🙋│image-help

[rain ivy]

Hmm not working image link if i'm missing something in my settings. I'll try with bolder text

[livid inlet]

Give me a few minutes to get my pc on and I'll see if I can find out what's going on

[rain ivy]

Thank you

Bolder text do not work either.I must be missing something in the settings.

I was using Leonardo 6 or 7 months ago, the interface has changed a lot, I'm a bit lost.

Thank you anyway for your time 🙂

[hardy sentinel]

You can also use Realtime Canvas

[livid inlet]

Its working for me, have you clicked this button(see image 1) to make sure the image sizes match up? @rain ivy

[rain ivy]

Yes

Maybe this part ?

[rain ivy]

i've never used canvas xD

[hardy sentinel]

ok

[livid inlet]

Have you tried with a different model, some are better with text than others. I like using Leo Diff XL

[rain ivy]

Oh... looks like it works with Leo Diff XL

Well thank you, i'll explore that

[livid inlet]

No worries, I'm glad I could help. I hope you enjoy all you can do with it ❤️‍🔥

[night vault]

glad it worked out, try lighting or anime xl, 2 latest model are cheaper and performing well

[rain ivy]

Thanks! 🙂

[rain ivy]

Do you mind sending your prompt and model ?

[livid inlet]

Its an old messy prompt, you may want to refine it a bit lol

[hardy sentinel]

that is awesome

[haughty eagle]

Maybe Just Learning Photoshop Text Editing isn't a Bad Idea

I Don't Think is That Difficult

[open echo]

Your Prompt Magic is turned on in the Alchemy settings, put it at « Off » as PM is not compatible with Image Guidance features 😉

[rain ivy]

rly ?

[open echo]

(It accidentally worked because XL models’s Alchemy also don’t support Prompt Magic 😜)

[open echo]

It’s this one here 😜

[rain ivy]

Thank you, it seems to work but it generate 4 almost identical images

[real lynx]

Create 1 image and change the strength.

Same prompt different forces and different elements

[hardy sentinel]

great example

[open echo]

Yes, since it’s passing through Alchemy at a certain point, you can see differences in the shapes themselves (faint changes) but then Alchemy takes over and apply the same « seed » to each of the variations

[rain ivy]

Ok ok

[hardy sentinel]

because the weight of your reference is very high

[real lynx]

I don't use premium tools, I only use free tools.

[open echo]

Eh! To be fair since the XL models, I don’t really use Alchemy myself 😅

[night vault]

yeah that what I observed too. alchemy and controlnet works but you need to give up diversity of output

[hardy sentinel]

I mostly use free tools too to help others.

[haughty eagle]

Freemium Gang

[feral bough]

🧐

[prisma spade]

from where @real lynx

[real lynx]

Hi

I from argentina

@prisma spade How can I help you

My name is Camila, nice to meet you.

[lavish whale]

is this the same as the alchemy "dynamic"?

[open echo]

@west pivot Get over here and it'll be easier to workshop your idea 😄

What have you tried so far?

[open echo]

Good question 😄 It's true that some terms can overlap between functions and purposes.

the "dynamic lighting" here is a keyword you can use in your prompt to tell the system to incorporate patterns, colors and shapes that have more light sources and not the "Dynamic" preset of Alchemy

For Alchemy, they will have indeed have an pre-set effect on the generation. Well.....maybe I didn't create the best exemple as here the only difference is that "Dynamic" has two more leaves in her back than "None"... 😅

[west pivot]

I wanted a black angel and a white angel fighting against each other, but this is what i get.

[open echo]

Yeah... not much fighting going on 😄

What was your prompt and model?

[west pivot]

It's easy to generate images when prompted with hugging and kissing,😂😂 but when it comes to violent actions, it's much harder to generate them. Is this difficulty due to the filter?

[west pivot]

prompt : aerial shot and fullbody, an angel with white long hair and feathered wings and a devil with a black feathered wings, face to face, holding a sword against each other, phantasm, with big expressive eyes, superimposed, chimerical, surreal, fantasy-like, perfect anatomy, symmetrical, Intricately detailed, ultra-detailed, high quality, high resolution, 8k
Anime XL

[open echo]

Getting somewhere...

(Also got a cool accidental pic :D)

There's also always the posibility of fixing issues and details after the fact in the leonardo canvas

[west pivot]

this one is pretty

[real lynx]

Fusion

[ashen coral]

No, dynamic is a preset of alchemy

[lavish whale]

are there any docs for what these presets actually do

i cant find any

[ashen coral]

Ask to @open echo

[night vault]

i’m impressed you manage to fit everything without exploding the model 😆

[real lynx]

New super power for the world cup :3

[craggy gull]

how do I keep getting this style of image. You know, the height and the body overall.

How to control the height and body shape?

[night vault]

chibi - for body proportion, and use render size as square or landscape
vector art, cel-shading - for color palette (also try flat illustration, folk illustration, simple icon element may help here (pick one try like 0.5 then tune it))

[craggy gull]

thank you! I'll try

How to get body proportion like this? @night vault 🧐

Or what is the overall style?

[night vault]

just use chibi in general and prompt for full body and hope it will give you

chibi has many ratio. this one is 1:1. I’m not sure if model can understand ratio specificity though so i just recommend chibi

[craggy gull]

😆 Nice. This is so helpful

Are you a staff from Leonardo?

[night vault]

i am not. just helpful guy 😁

@craggy gull chibi 1:1 are not very common. maybe big cute head may help with that

this is 1:2 (head to body ratio)

[craggy gull]

ohhhh

I get it

[night vault]

1:1

[night vault]

1:2 i think

[hardy sentinel]

Very cute

[craggy gull]

I think this ratio is perfect but the muscle etc is too detailed. Should I just use prompt "less detail"?

I'm using this for creating mobile games but we can't have the same muscle

seems like turning off Alchemy is better

[night vault]

add simple icon element helps

[craggy gull]

OMG

Thank you again！🥰

[severe bobcat]

when I try to generate an anime character there's something wrong with their eyes. They aren't clear.

[night vault]

can you show some example and prompt? I will check in few hours

[severe bobcat]

A boy with blue hairs sitting on a dragon throne in king's posture. Background has space.

[night vault]

those can be fixed with upscale, you can try refiner upscale

[severe bobcat]

I tried upscale. But they were still same. I'll try refiner once again if you suggest.

[hardy sentinel]

how about fixing it in canvas editor? you can use sketch to image so that you can choose which color to use.

[night vault]

@wanton belfry could you help?

[wanton belfry]

Alchemy Refiner Strength High

if it turn girly, might tone down

by the way, how about this?

[livid inlet]

https://discord.gg/AFCPdq66?event=1229445194802270308

[analog stream]

I am creating an image. When I click on generate I have the following error message: "Error
Invalid init generation image id"

What is it?

It is not possible to create anything. My prompt: "Fauvism, naturalism, pencil drawing, pale pastels, extremely fine drawing, hyper detailed. Full body pose from head to toe of a standing barmaid. The barmaid, named Lora, wears very low cut lace neon red dress, very thin, transparent. Lora has a tanned skin, symmetrical neon blue eyes, thin nose, pink thin lips, black stockings and garter belt, leather belt with gold buckle. Lora comes to serve the coffee on the terrace. "

[severe bobcat]

Hey! It's muchh better now

[severe bobcat]

Alright, I'll do it.

[hardy sentinel]

Ok set the image generated per click to one, so that you save tokens

[severe bobcat]

Okay! Thanks a lot.

[hardy sentinel]

You're welcome

[dim zodiac]

This sounds like there is a seed or image input that shouldn't be there. If this isn't the case and both are clear, I would clear the browser cache and log back in. If that still doesn't work, I'd talk on the intercom with a solution representative

[west pivot]

i need help plsss, how can i make a 1st image just have only two fangs, it always turn out like this 2nd pic.

[tall elbow]

try "showing a little white teeth from her smile" or some like that

[west pivot]

now it have two fangs 😆😆

[west pivot]

i got it through image guidance, image-to-image,

[tall elbow]

great!

[west pivot]

thank you so much for your help and tip😊❤️

[tall elbow]

I hope you can continue growing creatively!

[dense vector]

i dont get why leonardo is totally ignoring the description i give it and just does whatever it wants. it should take in account when size is being mentioned.

is it a bug ?

[hardy sentinel]

Hi, nice to meet you. You can adjust size with sliders

or type in the exact pixel in 8 pixel steps. For beginner I would recommend using the slider or use recommended sizes.

Number 3 (light green) is where the recommended sizes.

[thin tinsel]

why would you be putting the size of the image in the text prompt?

[hardy sentinel]

Hi Willow, this is customary in MidJourney

@dense vector I have marked where you can adjust the aspect ratio

a pull down menu will appear as soon as you left click that area

you can use the slider (here shown as 640 x 832) or if you know what works you can type in the exact number

Be careful if you type in yourself. Check it at least twice, before you click generate.

[hardy sentinel]

as for description, if you need help, you need to post your image, prompt, model and parameters so that one of members here can try to help you.

[warm forge]

Hi, I really want this img to be redrawn in the following style (same pose, colours ect)

Style imgs:

[thin tinsel]

aha, I see, thank you Steven. 👍

[hardy sentinel]

you are welcome

[rich rivet]

Hey

I want some help

Help me create this

What kinds of prompt would it take?

[subtle gulch]

You can do that with a img2img with a strength of around 0.4-0.5

Then you can add description of the style in the image you provide.

