login section

[subtle gulch]

Perhaps you can clear the cache/ history/ cookies of your browser first?

[subtle gulch]

Stencil style tattoo design pattern, bold intricate Māori Tribal pattern, giant skull, asymmetrical, extremely sharp focus (with transparent setting)

[wise drift]

What im trying to avoid lol

[wise drift]

Nevermind I got it figured out. Had tried logging on with wrong email and I guess it logged in as that email after signing up, but since it wasn’t an existing account it was stuck on select username page. Had to just finish filling it out then log out so I could log in right one

[low bear]

Hey there! Are you sure you're logging in with the same email address and the same login method you used to create your account?

[low bear]

Whoops, missed this message

Glad you got it resolved!

[wise drift]

lol thanks anyways. Missed by about 5 minutes lol

[wise drift]

What’s best model for like typical instagram photos?

[thin tinsel]

Do you have any example of what you want to create?

[wise drift]

Generic AI influencer

[thin tinsel]

Nope, no idea what you mean.

[wise drift]

Think its basically just a camera phone

[thin tinsel]

Ok so now I've seen them you can delete the post,

An instagram Model fashion photography?

[wise drift]

I suppose that would be accurate. Doesn’t look like actual modeling photos, though. Just like amateur level with probably a camera phone. Basically most of IG photos out there all have a similar sort of photography to them. Granted, in that example it was a pretty girl’s profile, but the photography itself is the same as most of IG as a whole

[thin tinsel]

So you want to the style of photo taken with a phone.

Or you want the subject of a beautiful woman.

Or both?

[wise drift]

First one

Like, would that be Phoenix, or Lightning, etc?

[thin tinsel]

Personally I would go with 'Perfect Portrait' for the model

And the Portrait Fashion

But Portrait Retro might be worth a try too

[wise drift]

Okay I’ll try both. Thanks

[thin tinsel]

You've also got the Vintage Photography Element, which you can put on a low weight, if the image looks too glossy and professional.

--or too AI looking

[wise drift]

Wanting to do something similar for my godson to put him in stories he can read and such. Did you use a raw photo for the character reference? Or was the character reference from stock that you made with Leo?

[real lynx]

Use character reference with a photo of myself

Style and type: The image is a 3D digital rendering. Layout: The image Woman Blonde Hair features six identical animated characters arranged in a horizontal line, each in a different pose, showcasing a 360-degree view. Subject: The character is a young person with light skin, short brown hair, and a casual outfit. They are wearing a purple cap, a pink short-sleeved shirt, purple shorts, and white shoes. The character also has a purple backpack. The background is plain white, emphasizing the character's design and attire.

The prompt is this if you modify it you can create other characters

[wise drift]

That example I replied to went directly from photo of yourself to character reference, though? My character references always seem way off, but haven’t tried on basically cartoons yet

[real lynx]

If only in the prompt I put woman blonde hair

A little description helps

[wise drift]

Ah okay thanks

[real lynx]

Help is urgent, my face is half round and it is easy to imitate

[wise drift]

What are you holding onto lol

[manic root]

Desire to level up your skills, take on challenging projects, or hit your learning targets? Our mentors have got your back! Just swing by #👋│accademia and drop a /mentoring to get one-on-one help when you need it. 🚀

[warped locust]

Thanks Siloa ❤️

[fierce hamlet]

Anyone has some tips to prompt a 2-person fight scene? For example, "batman vs superman". When I used that prompt, the platform only took 1 character, not both batman & superman.

[storm jungle]

Did you try note where they stand, for example Batman standing on the left, Superman standing on the right

[fierce hamlet]

No, i didn't try that. I only did "batman vs superman". Ok, I will try that .. thank you for the suggestion

[storm jungle]

Example quick prompt:
A battle stance with batman on the left and superman on the right

[fierce hamlet]

Perfect. Thanks!!

[storm jungle]

Your welcome

[storm jungle]

With prompting it is best to explain each character differently and note which character has or doing what, another example, a girl on the left and a guy on the right, the guy has a blue shirt and pants, the girl has yellow dress........ you can also name characters too and describe each character for each name......

[fierce hamlet]

Ah, I didn't know we could name it. So for example, "a girl named jane, a boy named joe. Joe is on the right, Jane is on the left". Something like that ?

[storm jungle]

Exactly

[fierce hamlet]

Got it. Awesome. Thank you so much!

[storm jungle]

Anytime : )

[fierce hamlet]

Btw, does the prompt remember the name characteristics? For example, i named a girl "Jane", and generated an image. And for the next promt, can i say something like "Jane is now standing up"? Does it remember "who" Jane is ?

[storm jungle]

Short answer with a second prompt, no, you will need to explain all character names again, if your using edit with ai (Phoenix only) it may work, but a separate generated image it won't work, it might but as an Ai guess though.

[real lynx]

Hi

There is a structure that reminds us of the comic type in Phoenix

A 3-panel gag comic
Person=chibi cute 1Lolita 1girl Blonde Hair, studio giblhi style
Panel 1:
Person: looking at a blank piece of paper "I'm going to write the best story in the world"

Panel 2:
Person....

[fierce hamlet]

Thanks! hmm.. this may be a stupid question .. .what is "Phoenix" ? 🙂

[real lynx]

For example

It is a preset that specializes in texts, posters, comics, logos.

[fierce hamlet]

Ah.. i see it under the "Preset" .. thanks!

[real lynx]

☝️

[fierce hamlet]

thank you

[storm jungle]

There are no stupid questions when it comes to Leonardo : )

[real lynx]

By asking you learn, stupid is the one who does not ask because he does not learn ♥️

[pure geyser]

"Wow, that's deep. Maybe you should consider writing a motivational poster."😂

[thin tinsel]

There should be no 'memory' between prompts, as the AI is trying to offer variety and randomness, otherwise it would be a bit disappointing if we ended up with images that all looked rather too similar.
However, if you are trying to create a character, I normally use the name of a lesser-known tv actor in double-quotations. It won't look too much like that person, but the face will look consistent across images. #🙋│image-help message

[real lynx]

It is very likely that the phrase has already been invented xD

[storm jungle]

But in general, I believe there are no stupid people either, even if they don't ask, I don't like to judge : P

[wise drift]

It shows up on some tutorial videos on YouTube for creating consistent characters, sometimes emphasizing using unique names. Probably they got the idea from. E.g., make like 100 images of Hojiku Pojinara (made up name) with the prompts describing their looks, delete all images that don’t match what you want, AI then recognizes “Hojiku Pojinara” as though it’s a real person and proceeds to try and mimic such

[wise drift]

Have you looked into training models any? For like long term projects with set fictional characters, I believe that’s the route you would go

[storm jungle]

It may be referring to other Ai generations, as each have their own ways of dealing with prompts, MidJouney, Stylar/Dzine, Dall-E etc

[thin tinsel]

the issue with training a model is that you need 10-15 images. and so if people can't even create 2 similar enough images, they certainly can't generate enough to train a model with.
The actual key to character consistency is to use the character reference feature.

[wise drift]

I checked. It’s Leonardo.

[wise drift]

Isn’t it use character reference until you get 10-15 that are close, then swap over to trained model?

[thin tinsel]

Then what's the point of the trained model?

I know what you mean though. I have also seen YT videos that just use a mde up name.

