🙋│image-help

[dim zodiac]

I would call this "hyperrealistic illustration" with "cute" somewhere in the prompt

it's not really anime but if you use the term "anime" you do often get people with larger eyes who look more Asian

[gritty sky]

so which model could best generate such a style?

[dim zodiac]

Dreamshaper is good, AnimeXL is good, and AlbedoBase is good for this sort of soft illustration

RPGv5 can do it too with the right prompt

[worldly lotus]

can anyone help me regarding how can we generate realistic portrait and full body shots of your self

[hardy sentinel]

at this moment only using a dataset (own model)

as a free user you can only try it once

to make your own dataset you need 40 photographs of yourself

[vestal lake]

anyone have any advice why this keeps happening with the swords? and how to fix it , thanks

[hardy sentinel]

Dear Mr. Coleman, you can use canvas editor to remove the swords

[vestal lake]

i tried that but it not working

[hardy sentinel]

i have difficulties of prompting swordsman with a single sword (initial generation)

[vestal lake]

for me it like 1/10that come out ok

[hardy sentinel]

unfortunately

@vestal lake realtime canvas

no reference is used

using reference is not always better

i have used minimal inpainting to correct the image (within realtime canvas)

[vestal lake]

fingers and swords ,something Ai needs to work on

[hardy sentinel]

yes

[vestal lake]

ill give realtime a try

[hardy sentinel]

guitar neck as well

[hardy sentinel]

it cost 0 token

[vestal lake]

not had that one ..yet guitar neck

[hardy sentinel]

ok

instant refine = 0 token, alchemy refine = 8 tokens @vestal lake

[vestal lake]

canvas editor did work it just took a few times to remove it , thanks

[hardy sentinel]

you are welcome

[worldly lotus]

can we copy pose of another image into your image

[hardy sentinel]

canvas editor could be expensive sometimes

[worldly lotus]

replicate it same exactly ?

[hardy sentinel]

with pose to image?

but then it is not your face

[hardy sentinel]

if you want exact then please use edge to image

[worldly lotus]

yes everything but except face

[worldly lotus]

is it free for trial one's ?

[hardy sentinel]

yes, face would be changed

[hardy sentinel]

you can use it, it will cost token

[worldly lotus]

ok how many times free users can use ?

[hardy sentinel]

as many times as your token allow you to 😄

about the dataset, the making of the dataset is limited, but using it only depends on your token @worldly lotus

[worldly lotus]

ok thankss

[hardy sentinel]

have fun

[idle remnant]

I want to use some specific materials of some company in the render I created. Is there a way to do this?

[real lynx]

Yes

[lofty vortex]

wondering if it's possible to create product images? the product is this site: https://rayndeodorant.com/ , Since I don't have an idea of the prompt, I think if Leo can help me coome up with some ideas for the product placement

[marsh quarry]

Guys, I'm trying to upload an image from my local drive to generate Motion but don't see this feature available. I'm using a paid plan. The feature is supposed to be available based on the info from the user guide. Any comments on this?

[hardy sentinel]

Hi Anna, you can access motion from your personal feed

click on your chosen image then click image2motion at the right side

[dim zodiac]

you can certainly create product placement images

you will want to figure out Canvas, so that you can place an image of your deodorant with the proper text and branding etc into the image

[hardy sentinel]

after clicking your image, a more detailed window will pop up, image2motion is marked with number 1 (red arrow)

[worldly blade]

Can anyone help me with a prompt. I'm trying to create a female centaur and it just keeps spitting out images of women beside horses or just a woman.

[dim zodiac]

yeahhhh centaurs are hard for the AI 😄

[hardy sentinel]

how about replacing the head of the horse with a woman's head using canvas editor ?

Good morning @dim zodiac

[worldly blade]

Not too sure on how that works tbh

[hardy sentinel]

ok do you have any favourite image of a horse?

[dim zodiac]

AI says: female centaur... SURELY you meant "woman riding unicorn"!!

[hardy sentinel]

a good image though

[dim zodiac]

yes the image itself turned out great, just no centaur 😄

[hardy sentinel]

do you know how to select canvas editor?

[worldly blade]

no

[hardy sentinel]

this is how

this is screenshot of "home" at leonardo ai

please select number 1 (marked with red circle)

[worldly blade]

It's not as simple as just making a basic centaur. It's Set of images based on the Zodiac. I am doing a sort of Tarot Deck inspired Art Nouveau.

[worldly blade]

Ok I have it Canvas editor open

[hardy sentinel]

ok, wait a minute, I will post another screenshot

[hardy sentinel]

click on number 5 (blue arrow) to upload your image

please if possible post your screenshot @worldly blade

are you still online?

[worldly blade]

yes I'm trying to figure out

I don't have a base image yet

[hardy sentinel]

are you working with a pc ?

[worldly blade]

yes

[hardy sentinel]

ok, let's wait for your tarot deck of horse then

[worldly blade]

It keeps giving me images with a woman in the foreground and a horse behind her even though my prompt doesn't say anything about a woman at all. or it gives me a horse head on a human torso.

[hardy sentinel]

no woman, only horse

you need a prompt that begin with "horse" 😄

[worldly blade]

Beautiful Elegant HORSE. (((elaborate ART DECO and Art Nouveau background)))

[hardy sentinel]

maybe even put "human" at negative

[worldly blade]

this is my prompt and i keep getting

[hardy sentinel]

do you have anything that indicative, like clothing or something?

[hardy sentinel]

ok beautiful images

oh ok, maybe you don't need "beautiful elegant"

[worldly blade]

yes this is the full prompt - (((FOCUS color: BLUE TOPAZ))) hyper-photorealistic typographic (((Horse))) DYNAMIC fantasy art. Beautiful Elegant HORSE. (((elaborate ART DECO and Art Nouveau background)))

[hardy sentinel]

I see that you probably also use sparklecore element?

[worldly blade]

yes for the gemstone associated with Sagittarius is Blue Topaz

[hardy sentinel]

oh ok, let me try myself with my own prompt then.

[marsh quarry]

Hi Steven, thanks, this part is clear. I'm trying to upload an image from my local drive, but this option seems not available (I'm on a paid plan). Any advice?

[hardy sentinel]

without sparklecore @worldly blade

[hardy sentinel]

it is clear, I can only advise you to do it from personal feed. Thank you.

[worldly blade]

ok looks good for a start

[hardy sentinel]

it needs some polishing, sparklecore element coming up

[marsh quarry]

I see, thanks. Will ask the support team

[hardy sentinel]

yes, thank you

[hardy sentinel]

a bit sparklecore and too many legs 😄

[worldly blade]

a leg can easily be removed

[hardy sentinel]

yes, now the challenge is to replace the head and neck of the horse to be head and upper body of a woman

[worldly blade]

I'll give it a go

[hardy sentinel]

ok good luck. Please tag me if you have any screenshot or result

I would recommend using sketch to image mode

To change the mode,pls click at the marked green arrow

alternatively you can use inpainting

[haughty fog]

So I just been trying to figure this out and I'm having a hard time on trying to figure out how to do the prompts to get what I'm looking for so I told it this was a alien venus flytrap and this is the texture that I got So if there's any tutorials or anything I can check out to learn how to get this to work for the way I want it to

[stable carbon]

did you use the texture generator for that?
I dont know anyone who's used that enough to be able to help I'm afraid. note that its also still in alpha..

[mortal marlin]

I want to generate some images of the specific style of an anime/manga (Nana), but I really don't know what to put or what to use(albedo, leonardo xl, etc) , and the image search engine doesn't work to search for examples. Any ideas or basic info to search for a specific animation style like nana?

[wooden pawn]

Hey everyone!

I have a question about using Leonardo for a storyline I'm creating. The story involves a family of five characters: a mom, a dad, and their three children. I've already written prompts describing their facial and physical features, but I'm unsure how to consistently create images of this family across different scenarios. For instance, I want to depict them during their happy times as well as when tragedy strikes and the father is bedridden in the hospital. Any tips or advice would be greatly appreciated! 😊

[vapid swift]

Im trying to outpaint but lately im having bad results,

this is my prompt
"dark fantasy art, landscape, background, continue"

i dont know if u have any tips, i used it 1 year ago and i have very good results combining multiple images
(None of these illustrations are mine)

[lusty chasm]

Hi I like to know which models are perfect for Fashion / Jewerly / Architecture / Interior Designs???

[open echo]

Hey there!