[rich rivet]

Can you do?

I am not getting it

[subtle gulch]

I used your image and did an image to image of strength 0.45 with my prompts for style.
(Spoiler: mild gore)

[rich rivet]

@siola

can you generate for me

[subtle gulch]

You can trying promoting “Cinematic movie still, high angle, dutch angle, vanishing point, an Aisle of an aeroplane with passengers on two sides looking shocked, light from plane windows”

[real lynx]

Hello how are you

Nice image to transform into an elf

💞

Only with free tools

I don't have any more tokens :3

2 tokens

[ionic cobalt]

Passengers aboard an infinite plane, faces illuminated by a mysterious cosmic glow, gaze in awe and wonder as they travel through a vortex of time and space, embarking on a journey beyond the stars. ✈️🌌🌀

[warm forge]

Nearly perfect. Just remove the weird shadow please. And I want red eyes like original. Other than that the style is perfect

[thin tinsel]

I'm pretty sure he was giving you an example of how YOU can do it.

[stark zenith]

How can I remove those random drawings from my character's face in Photoreal mode?

[thin tinsel]

Do you mean how can you stop it happening?

[stark zenith]

Yeah. I apologize, I wasn't clear. You can delete it later. But how to prevent it from appearing in the Prompt?

[thin tinsel]

May we see your Prompt?

and what model did you use?

[stark zenith]

A seductively fierce elven warrior, her fiery braided hair cascading down her armored back. This digital painting captures her in a dynamic battle pose. The artist expertly highlights the intense expression in her piercing emerald eyes, reflecting determination and strength. The attention to detail in this stunning piece immerses viewers in the fantasy world of this captivating character.

Leo Photoreal v2

[thin tinsel]

For the model, do you mean Leonardo Diffision XL ?

[stark zenith]

It seems to me that parts of the armor are displayed on the character's face.

[thin tinsel]

My suggestion is to remove the word "Elven" and just have warrior.
Then in the negative prompt put:
Tattoo
Facial-markings

[stark zenith]

But there's no way to write a negative prompt there 🙂

[feral bough]

Can't have negative prompts in Photoreal🤐

[thin tinsel]

Maybe try Celtic Warrior as an alternative

[feral bough]

How about "smooth skin"?

You can also quite easily edit out the markings with Canvas.

[thin tinsel]

Do square brackets work as an alternative?

[feral bough]

I'm not sure

[thin tinsel]

Ok I'll try it. Thanks

[feral bough]

I think brackets are only in Alchemy but I don't know if Photoreal uses Alchemy as well

[stark zenith]

With Photoshop or additional processing, you can clean it up. I just thought there was something I could do to prevent it from appearing

[thin tinsel]

That's a shame the prompt changing didn't help. Did you get the same without Photoreal.

[stark zenith]

Without Alchemy

Even with a negative prompt showing up

[thin tinsel]

Yes. But in those you got 3 of 4 without , whereas previously it was 3 of 4 with.

[stark zenith]

Yeah, but I don't know what it is. But it looks like there's no way to get rid of this problem completely.

[thin tinsel]

You could do some with Lightening XL and then use the Upscaler to acheive more photorealism?

[stark zenith]

I'll edit the finished versions later.

That's not the point. In photorealism, the model draws the plastic anatomy of the face that I want. 🙂

[feral bough]

@stark zenith Don't put "No tattoos" in your prompt, it works counter-productively

[thin tinsel]

There will always be trial and error with ai images at the moment.
I was just looking back through my celtic warrior images and I can see some of them had facial markings too.
So you might want to go down more of a descriptive route.
Lose warrior altogether and say green eyes, red hair, wearing green medieval armour etc.

Did you use Leonardo Prompt Improver to write the prompt?

[feral bough]

Celtic is associated with markings, runes etc. If you use that term you have a higher chance of getting those kinds of markings.

In my experience at least

[stark zenith]

I'll give it a try

In the beginning it was an elven warrior 🙂

[thin tinsel]

Yeah, I thought that's where the markings were coming from.
But seems celtic warriors have them too. -- I was trying to keep the red hair with that suggestion

[stark zenith]

Thanks for the help 🙂

[thin tinsel]

You're welcome.

@stark zenith You're welcome to try out my prompt:

[thin tinsel]

What do you think?

We're doing 'Knight' images in the live stream event.
You could try Knight instead of Warrior. Just as long as there is something else in the prompt that indicates it should be female

[stark zenith]

Yes, problem solved with this prompt 🙂

[thin tinsel]

Happy to hear it.

[stark zenith]

Thanks again!

[thin tinsel]

You're VERY welcome!

[dense vector]

https://cdn.leonardo.ai/users/8911d5b6-aff8-4757-bc66-478c7c8e8b2e/generations/b3a5f279-9c4d-4af7-8a19-2fa61ed2b73f/Default_Generate_a_tiltshift_photo_of_a_miniature_real_life_LE_0.jpg

[hardy sentinel]

ok I am at a live stream right now, you can join at the stage.

[dense vector]

https://cdn.leonardo.ai/users/8911d5b6-aff8-4757-bc66-478c7c8e8b2e/generations/53cc1c46-bba8-43ff-aca4-77bbd992836d/Default_Generate_a_tiltshift_photo_of_a_miniature_real_life_co_0.jpg

https://cdn.leonardo.ai/users/8911d5b6-aff8-4757-bc66-478c7c8e8b2e/generations/83210957-8e93-4d18-9906-4f66a9543df0/Default_Generate_a_tiltshift_photo_of_a_miniature_LEGO_Star_Wa_0.jpg

https://cdn.leonardo.ai/users/8911d5b6-aff8-4757-bc66-478c7c8e8b2e/generations/83210957-8e93-4d18-9906-4f66a9543df0/Default_Generate_a_tiltshift_photo_of_a_miniature_LEGO_Star_Wa_0.jpg
in this one Kirby is nowhere to be found
https://cdn.leonardo.ai/users/8911d5b6-aff8-4757-bc66-478c7c8e8b2e/generations/e07431cc-e013-45e8-90bc-0c19d651e8ad/Default_Depict_tilt_shift_photo_of_a_real_life_Super_Mario_Kar_0.jpg
https://cdn.leonardo.ai/users/8911d5b6-aff8-4757-bc66-478c7c8e8b2e/generations/e07431cc-e013-45e8-90bc-0c19d651e8ad/Default_Depict_tilt_shift_photo_of_a_real_life_Super_Mario_Kar_0.jpg
https://cdn.leonardo.ai/users/8911d5b6-aff8-4757-bc66-478c7c8e8b2e/generations/d63a3162-5290-401b-b9c0-b03228497be3/Default_Create_an_enchanting_tiltshift_photo_that_transports_v_0.jpg
https://cdn.leonardo.ai/users/8911d5b6-aff8-4757-bc66-478c7c8e8b2e/generations/9fc8d197-ba9c-4d6c-80af-bcf3a88b3780/Default_Create_a_captivating_tiltshift_photo_that_transports_v_0.jpg

[hardy sentinel]

ok, let's wait after the live stream is over @dense vector

[hardy sentinel]

Would you review the warning pelase? @stable carbon @feral bough it think the bot misunderstood the intention of @dense vector , it was part of prompt helping session.

Please kindly review the warning points

are you online @dense vector ?

[dense vector]

yes

[feral bough]

Thanks Steven but I dont think removing those points is possible because it's automated by the Wick bot. But there are no negative consequences for @dense vector in any case. Just please in the future copy-paste your images into Discord directly instead of sending a bunch of image links at once because the bot doesnt like it and it takes up a lot of space.

[dense vector]

ok will do

[hardy sentinel]

Thank you , Dank je wel 🙏

[dense vector]

im new to discord aand its very confusing

[hardy sentinel]

ok let start with your questions now that I saw your images.

Does it has to be that aspect ratio and size?

[dense vector]

well yes. i donmt understand why (not that particular image) leonardo compleatly ignores prompts. like miniature and tilt shift and such and just gives me the exact oposite

[hardy sentinel]

ok, let us start with one image first.

So do I understand correctly, the image has to be that size and 9:16 aspect ratio?

I mean there is no other size?

because this aspect ratio limit the number of subject you can display optimally

ideally only one subject for 9:16

[dense vector]

it can fit two if they are miniature

[hardy sentinel]

yes, do you need this for Tiktok or YouTube Shorts? why it has to be 9:16?

[hardy sentinel]

ok but it is not optimal

[dense vector]

just standard mobile wallpaper

[hardy sentinel]

oh ok, then this explain why Kirby is nowhere to be found.

There is not enough space to include Kirby

@dense vector I have not read your whole prompt but it seems a bit wordy and using complete sentences?

[dense vector]

it was just an example, i made it much much smaller and to the point using simple english

[hardy sentinel]

ok, not everything is covered by prompt, i see that Super Mario and the first R2D2 are minature and using tilt shift

[dense vector]

they all are.

im ina tilt-shift mood since a few days

[hardy sentinel]

yes, but your aspect ratio and subjects are not always make it possible

[dense vector]

kinda true but sitll i had some good result but thats one out of 30 and there's the problem. ive noticed that as leonardo goe further and deeper the more messed up the images become and he doesnt recognise prompts anymore nor does it listen to negative prompts

it just adds stuff. ive it that it cant do two animals in the same frame....at all

[hardy sentinel]

well, ok I can not help you with that.