[wise drift]

Fine-tuning, I suppose. Like if character reference weight went up to “we hijacked this person’s existence” lol

[thin tinsel]

I'm not sure the tuning is that fine.

I trained a model just to get the art style and general composition the same, but actually to create different characters.

[wise drift]

Idk videos I watched on it were for creating artificial people, with the trained model eventually being able to be prompted with whatever and it still look like exactly same person

[night vault]

I think Leo’s model training is outdated and quality is not good enough to my taste. I’m happier using new tools such as style reference, content reference.

[wise drift]

Probably less important with cartoons

[wise drift]

What were ya using em for

[storm jungle]

I'm the same, I much prefer content and style references

[night vault]

mostly hobbie and consistency is not issue. i can understand if you using for comics then LoRA is probably the best method but that is not available on Leo yet

[wise drift]

Ima flood social media with stories about Elon Musk’s secret estranged identical twin: Ebon Musk. He’s been on meth for years, so doesn’t look exactly identical, with the missing teeth and whatnot, but he still owns half of Elon’s stocks.

[wise drift]

Thought the elements were basically loras

[night vault]

yes it is. They are not let us train them though.

[wise drift]

Ah okay

[manic root]

Ready to level up your skills? 🚀 For personal guidance on your projects or learning objectives, our mentori team is eager to assist! Just pop into #👋│accademia and use /mentoring to connect. Let's grow together! 🌱

[potent basalt]

how do I make an image like this?

[thin tinsel]

Use describe with AI

[potent basalt]

and what preset and preset style you think is best to use for recreating this image

[thin tinsel]

@potent basalt Graphic Design > Graphic Design Vector

[potent basalt]

generates images like this. Its looking better then before but it still doesn't really look like it.. Maybe a setting I should turn on to improve it?

[thin tinsel]

What about these images differs from your example?

[thin tinsel]

I guess he's gone. If he comes back when I'm offline. I've done some images in the Team Account if anyone else wants to share my prompt or any example images.

[pure geyser]

what preset and preset style you think is best to use for recreating this image

[split umbra]

When I like one particular image of the 4 that were created, how do I keep working on only that specific image and disregard the other 3? By 'working on', I mean Edit with AI for example. (sorry for double post)

[thin tinsel]

Hi StiKy,
There is no exact answer. I think there are a few models that will do photography style.

My first try at this style would be;
Cinematic Kino > Dynamic.
And then try some of the other Preset Styles.

But I'd also use that image for Style Reference on Low. (if you have access to that).

[thin tinsel]

Hi Egrin,
It depends what you mean by 'keep working on'. If you means you want to make only minor alterations then you can use 'Edit with AI'.

However, in that case, you're not picking one image from the batch of 4.

It depends what you like about the one image. So for example, if you like the composition, then use that as content reference in the next generation. Or if you liked the style/colours, then use that as Style Ref to iterate from.

Let us know if you want someone to walk through the steps with you.

NOTE:
Edit with AI is only available with Phoenix.
Content and Style refence are not available with Phoenix.

[wise drift]

Does he maybe mean canvas?

[thin tinsel]

I expect they have been using MJ.

[night vault]

Except he is asking for edit with AI which does not how canvas works

[wise drift]

What Model on the mobile app is “Portrait Perfect”?

[thin tinsel]

Kino XL

[potent basalt]

what settings did you use?

[night vault]

Default is Kino. But you can change it with advance setting if you like.

[thin tinsel]

The ones I gave you.

[thin tinsel]

Did you want the white outline?
Did you want the side-profile?
Did you reddish tone to his hair?

[potent basalt]

and the model, does that matter?

[potent basalt]

yes

[thin tinsel]

#🙋│image-help message

Yes it matters.

[thin tinsel]

Then add these things to your prompt:
At the beginning.

"STICKER, White outline"

"Side profile"

and somewhere that is mentioned his hair, I just added "Reddish"

@potent basalt Are you a free user or paid plan?

[potent basalt]

paid

[thin tinsel]

Ok cool!
So you can also use the Style Reference feature. Do you know where it is?

[potent basalt]

no

[night vault]

Style Reference is also for free user anyway.

[thin tinsel]

Oh thanks. I didn't think the newer ones were.
We need a list somewhere handy!

[potent basalt]

where?

[night vault]

#💬│general message

[thin tinsel]

get it next to the promp. top one is style reference.

[split umbra]

Thanks, yeah i meant it like that, if i like the composition of one them and want to keep that for the next iteration working with Edit with AI. Thanks!

ehm but how do i put in the content reference again? 😀

ah ok not possible with phoenix...hmm damn

So lets say I have an image of a city created with phoenix and I want to add mountains in the background to that exact same image. Is that possible?

[real lynx]

Hi you can use albedo

It is the predecessor of Phoenix

It's not as good with the lyrics but it will keep a very similar structure and you will be able to modify it.

[wise drift]

Okay, Phoenix pretty cool lol

Crap it misspelled nuts lol

[thin tinsel]

You made the image with Phoenix, then you use Edit with AI, and write, "Please add mountains into the background".
Then run the generation again.
See my video here if you need to see how it works:
#💞│self-promotion message

[real lynx]

I understood everything I think xD I'm asleep

[wise drift]

Trained model works pretty well. Need to get all the base images to be same size though. What’s Input Size for? Tutorial recommends 1:1 ratio for the training images, but none of the Input Sizes are on a 1:1 ratio

[wise drift]

Also, would a character’s body/figure be a character reference or one of the other references?

[fleet hill]

Hi there, I'm struggling with fine tuned models, would really appreciate some of your help. In my dataset, I've uploaded 12 pictures of playing cards (all issued from the same classical "Parisian" style, so it should be very consistent).
But regardless of the category I selected (I've tried so far with General, Illustrations, or Vector), the result is completely off. See here result with prompt "King of hearts playing card".
Any idea what I'm doing wrong? Thank you!

[wise drift]

What are your training settings?

[fleet hill]

First one:

Second one:

[wise drift]

The Instance Prompt is what you have to include in your actual prompt. So you could just do like “Cards1” for the instance prompt for the first set. Then in your actual prompt for the generation do “Cards1. King of hearts”

[fleet hill]

oh, ok thanks, I'll try right now

[wise drift]

So you have to both select your trained model and also call upon it in the prompt

[fleet hill]

Does the "category" matter?

General, Illustrations, Vector... Is it just used to classify the fine tuned model or does it actually impact the generation of pictures?

[wise drift]

I’m on my phone right now so can’t check. Screenshot the whole page

[fleet hill]

Hm, putting the Instance Prompt didn't seem to work, I'll retrain a model with "Cards1" like you said.

better v1.5 or 2.1 by the way?

[wise drift]

Er, can’t see the rest of the page with like options and whatnot

[fleet hill]

ah sorry

I'm retraining the model now, do you think this is alright?

[wise drift]

I’m learning this at the same time. Haven’t got that far yet. Just started messing with it today. I did read, though, that you want the base images to be as close to your chosen resolution (512x512) as possible

[fleet hill]

Ok, good to know

[wise drift]

SS the list of categories

[fleet hill]

he, it's stupid but I can't screenshot while opening it...