So the models don’t necessarily have trainings on all styles and models. The impact is that some titles may lack precision (or even any relevant results)

The best way would be to use the new “Style Reference” so you can mix your prompt with an exemple of the style you are aiming for.

Leonardo Anime XL will be a good bet for any anime style.

I used this prompt:

Style of Nana, a young man holding a red plastic cup, bored, at an underground party

And I did two exemples using Leonardo Anime XL and Alchemy:None

First one is the prompt only, second one is the prompt with the style reference.

[open echo]

Hey there!

Ah! Character consistency… one of the holy grails of image generation 😁

Stable Diffusion (which Leonardo is based on), works by leveraging randomness. It’s a double edged sword as that is what makes it possible to create art that never existed… but at the same time make it volatile from one generation to the next.

There are some techniques to make prompts more resilient to randomness: make sure the most important parts are at the front, avoid metaphors and use simple nouns and adjectives, keep relevant terms together (ie: clothing with clothing and background with background) and try to make the prompt as short as possible.

There are others techniques of course, but it’s a least a quick primer to maximize your chances of consistency.

(And when all else fails, bring the pic to the Leonardo Canvas and change the details there 😜)

One technique I used when I did children books: making sure the description stayed the same from scene to scene and generating about 20-30 pictures for each page.

I chose the one with the best resemblance, and if a piece or clothing or small accessory would ruin an otherwise perfect picture, I used a photo editing program to correct it.

[haughty fog]

Yes that's exactly what I used Once that's up and running to where I can use it I plan on getting a subscription plan

[mortal marlin]

Thank you so much♡

[white delta]

i need help creating something like this

but i want the roost to have no shirt on big biceps and six pack abs

[river sky]

Hi eveyone, I am beginner using Leonardo and I would like to ask for help in my use case. I would like to learn how to generate, for example, my portrait image in another styles keeping my face expression. I have used Style reference setting the reference style I want to converts my image and the style is quite good however my face is not the same, It is not me. I want to avoid train a model with my images. There is any way to do that?

[white delta]

anyone?

[stable carbon]

hiya. there is a new function coming hopefully very soon, that will be able to do exactly that. its available on some accounts already. it allows you to upload or choose an image and generate that face in different situations and styles. its called Character Reference. a bit of patience 🙏

you mentioned style reference, thats not really useful for consistent faces. by chance, do you have paid membership?

[stable carbon]

so basically you want a topless bodybuilder parachuting..? 🤔

[river sky]

Characters Reference 🤔 that could sound interesting .
I've a paid membership, sure.
So with this new release, Could I get my portrait image in another styles? Do you know when this will be released?

[stable carbon]

all i know is "soon". its already available to some accounts. could be this week, could be next..
and yes. it does exactly that. here's an image using that feature to get Kit Harington (Jon Snow from game of thrones)

as a paid user, you get access to some more channels in our discord. should I let the others know, so they can hook you up with the paid user perks?

[river sky]

Yes, you should.

Regarding the image, I understand that generating images in various styles from a image of a well-known actor is easier than doing so with my face, since I'm not famous. Is that correct?

[stable carbon]

i did the same thing with my own face too. and got very good results.
with that image there however, i just prompted for a generic "man". and got exactly the face i provided.

ofc, if you prompt the name of a famous person, you get that famous person fairly well. but this new feature will let you redo any face you want, without needing to describe it even.

[stable carbon]

@vocal portal found a Patron for you, if you got a min to set it up

[river sky]

Wow Thom, you are giving me a lot of hope with this new feature. I might be a little envious because you've been able to try it out and I haven't yet... I'll have to wait

[vocal portal]

Hello mr.bandicut. Please DM me and I will help you with the patron role!

[real lynx]

Imitates things and letters too and modifies them

Something very magical is coming

I was delighted

[river sky]

that's promising

[vocal portal]

hi there! this is not allowed.

Please do not do this again @calm ivy

[real lynx]

@pure skiff

You can ask for help

[hot pivot]

hey I have a question, I have tried uploading images that i have downloaded from the image generator in order to share it. However when I upload them, They are not seen. Instead they are a different type of file. Is there a processing time for the file in order to be seen? or am i doing something wrong?

[real lynx]

You download them and you can upload them

Maybe you have the rank delay when uploading images you earn points so you can't upload as many as you want

[hot pivot]

thats what I was doing but for some reason It was just appearing as a document instead of an image when i was uploading the jpeg

[real lynx]

I remember it happened to me once, it fixed itself I think What is a discord problem

I would try uninstalling it

[hot pivot]

ok thank you

[livid inlet]

https://discord.gg/leonardo-ai?event=1242527865728274604

[sullen sedge]

Hello com, how do I get Leonardo to create an image about an specific object made out of an specific material? Lets say, a balloon made out of transparent glass? or a skateboard made out of butter?

Like this, I couldn't make it

[open echo]

Hello there 😄

I understand you are attempting to generate items made of materials not associated with them.

I would suggest trying first with a minimal prompt, just to get the concept going

Ex: shoes made of lead

And then to get even crazier, if you have a Premium account you can use the new Style Reference where you add a base image (to get the style) and prompt normally

That was the base image used for Style Reference:

[sullen sedge]

hmmmm

interesting

I'll give it a try without the base image, as I don't have premium

[open echo]

Also keep in mind this isn’t ChatGPT, so don’t use too much flair and turn of phrases when aiming for something specific 😄

[sullen sedge]

hahahah so I don't use phrases to aim for something specific?

[open echo]

exactly 😄 Using phrases is great for experimentations, surprises and discoveries... but a little too volatile when you need something precise 😄

[wanton belfry]

if that specific is understood by the model and AI its okay, if not then you aren't going to get it

its a unfortunate limitation

also more than one subject is tricky

[livid inlet]

@latent topaz I can get sad looking but tear, no. Let me take this to the next step and see what canvas can do.

[latent topaz]

That would be perfect, thanks

BTW, as suggested by @stable carbon I'll also post what I have done so far: So far. I have tried creating images with many prompts, for example I tried generating images for the prompts "A cute, crying dog" and "A cute, crying kitten" in all Leonardo models. I also tried variations of this prompt, sometimes with more detail, but basically all images resulted in normal cats or dogs being created. I provided two example reference images with the prompts "A cute, crying kitten" and "A cute, crying dog".

[stable carbon]

do you want them to be photorealistic, or do you want them as illustrations (drawn/comic, etc..?)

[latent topaz]

Photorealistic would be preferable, but I otherwise would also be open to generate these images in drawn / comic style.

[stable carbon]

that might be the harder or the two options, since the AI is trained on whats available on the internet. and i dont think i've seen many photorealistic crying animals tbh.. 🤔
but good for the helping people here to know what exactly you're looking for

[livid inlet]

I added tears with the Canvas Editor Tool, wasn't easy lol. Let me see if I can do the same with one of yr images quick...

[latent topaz]

Ah, that makes sense, well as stated I also would be very open to creating illustrations, if that is easier

[latent topaz]

Thanks for the image, looks good 😄

[livid inlet]

Using the canvas tool

[latent topaz]

That looks even better, because in this image the tears are all in the right place! Thanks a lot. Btw is the canvas editor tool only available for premium accounts?

[livid inlet]

I had to sketch in the tears... The Canvas tool is for everyone I think

[latent topaz]

Great, I'll try sketching my own tears in a couple of the images, thanks for taking the time to help 😁

[livid inlet]

Always a pleasure 🫂

[real lynx]

Hello, use an account without premiums tool

After several failed attempts

I think I found what you were looking for

Use all of these items

Of elements

And this is the prompt

Skate Translucent Ice Sculpture Hyper Real Masterpiece Intricate Meticulous Ice Statue of a Translucent Skate

I hope it helps you 👋

[warped igloo]

Me encantan que 😍

[thin tinsel]

@latent topaz done in canvas editor

[real lynx]

How beautiful it reminded me of a famous painting of crying children.

[latent topaz]

Thanks, this image looks great, too. Do you happen to have the prompt?

[thin tinsel]

A watercolour painting of a very SAD LITTLE fluffy grey kitten, (((tears SPURTING from her eyes))), VERY sad sweet grey kitten.

You know that I edited the image in Canvas Editor right? So there's not just one prompt.

https://cdn.leonardo.ai/users/fa466aaa-cdd2-46ea-a529-92bc0a96898d/generations/c5ec2ae2-56b9-41c5-8a74-7fd5b6d3ce29/Default_A_watercolour_painting_of_a_very_SAD_LITTLE_fluffy_gre_2.jpg

[latent topaz]

Perfect , thanks 😃 , and yes I know that the image was edited with Canvas Editor, I was especially wondering about the base image

[velvet barn]

I can get sad looking but tear, no. Let me take this to the next step and see what canvas can do.