[hardy sentinel]

two or more subjects are tend to be more problematic than one

[hardy sentinel]

it is almost midnight in Europe, so I need to go. Maybe somebody else could help you.

[dense vector]

yes its 12:36 here. have a good night and thx

[hardy sentinel]

Good night, thank you

[rough hedge]

hey i kinda need some help, I dont know how to generate images like this one, i've been trying for a while but cant get anything remotely close to this style

[open echo]

hmmm... that looks to be a very precise style.. do you have an artist name?

trying with "minimalist vector art, cartoon" I can get this for a "business deal"

but "sleeping lion, savanah" clearly borrows from another set of images 😅

[rough hedge]

yes yes this is almost exactly what i wanted haha

Wow lol

Thanks ahah

[frank juniper]

hi there,
Quick question, what is the best way to "always" get a full-body image?
Adding "full-body picture" in the prompts always do it, most of the time, it will zoom in.

[dim zodiac]

it's more consistent to describe footwear, or what the character is standing on

this forces the AI to consider the full person

[frank juniper]

thanks you 🙂
I did use the "trick" to tell it what she is standing on, but I was still getting zoomed in pictures

[dim zodiac]

what aspect ratio are you generating?

taller aspect ratios will often get better results for full person images too

but yes, the AI does like to generate portraits from waist up

[hardy sentinel]

Hi there, try to change aspect ratio to 9:16 or 1:2

[floral geyser]

hello hello, is the image you use for image guidance public? been trying this feature recently and I was curious about the outputs if I used a picture of myself

[wise canyon]

Hi there, the images you upload for image guidance are not public. The generated output images from it will be public though..
(Paid subscriptions can generate private image outputs.)

[floral geyser]

thank you

[wise canyon]

Also, with a paid subscription users can delete any of the images they upload after using them, if needed.

[floral geyser]

sounds convenient and useful thanks for the info

[hardy sentinel]

I will suggest that you pick one image that you like and still has some space, then you can add Kirby with canvas editor, using inpainting.

you can change the amount of images generated per click to one to save token.

[thin tinsel]

You Can use the Simple Flat illustration element with any of the XL models.

[rough hedge]

Yes i did that and its perfect, i use 0.10 and my images are great

i made this one which isnt really how i wanted it but it turned out cool lol

[thin tinsel]

Yes, You can see it has the Vector Illustration style.

[open echo]

WHY DO I ALWAYS FORGET ELEMENTS EXISTS?!?!?! 🤣

Thanks Willow!

[thin tinsel]

LOL. Probably because your Prompting Prowess gets you to amazing places without elements. 🤗

[open echo]

https://tenor.com/view/oh-stop-it-gif-24046694

[rough hedge]

yeah its very good thats the style i wanted lol, its just very hard making prompts

[thin tinsel]

You're welcome to go and look at my prompts. Click on my name and you'll see an link to my profile. You'll see some of the keywords I used.

[rough hedge]

Cool! As of right now, i ask AI to write me prompts lol, thats how bad I am lol

[hardy sentinel]

beautiful, looks like an eye

[rough hedge]

Hmm yeah it kinda does, Wanted it to look like a black hole from interstellar

close enough i guess lol

[hardy sentinel]

it is black and looks like a hole = blackhole 😄

[rough hedge]

Haha yeah true

I made one that was supposed to look like a nebula

It kinda does

[zenith tiger]

@stable carbon

[subtle gulch]

If you want to create image of that style, Leonardo has just released the “style reference” feature. Just put the image in image guidance and choose “style reference”.

Just be aware you won’t get exactly the same character, but just “the same style” as the image.

[zenith tiger]

ok thank you

[little moss]

@iron steeple BORA TRABALHAR HOMEM. Vou já fazer uma cota aqui pra comprar um cel que grude em você

[rough hedge]

What does this mean?

I cant generate images

[night vault]

some setting is broken, try login again

[lavish whale]

Does anybody know what the difference between alchemy general and dynamic are?

[subtle gulch]

#❓│platform-questions message

[west pivot]

Hello, I need help pls. Can someone give me a trick or something to combine these two? I'm having a hard time with this 🥴🤨

i just wanted a fangs of the first image to this second image.. i tried image to image, depth, and style transfer but it didn't work the way i want

[warm forge]

How to change all the blue to red?

[night vault]

have you tried pose? and add fang to prompt (not combine image though)

[night vault]

edit from this? you can use edge2img then put another style reference of red you like

[west pivot]

yup, i tried that at the same time, pose, depth and style reference

[night vault]

what are strength level used?

[west pivot]

all 1

this is the result, but the head

[night vault]

you can try 0.7-0.9, btw looks great already

i think depth maybe better imo

[west pivot]

ok ill try it thanks😊, sometimes, maybe i should try this one, for style references

[rough hedge]

This is my image, i will show my prompt below but i keep telling leo to make the penguing stand on top of the iceberg not on the seawater, i want the penguing standing on the edge of the iceberg with the seawater below him but it keeps putting the penguing there on the water:

prompt:

A serene winter antarctic scene bathed in shades of cold. Large icebergs floating on the deep wave seawater, and misty background. A solitary small black penguin stands prominently on the edge at the top of a iceberg overlooking the antarctic in a majestic pose, its form a striking silhouette against the icy landscape. The dreamlike atmosphere is enhanced by a hazy, ethereal light shining, creating a sense of mystery and magic. In the distance, faint spires and peaks hint at a fantastical frozen realm beyond the antarctic. The overall aesthetic captures the tranquil beauty of an enchanted wintry arctic wilderness, inviting the viewer to lose themselves in its cold lonely hushed, mystical ambiance, Illustration, wallpaper, aesthetic, add seawaves,

[real lynx]

He didn't want to open his mouth to show me his fangs.

The man in green has the fangs xD

I liked the shine in his eyes

[dim zodiac]

Your prompt is very detailed but is 118 words. Rule of thumb is that the AI will start ignoring after 75

Start with your sentence that describes the penguin at the edge of the iceberg, then follow with the environmental description

This will help the AI focus on the penguin and what it should be doing

Consider using real-time generation to get the composition right using the prompt and potentially even a sketch2image from canvas or realtime canvas

[rough hedge]

Oh wow this is a lot of things to consider, is it really true that it ignores after 75 words? Also im afraid that if i change the prompt the images will look different i used this prompt for various of animals and envirnoments and it worked well but not with this penguing and iceberg lol.

Also im not really that good with using all the different tools you mentioned at the end lol

[warm forge]

How Did you make it?

[subtle gulch]

I told you already. Use img2img of strength 0.45 and use prompt to describe the style you want

I did use alchemy though, which is a paid feature that enhance details and quality.

[warm forge]

I get this menu

[subtle gulch]

Do you see “Image guidance”?

Just next to generation history

[warm forge]

I need img to style img

[subtle gulch]

Upload your reference image and choose the image to image function

[warm forge]

Bit confused, which img is original and which is style

[rough hedge]

Kinda like this?

In the heart of the Antarctic expanse, a vast canvas of icy serenity unfolds. Towering icebergs glide atop the frigid, undulating waves, shrouded in a veil of mist. Amidst this frozen kingdom stands a solitary figure - a diminutive black penguin, perched proudly upon the pinnacle of an iceberg. Its silhouette stark against the glacial backdrop, exuding an air of regal poise. A soft, ethereal light bathes the scene, casting an aura of enchantment and intrigue. Beyond stretches a realm of surreal beauty, where distant spires and peaks hint at an otherworldly frozen domain. This captivating tableau invites immersion into a world of tranquil solitude and mystic allure, where the chill whispers of the sea beckon exploration.

I asked gpt to help me lol

[wanton lark]

There is a toggle drop down where it says image to image. You'll see it says style reference

[dim zodiac]

Give it a go! I would open with the "a diminutive black penguin perched on the PRECIPICE of an iceberg" sentence. I changed pinnacle to precipice as pinnacle means peak, and you asked for edge

[rough hedge]

?

#🖼│image-share

[feral bough]

In addition to what Eric said about the 75 words, it is better not to use ChatGPT-generated prompts because they are overly verbose and have a lot of extra words. And if you have extra words, the later parts of your prompt are more likely to be ignored due to the limit of what AI can understand of the prompt. Try to remove fluffy description like "in the heart of ... expanse", "captivating tableau invites immersion" etc. Instead, use keywords, like "black pinguin standing on an iceberg, in the antarctic, mist, frozen waves, regal poise" etc.

[rough hedge]

Oh, I get it, yeah I saw those weird fillers words but didnt want to remove them because i thought they served a purpose.

[dim zodiac]

yes, GPT even when told to format for Stable Diffusion, seems to be extra verbose

I was trying to help from my phone 😄 but now I'm at a PC so could suggest an edit:

a diminutive black penguin, perched proudly upon the edge of an iceberg in the heart of the Antarctic expanse. Towering icebergs, undulating waves, shrouded in a veil of mist. Silhouetted against glacial backdrop, with regal poise. Soft, ethereal light, aura of enchantment and intrigue. Surreal beauty, distant spires and peaks, otherworldly. Tranquil solitude, mystic allure, icy serenity

This would be a better Stable Diffusion prompt, that would hopefully contain enough of the same keywords as what you've done before to keep consistency

[steady shale]

Hi everyone. I am new to Leonardo. This is an image I generated, but I want to change the color of the hair to match the scales. How can I change it?