[wise drift]

I think I’d also need to know what your “correct” image generation would look like, and how it would look compared to the originals

[fleet hill]

I can select General, Buildings, Environment, Fashion, Illustrations, Game items, Graphical elements, Photography, Pixel Art, Product design, Textures, UI elements, Vector

[fleet hill]

very similar to the original ones, but with different accessories

like this:

[manic root]

Seeking to level up your skillset, conquer big projects, or hit your learning targets? 🎯 Our mentors are ready to lend a hand. Just pop into #👋│accademia and type /mentoring whenever you need personalized guidance. Let's grow together!

[wise drift]

That almost seems like it would be graphical elements or UI elements, given the nature of playing cards

[fleet hill]

See the rabbit in the hat? That's the kind of modification I want to create. But I don't mind a bit of change on the rest of the illustration

[wise drift]

Did you already try Image to Image?

[fleet hill]

Yes, and style as well

[wise drift]

What was Image to Image result at max weight?

[fleet hill]

I'll try again

[wise drift]

For Image to Image, for something like that, you’d max out the weight and just prompt what you want different and nothing else

Aside from that, you could do the thing that generates a prompt from and image, generate with Image to Image, and then Edit with AI. There’s also Inpainting in the canvas.

[fleet hill]

Thank you so much

Didn't know about inpainting, sounds like a great fit indeed

Bottom ones are created with 0.3 weight, top one is maxed out. Let's try to describe the rabbit now... I'm not very hopeful haha

[wise drift]

For the inpainting, you’d set the box appropriate for the area you’re wanting to change and prompt what you’re wanting

Then just roll the dice and fiddle with it until you get something you like

[fleet hill]

For the inpainting, can I just upload the original picture directly? Or does it have to be the result of a prompt?

my bad, please ignore, just got the answer

[wise drift]

I honestly don’t know lol. The mentors are all asleep or hiding lol

[fleet hill]

haha

Not so easy either haha
Input

Output:

Still completely different result, even with the instance prompt :/
Thanks a lot for all your help though SilentSagas!

[wise drift]

Well, I tried, at least lol. The art style of the cards is pretty archaic. I doubt there’s enough info to just do it like that.

Did you try Illustrative model, with that as a style and content reference?

[fleet hill]

oh wait, I'm almost onto something

[wise drift]

Neato

[fleet hill]

You were completely right about the importance of image dimension

It's still messy, but the influence of the custom model is clearly visible now. Both in colors, composition etc

[wise drift]

What were they before?

Try to make them the 512x512 with just blank background behind the cards. Can crop all that nonsense out later.

[fleet hill]

I believe those were initial settings leading to the model being ignored:

[wise drift]

Both the circled ones are the default

[fleet hill]

but when you train a custom model, you can only train either 512x512 or 7xx x 7xx

other finding: the instance prompt isn't recognized if it's made of 4 words. My example "a court playing card" is not recognized. But "cards1" is

[wise drift]

Neato

It does say like 2-3 words on the tooltip

[fleet hill]

I should correct myself, I think I understand it better now:
What was breaking the model to be taken into account wasn't resolution nor the 4 words prompt instance. Actually both work now.
The main issue was that "Alchemy" was selected.

Nevertheless, resolution still matters: On top is the wrong resolution. At the bottom the correct one. There is a clear impact

The same everything except with Alchemy enabled:

[chrome kiln]

I want something crazier, my prompt just renders a ship in choppy water: Cinematic dark fantasy illustration. Medium shot. A pirate ship listing through stormy sea. Wide angle lens. Dramatic lighting with lightning strikes. Vivid colors with shades of indigo and grey. Foamy sea water splashing on the hull. Ripped black sails with a skull and bones logo. Kraken monster emerging from the depths. Best quality. High contrast. Moody atmosphere. Fierce dynamic lighting. Wild sea. Dark ominous clouds. Fantasy setting. Stormy weather.

[wise drift]

Gotta go to sleep. Good luck.

[pure geyser]

This is adorable @real lynx in card (♠️ ♦️) universe.

[marble cape]

yeah working on these two design i did on my ipad years ago trying to generate them

[real lynx]

What a great idea it reminded me of a character's eye

@nova turret

A modern, gritty, and vibrant street-style graffiti artwork sprawls across a weathered, textured beige wall, reminiscent of a New York City alleyway. The graffiti, done in a bold, fluorescent orange hue with crisp black outlines, takes center stage, evoking the dynamic aesthetic of a PlayStation video game. The phrase "Welcome Prelude" is scrawled across the wall in a stylized, urban script, with chunky, 3D-like letters that appear to pop off the surface. The surrounding wall is adorned with subtle, faded tags and splatters, adding to the urban, rebellious vibe. The graffiti itself appears fresh, with bold, expressive brushstrokes and a sense of energy, as if the artist just finished adding the final touches. The overall effect is a dynamic, high-contrast image that bursts with color and attitude.

Use the Phoenix preset

[nova turret]

i created an Image
then put it in motion
it took 4 -5 min but still generating

[karmic pulsar]

Now that we have a daily themes to create a movie poster, I was curious about adding texts to your generated image. I do put that in the prompt, but for some reason it either comes out all weird or not at all. Should one use specific prompts to add texts as preferred?

[real lynx]

Hi

Do you use our Phoenix preset?

[nova turret]

mentors will guide you they have a great experience in prompting

[karmic pulsar]

Haven't tried that one yet. Using Cinematic Kino and Lifelike Vision mostly. Will have a look at that Phoenix one.

[real lynx]

If our specialist in posters and signs is Phoenix

The other presets need another way to add text

You would need to already have a banner created with text to add to the image

In Phoenix just by typing what you want in quotes, it will appear

[karmic pulsar]

Ah, made a test generation with that Phoenix preset. Works like a charm! Thanks

[wise drift]

What’s “daily themes”

[karmic pulsar]

#1108130218260041758

[balmy prairie]

Hello, good to be here!
Can i use a description about something from a website? Can be anything within regulations ofcourse, I don’t mean to publish it, just to see what images it generates.

An example, a description of an hydrogen atom

[wise drift]

() is used for emphasis right?

[real lynx]

I think there was no problem

I would like to see what image comes out

[balmy prairie]

The generated language is great 😃 Hydrogen is the lightest chemical element with its most common isotope consisting of just a single proton and electron. At standard temperature and pressure conditions, hydrogen forms a diatomic gas, H2, with a boiling point of only 20.27 K and a melting point of 14.02 K. Under exceedingly high pressures, like those found at the center of gas giants, the molecules lose their identity and the hydrogen becomes a liquid metal (see metallic hydrogen). Under the exceedingly low pressure conditions found in space, hydrogen tends to exist as individual atoms, simply because there is no way for them to combine; clouds of H2 form and are associated with star formation.

The hydrogen atom A hydrogen atom is an atom of the element hydrogen. It is composed of a single negatively charged electron, moving around the positively charged proton which is the nucleus of the hydrogen atom. The electron is bound to the proton by the Coulomb force.

lol, not how I want it go, but there it is

[warm skiff]

If I want to make a series of images using the same face what is the absolute best way? I've done the whole fix seed and then trained based on them and got results that are less than... good.

[real lynx]

The best way is using image character reference

It is a premium tool

[warm skiff]

@real lynxI've tried that but I can't get it to not use the same kind of expression and angle.

[real lynx]

Let's start from the beginning again

Do you have premium tools available?