[livid inlet]

https://tenor.com/view/talk-about-deja-vu-ive-been-here-before-experienced-this-before-deja-vu-looks-familiar-gif-16490454

[sand kernel]

Hi, I try to solve a problem with hands, faces and multiple arms etc. negative prompt doesnt work here. what I must doing to solve the problems easy and fast?

[wise canyon]

Hi deaneau, some models are better than others with hands and faces, and it's not uncommon for XL models to ignore the neg. prompt completely too. That said, the quickest fix for your example above would be to apply the Creative Upscale-set to med or high, and possibly needing Smooth toggle turned on as well. This can improve the hands and faces considerably, however the additional arm would need to be removed using the Canvas Editor.

[sand kernel]

ok Ill try this

[sand kernel]

It seems like Leonardo has problems with faces that don't look into the "camera". The lady on the left and right look strange. I would even say they are cross-eyed. It's possible that this will become a problem the smaller the faces become. the problem with the arm I try later.

[robust gulch]

It's less a problem with which way they are looking and more a distance issue. Prompting multiple people often creates issues because the face just isn't close enough to the camera for it to provide all the details. You can use canvas to fix it up or adding "clear facial features" to the prompt has helped me in the past.

[wise canyon]

Yes deaneau, this is true, great observation. The smaller the faces are-the less pixels the AI has to work with, so some facial distortion occurs. Fortunately a tool was added to the Canvas Editor to help repair these smaller detail areas, and it's the "Density Slider". This slider increases the pixel density within the Canvas edit box, the box will appear smaller to indicate this. Masking the face area, placing the higher density edit box over it, and setting the inpaint strength low (so it looks like the original face), you can then prompt just for a "beautiful face" in Canvas prompt field, and the faces will be better.

[sand kernel]

work the "Density Slider with hands,too? I have try to remove the arm here the result.

[wise canyon]

Firstly turn off the outpaint toggle,
Put a mask over the extra arm, inpaint strength at full 1.0, and prompt "match background", this may require a few gens to get a good result to choose from. A further advanced method is to change the Canvas mode to "sketch to image", then use the pen to color over the arm using the same colors as the background (the eye dropper can match the pen color to the background), and then prompt as before "match background".

[thin tinsel]

I don't think you can give commands such as match background, you can only provide a descriptive prompt, such as "trees in the background" or forest background.
Personally I would use the erase tool to erase the arm, the colour picker to draw in the background, leave outpaint on and and have no prompt at all (just a space).

[regal edge]

What problem in my video? Help

[thin tinsel]

No problem. That's just how it works.

[hardy sentinel]

mksdnya anda apa?

animasi skrg terbatas 4 detik

dan skrg msh blm bisa dikontrol selain angka yg anda tentukan dari 1 smp 10

saya anjurkan 3

[thin tinsel]

what language are you speaking?

[hardy sentinel]

Indonesian, the person has posted in indonesian channel in indonesian language

[thin tinsel]

is it a romanisation because it uses it's own alphabet?

[hardy sentinel]

since 1901 it is written in roman alphabet

[thin tinsel]

oh wow!
why? 😮
And how did everyone adapt?

what was it like before?

[hardy sentinel]

please ask the Dutch 😄

[thin tinsel]

lol, I don't know any.

[hardy sentinel]

Stef, Ruby, Angel, Victor, Maximus they are all Dutch 😄

[thin tinsel]

🤯

[hardy sentinel]

to be exact the romanisation was designed by Prof. Charles van Ophujisen with help from ngku Nawawi gelar Sutan Makmur and Mohd. Taib Sultan Ibrahim

[thin tinsel]

ok, but what was it before?

[hardy sentinel]

I guess lierenaar, Dark Tulip could be Dutch too

[hardy sentinel]

depends what time and which language (Indonesian is relatively a new language based on Malay )

way back then was mostly Pallawa or Devanagari script

then through medieval times (1440 AD) it started changed into Arabic script

Javanese language actually have its own script

[thin tinsel]

OK I see. Thank you.

[hardy sentinel]

Indonesia right now has 712 languages

but the lingua franca is called "bahasa Indonesia" or Indonesian language

[thin tinsel]

Ok cool. really interesting

[hardy sentinel]

to be exact it is based on Melayu Riau-Lingga. The Malaysian has their own Malay language which is based on the peninsular Malay.

[feral bough]

I feel left out

😢

[hardy sentinel]

I am so sorry. Stand corrected.

[feral bough]

Yayyy

[fringe crater]

Oh wow this new layout on website is gonna take some getting used too

Looking good with improvements though

[drowsy sage]

This is awesome!

[wise canyon]

Hi WillowB, our methods here are similar, it's great that there are many options to use in the Canvas Editor, I love using Canvas for image repairs.
Usually I don't need to erase first unless Leo is being stubborn and and still generating the underlying image content, then I do erase before painting.
"match background" has worked well for me for over a year, but if you put a capital M it generates sports backgrounds, haha.
"match skin tone" and "blend skin tone" are also very useful when doing repairs in Canvas. lol

[sand kernel]

sorry for my late answer. I have removed the arm via photoshop in one step. leonardo need ,how you said, more steps. the problems with faces I have not really solved. by the way. the women in front looks like twins 😉

[wise canyon]

no worries, and yes it's a great looking image, I liked also how the UU versions looked. 🤟

[sand kernel]

I would wish leonardo can solve the problems with faces and hands, regardless of the size of the image and the size of the hands and faces as well as the orientation to the “camera”.

[wise canyon]

I find Lightning XL a little behind the other models with hand and face quality, the photorealistic models and PhotoReal give much better results, with hands especially, just imo.

[sand kernel]

there are many possible combinations. you just have to find the best combination. I had mostly used KinoXL so far. and then continued with color sketch. It's not uncommon for me to get a style that doesn't look bad. As a beginner it is a phase of “trail and error”.

[wise canyon]

Absolutely, this is the best way to explore Leonardo's capabilities.

[real lynx]

I would like to make a Sketch, lend me your prompt :3

[sand kernel]

@Cami you can use a reference photo too. I wold send you this pic in PM however I can´t send messages to you

[thin tinsel]

awesome, thank you for your tips, I will try them too. 🙏

[sand kernel]

here other examples

[uneven terrace]

If a model say that it best to use for 512x512 image and I use it to generate 1920x1080 will it effect generated image quality?

[wanton belfry]

very much, largly base SD models aren't good with that, XL is way better at handling that. Only option you have is to upscale to ge that

even with XL you might have to pull Alchemy refiner or Alchemy to get great output

well it mostly depends on subject

[uneven terrace]

I don't know name of the model but it starts with A. Is that model good with this XL thing?

[wanton belfry]

Albedo?

[uneven terrace]

Can't afford paid version. Isn't there any free option?

[uneven terrace]

YESSIR

[wanton belfry]

Albedo is free, I have made great stuff with it

however going too big might lead to bit of dissappointment

[uneven terrace]

I want to use it for 1920x1080 generation. Do u think it will be able to make good art with that setting?

[uneven terrace]

Actually man I need 1920x1080. I have tried Canvas but it is too hard for me to get that stuff yk.

[wanton belfry]

base for XL is 1024x1024, so its generally okay. Its good around that range

[uneven terrace]

I will give it a shot. One more ques. How I can use seed to generate same old image with high dimensions?

[wanton belfry]

you can actually calculate the ratio and try it on lower size and then upscale

you can't actully do 1920x1080 directly. 1536 is upper limit

and 512 the lower one

[wanton belfry]

by upscale

man the upscalers in XL don't actually upscale at all, wth

size is same

difference is miniscule too

[feral bough]

Upscaler is really a misnomer, it used to be called Refiner if you'll remember, that is a more accurate description

[uneven terrace]

Is there any tutorial abt this?

[wanton belfry]

refiner is different

[uneven terrace]

That sucks

[wanton belfry]

I was looking at smooth and crisp ones...

[wanton belfry]

well its sad but there are multiple places to go on net if you want to just increase the size of image, so just focus on getting image in right ratio and getting good output and then upscale

maybe even in Base SD

[uneven terrace]

THAT SMART

What ratio u recommend man?

[wanton belfry]

16:9

[uneven terrace]

I will give it a try

How seeds works in Leo?