[open echo]

Hello hello 😄

You can use the Canvas Editor to make changes to an existing image. Note that since it's already VERY large, you will need to either shrink the base image or raise the generation box (Maybe even both :D)

So you can use the Mask tool to cover the strands of hair, then using a very simple prompt apply it and see if the results are to you liking (Sorry for the bad quality... I did a really sloppy exemple just to show how to do it haha)

[steady shale]

Thank you! I tried but had no luck. I will go back and give it a try. Thanks again.

[open echo]

Another option to try would be to use the image in Image Guidance’s « Depth to Image », enter the same prompt, same settings but add « golden hair » in the dragon description.

It might just work by itself like this 😅

[vital crescent]

Trying to change up a picture using image to image using this prompt for something more gory ( In the haunting atmosphere of Lavender Town, Pikachu stands amidst the mist, its once-vibrant yellow fur now soaked in the deep red of spilled blood. Its eyes emit a sinister crimson glow, reflecting the darkness that has consumed it, while its mouth twists into a malevolent grin, revealing rows of razor-sharp teeth stained with the crimson evidence of its gruesome deeds. An aura of palpable bloodlust emanates from the creature, mingling with the metallic scent of fresh blood that hangs heavy in the air. The silence is punctuated by the squelching sound of each step Pikachu takes, leaving a trail of crimson footprints behind, further cementing its terrifying presence in the minds of those unfortunate enough to witness this chilling scene. ) I’m using the anime and lightning models for the low token cost but never comes out with the level of detail I imagine can somebody help is the trick to using a shorter prompt?

I’m not looking to use it on one image but us but a number of images

I’m currently limited to the daily token allowance and trying to get a few pictures to come out perfect is getting exhausting

And I might in the future considered tweaking the prompt to make it more universal

[dim zodiac]

Yeah, I mean... it's a short story 😄

it's very well written, but it's so much for the AI

GPT always loves to give things "auras" too

But in a more helpful response... try something like:

A blood-drenched Pikachu grinning malevolently, glowing eyes, terrifying, Lavender town background

and add more detail words depending on if you want it more cartoon or realistic

tested the prompt a few times and it doesn't really "get" Lavender town... just adds a purple hue to everything

[vital crescent]

I used up the tokens for the day, so I’ll have to wait till tomorrow to try it but thanks for the response

Ps I suppose if I want to make it universal all I have to do is replace Pikachu with something like creature or figure and possibly change locations

[oak nest]

It's still very difficult to control precisely.

[real lynx]

I'm going to see what I can do for your pikachu.

[thin tinsel]

Because the new colour looks quite flat, you now need to take this image back into Leonardo with Image Guidance, Image to Image set to about 0.75 - 0.8. Run it with your original prompt ... being sure to say GOLD MANE in all caps and in round brackets somewhere towards the front of your prompt. Let me know how you get on.

[wanton belfry]

ey yo...wtf....evil Pikachu

[hardy sentinel]

@mortal prairie it seems that you have questions about image creations? please post it here.

[fringe crater]

Real cool images. Pokemon in real life with people continually trying to catch them 😅

[pure sluice]

Is this normal?

[thin tinsel]

I think there's a conversation in #❓│platform-questions about this. see if it's the same thing.

[dim zodiac]

Every now and then we have a bit of a server slowdown, I wonder if that's happening. You're the second to be reporting this today. No, generations shouldn't take that long

You can submit a report through the intercom, or hang tight and see if it clears up shortly

[pure sluice]

Ok ty ^^

It just stabilized and I got tokens back automatically so all good

[real lynx]

@celest fern

@night vault I borrowed your image :3

Super rabbit power

[night vault]

glad you like it, make a nice spinoff 😍

[celest fern]

These look super good

[glad relic]

Hi folks, I'm a portrait artist (painter) and I'd like to use Leonardo to help me add more color and excitement to my work. I don't want the likenesses to be affected in the process. Is there a way to prevent this happening? I'm on the free membership at the moment, hope to be able to affort to upgrade later. Thanks for any advice

[night vault]

so you want to not deviate much from original image i suppose?

[glad relic]

I don't mind that, just don't want the faces to be changed

[gloomy lodge]

@fringe crater Image that I got (left), and I kind of wanted it to be in this style (right). I wanted to keep the same scene on the left image just to add another guy, because I am trying to create short film with motion

[fringe crater]

Left a message in #💬│general 😁

[night vault]

face is a only thing that is not great at consistency though. Tell me a bit more how your result so far, which part you dont like about it

[gloomy lodge]

thanks

I'll try to see what I can do

[glad relic]

Well as an example, I have a photo of my niece which is b&w and I wanted to colorize it and add some jewelry and a medieval look

[thin tinsel]

Did you try depth to image?

[night vault]

colorize can be done using edge/depth to img as first step

make it look different (overhaul style) though without face change is unlikely.

[glad relic]

https://cdn.leonardo.ai/users/8de9335b-c6d6-4537-958b-5b3f95c16bad/generations/a0be5777-ae39-42c1-8c16-d02def50b522/Default_lovely_young_woman_mulatto_warrior_bard_thick_wavy_hai_3.jpg

see, it's a nice interpretation, but doesn't look like her

so it's not possible to preserve the original face?

[thin tinsel]

Shall I repeat my question?

[glad relic]

oh no, sorry Willow

I wasn't aware of that option, I'm really new to Leonardo

[night vault]

I think maybe possible with canvas. paint mask everything except face

[thin tinsel]

I would say that what you're trying to do IS possible but it's pretty advanced and therefore you might find it easy to explore and learn more about Leonardo as it might take quite a few steps and re-generating to get something you're happy with.

[glad relic]

so I need to slow down and have patience 🙂

I will do that, because I'm excited by the possibilities

[thin tinsel]

Yes to some degree. 🙂
AI images are absolutely amazing and Leonardo tools and features will be able to give you a huge amount of control. But there are still some limitations we need to work around

[livid inlet]

https://discord.gg/p4Fns5FT?event=1231998558928896031

[thin tinsel]

If you're happy for me to, I don't mind trying a few things with your photo? @glad relic

[glad relic]

I'll try that Zorgi, nothing to lose and thanks to both you and Willow!

sure go ahead Willow

[hardy sentinel]

it is heartwarming combined with the previous image

[thin tinsel]

Ok! 🙂
the best result for keeping her likeness was with LINE to IMAGE on 1.83
What might be an idea is if you have a black fine-liner is to trace over the photo. Then use that (make sure you have closed lines).

I did massively increase the contrast on your original image to start with.
So bear that in mind if you want to try the photo with line to image.

[glad relic]

ok I will try it. Can you show me how yours came out?

[thin tinsel]

Sure!
Click on my Name.
Click on the link to my profile
Where it says Trending, change it to NEW
Click to open the image, and at the bottom on the right, you'll see the control net option I used and at what weight. ie. Line Art at 1.20

[glad relic]

sorry which image is it?

oh wait

Oh I see

[thin tinsel]

Download any that you want. I'll probably delete them after a couple of days

[glad relic]

Yes that does look more like her and I can always tweak it back when I paint it

thanks Willow, I'll play around with it

[thin tinsel]

Cool. That's just the first step in the process though. We turned a black and white photo into colour. 🙂

[glad relic]

yes it looks pretty good already! I will try to duplicate your results

[thin tinsel]

The next step is to learn how to use the Canvas Editor.
Check out the guide here when you're ready: https://discord.com/channels/989166677390426132/1126679661741609120

[glad relic]

Thanks, I will 🙂

[wanton belfry]

it does happen from time to time...

[steady shale]

Thank you. I will try that.

[steady shale]

What do you think?

[thin tinsel]

Gorgeous!

What do you think? 🙂

[steady shale]

This is a main character I am creating for one of my novels. FYI

[feral bough]

Adrean, it is asked that you speak English in the public rooms, otherwise fewer people will understand you and fewer will be able to assist you

[tall elbow]

ah well, it doesn't matter... I'm tired today to be bilingual... thanks anyway

[feral bough]

Anyways I'm going to bed but someone will be able to help with your issue im sure 🙂

[feral bough]

Haha, I think a lot of people use Google Translate for it

[tall elbow]

sometimes i use that...

[feral bough]

ChatGPT also does a good job at translation

[smoky wave]

I'm having an issue with a prompt (and it includes Midjourney and DreamStudio not just LeonardoAI). I'm trying to get a clouded leopard, which has a considerably different coat pattern from a regular leopard, plus they are smaller and built more stocky. But most of the time, the generators give me a regular leopard. I've been trying to set it off by running the words together: cloudedleopard, or to set it off in parenthesis ((clouded leopard)) but still, they keep giving me regular leopards. Of the 3, Midjourney is the best at getting me a clouded leopard, but only about 25% of the time.
Can anyone give me any advice on how to get the generators to understand that I want, specifically, a clouded leopard?