For example this angle

You can achieve this if you just put a distant bird's eye view aerial view at the beginning of your prompt

[marble cape]

Is that some sort of prompt you said?

[marble cape]

I did those years ago

[real lynx]

And I don't change the angle?

[storm jungle]

So your no longer Batman : P

[marble cape]

What?

[marble cape]

?

[real lynx]

Let's start over

What did you need?

Don't you want to change the angles of your images?

[wise drift]

Would it be Image to Image that I’d use to turn a slightly cartoonified image into a photo? Sometimes I get images I really like, but they look like 75% real and 25% AI as far as like textures and smoothness goes

[wise drift]

You sure you used the trained models correctly? The reference images should be 1:1 ratio and like 712x712 pixels (or whatever it says in the train model area). You then have to set the Prompt in the training options to be a keyword, like “JimmyDean”. Then when you go to generate images, you have to select your trained model and use the keyword at the start of your prompt like “JimmyDean. Man dressed as a giant hotdog”

[wise drift]

You mixed up two different people lol

[wise drift]

Cami you know answer to my last question? You tend to cartoonist things, wanting to know the reverse

Example. It’s like 25% cartoonified.

[night vault]

i think use any realistic model should work in most case. try lightningXL

[wise drift]

Image to image, though?

[nova turret]

uffff uffff

[wise drift]

Was an example I saved as an example lol. Pretty positive that’s just Madison Beer wet lol

[night vault]

if you want same skin tone, yes. edge to image or content ref (premium) in case you want same structure but change color

[wise drift]

I’m premium and have like 29k points or whatever. Don’t really understand edge to image or the other similar one, to be honest. Examples struck me as wildly different workings than the other references, enough so that it seemed they should be labeled and categorized as something else entirely

Everything is fine with the source image, aside from looking slightly cartoonized.

[night vault]

content reference is better then. edge are quite primitive version on content so it hit or missed a lot but some might figure out how to use then with alternate purpose.

[storm jungle]

If you are on the Artisan plan or higher, if your worried about tokens, you have unlimited, after you run out, you will go to 'relaxed generations' mode, the tokens will go negative to let you know your now in this mode, it resets to 150 each day and repeats until your new month starts again.

[wise drift]

Better than image to image, though?

[wise drift]

Well, ran out on MJ today, but that’s where my source images are from lol, before I refine on Leo of course

Zro, got any inputs on the uncartoonification?

[storm jungle]

You mean you want to turn a cartoon image into a realistic one ?

[wise drift]

For Leo it’s more of a time resource matter, due to PC times

[wise drift]

The example I posted higher up and similar, where it’s like 75% realistic looking and 25% fake looking, mainly due to textures, reflections, smoothness, etc.

[night vault]

as i said, if you want same skin tone. img2img is the right tool.

[wise drift]

Another example. Looks damn close to reality, but also very clearly not lol

[wise drift]

Ah okay. Just realized I read that wrong. Makes more sense now lol

[night vault]

what is prompt, preset and model used?

[wise drift]

Would take too long to scroll back that far on my phone, tbh. Believe it was from an MJ character reference, though.

[storm jungle]

You want something like this ?

[wise drift]

Minus the added ugly, yes lol

When I tried upscale and increased the creativity or whatever, it basically increased drug usage lol

Looks like that one just got home from a bender, if you understand the terminology and if that makes sense

[night vault]

haha, i see what you get into

[wise drift]

Sorry for the pickiness, btw

[wise drift]

I am well-versed lol

[storm jungle]

I was only testing it, so I didn't worry about image size/ratio at first

[wise drift]

Is it a clock, or a watch, out of curiosity.

[marble cape]

What do you think

…

[wise drift]

Yeah, that’s closer. Cheekbones slightly more angular or drawn in, and like she just isn’t wearing makeup besides the lipstick

[marble cape]

What do you see…

[wise drift]

Looks like a fkd up watch, tbh

[marble cape]

A watch is something on your wrist does that come close to it?

I don’t know if your seeing clearly

[wise drift]

The shape looks closer to a watch than a clock. You could just generate the right face you want, generate the right backing that you want, and then use canvas to stick em together

[marble cape]

It’s mounted on a wall…

[storm jungle]

I just used your image as an Img2Img, set the weight to 0.50, and use the 'Portrait Perfect' preset to generate......

[wise drift]

And it looks like a model of a watch lol

[wise drift]

Okay. Thanks. I can work with that, then.

[marble cape]

Is a watch that big of a size to you?

[storm jungle]

No problem, hope this helps to get what you are after 👍

[wise drift]

Dude I’m just telling you what it looks like lol. It looks like you took a basic watch and tried to turn it into a cuckoo clock lol

[wise drift]

Yeah, probably just do that and slowly roll the dice up until it gets close

[storm jungle]

Since your using image guidance, you don't need a prompt, other than a starting prompt, I just used 'Portrait photo" 😛

[wise drift]

Roll, save best, roll on that, etc

[marble cape]

It’s an actual real clock…

You seemed confused

[wise drift]

Happen to know if that works well in legacy mode with trained model? If not, it’s cool

[wise drift]

Sorry, I’m neurotypical.

[storm jungle]

You could try, not everyone likes Legacy, I admit I didn't like Legacy at the beginning, but now it's great, mainly because you need to do some trial an error and see what results work, people who don't' like it are the ones that don't know 100% of its settings etc

[wise drift]

Do the image references on it actually work with trained models? Was kinda confused on how those two interact

[storm jungle]

I use references all the time with it, it still depends on the strength and presets etc, I mostly use the text image input for creating logos etc

This was done in Legacy mode

[manic root]

Eager to enhance your abilities, take on big challenges, or hit your learning objectives? 🎯 Our mentors can lend a hand! Just pop over to #👋│accademia and type /mentoring to get one-on-one help when you need it.

[wise drift]

Ah, okay. Seems like they should roll all that into an upgraded “advanced” mode, tbh, as counterintuitive as that seems

Guess that other dude went to sleep

[storm jungle]

lol

[wise drift]

Any other just generic advice?

[night vault]

what generic are we speaking?

[wise drift]

Just whatever.

[night vault]

for me, use describe by ai helped gets subject i like from source. anything but style which you need other means like preset and elements

[real lynx]

First I made a cartoon with my photo

Then I use different reality presets

My drawing became a little more real and if I want it more realistic I use elements

[karmic pulsar]

That is absolutely adorable

[real lynx]

I love elf ears

[wise drift]

Friend of mine has a 3 yr old daughter who’s ears very partially folded during pregnancy, so it looks like she has ears of a Half Elf or Quarter Elf

[thin tinsel]

Content reference looks at everything in the scene and creates the same shapes. So if you want the person to stay in the same pose, then you can use Content Reference.
If you want the person's face to stay the same and everything else to change, then you'd pick Character Reference.

[wise drift]

Moreso meant, if I divided up all body types into like 20 different types, and CharacterX has BodyTypeY, and I want BodyTypeY consistency, regardless of poses or clothing, which would reference would I use?

[thin tinsel]

I've pinged you in the other channel

[hallow sable]

hi

i need help

[real lynx]

Hi

How can I help you?

[outer fiber]

Hello, i need help with an image i am working upon.

[nova turret]

what help do you need?