[wanton belfry]

https://discord.com/channels/989166677390426132/1093295907358900284

he made a guide on that

its better than what I can explain

I am honestly very poor at explaining the seed thing

despite using it multiple times, its not my favorite thing

[uneven terrace]

Ill check that

[uneven terrace]

Ahh

[wanton belfry]

exactly this or with some change

[wise hazel]

something as cute as that

mine made this

and the mud doesnt stick on the cat

[wise hazel]

not 2d style, i got a yt and needed to switch from bing ai to leonardo ai

[wise hazel]

yeah, can u share ur prompt and model used?

[wanton belfry]

anime xl, alchemy(3d render), Prompt: 3d, Cute cat in Mud, looking in front, Adorable, blush, big eyes, fur, mud, farm, crops, stunning enviroment

[wise hazel]

👍

ty

[wanton belfry]

argh cute...

https://tenor.com/view/theoffice-kevin-cute-cuteness-overloaded-gif-20377186

[copper warren]

hahaha just have to say, describing the kitten as being muddy results in a very muddy kitten lol

[wanton belfry]

when the mud and kitten becomes 1

behold, lying down technique

[wise hazel]

u guys generations are insanely good ngl

[wanton belfry]

just a matter of being used to the tools

[copper warren]

k, this isn't the requested pose or anything, but his expression was just so cute I had to share lol!

[wanton belfry]

your not the only one, lol

[copper warren]

and then came the day that image help became all kittens 😂

[wanton belfry]

not just kittens, Pirate Kittens, with alot of gold

[copper warren]

haha it seems leo is predisposed to create cute kittens 😂

[wise hazel]

lol

[wise hazel]

how u put the mud on him? like that?

[copper warren]

with muddy feet lying on his belly in mucky mud, with his legs outstretched in front and behind him. sinking into the muck.

[wanton belfry]

"Covered in mud" "Dirty" "Muddy fur" these prompts can help

[wise hazel]

💀

[copper warren]

"covered in mud" with the rest of my prompt and settings got me the kitten who was basically all mud lol 🤣 specifying his feet being muddy, his environment being muddy, and that he was sinking in the mud, that got him more muddy for me hahaha

[wanton belfry]

https://tenor.com/view/monkey-ape-smile-digibyte-dgb-gif-9628022257971076819

[copper warren]

okay, what model and settings?

[wanton belfry]

oh wait contrast is available now?

[wise hazel]

wat model and settings?

[wise hazel]

pirate 😄

[copper warren]

photoreal, kino, cinematic,
elements: Kid's Illustration 0.8, 3D sculpt 0.2
cgi cartoon ginger kitten with muddy feet lying on his belly in mucky mud, with his legs outstretched in front and behind him. sinking into the muck. oozing mud. wearing shirt and trousers. happy sweet hopeful curious gentle expression. muddy pirate kitten.

[wanton belfry]

personally I prefer Anime XL

[wise hazel]

im having a hard time finding "elements: Kid's Illustration 0.8, 3D sculpt 0.2"

[wanton belfry]

unless its for realistic

[wise hazel]

oh

[copper warren]

to the left of the prompt box, there's an icon like an image, to click on. you should be able to select elements there

[wanton belfry]

Elements can actually do opposite of help on high, keep it decenly balanced or low unless you can counterbalance

[wise hazel]

oh lmao

[wanton belfry]

you dont particularly need kid's illustration with this

[copper warren]

yeah, I usually start elements out at 0.1

[copper warren]

hahaha you help your way, I'll help mine

[wanton belfry]

you can put weight on 3d render

[copper warren]

start elements out at 0.1, increase as needed

Kentox make your own images 😂

[wanton belfry]

which is also okay, prompt is still strong in that, as long as you dont go over .4 or .5

[copper warren]

everybody has their own methods 😂

[wanton belfry]

unless its something like glowwave, it literally shifts the whole theme with even .1 or .2

[copper warren]

yes this is truth. or psychedelic

[wanton belfry]

true true

[copper warren]

so hopefully that gives you a few ideas of what to try. the key is to experiment until you find what you're looking for. we can all guide you but there's no one solution. 😁

[wise hazel]

thank you all

much love

[copper warren]

I have found the glowwave can still be made to act politely with the right balance of settings, but psychedelic takes over everything it touches lol

[open echo]

https://tenor.com/view/buckle-up-hold-on-get-in-fear-and-loathing-in-las-vegas-gif-13669565

[lavish yew]

Does anyone create prompts with chatgpt?

I need help with it

Any1?

[dim zodiac]

it works, but I wouldn't recommend. It tends to be too verbose, it doesn't seem to understand the Stable Diffusion prompt style

[lavish yew]

So how can i make detailed images without chatgpt?

Any tips?

Smth like this

Or this

@stable carbon

[stable carbon]

hello?

[lavish yew]

Can u help me?

[stable carbon]

not right now, no. but you're welcome to check out my stream, which starts in 15min. you might learn something from there

[lavish yew]

Aight

[lavish yew]

Can u try to make such images on stream?

[stable carbon]

I'll be doing various things. you're welcome to ask questions or make suggestions in the chat. but i cant guarantee I'll see them all or decide to do any.

[lavish yew]

Ok

[lavish yew]

@dim zodiac

[dim zodiac]

how do you learn better

wikis or videos

[lavish yew]

Videos i guess

[dim zodiac]

https://www.youtube.com/watch?v=-L9bvr7jLcE

this will help you out then

between the prompt structure that they walk through in this video, and learning some good keywords - you'll be well on your way

what are you trying to make images of?

I could suggest a starting prompt

[lavish yew]

I make images for my videos

[dim zodiac]

also @stable carbon will be doing a stream in... 3 minutes, so come on by and get some help live!

[lavish yew]

Ill join it

@stable carbon try to make images that ive suggested plsss

[dim zodiac]

you can ask for stuff, but he'll do his own thing

[lavish yew]

Aight

[open echo]

Hey there 😄

What have you tried so far? Could you share your prompt and the results you got?

This way we'll be able to help you get better results

[thorn garnet]

Hi I have some question, could you help me improve my image generation, I want to generate similar style of image 1 ; I try some preset and element but in vain, like fur detail and eye detail, need some suggestion from you expert thanks. (I am none English speaker, sorry for the bad gramma)

[lofty vortex]

Wondering if anyone know how to use your own likeness and get it exactly like the photo. When I've tried this, it's always had butchered results and never close or nearly close or there was always something off

[dim zodiac]

hey Samkuo! very cute creations, are you trying to make them more realistic but still fantasy and cute?

[dim zodiac]

exactly like the photo? that's what a photocopier is for 😉

but the AI can give you things that are very very similar if you feed it an image as input, along with a good prompt

can you give some examples of things you tried and how they went wrong?

[lofty vortex]

Whenever I try to upload my own photo and get it to put me in a different outfit, it makes a whole another person

If I share a photo of me, do you think you can give me an example what's possible?

[real lynx]

@lyric latch

I brought you a fox

[lyric latch]

Thx Cami 🙂

[real lynx]

If you prefer a bird fox

Or a blue fox

These tools are available to everyone up to 30 and are premium character image tools.

That's why the character looks like me

[lyric latch]

wait, you can use your picture?

[real lynx]

use my photo

[lyric latch]

I thought you couldn't use a picture in leonardo

photo, sorry

[real lynx]

Yea

[lyric latch]

and can you enhanced an old photo?

[real lynx]

Yeah

Highly detailed shot of an (((iridescence))),Hugging a beautiful little fox, 1Lolita 1girl crystal sculpture in the Shape of a miniature Elderly firebender having tea, on a crystal tree branch, often associated with the arrival of spring. In folklore, the robin is sometimes depicted as a symbol of good luck or as a harbinger of news from the spirit world, vibrant background, full motion effects, diagonal view, crystal particles glittering, back light, ultra sharp focus, high speed shot, vibrant color, Bioluminescence, high quality

This is the Crystal thumbnail prompt

[lyric latch]

thx a lot Cami

[real lynx]

Character image It's called the photos of yourself tool.

[lyric latch]

With just this one prompt, you generate 3 differents pictures with different moods

[real lynx]

At the prompt I added your fox and changed my hair color to blue

With this tool you improve the photo

I would give it a strength of 3 or 4 more, it does not depend on the state of the photo.

[lofty vortex]

@real lynx that's awesome! I don't even know where to start

here's one of me

[real lynx]

Hi

[lofty vortex]

I tried the image reference

I can't get anything to look like me

how did you do the ones of yourself with Leo?