[real lynx]

I'm going to Google what a clouded leopard is to give me ideas 👋

[real lynx]

This is the specimen we are looking for

[slender knoll]

i made a reverse promting of your picture. The down below has this prompt :

''A rare clouded leopard is shown crouching in a misty forest. The scene is full of detail, depicting a lush, overgrown environment with tall trees and dense underbrush. The leopard has a beautifully patterned coat with intricate spots and piercing golden eyes. It is perched on a fallen tree trunk, surrounded by foliage and tiny specks of light resembling floating dust or snowflakes. The atmosphere is mystical, with a soft fog permeating the forest, casting a muted glow across the scene. The overall mood is serene yet slightly eerie, evoking a sense of wilderness and isolation.""

the first one this prompt :

""A clouded leopard is seen sitting in a natural setting. The scene shows a lush environment with a variety of green foliage and a tree trunk in the background. The leopard's distinctive coat has large black cloud-like spots and a long, striped tail. It is seated on the ground, surrounded by greenery and dirt, with its head turned to the side, looking off into the distance. The expression on its face is calm, and the atmosphere is serene, providing a peaceful depiction of this rare and majestic animal in its natural habitat.""

maybe you can play around with the differences

🤖

[slender knoll]

@here
Hi, Can you give me a tip on how to create 4 images in Leonardo that generate a character that looks exactly the same but just represents a different position.
Exactly: I want to embarrass a skateboarder who always looks the same, but always stands differently or has a different position. How does it work? He has to wear same clothes... and same skateboard in his hand.

Thank you for the help and have a productive day

[craggy gull]

I want Leonardo to give color to this picture. How can I do it?

[wise canyon]

You can upload the image into "Image Guidance" along with your prompt, and select the "Depth to Image" setting as a starting point. Try different strengths to stay close to the original image.
Be sure to prompt for a color image too

[craggy gull]

ohhhhh thanks!

[wise canyon]

No probs, also note the image guidance options are different when/if you have Alchemy off.

[feral bough]

Please don't use that tag as there are over 1.7 million people on this server. Luckily the ping doesnt work so you didn't ping 1.7 million people.
As to your question: what you're describing is currently very hard, if not impossible. One thing you can try to do is give your character a name, but the face and clothes will never be fully the same because AI is based on randomness.

[slender knoll]

@feral bough
Hey Max ok sorry, wont use the tag anymore 👍 . My Fault, Sorry.
Oh i find image generator which are able to do this. I talked to some Agencies and AI Models and they have done this.
For example this a agency from LA which are also generating with AI, and i asked them how they do it and they told me they have trained theier model so that the model reacts on small prompts.
https://www.instagram.com/p/C2lnEEatGEP/?img_index=1

this is a perfect example how i want it

and i talked to an agency in Tel Aviv with are using promtchen.ai and there it seems to be possible, too

and in my opinion leonardo makes the best pictures from all ai tools i used.
And i know my prompts there meanwhile very good so i really dont want to change my generator.... it is soooo powerful 👍

[real lynx]

I think this doesn't belong here

[wise canyon]

I agree Cami, 👌

Done.

[slender knoll]

@real lynx
oki...👍

@real lynx
Please let me know if theres a channel where i can discuss such topics if there is one.
Thx so much and wish you a beautiful day

[west pivot]

hello, i have a question, i just tried to used solarpunk, and i notice almost all my images have a solar panel? why is that? weight is 0.30, that's why i lowered it to 0.10 or negative

like this

[hardy sentinel]

it is in the nature of the LoRA (element) , so I would write solar panel in the negative

[west pivot]

thanks😊 should also include that in negative prompt when using solarpunk

[hardy sentinel]

you are welcome.

[uneven terrace]

Im helping a frnd with a logo for some project he is working on. He needs a logo that have specific colour n have isometric or geometric style. Is it possible to create such things with Leo?

[uneven terrace]

Can you help? @real lynx

[real lynx]

I think so but I am not at all an expert working on logos.

[real lynx]

It's a drawing but I see the spots, they look more like the original photo I found of the leopard.

[thin tinsel]

https://cdn.leonardo.ai/users/fa466aaa-cdd2-46ea-a529-92bc0a96898d/generations/f4e1d5b9-4019-4175-a6b2-c1867dc24f7e/Default_A_grey_cloudedleopard_1.jpg

Preset Leonardo Style
Prompt Magic off
AlbedoBase XL

Prompt: A grey "clouded-leopard"
@smoky wave

[uneven terrace]

You can still guide me

I am pretty confident I willl figure out something with your guidance

[uneven terrace]

Im trying to convince him to get mascot

But his is not listening

💀

[wide kayak]

How would I get Leonardo AI to create a coloring page? So far, it keeps giving me anything but black and white.

[thin tinsel]

have you tried Realtime generation

[wide kayak]

Is that a generative model?

[dim zodiac]

Heya Solomon! Use AlbedobaseXL, it works well for illustrations

Then start your prompt with Logo, Vector Art. Describe after that what you want in the logo

If it is still giving too much detail, you can add Simplistic, or Minimalist

Please note that if this logo is to include text, you might want to do that separately as the AI still doesn't interpret text well

You can even generate with Transparency active if you want it to appear like an icon!

[thin tinsel]

I change my advice.
Actually I've never done a colouring page.
But if you do AlbedoBase XL, and use the Colouring Book Element on 1.00
And make sure to put Colouring Page at the beginning of your prompt

[dim zodiac]

And lastly, the Simple Icons element works well to make very simple logos, but usually only in small square aspect ratios!

[thin tinsel]

@wide kayak , here is my attempt at a colouring page following my own advice:

https://cdn.leonardo.ai/users/fa466aaa-cdd2-46ea-a529-92bc0a96898d/generations/43856532-eaa8-4cd0-ba27-294d572fc96b/Default_A_COLORING_PAGE_RedPanda_reading_a_book_Line_art_kawai_1.jpg

[wide kayak]

That's amazing! Thank you for the advice!

[thin tinsel]

let me know how you get on.

[wide kayak]

It worked pretty well! Now I just got to fine tune the prompt to get what I want.

[real lynx]

I arrived late xD

[dim zodiac]

Huzzah!!

Well done @thin tinsel and never too late @real lynx

[livid inlet]

https://discord.gg/leonardo-ai?event=1232488852971716658

[uneven terrace]

Dude ur this line have helped me understand prompt writing

I bet there is more to learn

But to sounds very help

TYSM @dim zodiac

[dim zodiac]

Excellent!! Good luck!

[craggy gull]

It seems Leonardo has never watched DC series Peacemaker.
I tried multiple times but none of the results is slightly related to the actual Peacemaker
How sad. Is there a way to solve this?

I CAN'T BELIEVE LEONARDO KNOWS NOTHING ABOUT PEACEMAKER

IT WAS SO HOT LAST YEAR

[real lynx]

Don't worry, there's no need to get angry like that.

Everything has a solution

I'm going to google the series and get info to create a prompt

[uneven terrace]

YOU PROMPT GUIDELINE HELPED ME :D

[uneven terrace]

Is there any template for creating prompts?

[real lynx]

This one doesn't look like

[craggy gull]

Nice

I just don't get how such a smart ai don't know about peacemaker

like, it's everywhere

[real lynx]

I don't know him

Search for the actor's name

And he already brings out similar things

[feral bough]

No need to get angry, there's a very good reason. Answered you in #❓│platform-questions

[faint rain]

Guys

[faint rain]

Can anyone tell me how to make better then this?

I feel like something is missing

[real lynx]

Mmm

I like kittens

[faint rain]

Now way💀🙏🏽🙏🏽🙏🏽

[faint rain]

How did u do it 💀

[real lynx]

Tome tu imagen y la metí en imagen guidance

Y la mescle con un prompt que tenía de gatos

[faint rain]

Ooooohhhh

Ok ok

[real lynx]

L
I gave more strength to your photo

[faint rain]

That's useful to know

[real lynx]

And they mixed

[faint rain]

Yea and how did u make the quality way better?

[faint rain]

Yess

[real lynx]

Utilice alquimia

use alchemy

Is your account free?

[dim zodiac]

Peacemaker is from 2022 (at least when the show got popular). Consider that most of the training data from these models is pre-2022. SDXL is the most recent base model that Leonardo is working from and it was released in 2023, meaning the training data had to be much older

In addition, Leonardo tries to ensure that everyone's images are commercially viable, meaning they vet the models that become officially added to the site. Something that was trained on recent images from a show might infringe on copyrights, so that gets considered when models get uploaded

These are just personal thoughts, but that's what I would suspect.

[uneven terrace]

Cami u accidentally pinged me lmfao

[dim zodiac]

We are working on this, and there is going to be a public stream in the near future that will go over it too

The main pieces of a prompt could be considered: Style, Subject, Scene and Details

What I mean by that is Style - what kind of image is it, Anime illustration, Hyperrealistic photograph, 3D Render, etc. Subject - describe what you want in the scene, your character and activities, Scene - describe the background or setting that your subject is in, Details - here you can add lighting or camera angle or other specific details

You can play with the order of the prompt, but typically Style or Subject would be the first thing as it the AI puts the most impact on items earlier in the prompt

[terse void]

Hey Guys, i tried the Canvas Editor at Leonardo. As i trief there is an Error like "Invalid initDataUrlImage prefix" how can i fix it

[dim zodiac]

Clear your cache and log in again

If that doesn’t fix, submit a bug report on the Intercom on the site

[faint rain]

I don't have the prime

[uneven terrace]

Man can you send that in private chat? Ill be easy for me revisit it later

[haughty acorn]

Hi guys! How do I create art at this level with Leonardo (it doesn't have to be anime, I'm referring to art in general, including cartoons)? And also how would I create 3d models or game designs?

[real lynx]

Oi

Você pode procurar personagens de anime conhecidos para começar.

P

Ang avatar

[haughty acorn]

pensei q o chat fosse só in english

[haughty acorn]

It's my fault that I didn't specify, but I was talking about filters/styles like Kino XL that people use for realistic medieval themes

[real lynx]

It is spoken in English so that all the people who want to help you can try.