[outer fiber]

I am having trouble with the prompt not generating the intended result. The settings are :

Model : Leonardo lightning XL, Leonardo style, Using Negative prompts

I am trying to change the color of the object, it isnt working.

with few other issues

Ready whenever you are

[storm jungle]

Hey there, what are you trying to colour, and what results are you getting?

[outer fiber]

@storm jungle , I am trying to change a car color from Red to Blacl, along with it's alloy wheels color to black.

Black*

[storm jungle]

Are you using a prompt only or image guidance as well ?

[outer fiber]

Both

[storm jungle]

Not sure which guiidance you are using, but try using content reference, on med or low, the colour changes should work

[outer fiber]

Got it, ill tweak the stength, and update. Thanks.

[storm jungle]

No problem, if your still not getting the results you need, let us know

[nova turret]

done or not

[outer fiber]

Not working.

[storm jungle]

Example, changed red ferrarri to black with green wheels..... This was done on High setting, sorry, not low

[outer fiber]

Okay, however, i lowered the setting to low, only after your recommendation. It wasnt working at 97 either

[storm jungle]

This was with content reference only, content looks for the shape in the image only, so if you note the colours that you want it will change.

[outer fiber]

Do you mean 'Color' when you mention "content looks for the shape in the image only" ? Sharing the prompts: "Change Car color from Red to Black. Change alloy wheel color to black", Negative prompts "Don't change the background vegetation and the road".

[storm jungle]

Content reference is one of the image guidances

[outer fiber]

Sir, I have used an image from internet, used it as a Content reference to generate a new one, which was successful. However, further iterations are proving to be challenge..which is to say, color not being changed.

[hardy sentinel]

Hi there, delete your prompt and type from scratch : black car, black alloy wheels

[outer fiber]

Thanks, will update.

[storm jungle]

Can you provide a screen shot

[hardy sentinel]

You are welcome

[outer fiber]

Sure

[outer fiber]

Only 2 tokens left, exhausted the freeo ones trying to fix and iterate. WWill try after reset and update in here. CHeers!

[storm jungle]

Remove the notes about changing, as Leo might read this as a gradient with red to black, know what I mean... use notes like, car black, wheels white.. etc

[outer fiber]

Understood, ty for your time.

[storm jungle]

Your welcome, if you wish to wait until tomorrow to get fresh tokens, we will be still here....

@outer fiber Though I am not a fan of negative prompts, just use basic words, things like "text" or "words" instead of "don't put text in image" etc, just put a comma between words-----> text, missing fingers, extra limbs, poorly drawn, duplicate faces........... etc

[outer fiber]

Ill keep that in mind. Since you mention negative prompts, i tried it as the road and back ground vegetation was constantly changing. After negative prompt, it addressed that bit.

[hardy sentinel]

ok, let me try using realtime generation

how about this ?

[outer fiber]

Looks pretty good. I dont have to use the image for any specific purpose, i am just trying my hands on the app, and learn.

[storm jungle]

My example earlier with the red Ferrari to black and green wheels, this is all I used:

Black Ferrari and green wheels

[hardy sentinel]

ok

[hardy sentinel]

another black car

[storm jungle]

He wants to use an original image of car which is red, but change it to black, keeping the same background, but not a random black car in a random location

[thin tinsel]

Content reference can't do that.
Why not suggest canvas editor?

[hardy sentinel]

i see, but he is out of token, so I show him realtime generation

[karmic pulsar]

Nice set of wheels

[hardy sentinel]

great point, canvas editor can do that

[storm jungle]

This is what I got with content reference

I changed it from red to black, with green wheels

[thin tinsel]

Does that person's image just have a plain background?

[storm jungle]

His example is here:
#🙋│image-help message

[thin tinsel]

Oh OK, so it's their prompt that is the problem rather than the setttings.
plus the source image isn't very well contrasted.

[storm jungle]

Yeah, I told him not to describe it as change the car from red to black, as it might get looked at as a gradient from starting red and ending in black, but just to use simple words, black car black wheels

[hardy sentinel]

fantastic

[storm jungle]

@hardy sentinel @thin tinsel This was just content reference only as well.

Slight difference in the front of the car model, but colour wise turned out great

[hardy sentinel]

awesome

[hardy sentinel]

the difference is minimal, it looks like you set it to "low" ?

[storm jungle]

It is set to high

There needs to be an ultra option : P

[hardy sentinel]

awesome

[night vault]

you are using type=image to image which is not correct.
btw, are you on paid plan? because for content reference is premium otherwise you will need some alterative that work a bit more difficult but free one

[outer fiber]

Hi, i am on free plan. Alternates within Leo or third party tools. I've read the messages which were posted later after my last messages. What should be the 'type' ?

[outer fiber]

And by alternate, are you hinting at Work arounds ?

[night vault]

No, it is within leo. Just previous and less powerful iteration.

a moment

[night vault]

Click on view more, then either depth2image or edge2image

[outer fiber]

Thanks, will surely keep the above in mind. Waiting on the token reset 😭.

[manic root]

Hey there! 🚀 If you're eager to enhance your abilities, dive into awesome projects, or conquer personal learning objectives, we've got your back. ✨ Swing by #👋│accademia and type /mentoring for one-on-one guidance from our fantastic mentori.

[chrome kiln]

Can someone help me make this prompt better: anthropomorphized kajun frogs jamming on guitar, microphone, one bass violin, a 10 piece drum set, one grand piano, playing classical jazz, wearing tuxedos, in a swampy bayou setting.

on free account

[storm jungle]

Try this:

Anthropomorphized Cajun frogs jamming in an electrifying bayou jazz band. The scene is set in a swampy bayou with moss-draped cypress trees and glowing fireflies. Each frog is dressed in a sharp, tailored tuxedo, complete with bow ties, exuding a sense of elegance. The lead frog is passionately strumming a vintage guitar, while another frog croons into an old-school microphone, its voice filling the air with soulful jazz melodies. A third frog is rocking a stand-up bass violin, its fingers expertly dancing across the strings, adding depth to the music. The percussion section features a 10-piece drum set, where a frog with sunglasses and a fedora is energetically drumming, creating a rhythmic heartbeat for the ensemble. To the side, a grand piano sits on a small patch of dry land, where a frog with a top hat plays with effortless grace, its fingers gliding over the keys. The atmosphere is alive with the vibrant energy of classical jazz, and the entire scene is illuminated by the warm, golden light of the setting sun, reflecting off the water. The backdrop features a crescent moon slowly rising, adding a mystical aura to this lively Cajun jam session.

[chrome kiln]

super classy

[chrome kiln]

what is this witchcraft? I couldn't get any frog to understand what was in its hands LOL

[real lynx]

😅

[chrome kiln]

the lighting in this one is amazing

what style did you use to get it that sublime anime like?

[real lynx]

Frog mentoring :3

[chrome kiln]

it reminds me of rotoscope

[storm jungle]

Like that Keanu Reeves movie 'A Scanner Darkly' done entirely by rotoscoping, its a cool effect.

[chrome kiln]

can i get that effect by saying style rotoscope?

[storm jungle]

Not that I know of, don't think it is that simple

[real lynx]

((((scene inspired by Amanda Blake,Steve Cutts and Stephen Hillenburg.art))))).anthropomorphized kajun Mariachi band frogs jamming on guitar, microphone, one bass violin, a 10 piece drum set, one grand piano, playing classical jazz, wearing tuxedos,

[karmic pulsar]

Hey, is artsy nipple allowed? I'm trying to do this thang and Leo gave the woman a very abstract nipple....