[real lynx]

You need to activate the new interface

Upload your photo there

I used this tool

But you can choose the one you want

and create other styles yourself

Remember that the prompt modifies your photo

If you do not modify in the prompt that you are a man or a woman, no matter how much you post the photo, in my case, they will try to make me a man. Because I did not modify it xD

[dim zodiac]

depends on how much you prompt:

[real lynx]

Exactly the point, there is a prompt that will fit you better than others.

[dim zodiac]

I may have been too strong in the wording I used, looks like I added a bit of age

[real lynx]

I liked this one

[dim zodiac]

@lofty vortex is that what you're looking for, or you want to really really just copy an image

[real lynx]

I got you the prompt I wanted about foxes :3

[lofty vortex]

I don't get that option

is it because you're using mobile?

[dim zodiac]

start by going to "Image Generation" on the left

[lofty vortex]

also I'm on desktop, and I did go into image creation and saw no option there for htat

[dim zodiac]

I'm on Desktop too

[real lynx]

Show me with another photo that says

[real lynx]

Astral Translucent Chibi Redpanda, floating, extremely saturated, detailed, best quality

[lofty vortex]

right, my interface still doesn't show that, things have changed on it and it did say somethings about new presets and character consistency

[real lynx]

Choose a style

and close

[lofty vortex]

right, I have, but even though the photos you've shown me, doesn't match my likeness

I've tried multiple ones

[real lynx]

I'm on my cell phone so it's different.

[dim zodiac]

I can show the steps on PC

[real lynx]

You have to find a better prompt to get your face

For example, he looks like me but not that much.

Fail

Better

[dim zodiac]

@lofty vortex you can get to this generation screen?

[lofty vortex]

https://tenor.com/view/cry-hello-thank-you-gif-11698623

[dim zodiac]

nice! if you can get to that screen, on the side of the prompt box is an image icon

that will give you ALL the image input options

for the examples that @real lynx and I were giving, we were using Character Reference

that will allow you to keep your characteristics, but add different things with your prompt - but with the caveat that some prompt terms will make it look less like you

if you want a more exact copy, you can click the More Options button below content reference

Image to Image will give you a more exact copy of the input image, that you can loosely change... with the image you gave, you would have a hard time making it be anything other than a green background, that's a very strong image

Character reference left, Image to Image on right - same prompt!

so Image to Image is going to use far more of the image colour and content as input

[lofty vortex]

i had to use another app to make a cartoon version of my likeness

as you can see by my avatar

[edgy nebula]

how can I generate a photo here

[dim zodiac]

hey dragon, gotta go online and use the web app!

[real lynx]

#🙋│image-help message

[dim zodiac]

https://app.leonardo.ai/

[real lynx]

😅

Hi

[edgy nebula]

so here is only for communications?

[dim zodiac]

that's right, for learning, sharing, some contests, fun and help!

[real lynx]

Coffee?

[thorn garnet]

Hi thank you for reply,yes you're right, more realistic and still keep the cute factor with detail fur and eyes, so far I tried like 100s image but couldn't achieve.
Do you have any suggestion to fine tune? yesterday I did tried some preset.
things I tried : model related to animate(x) , now I keep Leonardo diffusionXL or Lighting
It seems like select element (3D,Kids,Toons) and keep it low strength could do some hyper-realistic things.
but once you select element the detail deform(don't know how to explain better, like eye ball pupil deform, fur detail is gone,color dilution..etc)
thank you.

[lofty vortex]

https://tenor.com/view/hhgf-gif-25031041

what ever happened to the part where you can train your own model? Where you can provide Leo and bunch of your own photos so it can create a better result for you? Is that legacy? Has anyone used it successfully (with their own photos??)

[elfin nova]

Is it possible to train a model off one image?

[lofty vortex]

This looks soo much like you

[dim zodiac]

it still exists, they shifted it into paid tiers though

[dim zodiac]

no, but that's where Character Reference would come in

or Style Reference/etc

there are Controlnets to take elements from specific images and generate off them

and actually, I should change what I said. It's POSSIBLE to train using a single image, but it would not be a good dataset

[elfin nova]

I really like the 1 image 😅

[dim zodiac]

if you trained a dataset on the one image, it would only be able to output that same image over and over

I would suggest content or style reference to get that type of image again, using a similar prompt 😄

[dim zodiac]

something more like this?

or do you still want the eyes to be cartoony big

[thorn garnet]

yes,and plus cartoon style, attached image is what I currently achieve

[thorn garnet]

it hard to find the balance to have more cute factor but keep the realistic feeling, do you have and suggestion, wihich model or ssetting I should aim for fine tune?

[dim zodiac]

I think you're already well on the way

it really is going to be all about finding that balance

because realism is going to want to make the eyes normal, but asking for cartoony will lose realism

I tried an experiment with Style Reference

so I could use a more cartoony prompt, but then also get fur, I put in an image of a fur texture

what do you think of that one

[thorn garnet]

better than mine, I am reading first, but you gave many valuable info.

I am not reading first

[thorn garnet]

so you mean it's better add cutey facor into prompt, not using element? is that what you mean?

[dim zodiac]

so what I'm trying here is using a very cute prompt, and then using a realistic image of a cat as Style Reference

to add back in the more realistic fur

it's having pretty good results

This is using the Lifelike Vision preset (so leonardo VisionXL)

but I think DiffusionXL could do the same, if you like what you get from DiffusionXL

this is the same prompt in DiffusionXL

so it lost a bit more of the cartoony eyes

it can still go really cute, but the eyes stay more realistic in Diffusion it seems

[thorn garnet]

currently I main use three element to fine tune (3D + Cartoon +Kids), and use basic prompt, I also use Lifelike present, and model I feel diffusionXL gave more realistic result, and style do not impact much, I keep cinematic

[dim zodiac]

yes this was all with Cinematic

ok I need to go to bed, it's 1:13AM for me

I am happy to pick this up again tomorrow though, there's also #👋│accademia as a spot to pick up more specific help from some very talented people if you want to keep getting mentorship on this type of thing

[thorn garnet]

may I ask in this image for example, do you use any element ?

good nite thank you bro

[dim zodiac]

No, no elements just a prompt and style reference from a realistic cat

One sec I can share the prompt if you want

[thorn garnet]

yes please

but we talk tomorrow, you have good sleep

[dim zodiac]

prompt: Professional cartoon photorender, ultra-cute kitten with (exaggerated cartoonish doe-eyes: 1.3), dressed in suit and tie, (fur rendered in exquisite detail), eyes watery, kawaii, UHD, 32k, hyperrealism

and was using this image #🙋│image-help message as the style reference, on High in VisionXL, and medium in DiffusionXL

good luck! Hope you get what you're looking for, please tag me in results or if you want to troubleshoot/brainstorm more

[thorn garnet]

sure,much appreciated ur help

[real lynx]

It's a photo from a few years ago

[lyric latch]

@real lynx Thx for all the foxes. Waht a night bird.

[white delta]

can someone give me a prompt for a boss fight?

[robust gulch]

I don't have a prompt but I can recommend Albedo for that. It has a habit of making a lone person walk towards a big monster and similar.

[pine star]

I have a not working prompt, and I don't know how I can get it work: "A street scene in a post-apocalyptic city. The ruins of the buildings were built in Art Nouveau style. It is drizzling, there are impressive deep red clouds in the sky, children dressed in rags are playing in the puddles on the street, a caravan of traders with goods on sick donkeys and dressed in tattered clothes are on the road. There are piles of rubbish on the side of the road. Barrels of toxic waste stand next to the house entrances. Creepy figures in the entrances to houses. Maximum details, dark mood, intense colors."

The Art Nouveau and the red clouds are ignored every time and the other things most time. Only Street scene and drizzlings is there every time.

Have any one of you an idea how I can get an Image like that?

Oh, and ruins i had not a single time too.

[dim zodiac]

There are lots of good details obscured by too much text here

I am going to take an attempt at it, might take a bit before I respond again 😄

yeah, this was my worry - so many details that the AI starts to blend a lot of them

I like this one, although it has very little imagery of the children playing

[pine star]

But you get red clouds. 🙂 But not Art Noveau architecture too. 😦

And you get ruines.