[hardy sentinel]

@last rover please repost your question #🖼│image-share message here

[open echo]

@last rover Well... I would wager that "...rich hues of blue and green...." might impact the colors....... Just a hunch though 😅

[livid inlet]

https://discord.gg/hwXpErp4?event=1233099295683776642

[smoky wave]

All my images are coming out at 96 dpi. How do I get them to be 300 dpi? I've tried changing the image size, but that doesn't change the resolution. I've also googled the heck out of it, and... yeah, I still don't get it 😦

[dim zodiac]

Heya Trinket - I've had this discussion with quite a few people... DPI is a function of a printer and thus would require you to know what physical size you wanted to print in order to convert that to a resolution

but ultimately it will always come down to Upscaling

an image at 720 x 1080 printed at 4 x 6 inches is 180 pixels per inch... but print it at 2 x 3 inches, or sticker size, and it's suddenly 254 pixels per inch

in Leonardo you can use Universal Upscaler to get your generations up into the MegaPixel range (5-7MP)

that would be 300 pixels per inch up to 8x10 inches

I hope that helps?

[karmic pulsar]

#🙋

https://cdn.leonardo.ai/users/e2e95c14-79e2-4a42-92a2-9cfd821b25c6/generations/c32c928b-136f-48ce-8462-d228a48117b6/Default_Render_a_scene_inside_a_car_where_both_Dwayne_Johnson_1.jpg

I am not getting the desired pictures

The prompt was:-Render a scene inside a car where both Dwayne Johnson and the cab driver are seated facing forward. Dwayne is in the back seat while the cab driver is in the front seat. The interior of the car should be realistically depicted, showcasing details like the dashboard, seats, and window.Ensure both characters are portrayed with realistic proportions and expressions. Dwayne can be seen looking out the window or engaged in conversation with the cab driver. The cab driver should be focused on driving, adding a dynamic element to the scene.Avoid any distortions or unrealistic features in the rendering, maintaining a high level of detail and quality. Emphasize the interaction and dynamics between the characters, creating a believable and engaging composition

Pls help

@dim zodiac

Pls help

@dim zodiac https://cdn.leonardo.ai/users/e2e95c14-79e2-4a42-92a2-9cfd821b25c6/generations/c32c928b-136f-48ce-8462-d228a48117b6/Default_Render_a_scene_inside_a_car_where_both_Dwayne_Johnson_1.jpg

I am not getting the desired pictures. The prompt was:-Render a scene inside a car where both Dwayne Johnson and the cab driver are seated facing forward. Dwayne is in the back seat while the cab driver is in the front seat. The interior of the car should be realistically depicted, showcasing details like the dashboard, seats, and window.Ensure both characters are portrayed with realistic proportions and expressions. Dwayne can be seen looking out the window or engaged in conversation with the cab driver. The cab driver should be focused on driving, adding a dynamic element to the scene.Avoid any distortions or unrealistic features in the rendering, maintaining a high level of detail and quality. Emphasize the interaction and dynamics between the characters, creating a believable and engaging composition

[dim zodiac]

Hey!

so... I am not on 24/7 😄

I have a full time job, with this on the side - I see you!

@karmic pulsar looks like you used GPT for your prompt? It's not bad, but it's used a ton of words that aren't going to give you anything

are you paid or free on the site, will just change what I recommend

[karmic pulsar]

Free

[dim zodiac]

and can you tell me what model and settings you were using to get this image?

[karmic pulsar]

AlbedoBase XL and Alchemy V2

[dim zodiac]

oh ok! You still have your free alchemy trial active, that's good to know

[karmic pulsar]

Yeah

[dim zodiac]

but you want this to be almost Photograph realism, yes?

[karmic pulsar]

Yes

[dim zodiac]

OK, let's swap to KinoXL and the Photography alchemy pipeline

[karmic pulsar]

Sure

I have to use the same prompt

[dim zodiac]

the prompt is MASSIVE - you lose the AI adherence after about 75 words

give me a moment to edit and I'll suggest a prompt to start with

[karmic pulsar]

Ok

Sure

[dim zodiac]

I'm getting closer to something I think you'd like

[karmic pulsar]

But Dwayne is sitting at the back seat of the cab

[dim zodiac]

yeah yeah

you are asking the AI to do something difficult - a composition with 2 people, one who is a celebrity

[karmic pulsar]

Umm u r still trying

[dim zodiac]

it will do a lot of blending, I'm trying to write you a prompt that will avoid this

brother have some patience

[karmic pulsar]

No problem

[dim zodiac]

😄 I can send every step of my process if you want

[karmic pulsar]

Sure! I would like to see the progress

[dim zodiac]

the above image was created with this prompt: Interior cinematic still of two people in a taxi cab, (young driver) with Dwayne Johnson in (back seat) as passenger.

[karmic pulsar]

At this point u r generating my pictures

But the cab looks ugly

[dim zodiac]

part of the issue is that I bet most of the images of Dwayne that the AI would have been trained on would have him driving

so it doesn't really want to put him in the back seat

[karmic pulsar]

If u can make the cab look good that would be great

[dim zodiac]

right, which is why I wasn't even going to share that photo

I know what a good generation looks like 😄

[karmic pulsar]

So how can i make him go back seat

[dim zodiac]

😄 nearly

but backwards

[karmic pulsar]

U can take help from this photo

But this is animated

@dim zodiac yup

Way to success

[dim zodiac]

Yeah having an image input could help, or you could start it in Realtime to get the composition

this one with the text will make things tricky again

[karmic pulsar]

Ok

Let me find a good image guidance for it

[dim zodiac]

I could get this: using the Family Guy image, on a very loose depth to image guidance

[karmic pulsar]

Hmm

The person at the back seat is not sitting in the car

[dim zodiac]

now I'm using that image to try and get the people right

[karmic pulsar]

Ooh

Let's e

Let's go

[dim zodiac]

closer, just need to get the driver and Dwayne to swap

[karmic pulsar]

Yes

Can u pls prompts also which u r using to generate images

Provide with the*

[dim zodiac]

yes at this point I'm using: Interior cinematic still of driver and passenger in a taxi cab, (diverse driver up front: 1.3) with (Dwayne Johnson in back seat: 1.3) as passenger.

the brackets and numbers are weighting that is used in Alchemy

it puts extra emphasis on those terms in the brackets

[karmic pulsar]

Can u provide a screenshot of the settings which u have implemented. It's really hard to understand in context.

[dim zodiac]

trying to get the AI to recognize where to put the people

[karmic pulsar]

Ok

[dim zodiac]

At this point, it might be easier to get the rest in Canvas

[karmic pulsar]

Thnks bro

[dim zodiac]

So the previous image is uploaded under the Image Guidance tab

[karmic pulsar]

U do a really great job

Ok

[dim zodiac]

I was trying now that it was a satisfactory composition to just use image to image with very low weight, but it still sees Dwayne as too recognizable

[karmic pulsar]

Yes

[dim zodiac]

I'm going to try one more with Depth to Image, which should conserve the foreground and background elements

with more weight on the "Backseat part"

but still no luck

[karmic pulsar]

So how will u use it to make a good picture in the canvas

[dim zodiac]

I like this composition the best so far, so I'll take this into Canvas

this seems the most like a cab

[karmic pulsar]

Sure

[dim zodiac]

So I'll download this image onto my computer and upload in canvas

[karmic pulsar]

Let's go

[karmic pulsar]

Now what r u going to do

How will you Photoshop it

[karmic pulsar]

What is ur progress till now

Ok

[dim zodiac]

using inpaint, I'll mask the back passenger and prompt for him to be Dwayne Johnson

max inpaint strength on the right

[karmic pulsar]

Ok

[dim zodiac]

yes, he looks out of place, don't worry

[karmic pulsar]

Great

U r really good at this

[dim zodiac]

Now I've deleted the driver and prompted for a new york taxi driver and Outpaint mode

[karmic pulsar]

Very well done

[dim zodiac]

it wasn't giving anything good, I'll try just inpainting it again

[karmic pulsar]

Can u show me the progress

[karmic pulsar]

Can u pls add a man as a cab driver

That will be great of u

[dim zodiac]

OK well I've shown you all the steps, you can take over 😄

[karmic pulsar]

Hmm

No pls

[karmic pulsar]

U r great at this

[dim zodiac]

that's the one I was working from

[karmic pulsar]

I know

Can u give me the last image which processed

It will be much easier for me

[dim zodiac]

One sec

[karmic pulsar]

And the prompt which u have used

[karmic pulsar]

Great

[dim zodiac]

I think you can then take this back into KinoXL

[karmic pulsar]

Excellent

[dim zodiac]

use it as Image input in the Image Guidance tab

[karmic pulsar]

Ok

[dim zodiac]

use Image to Image with like... 0.7 strength

[karmic pulsar]

Ok

[dim zodiac]

Interior cinematic still of driver and passenger in a taxi cab, (diverse driver up front: 1.3) with (Dwayne Johnson in back seat: 1.3) as passenger.

as prompt

with the Photography pipeline

it should make the style all consistent

[karmic pulsar]

Should I copy this whole and paste it or should I copy it till taxi cab

[dim zodiac]

no use the whole thing

[karmic pulsar]

Okk

Thnks buddy

[karmic pulsar]

Thanks a lot buddy

U r really good at this work of art

Btw can u pls how to make pictures consistent and change some facial expressions of the characters

Tell*

[dim zodiac]

can you explain what you mean by "make them consistent"?