[night vault]

you can do in your Leo account. Not sure if it will be welcome to post here though.

[karmic pulsar]

Kinda what I thought. Thanks 🙂

@clever fiber canva has a color wheel page, that's where I learned about color combos, fyi

[clever fiber]

Thank you!

[outer fiber]

Hi, how is your day/evening. What settings are required in the Canvas mode for the desired result under discussion.

[hardy sentinel]

Good morning

[hardy sentinel]

you can use both inpainting and sketch to image

[ocean bear]

How do you use unzoom?

[outer fiber]

Hi

[outer fiber]

Is there any settings that i need to keep in mind, under the 2 while trying?

[prisma sand]

Hello everyone! I'm working on a project where I want to combine two different images into one. The goal is to create an optical illusion: at first glance, the image should show a specific scene (like a landscape or cityscape), but when you slightly squint your eyes or look more closely, a hidden figure, such as Homer from Simsons, should become visible.

I'm looking for advice on how to best achieve this using AI tools like Leonardo AI. Are there specific techniques or settings I should use to ensure the figure is subtly integrated into the image, so that it only becomes apparent upon closer inspection? I would greatly appreciate any tips or guidance on how to approach this. Thank you in advance!

[hardy sentinel]

keep the working area within your image,

if you can , choose to generate one image per generation

[outer fiber]

At my end, the settings are wild, i can unchoose 4, everything else in premium

cant*

[hardy sentinel]

ok, good luck, you have positive prompt and negative prompt as well

[storm jungle]

Nice idea, though wouldn't be an easy thing for Leonardo to create an optical illusion or squint imagery as it is also called, wouldn't mind knowing how this is done too, I think basic optical illusions may work with spiraling shapes etc, but that's about it.

[manic root]

Want to level up your skills or dive into a cool project? 🚀 Our mentors will have your back! Swing by #👋│accademia and type /mentoring for that one-on-one help.

[uncut heath]

Is it possible to create this type of image with leonardo 🤔

[wanton belfry]

yes

but no blood

I'll reccomend you to find a good pose pic and then run it in anime XL(illustration) with your prompt

[uncut heath]

im not talking about just the effects, the character too

[wanton belfry]

you dont neccesarily even need pose, but its just for getting better output somewhat

anime XL is trained on it, so yes

[uncut heath]

is it trained on all anime characters?

[wanton belfry]

no, just few main ones

like Naruto, bleach, one piece etc

[uncut heath]

thats what i meant, thank you

[wanton belfry]

of course there are also other series involved too, as long as it was famous generally

even in the series, only famous characters

you can't make Fourth Hokage...

[night vault]

how about grown up naruto?

[wanton belfry]

it usually gives shippuden version

I tried very hard to make minato, doesn't happen...it keeps making naruto

it does understand Orochimaru, Zoro and alot of others though

also it mixes up Gintoki with Bleach style lol

[wanton belfry]

edgelord

[uncut heath]

i have some exp on ai anime with leonardo, i think it is quite impossible to creat image that lvl of details espcially the character

[night vault]

does it work with Super Saiyan 4 ?

[wanton belfry]

it has to be trained on it

[rigid folio]

Hi Everyone, i cant figure out how to turn a sketch in to an image using the APP. is this even possilbe?

[wanton belfry]

Yes

[wanton belfry]

doesn't seem to understand ssj4 form

probably not trained on it

[rigid folio]

thanks. sorry for the newbie question. probably just a little too excited.

[wanton belfry]

nevertheless cool image

[wanton belfry]

this is exactly a place for newbie questions

so no issues at all

[uncut heath]

cuz it is not canon hhh

[wanton belfry]

hehe, it doesn't matter much to AI...as long as it has some pics for it

[uncut heath]

btw i cant login into my leonardo account and i checked on reddit and it seems it is not just me

[rigid folio]

is there some documentation on this

[wanton belfry]

there might be, but frankly I've been off for long and I cant remmember on the top of my head about where they stored up new guides

but as a free member, img2img is probably your best bet

[subtle gulch]

Perhaps you can think outside the box and prompt for a female cosplaying Naruto? 🤔

I got this by accident 👆🏽

[wanton belfry]

still very hard, it always gets those whiskers

[subtle gulch]

Whiskers can be removed quite easily, at least easier than changing the whole hairstyle.

[wanton belfry]

it keeps around naruto/boruto

this is the closest I get

[rigid folio]

all good thanks!!

[subtle gulch]

Edited version

[storm jungle]

@pure ivy Can you forward this to the moderators please

[pure ivy]

I have forwarded your issue to the moderators. They should get back to you soon.
-# Our apologies if this message did not answer your question.

[elder glade]

Hey I just got the plan on leonardo. I want to make portrait headshots, etc. But all the pictures don't look like me. Any tips?

[subtle gulch]

Have you tried character reference?

[elder glade]

Yes i've uploaded the photo and tried several others, also changed the strength around but still can't get it to look like me

[pure geyser]

Share me the prompt with model preset. Thank you

[pure ivy]

Unfortunately I cannot answer that question.
-# Our apologies if this message did not answer your question.

[pure geyser]

Tell me Model preset with prompt

[wanton belfry]

Goku, super saiyan 4, tail, fur, energy, Yellow aura, Purple smoke, illumination, sparks, Strong muscles, torn shirt, Destroyed city, rubble, ruins, stunning enviorment

Anime XL, Illustration

[wanton belfry]

ah ooops I tagged wrong person

you can remove 4 and fur/tail from it

[pure geyser]

I forgot Improve Prompt(improve your current prompt)
Work on subscription or any other model preset

[pure geyser]

👍

[wanton belfry]

since it doesn't know ssj4

https://tenor.com/view/breno-bruhzil-brener-monkey-brenooo-gif-22109223

[pure geyser]

ssj4

[warm skiff]

Anyone have any tips on how to make... 'action shots'? LIke someone swinging a sword or charging into battle?

[wanton belfry]

for best results, its best you have reference images that you can use in Depth2img or img2img

with 1-2 subjects only

since you can't do very specific poses, you can get near it to a extent

[nova turret]

I want an image of a soldier (can be any like SAS ,SWAT, SEAL team) in dark and there is lightening at his right side in sky a very aesthetic image

[nova turret]

?

[manic root]

Eager to boost your knowledge, conquer major projects, or hit your learning targets? 🚀 Our mentors have got your back! Swing by #👋│accademia and type /mentoring for one-on-one support whenever you need it!

[feral bough]

What is your question exactly?

[real lynx]

The user used mentoring and we are already helping him 🫰

[nova turret]

yes

[nova turret]

i should put this soldier as my laptop bg 😃

[feral bough]

Okay👍

[wise drift]

Tada lol

Gettin better lol

[ocean bear]

Is there a leonardo ai bot for /imagine?

[storm jungle]

No sorry, images can only be made using the actual Leonardo Ai app.

[ocean bear]

Oh

😢

[storm jungle]

I know other Ai's can on Discord be here no, kind of a good thing, because it stops people creating an inappropriate image, I much prefer to use Leonardo app anyways

[ocean bear]

Yeah

same

just wondering

for my school project

we have a discord

so...

[storm jungle]

I see, fair enough.