[dim zodiac]

this is the prompt I had edited it down to: (Art Nouveau) style painting, Street scene in a post-apocalyptic ruined city. Deep red clouds in the sky beginning to rain, children dressed in rags playing in puddles on the street. Caravan of traders on sick donkeys, dressed in tattered clothes on the road. Maximum details, dark mood, intense colors. Creepy, toxic waste, sinister.

using Illustrative Albedo preset

[pine star]

@dim zodiac Thank you!

Then I will play now with the new prompt. 😄

[pine star]

Today I ran out of tokens before I got the image I wanted, but I got a few new ideas on how to further improve the image, which I'll try out tomorrow. 🙂 Thanks again @dim zodiac , you helped a lot and the images got a little better now.

[dim zodiac]

awesome glad to hear it! hopefully the new prompt direction can help focus those tokens 😄

[elfin nova]

Hey all, had a question. I'm working on a logo for my brand. I have a few I like. Do I add all the ones I like to a dataset?

I should clarify. My brand is The Armchair CEO. I'm doing some content about business.

So I'm thinking I have the logo, but I also use this mascot, if you will, in other settings.

[dim zodiac]

you could add it to a dataset, yes

[elfin nova]

anyone have a tutorial of using datasets I can read or watch?

looking at YT now

I know it's user ignorance, I see y'all do some amazing things with it.

[wise canyon]

Hi Jeremy D, on the Leonardo YT Community channel you will find many useful tutorial videos, including this livestream about Datasets/Model training that was recorded here in our Discord community: https://youtu.be/YvHn4Kw4beg?si=h7wm6juNDvwKZlYD , hope you find it helpful 🙂

[elfin nova]

You're a hero! Thank you.

[wise canyon]

You're most welcome, lol 👍

[uneven terrace]

Y it is hard to make hands using AI even after so many yrs??

[lyric latch]

so many years? How old are you?

[robust gulch]

I think it's because hands move in so many different ways it's difficult for the ai to decide exactly what they should be doing.

[hardy sentinel]

Yes, it is still difficult

[white delta]

can someone hop on voice channel and help me please

anyone?

[feral bough]

Proper AI image generation is like less than 2 years old, but you can really see that the hands have gotten better (if you use newer models etc). But it's still not perfect because hands are a very complex part of human anatomy and it's a lot of complexity in a small amount of pixels which is always hard for AI

[white delta]

can you help me?

[feral bough]

I don't have special suggestions for boss batlle images except using "boss battle" and maybe "dynamic movement" in your prompt, and consider using the AnimeXL model because it gives cool creatures

[white delta]

no it's about manga generations

[feral bough]

Okay well I see Kentox will help you

[spiral axle]

I heard Manga 😮

I cant talk atm ... but I am glad if I cam help you

[wanton belfry]

already solved

[spiral axle]

❤️

what is your opinion on what Anime Setting is closest to the old "Dynamic" Preset @wanton belfry ?

[wanton belfry]

haven't tested whats the most dynamic one yet

[spiral axle]

seems with Non-Anime Styles that either Flat Illu or General is the best one imho

trying to merge some Hades II Art in there

[wanton belfry]

I like manga and monochrome, even though they might not work properly alot of times

[spiral axle]

I like Flat Illustration (Overwatch Merge)

Monochrome is not monochrome xD

[wanton belfry]

the irony

[white delta]

check dms

[white delta]

prompt?

[hardy sentinel]

@robust iris pls post your question about creating image here

[white delta]

steven

[hardy sentinel]

I mean this question #🖼│image-share message

[white delta]

is it possible to make a manga series?

[hardy sentinel]

Greetings

[white delta]

with leonardo?

[robust iris]

How can i create imag like this

[hardy sentinel]

to create a manga series with how many characters?

[white delta]

1

[hardy sentinel]

one is possible

[white delta]

what would you say is the limit?

[hardy sentinel]

one character in a scene

[white delta]

is 3 characters possible?

[hardy sentinel]

in 3 different scenes is possible

I will recommend one character per scene

[robust iris]

From where i can learn

[white delta]

can i dm you?

[robust iris]

I have upgrade plan

[white delta]

@robust iris if your up for it

we can learn together

[hardy sentinel]

I am away at this moment

[white delta]

generate some images together

[white delta]

oh alright

[hardy sentinel]

maybe somebody else can help you, so it is best to post here

[spiral axle]

What is needed ?

[white delta]

i want to create a manga series

i have a character reference

[spiral axle]

What especially do you want to create ? Scenes? Panels ? Covers ? Character Sheets ?

[white delta]

scenes

this character

[robust iris]

Can you help me then?

[white delta]

im not a professionakl

professional

but

if you want to create some images together

we can do that

best way to learn

[robust iris]

Okay sur e

[white delta]

i can't speak but do you know how to screenshare?

@robust iris join voice channel

i will type

and screenshare so i can guide you a little

[spiral axle]

I can only type too but I try to help

[white delta]

yh brother hop in

we can speak in the voice channel chat

[robust iris]

I just create the prompt

Now i want to know where i can put it

{
"type": "3D render",
"subject": "Asif scheming with a greedy look, thinking of money",
"environment": "dimly lit office with a desk cluttered with money, documents, and a computer",
"style": "hyperrealistic",
"camera": "medium shot",
"lighting": "dramatic lighting with strong shadows",
"resolution": "8K",
"details": "detailed facial expressions, realistic textures, reflective surfaces"
}

Like this

[hardy sentinel]

who is Asif?

[robust iris]

Random@name

Random guy name

[hardy sentinel]

it only works with celebrity

[robust iris]

Where i can put this whole Prompt?

[hardy sentinel]

wait a minute. Need to start my computer

[hardy sentinel]

after you arrive at image generation site (for version 2) you type in the prompt into the empty space that I marked with number one (red arrow)

the number of characters is limited

if you want to work with older user interface, please click at legacy mode at upper right corner

[robust iris]

But it needs simple prompt

[hardy sentinel]

I will omit Asif and replace it with "man"

[robust iris]

I want to generate image like this

[hardy sentinel]

yes, indeed

[hardy sentinel]

first please consider the resolution size, (image dimension)

and aspect ratio

[hardy sentinel]

but in your reference image, there is no documents nor computer?

[robust iris]

This is the style i want

[hardy sentinel]

yes, I understand

ok, this is what you need to set up first:

change the aspect ratio to 9:16 (Tiktok)

on the left side @robust iris

you can adjust your aspect ratio

click on the number 1 (marked with red circle)

pls pick the tiktok aspect ratio just as your reference image

marked with number 1 ( red arrow)

then adjust the number of images you generate at once

here please click number 3 and pick 1 image

this way you can save token while you are still try out your prompt

are you still online? @robust iris

[robust iris]

Yes i am looking into things you are sending

[hardy sentinel]

ok that is fantastic, are you working with pc or mobile phone? because I am sending the screenshot of pc

[robust iris]

Mobile

And I am just taking out the right style

[hardy sentinel]

ok, the mobile phone has a different layout

[robust iris]

I am getting screenshot

No worries

[hardy sentinel]

android or ios ?

[robust iris]

I am just trying to figure out the right style to select

Ios

[hardy sentinel]

ok could you share your screenshot so that I can guide you pls?

i don't have ios based mobile phone device

[robust iris]

Screenshot of what

[hardy sentinel]

of your mobile device screen right now (while working at this image)

after you set up your aspect ratio you can go back to number 2

marked with green arrow

there you can open up the image guidance menu

from the menu pls pick content reference

marked with number one (red circle)

then you can upload your reference image

[uneven terrace]

Make sense

[hardy sentinel]

@robust iris you can upload and use the reference image as I show you here #🙋│image-help message

[robust iris]

I tried it but failed

[hardy sentinel]

pls show me your result

[hardy sentinel]

do you mean the upload has failed?

[robust iris]

No wait let me show h

U

I want to change the style

[hardy sentinel]

the upload of the reference image worked fine for me @robust iris

[hardy sentinel]

do you want to be the style exact like the reference image?

pls add your reference image into the style reference @robust iris

marked as number 2 (green circle)

adjust the style reference strength if the result is not satisfactory

[lavish yew]

Can somebody help to me to write prompts for images?

[real lynx]

What would you like to create?

[lavish yew]

Dms

[crude crescent]

When I use something special like anime models, with anime elements, should I stil reference anime style in prompts?

[thin tinsel]

If you want.