[karmic pulsar]

Continuous pictures but with different emotions

[dim zodiac]

Changing facial expressions you can do in Canvas by masking the face and prompting for a different expression

[karmic pulsar]

How?

Do u have video for that explaining it

It's going till 0.10

[dim zodiac]

https://www.youtube.com/watch?v=TkK1BvhV5yY

[karmic pulsar]

Thnks

[dim zodiac]

closer to 0.1 you will get random image output, closer to 0.9 you will get nearly the same image again

I said 0.7 so that the AI would make some changes to the style, but keep the image mostly the same

[karmic pulsar]

Ok

But it's not going back after 0.10

[dim zodiac]

it won't slide back the other way?

that's new to me

[karmic pulsar]

Ooh

So then how will I become a pro in this just like u

[dim zodiac]

practice 😄

I've been using it for more than a year

[karmic pulsar]

Ooh k

[dim zodiac]

also if you are around on Monday I'll be doing a live stream about prompting and to do some live help

https://discord.gg/leonardo-ai?event=1232836633334644757

not sure what time that is for you

[karmic pulsar]

Umm I live in India

[dim zodiac]

😄 then it's middle of night

whoops

[karmic pulsar]

Ooh at what time

[dim zodiac]

there will be more, hopefully in better times for you

[karmic pulsar]

Might be awake

[dim zodiac]

3:30AM I think in India?

[karmic pulsar]

Ooh no

[dim zodiac]

yeah, sorry

I could only schedule after my workday

but there will be more!

and at different times

[karmic pulsar]

Btw u gave me so much time of urs

I appreciate that

[dim zodiac]

no problem, I had the time to give today

[karmic pulsar]

I am generating 1 more picture

And needed 1 more time help

Render Dwayne Johnson standing confidently outside a taxi cab, gesturing towards the distant White House. He exudes determination, one hand on the cab door. The taxi's details are clear, and the White House looms in the background, its iconic architecture recognizable. Ensure Dwayne's likeness is accurate, with a sense of purpose in his expression. Emphasize realism and detail, capturing the moment with precision and depth

Can u pls tell me which model to use in this

To make a perfect out of it

[dim zodiac]

all the same - KinoXL with photography alchemy

but cut it all out aside from:

Dwayne Johnson stands confidently outside a taxi cab, one hand on door, White House in background, captured in professional DSLR

[karmic pulsar]

If I have to use canvas again then it will most of my time

[dim zodiac]

nah, shouldn't need

[karmic pulsar]

Ok

Let me try this one

[dim zodiac]

OK I do have to go now though

good luck!

[karmic pulsar]

Might be

Is it best to generate 4 pictures.

Or 2 pictures are enough

[karmic pulsar]

@dim zodiac pls help me again

The taxi cab is standing near the White House. The taxi driver is trying to say something. Dwayne Johnson is addressing the driver to go away.

[livid inlet]

https://discord.gg/bUsJCXwm?event=1231998558928896031

[karmic pulsar]

I can't join the stage the discord is Crashing again and again

[open echo]

@dull cobalt

Hey there! 😁

Paste your prompt and the results you are getting here and people will be able to help you workshop a solution

[open echo]

Good news at least for this event, it has concluded. But it can still be an issue for you when there will be new events.

I’d suggest trying with another device, and if Discord keeps crashing to check their support website

https://support.discord.com/

[dull cobalt]

Hi! 😀

Here is the prompt: "Embrace the spirit of Eid al-Fitr with a heartwarming image embracing in celebration. Let the imagery convey the sense of unity, happiness, and gratitude that defines this special occasion, while incorporating subtle elements such as crescent moons or traditional attire to evoke the cultural significance of Eid al-Fitr."

[open echo]

Hmmm... so indeed the prompt is quite vague 😄

What do you want the image to show?

("unity', "happiness" and "gratitude" are emotions that are best portrayed with human faces. So it will skew generation results to include people)

Ah! also you ask for "traditional attire" ... so all these tell the system to give you people :p

You can try to simplify your prompt by removing a lot of filler and references to people

(Final note... I promise I'm done after 😄
Prompts suggested by ChatGPT are VEERY verbose. They are great for making discoveries and just wanting to be surprised... but when you have a specific idea in mind, it will be way more precise to use a list of clear and concrete words)

[eternal portal]

How do I fix the extra fingers? Prompt: caucasian female in her mid-thirties, chocolate-brown long wavy hair, hazel eyes, athletic build wearing a feminine business suit, realistic anime style, A close-up shot of Sarah Johnson, Mia Jane talking on the phone, with a wide smile on her face, in the streets of New York. Negative prompt: Deformed features, Crossed eyes, Two heads, Weird colors, Morbid themes, Fused fingers, Missing limbs.

I have tried fixing it in canvas, but that didn't work (I'm probably doing it wrong, I'm very new to the pro features).

[thin tinsel]

if you are here now I can show you.

[eternal portal]

I'm having dinner right now...are you on later (1-2 hours)?

And thank you!🙏

[hardy sentinel]

you can use sketch to image and select the color using color picker

[eternal portal]

I'm not sure I follow....is there a video step-by-step tutorial you could point me to?

[eternal portal]

I'm not sure I follow....is there a video step-by-step tutorial you could point me to?

[hardy sentinel]

No, pls give me some time. I am not in front of my computer

[eternal portal]

Of course, no problem🙏🏻

[hardy sentinel]

Have you experience with photoshop or gimp?

[eternal portal]

A little experience with photoshop, but still a beginner

[hardy sentinel]

ok the layout is almost exactly like photoshop

ok I am back

ok are you online? @eternal portal

please click at the green marked area, then choose sketch to image

[eternal portal]

Yes, I am...

[hardy sentinel]

ok please click the area marked by the green arrow

[arctic path]

hey whats up creators maybe someone can help me out im trying to make a cartoon like gracies corner for my sons. just 2 basketball brothers with there dog

[eternal portal]

Yes, done...

[hardy sentinel]

Hello, I don't understand?

[hardy sentinel]

ok please show me your screenshot

please look at this and do the step 1, 2 and 4

[hardy sentinel]

ok ok, you want to make anime right? then you need to change the model to Leonardo Anime XL

select model at the area marked byt the red arrow

[eternal portal]

Yes, I have changed the model now...

[hardy sentinel]

ok perfect, now I will repost my screeshot

first step change the number of image to one

at this moment your setting is at 4

this will save you some tokens, especially at the beginning where you have to adjust often

[hardy sentinel]

awesome

give me a bit time to make a suitable screenshot

for this case you can skip to number two, only changing the render density.

if you know how to change your working space size by typing, you could type 856 x 856 but 768 x768 is close enough

@eternal portal are ready for the next step?

now you can set your tool to "pencil" icon

then step number two to is going to the top and activate the color picker

after the color picker is highlighted, pick a color from the finger

then kindly show me your screenshot

[hardy sentinel]

perfect, now the consideration is either to draw the finger as you want to (not right hand) or draw the shadow / hair to reduce the available fingers, in the later case then you have to pick the color of the hair.

now it is your turn to make the decision

@eternal portal do you choose to draw a finger or draw over the uppermost finger with hair ?

[eternal portal]

I think I will try with the hair.

[hardy sentinel]

fantastic

please change the prompt accordingly

[eternal portal]

Thank you, I will try that

[hardy sentinel]

if you have generated something please download it

and kindly show me your screenshot thereafter.

[eternal portal]

Not yet...wikk show screenshot afterwards...

[hardy sentinel]

ok

[eternal portal]

doesn't look that great....will have to play around with it some more, but thank you for walking me through it🙏🏻

[thin tinsel]

I'm also having dinner so I can meet you back here in about an hour.

[eternal portal]

Thank you...Steven just walked me through, but I'll try playing around with it some more and let you know if I still need help🙏🏻

[hardy sentinel]

you are welcome, you can cancel and try again, it cost you 2 token per try. Canvas Editor is expensive

that is why I only try it with one image per generation

[thin tinsel]

Ok. That's great.
It's not just the fingers that are the problem by the way though.

[hardy sentinel]

Bon appetit

if you need help please post it here @white delta

[eternal portal]

I know...the phone too 😆

[thin tinsel]

I'm doing a live stream next Friday about how I use Canvas. Check it out if you'd like to learn more. 🙂

https://discord.com/events/989166677390426132/1233099295683776642

[eternal portal]

I will definetely show up to watch your livestream...also I changed model (from Dreamshaper to Leonardo Anime XL) and now I get what I want from my prompts.🥳

[thin tinsel]

Yes the XL models are much better at hands!

[eternal portal]

Yes, I found that out the hard way, lol....look at it now:

[thin tinsel]

Yay, so much better!
And it will save you a lot of time and tokens if you don't have to make major changes.

... or any changes ... in Canvas

[eternal portal]

Yay, that is awesome

[hardy sentinel]

Beautiful

[thin tinsel]

You might also like this. 🙂
You'll need to run it through img-2-img on about 0.85, with your original prompt on Dreamshaper to finish it off.

[eternal portal]

Thank you!🙏🏻

[eternal portal]

Thank you, you're good at it!🙏🏻

[thin tinsel]

You're welcome. 🙂
I cut someone else's hand off

[eternal portal]

Ouch!🤣

[thin tinsel]

😆 Yeah it was in the right position and holding a phone, so I grabbed it and stuck it on.