[buoyant oasis]

Is there way to recreate spaceship from schematics in Leonardo AI? 🤔 I have old 1998 S-F ship schematics, wanted to create real color image of it in Leonardo... 🤔

( I can reverse colors to got white ship with black outlines - but how to change perspective with it? 🤔 )

[real lynx]

Hi

This is from some game ?

[wise drift]

Appears to be from a game called Trinity, published by White Wolf, and originally title Aeon

[buoyant oasis]

It's for TTRPG I run - Trinity series, Aeon game, published by White Wolf in 1999. 🙂

[outer fiber]

Hello Steve, how are you doing?

[hardy sentinel]

Hello, i am fine but sleepy how are you?

[outer fiber]

Ah, i understand. We can pick it up later. Sleep well

[hardy sentinel]

Hi, maybe combining content reference with character sheet preset from anime preset group.

[hardy sentinel]

Thank you it is morning here.

[outer fiber]

I figured, must have been up all night, take care

[hardy sentinel]

Thank you. What time is it in your place

[outer fiber]

Its 9 minutes to noon

[hardy sentinel]

I see

[storm jungle]

I was messing around with wood laser logos and got this... though the text was originally messed up, I had to edit it by hand in MSPaint, the left border of the middle image was all over the place, so I copied the right side and reversed it and place it over the messed up border on the left, I had two "FF"s so I cut that out move the "D" closer and roughly panted the background, doesn't matter how rough because Leo will fix it, so I used my edited version as a guidance, to regenerate it, and wallah, the end result, I love editing, been doing it for a while, especially when it comes to creating logos with other presets that don't like text..

[hardy sentinel]

Awesome

[real lynx]

According to what I read on Google, it's a game from 17 years ago. I was 6 years old so I don't know it 👋

[hardy sentinel]

looks great

[real lynx]

I have no idea what the ship was like.

[buoyant oasis]

**1E come out more like 26 years ago **- as it comes out in 1998. I run 2E of it, that came out in 2019. 🙂

[buoyant oasis]

This one seemed more faithfull to schematics in book, I think 🙂

[hardy sentinel]

are you using DOS Box ? or how do you play it ?

[hardy sentinel]

awesome

[buoyant oasis]

We play on Roll20, around 'virutal tabletop' - it's TTRPG, not cRPG. We play 'in person'. If we would be in same room, it would look like this - only it's SF game. 🙂 https://www.youtube.com/watch?v=_OENI3Lgytk

[hardy sentinel]

ok, I see

[manic root]

Eager to enhance your skills or need a hand with your learning journey? 🚀 The mentori are on standby to lend support. Just pop over to #👋│accademia and type /mentoring to get started!

[wise drift]

Didn’t say what you did tho lol

[frail vale]

Hello everyone

I am new here and a beginner my weak points are propmt writing & constent character building

needs some tips from the experts how to overcome

[storm jungle]

As for prompting, try using Leonardo's prompt enhance to create a better prompt from your original

@frail vale

As an example, I typed the following prompt just now:
An orange cat outside

Used enhance and turned into:
A vibrant orange cat with bright, inquisitive eyes and a soft, fluffy coat is situated outside in a lush green environment, surrounded by dense foliage and a few scattered leaves, basking in the warm sunlight that casts a gentle glow on its whiskers, with a subtle sheen on its fur, as if it has just been groomed, standing on a worn, weathered stone path or a natural terrain, with a blurred background to emphasize the cat's presence, and the overall atmosphere exudes a sense of serenity and peacefulness.

You can then even enhance that again and again until you see what you like, you can also see how it is written to learn from too.

[frail vale]

Thanks for your time, Yes Currently I am using prompt enhancement despite of that i want to develop a skill to create the best prompts as a Ai student this is my goal actually and seeking tips to create the best of my own prompts

[storm jungle]

Prompting is like writing a book, it takes practice, and you learn from other books/prompts, there are endless prompts to use, so it will be mostly different every time.

[frail vale]

I am getting your point

[storm jungle]

There are key words to remember depending on your look, for example, 'Create a 3D pixar style character" then your description of the character

[wise drift]

For character consistency and prompt enhancement, have you tried using Describe with AI for a picture of your subject(s)

[frail vale]

And what about the characters it changes every time, Lets say if i am following a story and generating character so it changes every time and sometimes it becomes very frustrating for me infect till now i am not able to follow a story with constant characters..

[storm jungle]

Get used to basic starting prompts like the following

Create a 3D pixar style city landscape
Create a watercolor painting of a city landscape
Create a coloring page style city landscape
Create a digital painting of a city landscape
Create an abstract of a city landscape

Best to use image guidance to keep the same face etc, this is more @real lynx strength, if you need help, there is always someone here that can help with what you need,

[frail vale]

No. how can i do that ?

[wise drift]

What type of characters are you trying to make? Realistic people, 3D characters, cartoons, etc.?

[frail vale]

Cinematic, 3D character's mostly

[wise drift]

Click on the little sparkles icon next to “generate”. Bottom option is describe with AI

[frail vale]

yes I know about this, I use keep it on all the time

[wise drift]

#🙋│image-help message

[storm jungle]

I haven't used that in a while but via the sparkle icon, I think it works with short prompts, if you already have a long prompt it won't enhance.

[wise drift]

#🙋│image-help message

Can try something like that, maybe, until you get one you like, then try using the result as a Character Reference.

[full leaf]

Hi
I have some questions about the consistent character feature in the image generation tool. Does the reference image of the character have to be human or can it be a cartoon character or dog for example and does it work better if the reference image of the character is photographic?

[dim zodiac]

it is designed to maintain facial features of humans

it doesn't work better or worse for cartoon vs realistic

people have seen some success with non-human characters, but it was designed around humans so is less successful with animals etc

[storm jungle]

And more images you use to create your model and how clear the images are the better it will be.

[full leaf]

Thanks you so much 🙂

[plain wolf]

Hi there,

I hope you're doing well. I have a question regarding how to create prompts for generating additional images using a trained AI model. Specifically, I'm looking to generate images where the face and figure are identical to those in the model that was trained.

Could you please advise on how I should structure the prompt to achieve this level of consistency in the generated images?

Thank you in advance for your help!

[wise drift]

Do you have prompt magic turned on?

[plain wolf]

where can i find it?

[wise drift]

You’re on Legacy mode, correct?

[plain wolf]

yes

[wise drift]

Do you have the trained model selected as your finetuned model?

[plain wolf]

yes, but i only have 3 Picutes. Thats because this AI didnt got me enough realisic Pictures of this one person.

[wise drift]

Do your prompts include the Instance Prompt words in them? Whatever your instance prompt is, you must include that word or phrase in the prompt. Cannot exceed three words. I just use like “Model1” and start prompt with “Model1 “

[plain wolf]

ohhh

one second i will look

okay now i will test it. My instance prompt is `a Model´ so i have to put it in?

[wise drift]

Yes

[plain wolf]

only? Or with something else?

[wise drift]

Can do the rest of the prompt as you wish

Beneath where you select the Alchemy option, the here’s a box with like high resolution, contrast boost, resonance, etc in it. Click on that. There’s options for Prompt Magic. Off, V2, or V3. When you have the correct prompt, run a test batch with each of those 3.