[spiral eagle]

Oh I haven’t been here in so long,

But, I need help with this image:

I wanted no gloves and the things sticking out of his back I wanted removed, but I’m low on tokens

Also the tip of his tail was supposed to be white

[ivory ravine]

PLEASE HELP how to turn off Alchemy in the new interface 😂

[thin tinsel]

select fast instead of quality

[spiral eagle]

Someone please help

And now I’m out of tokens

[ivory ravine]

Thanks @thin tinsel , anyone has an idea how to create an engraved phrase on the image of the gold pendant (Easiest way…i tried Text in canvas with no luck ) 🖋️

[thin tinsel]

Hi Paris, there are quite a few different ways to acheive that. I need to sleep now, so just wait to see if someone else picks up your question. If not, I will leave you some pointers here tomorrow.

[open echo]

This looks like a classic job for the Canvas 😄

https://intercom.help/leonardo-ai/en/articles/8092594-canvas-2-0#h_de1a7c19a6

Using the Inpaint/Outpaint tool you should be able to change the colors, change clothing and remove items

[spiral eagle]

I’m on mobile, does that change anything

[spiral eagle]

Uhh… why am I getting this??

It doesn’t even align with my prompt

[real lynx]

Canvas in Mobile It is very difficult to use

[spiral eagle]

Yeah, but, I just made another one

[real lynx]

Muy bueno

Very good, sorry for my Spanish xF

[real lynx]

I already took out the weapons that you didn't want to be behind you.

[spiral eagle]

Oh, ok cool

[real lynx]

XD

[spiral eagle]

Uhhhh

Did you purposely do that?

https://youtu.be/7ncHu9EvQik?si=kV30x-4EIY920VjC

[real lynx]

Yes sry

[spiral eagle]

It’s fine

This one’s easier to fix, only the gloves and a small back part

[real lynx]

I don't have tokens

[spiral eagle]

What’s character reference?

[spiral eagle]

I didn’t do sparkle core at all though, I did solarpunk of 0.20 and just described what I wanted. That was for the original character but for editing the bad parts out, (the back item on him, and then the gloves and pants, he’s a animal so he doesn’t really need clothes but the vest would look good alone)

[hardy sentinel]

Hi, character reference is to adapt the face from the reference image into your picture

For example this is the way to try to make an image which resemble your own face.

[ivory ravine]

Colleagues anyone has tips how to do engraving of the phrase on the image of the gold pendant in Easy way…i tried Text editor in canvas with no luck 🖋️

[spice oar]

this is what I input

[night vault]

what is your prompt and output?

[spice oar]

this is the output

[night vault]

to be honest, it does not look like a image to image to me. can you double check what you have selected?

[spice oar]

prompt: create a man warrior from the medieval years that is holding a sword and is inside the battlefield

[spice oar]

yes that's why I asked. It doesn't match

[night vault]

can you capture what is your selected settings?

[spice oar]

I didn't change anything in settings and it was by default because I don't understand what should I select there

[night vault]

check this https://youtu.be/M-wyBg8_bhU?si=AtAlJyhaBQg6mBBq

there is new ui just released. you can access “legacy mode” to have same ui as i sent to you

[spice oar]

thx I'll check now

[open echo]

Try with the Canvas editor. A quick mask, a short prompt, and you should be good

Your 150 tokens must have reset by now

[spiral eagle]

Oh ok

[open echo]

I suggest you use the Dreamshaper 7 model on the Canvas. I had the best results regarding adapting to the existing style of the picture

And you should be done in about 10 minutes 😄

(And before you say "but his hands are black!!"... well... Foxes paws are, indeed, black 🤣 )

[spiral eagle]

Wait yo that actually looks cool

[spiral eagle]

👍

[spiral eagle]

Is that the final product?

[open echo]

Straight from the canvas, yeah

[spiral eagle]

Ok send that to me

[open echo]

Oh... I reset the canvas when I was done with the screenshots :S

It's super easy to do, you'll get it in no time

[spiral eagle]

Could you give me the prompts?

[open echo]

"jungle, background, bokeh", "orange fur" and "white fur" 😄

And just mask each parts one by one

[spiral eagle]

Ok

Why is this happening?

[vestal lake]

why did this cost 50! credits? in universal upscaler. if you can tell the difference let me know cos i cant!

[feral bough]

I think you (accidentally?) uploaded the same image twice. Both of your files start with Default whereas images from the Upscaler start with UniversalUpscaler.
As to your question: the cost in Universal Upscaler is based on the size of the image. If you input a large image and want to upscale it to make it even larger, it will cost more tokens. But if you put the Creativity Strength on very low, it will not change much, even though it will cost the same amount of tokens. Because the token cost is based on the size of the output image.

If you want the image to change more, you should put the Creativity Strength higher, and if you want it to cost less, you can lower the Upscaler Multiplier

[vestal lake]

ok thanks , it still cost to much

[feral bough]

You can make that determination for yourself if it's worth it or not. However the cost is what it is because it is very expensive to run such a detailed upscaler. Token costs are always a reflection of the amount of 'computer power' needed for actions

[vestal lake]

is this why when i printed one the other day it came out pixelated? i have to stand 3 feet away from it not to noticed

[feral bough]

If you see a discrepancy in the quality of the image on your screen versus how it was printed, then it might have been a problem with image compression (due to the way you saved the image) or your system/printer not being able to handle such high resolutions. Because UU images definitely don't come out pixelated, in fact the amount of pixels is greatly increased (which is why it's expensive)

[vestal lake]

i didnt use uu last time , that may of been the problem

[thin tinsel]

https://cdn.leonardo.ai/users/fa466aaa-cdd2-46ea-a529-92bc0a96898d/generations/495fcaed-6d56-4408-8142-1042fba17dfa/Default_ONCE_UPON_A_TIME_text_ENGRAVED_on_a_gorgeous_ornate_GO_2.jpg

You have a couple of options for adding text. It's going to be another challenge for you make it look like an engraving.

To create the image above, I prepared the text outside of Leonardo, then used Text Image Input at 0.53.
There was some trial an error involved to get the text to read properly. The larger and bolder the text is, the more success you will have
I would recommend you to prepare white text on a black background.

It is advisable to ensure the image you upload with your text must be the same size you want the whole image to be.
For my image above I positioned the text and circle in the lower bottom half because that's where I wanted it positioned

@ivory ravine Another option is to use the Canvas Editor. You can view some information https://discord.com/channels/989166677390426132/1126679661741609120 about how to use it.

[open echo]

When using the Canvas editor, you can choose a number of variations that will be proposed (choose on the right toolbar, from 1 to 8).
Then as you cycle through, you will be able to « Accept » the one that corresponds best to your needs.

[spiral eagle]

Yeah but it never shows what I want, could you make a video or something for me to get a good example?

[open echo]

Straight from the big bosses themselves 😜

https://youtu.be/TkK1BvhV5yY?si=tRxbFeqDFzistV85

[spiral eagle]

Oh ok

[thin tinsel]

@ivory ravine I did these images using Albedo XL, Text to Image 0.58, Alchemy Dynamic, and style reference on Low.

I think a couple of those pass as being engraved

[thin tinsel]

I'm having too much fun with it now.

[marble fulcrum]

I was unable to find the "Guidance Scale" Option in the new Version, Is it Gone?, or any idea how to tweak this value in the new Version?
Thanks!!!!!!

[thin tinsel]

I don't think it's there I haven't seen it

[copper warren]

Lovely!

[copper warren]

Catclaw, to be clear, this channel is for getting advice about your images, from fellow community members. It is not for having other people make images to send to you. The people here are awesome and spend their time and their tokens to help users like you learn how to create amazing images. They are not image vending machines.

[spiral eagle]

Yeah I know I just want to know why I’m doing the mask thing and it just puts a variant of what is already there instead of what I described

[copper warren]

When your tokens reset, come back and we’ll help you solve this. But if you want to keep using this channel you’re going to need to show more consideration and appreciation for the people who are helping you. I don’t think I’ve ever seen you thank someone for their help, but I have seen you expect someone to simply send you the image they create. That’s going to have to change.

[spiral eagle]

I have said thank you many times, or at least I forgot to tell them thank you, I thank all who help me on here but just forget to say it because I’m busy with something else, don’t tell me to change my ways, I do show appreciation and respect for the people here

[copper warren]

I’m just letting you know the situation. Make sure if you’re asking for help, you don’t forget the part where you made sure you voiced your appreciation. And do not ask users to send you images they make as they help you. These are the basic requirements to continue to have access to this channel.