[eternal portal]

BTW, is this the right channel for prompt help?🤔

[thin tinsel]

so in seriousness my steps were;

1. Erase the original hand, draw the brown from the hair and prompt in the hair to fill the gap.
2. Use a picture from the internet where the hand was in the right position etc.
3. Erase around the new hand and put into position. Use sketch tool again to fill in any gaps and join up the wrist.
4. Without any prompt and inpaint VERY low, I ran the generation again.
5. I drew the highlight onto the left side of her hand, and generated one more time.
   Voila

[thin tinsel]

Yes indeed

[eternal portal]

Thank you, I will save this to a google doc 🙏🏻

[eternal portal]

Awesome, my character is sitting in her office:

And I need another image where she holds a photoframe in her hands of her father...is that possible and how would you prompt that?

[thin tinsel]

You know consistent characters is a whole big task?
How many different pictures of the same person would you need?
If it's just her looking at the photo, then I guess it could be the back of her head?

Have you already tried a prompt to have her looking at the photoframe?

[eternal portal]

I was thinking an image where you can only see a close-up of the photo frame held by female hands....I wouldn't even need to see the back of her head...this is her father:

[eternal portal]

No, not with this model...it went horribly one with the other one....

[thin tinsel]

If it were me, I'd take this into realtime generation.

Do you want me to show you?

[eternal portal]

Yes, please, I don't even know what that is, lol....

[thin tinsel]

Let's go the Voice Channel.

[eternal portal]

I don't have a microphone....is it okay if I just listen and write answers?

[thin tinsel]

No problem

Do you know how to come into the voice channel?

[eternal portal]

No..😅

[thin tinsel]

On the left-hand menu under Main, at the bottom you should see Voice Channel, and me in there

[eternal portal]

Ok, I see you...should I just click on your name?

[thin tinsel]

Oh sorry! Yes, I left thinking you couldn't access it

Ciick on Voice channel, I hope it will let you in.

[eternal portal]

Thank you so much for showing me that, Willow!

[thin tinsel]

All makes sense?

[eternal portal]

Yes, I think so...I will try playing with it tomorrow...right now I'm going to make some of the images that are more straight forward

[thin tinsel]

Ok. Enjoy.
There'll always be someone around in here to help.
(I don't know why it's so quiet right now).

[eternal portal]

Thank you so much and have a great evening

[thin tinsel]

and you !

[open echo]

Hey it's the weekend! I'm out at the park :p

[open echo]

@clever imp
Oh I'm not saying it can't be done, and I see you are getting good results from it also .. I just find it a bit more tedious 😄

[slender knoll]

morning. i wanna show a woman playing table tennis for our side. but the results are annoying. Maybe someone have an idear?
Love out there ❤️

Prompt :

Beautiful blonde with light curls plays table tennis. She plays on a stone slab in a park. She wears a mix of super sexy clothes and street clothes. She is relaxed and focused. There are people in the park in the background. She's beating right now. Ultra Realistic. Ultra natural. Ultra sexy. She wears sneakers. Whole body in the picture. Kodak 400 Gold. She is holding a standard table tennis bat in her hand. She jumps and punches. The small white table tennis ball flies over the table.

Results ;

[sullen loom]

I'm looking for someone to help me with something peculiar, there's a character with little artwork of them (maybe 10 images total?), i've been looking of a way to use leonardo AI to increase the art and stuff.

So if anyone got an idea of how to do that i'd be infinitely grateful

[karmic pulsar]

@open echo pls help me with this
https://cdn.leonardo.ai/users/e2e95c14-79e2-4a42-92a2-9cfd821b25c6/generations/2ad98064-5820-4bbc-9512-c76e6da2f857/Default_Render_a_ladys_face_with_a_shocked_expression_near_a_d_3.jpg

I want to have her pointing towards the camera and having the same expression.

[open echo]

Ah that something I did way back.. but it was a guy 😅

The prompt was, aptly enough:

Man pointing at the camera, surprised

Just by the start of the prompt, I’m guessing you have a lot of fluff that can be removed like « command » words 😜

Keep in mind any image generation systems based on Stable Diffusion are not ChatGPT. They are not large language models, so any text that « speaks to the system » will be meaningless when pairing keywords to Colors, shapes and objects.

[open echo]

That sounds like a job for model training… unfortunately that’s e much the only thing I never really practiced 😅

[open echo]

I have trouble visualizing how she should be viewed from. I think I understand she should our teammate and is looking intently towards the opposite side? so like a profile shot?

[open echo]

Ah yeah! I forgot! 🤣

Expect pointing fingers to not always be what you were aiming for (sorry… I had to add pun there 😜)
But normally within 8 version of the image you will get good results to choose from (here are 4)

[karmic pulsar]

Thnks

[slender knoll]

@open echo
Thx so much. yes i wanna make a side Profile for the Table Tennis. But the Size Relations of Ball and Tracker is alwas wrong so like the color of the ball

[tawdry fulcrum]

"Stupid" question: i have many signs for unreads messages (white number on a red circle background) on several places in different channels that obviously have not marked as read, BUT i can not even find any of these "unread" messages...?!? What is this please?

[slender knoll]

please please please help me. If i use PhotoReal is always looking in camera.
would be soooooooo nice if i can have some tips where i can define which she is conentrated for.
Thank you soooooooo much 💖

i use mostly photoreal and cinematic

[livid inlet]

I hope this helps

[slender knoll]

always looking in camera but i want her to talk to some friends

promt : talking to friends everwhere

even the friends are not on the picture

@livid inlet ahhh thx soooooo much.
Maybe you have some Tips for me

it is stupid if i wnat her drive her skateboard but she looks in the camera

[livid inlet]

Yes getting multiple people can still be an issue, I went with this...

[slender knoll]

@livid inlet 💖
thx you my Hero

on what i have to take care?

[livid inlet]

"middle distance" helps to get people looking away from the camera in general, but this isn't something I specialise in so hopefully some one else will see this and be able to help more.

[slender knoll]

thx so much

do you use photoreal and cinematic?

[livid inlet]

Those were with PhotoReal and None preset.

[slender knoll]

Thaaaaaaaaaaanks 💋

and what when i want a picture from backside?

or from top?

[livid inlet]

https://discord.gg/H9kwa52f?event=1232836633334644757

[slender knoll]

yeees. thx sooo much. Leonardo is the best from all tools i tried

[thin tinsel]

@normal shore , the Glow wave Element would get you some of the way there.

[normal shore]

I'm trying to come up with a way to describe cyberpunk/borderlands style images. Kind of retro neon comic-thick borders and can't seem to get anywhere close to it. Does anyone have suggestions on models/prompts for something like this?
Adobe Stock ID: 637172475 would be a decent example of what I'm looking to do.

[normal shore]

Oh, you're already on it. 😄
Sorry, I'm still new at this whole thing. Can you point me in the direction of where to start with that?

[night vault]

I think @open echo is the man for you

[thin tinsel]

excuse me ... 😭

[normal shore]

All help is welcome!

[thin tinsel]

do you want to say what you already tried and what about it you're not happy with?

I did this one

[night vault]

I mean you can be great help too 🥺

[real lynx]

Are you looking for something like that?

[night vault]

check these #🖼│image-share message

https://discord.com/channels/989166677390426132/1091383546314240180

[real lynx]

I would create the images first and then put them in comic style borders

I can generate several boxes but not modify each of them

[thin tinsel]

Will we ever find out what he was looking to create?

[hardy sentinel]

Any reference image?

Willow has a point, it would be helpful to know what you want first

[thin tinsel]

We both tried to help, but had different ideas about what style.
He said cyberpunk/borderlands, which are both videogames.

[normal shore]

This is the closest I can find to describe what I'm trying generate. I'm going to be building a website for a company in the video game industry and this was the style described to me. I've been trying to generate something like on the Leonardo AI site and haven't been able to come up with anything close.

Here's something similar

Both clearly AI, but I'm unsure of prompts to describe the style or what models would produce the best results.

[thin tinsel]

So what about the example image I sent above?

[normal shore]

This looks great! I have no idea how you matched it so closely.

[open echo]

This looks more like a "Makoto Shinkai" style anime scene then Borderlands though..

So with a prompt like:

Makoto Shinkai, synthwave, arcade center, neon glow, vintage arcades machines, symmetrical
Using the AlbedoBaseXL model, we can get something very close to it

I'll just add the stools and cracked floor in the prompt to get them too 😛

[vital crescent]

What would be the benefit and lack of a better word opposite be For setting an element to 2.00 and negative values can someone describe that to me
I posted this question under general. Someone suggested I come here.

[dim zodiac]

great questions

setting an element to 2.00 in most cases is going to be a bad time

they are meant to apply a style, and maxing them out will for the most part give very chaotic results, aside from a very few niche use cases

[wanton belfry]

part about negative value is bit unclear to me even, it get messy the more negative you go

but I dont particularly see what its doing specifically

[dim zodiac]

negative values on an element will actually pull the style OUT of the image. So if you're trying to make an image and it keeps coming up way to "cutsie" for example... you might want to put the Cute Emote element in at a low negative value

agreed, most elements like to be handled in tiny increments... try it at like -0.1 or -0.2 if you were putting them in

and some just break

[wanton belfry]

glowwave at -10

glowwave at -30

glowwave at -20

hmmmm its like being very subtle unlike in positive numbers

[normal shore]

This is a perfect place for me to start. Thanks so much for your help! I'll post what I come up with, but the quick tests so far look much better than what I was coming with.

[zealous dock]

how can i produce picture from text?

[real lynx]

On Leonardo's website

https://app.leonardo.ai/