I get closest accuracy for the model I’m currently working on by having it turned off. V3 gives results more like an average individual, and V2 gives somewhere inbetween Off and V3

[plain wolf]

okay i will test it thank you. I have testet it with only the instance Prompt, and there came out nothing like my model

[wise drift]

Had you selected “character” as the category when training?

[plain wolf]

yes

[wise drift]

Your three images, are they all 1:1 ratio and of the same resolution as the training resolution and the base model’s resolution?

https://faq.leonardo.ai/finetuning/getting-started

[manic root]

Ready to level up your skills with personalized guidance? 😄 Our mentori are eager to help! Swing by #👋│accademia and pop in a /mentoring request to get started on your journey to mastery! 🚀

[chrome kiln]

QUESTION: I have this prompt, "Shin-hanga art, Inspired by Impressionism, Western elements, the effects of light & expression of individual moods, evoke nostalgia for history, "

I added this at the end: a group of geishas observing cherry blossoms along the river
could you suggest something else, vintage Japanese? It seems to totally prep the ai what it's going to draw and that last sentence is the execution

[ocean bear]

How can I make the most realistic images? (Alchemy or photoreal etc.)

[wise drift]

From a source? Or scratch?

[ocean bear]

Scratch

[wise drift]

Portrait Real or whatever model, photoreal. Haven’t seen any AI candid shots thus far, though.

[real lynx]

Real photo and alchemy

[real lynx]

Want to try something special?

[ocean bear]

Sure

[prisma sand]

Hello everyone! I'm working on a project where I want to combine two different images into one. The goal is to create an optical illusion: at first glance, the image should show a specific scene (like a landscape or cityscape), but when you slightly squint your eyes or look more closely, a hidden figure, such as Homer from Simsons, should become visible.

I'm looking for advice on how to best achieve this using AI tools like Leonardo AI. Are there specific techniques or settings I should use to ensure the figure is subtly integrated into the image, so that it only becomes apparent upon closer inspection? I would greatly appreciate any tips or guidance on how to approach this. Thank you in advance!

[nova turret]

@copper warren
brother help him I am not so expert of this
help him here so I will see how to do such thing.

[copper warren]

Hi there! If you are looking for help with a project, you might be interested in having your own thread with our mentors. If so, type /mentoring to request a private mentoring thread.

[storm jungle]

I am about to get off-line for a few hours, so maybe someone else may be able to help you, I have played around with this once before, though it is not an easy task, I have been tweaking around with optical illusions with Homer, and it's not quite working 100%, trying to get rid of the black outline it gives me and use textures of the forest to create an illusion of Homers outline

See examples of where I am at so far:

[wise drift]

One of my trained models while using v2 prompt enhance had a prompt like “wearing a dress in the city at night”. For whatever reason, all the generations kept being very heavy in like vines, with the subject turning into vines the further from center it got, and vines all along the edges of it. Homer pics remind me of that

[dim zodiac]

@storm jungle there is a Controlnet that can do something along these lines with the non-XL models, so you have to be OK with slightly lower quality images

It’s called Pattern to Image, and you can find it in the Legacy view, in one of the non-XL models if you have Alchemy and Prompt Magic deselected

Otherwise I think your best bet is Edge to Image at very low weight on the image you want to have acting as the “illusion” and prompt for the “camouflage” image

[wise drift]

I’ve seen what Prompt Magic being on/off in Legacy does. Whats the practical effects of alchemy vs no

[nova turret]

Brother is this AI?

[wise drift]

But what is it lol

[thin tinsel]

I posted it because that person was asking about optical illusion and 3RIC said about uding pattern to image, and then I remembered this image, of a woman's face, inside a leaf, at the colosseum, with a giant woman at the side. 🤪

[thin tinsel]

I don't think Brother is the best form of address if you're not sure of someone's gender.
Yes it is AI, it is created with Leonardo AI. 🙂

[chrome kiln]

what prompt did you use?

[nova turret]

sorry my bad.
ok.

[thin tinsel]

No problem

[dim zodiac]

Alchemy adds a lot of gloss as well as some prompt adherence.

[dim zodiac]

Yes this is a great representation of Pattern2Image in action

[real lynx]

A Shin-Hanga art piece inspired by Impressionism, with Western elements, light effects, and the expression of individual mood states. The piece evokes nostalgia for history. There is a landscape with mountains, water, and trees. In the foreground, there is a wooden boat. The sky is cloudy, with a few trees near the water. The mountains have a few structures. The overall image has a soft, dreamy quality.Engine 5, Pitted Metal, Rust Green, Ashen Gray, Clear White, baked and emissive light, natural light, reflective and absorptive, contour and relief, light shadows, hyper-detailed, Shot on 70mm, Depth of view, F/22.p

[storm jungle]

Are there any examples to see this in action?

[storm jungle]

Right on sister : P

[dim zodiac]

Pattern to Image

Same prompt Edge to Image low weight in current models:

the edge to image strategy has similar pitfalls to what you were showing with the Homer Simpson... can also show up way too strong:

(all same prompt BTW, just for sampling sake)

and the pattern is a name as it's one of the only quick things I had on hand for demonstration in black and white

[manic root]

Hey there! 🌟 If you're itching to level up your abilities, dive into cool projects, or hit your personal learning milestones, our mentori have got your back! Drop by #👋│accademia and type /mentoring to get the one-on-one support you're after. Let's grow together! 💪📚

[wise drift]

So edge to image is basically like “pose” but for non-people?

[storm jungle]

Oh I thought there was something already done with optical illusion

Too me this is optical illusion, see the baby that takes up most the image.

[dim zodiac]

yes this is exactly what you can do with it, but all I had on hand for a quick example was a name

aha! I knew we did a challenge on this once upon a time: https://discord.com/channels/989166677390426132/1156285704444326049

[storm jungle]

Maybe @prisma sand could check it out and get some ideas from it

[dim zodiac]

works best if you have the pattern you want in high contrast black and white (so in your example, an image of a baby) and then you prompt for the "deception" (IE: the couple looking out at a seascape)

[storm jungle]

I've never done this style before either, so maybe I will some testing with it myself

[dim zodiac]

it is pretty fun! And there was another controlnet added specifically for QR codes, because that was a trend for a little bit, hiding QR codes in AI images

[storm jungle]

I've only done double exposure images

[dim zodiac]

another really cool effect when you get good at it!

[wise drift]

I’m confused. Is it Pattern to Image or Edge to Image y’all are talking about? Cuz I show both in legacy mode

[dim zodiac]

Pattern to Image is the one that is ONLY available in Non-XL (easiest to find in Legacy) with Prompt Magic disabled

and it is designed specifically for the illusion type of images

[wise drift]

Ah okay 👍

[dim zodiac]

Edge to Image can work everywhere, but is a bit more "hacked" to try and achieve the illusion effect

[wise drift]

They kinda like inverses of each other?

[storm jungle]

What model work with pattern, each model it says the ControlNet is not compatible

[dim zodiac]

any Non-XL model as long as you have Prompt Magic turned off

(alchemy by default has Prompt Magic V2, so either turn that off WITHIN the alchemy settings, or turn off alchemy)

[storm jungle]

Still nothing

it fixed itself for some reason : P all good now

[dim zodiac]

kinda hilarious result... same prompt as above

[wise drift]

Primeape photobomb

[dim zodiac]

😄 indeed