[spiral eagle]

Earlier I was asking for a way that someone else helped me with my image and I asked how did they do it, that’s what happened

[copper warren]

As long as you’re within the rules I’ve stated you should be fine. Let us know when your tokens reset so we can continue working on a solution with you

[spiral eagle]

It’s in a hour

[copper warren]

Okay see you then

[spiral eagle]

I have the tokens

[copper warren]

Okay cool. So where are you at in the process? Have you tried any of the advice you’ve received so far?

[copper warren]

I would recommend not doing the entire arm at once, unless you want a big change.

[spiral eagle]

Ok

[hardy sentinel]

It is not available at the version 2. You can use it by getting back to legacy mode (top right corner)

[silent crown]

/imagine little boy character, multiple poses and expressions, front view, pixar styled character, simple, cute, white background, white t-shirt, red pants, short brown hair -- ar 4:5 --v 4

[spiral eagle]

What would be my prompt be?

[copper warren]

describe what's in the generation frame, with a focus on what you want changed. so "orange fur", "white fur", and you could say something about the the armour or the backdrop. you can also use the sketch tool, at like half transparency, to hint at the orange colour you want. you'll need to play with the creativity strength to find the right balance. too high and it will start adding in odd things, and too low and it won't change much.

[spiral eagle]

Ok

[copper warren]

This might help explain how some of the aspects of masking work

[spiral eagle]

Ok

[marble fulcrum]

thanks for your answers!

[scarlet anchor]

is it possible to make maps using leo?

[wise canyon]

Hi, last year I saw some members made D&D maps, so it is possible. They were drawing sketches and using ControlNet "Image to Image". Now we have many more feature options and ControlNet options to use. Realtime Canvas is another possible method for maps.

[lyric latch]

can you use Leonardo to complete a design?

I made this, and I want to make a youtube banner with it.

But no matter how I try to redo the design on Leonardo, I won't obtain the same design. So I wonder if it is possible to make Leonardo complete the sides of the design

[wise canyon]

Hello, yes you can outpaint the sides in the Canvas Editor.
Load the image in the Canvas Editor, place the edit box half on the image and half outside the image, with the outpaint toggle on you can then prompt what you want in the outpainted areas.
Always try and do a small section at a time, so the AI can focus on matching the existing image content that's within half the edit box field.

[lyric latch]

and for the prompt, should I just say "complete the left of the design" and in another prompt "complete the right of the design" ? Thx for the tips anymway 🙂

[wise canyon]

Just try using the background description from your original image prompt, and also select, in Canvas, the same Model used as for original image generation, so the style is the same. lol

[hardy sentinel]

You are welcome

[lyric latch]

@wise canyon thx for the info 🙂

[wise canyon]

Any time, lol

[thin tinsel]

also works with no prompt at all, just a space.
but you must leave sufficient amount of the original image within the generation square.

[ebon rampart]

I've been playing with the character reference for about 15 minutes now and I can't seem to get my prompt to change the image to someone that is smiling, cunning, sly, etc. I only get the same stoic face that was on the original image. Any recommendations?

Honestly, other than that, I'm totally blown away. This is a game-changer!

[thin tinsel]

What strength did you have the setting on?

[amber dagger]

A delicately intricate flower, its petals bursting with vibrant colors and delicate textures, captured in stunning macro photography. The image showcases every tiny detail of the flower, from the intricate patterns on each petal. The vivid hues and exquisite details bring the flower to life in a way that is both mesmerizing and captivating. This high-quality photograph truly captures the beauty and essence of the floral subject, making it a truly enchanting piece of art.

It was much better on the fourth attempt, but the droplets were visible

[thin tinsel]

Did you use the built-in prompt improvement tool in Leo?

[amber dagger]

Yes

[thin tinsel]

my first instint would be to change "its petals bursting with vibrant colors and delicate textures" to "vibrant colors"
I feel like 'bursting' might have a watery feel

[thin tinsel]

Prompt improvement is great, I use it a lot, but sometimes it can be a little wordy and simplifying the prompt can be beneficial

[ebon rampart]

Medium or High. I'm guessing I should try low

Here's the original image, previously tuned from HeroForge. Then here's the image I just generated using it using the prompt: "a realistic female half-elf in a tavern. She is smiling as she drinks her ale. She has wild, frizzy hair" and Low on the setting

[thin tinsel]

well, I just wouldn't max it out.

[thin tinsel]

try using your created image (right), as the reference image.
I belive it works best when the source shows good quality image of head-and-shoulders, rather than a full length image.

[ebon rampart]

Using the other image as a reference and the prompt improvement: "A jubilant and beaming half-elf, his infectious joy evident in his wide-open mouth and toothy grin. This image of pure happiness captures the essence of a carefree spirit, radiating a warm glow in every pixel. The vibrant colors bring this scene to life, showcasing the half-elf's sparkling eyes and dimpled cheeks. The detailed brushstrokes in this exquisite painting showcase the artist's skill in capturing the essence of unbridled joy."

Still nothing 🙂

Looks like "unbridled joy" to me. She just keeps it inside

[night vault]

character references transfer facial expression as well, and more important than prompt itself. so you might need to have lower weight and few generation that she gradually transition into a bit happier

[jagged glacier]

I am having trouble generating a cartoon image of a coyote without a neckerchief, The prompt happy coyote with no human items in a desert, Disney's Jungle Book style cartoon produces this image

[thin tinsel]

Using my own prompt. Strength Low.

@ebon rampart

I recommend you to do a detailed description of what the person looks like.
Keep the clothing simple so you stand a better chance of replicating it.
Then keep the Reference on Mid.
oops, I just realised I did style reference instead of Character Reference

[thin tinsel]

Have you tried putting words such as neckcheif, or scarf, into the negative prompt?
Also put 'humans' in the negative prompt.

@ebon rampart Ok, here is this person with Character reference on Mid.
It made her hair green because I didn't take my own advice to put in details about her appearance.

[ebon rampart]

What did you use for the prompt on this?

[thin tinsel]

This is on high strength.

"Half-elf female laughing and drinking Ale in a tavern. "

[ebon rampart]

I wonder if it was the model I was using as well. I was using the new Image Generation v2 stuff, with Leonardo Lightning. Changed it to using Cinematic Kino and got this with the same prompts.

Thanks for the help - I think I'm heading in the right direction now

[thin tinsel]

I was using Albedo XL. That tends to be my go-to at the moment.
Enjoy the rest of your character creations. 🙂

[ebon rampart]

Yep - this is great - getting different expressions well now. Thanks, again, @thin tinsel !

[thin tinsel]

Wow, that's great 😄
Happy to have helped.

[jagged glacier]

I was confused zbout this, but I will try it again. Thanks

[ancient magnet]

Do you guys know any method in Leonardo (or another AI generator) to make multiple pictures of the same character but with varying expressions and poses?

Ideally a method where the same character can be revisited at a later time?

[wanton belfry]

Character reference

[ancient magnet]

Where is this?

... Oh geez, a tutorial for this feature is 1 day old

Was this a new add?

[wanton belfry]

well its almost a more than a week old, but yeah since everyone doesn't get content rolled out at same times so it can be still counted as new

[ancient magnet]

It seems to be changing the character's outfit a lot, is there a way to stop it from doing that?

[wanton belfry]

honestly not really, it will change according to prompt, either pigeon hold it with prompt or simply give in

[ancient magnet]

How would I pidgeon hold it with the prompt?

[wanton belfry]

something like a very simple prompt "Green robes" "Tuxedo" "Suit"

you can't specify intricate features, better it remain generic

[ancient magnet]

🤔

My intended purpose is to use them as effective stock photo's in a video game to represent expressions and such until either:
a) The project finishes and remains hobby level
b) The project is scrapped
c) The project evolves and I replace it with hired art

But I'd still like the stock art to not make it super obvious that it's AI art, so that a character could change expressions during a scene and still look to be in the same outfit. Would that be possible if kept to simple outfits?

[wanton belfry]

more or less yes...

or atleast it would become work-able atleast

few odd images you can fix in canvas

[ancient magnet]

Any tips and advice for succesfully assembling that?

[wanton belfry]

not sure what you are making, Visual novel?

[ancient magnet]

Also side question, is there a way to improve the no background option to also effect stuff within the hair?
Or is this something where I'd also need simpler hairstyles?

[ancient magnet]

Not really. But Visual Novel is apt enough for the purpose of this.
The narrative scenes would play out similiarly to how a VN works

[wanton belfry]

I would give you the same advice I gave another person to have maximum control, render Character in various emotions and positions as according to needs with transparency, Generate BG seperately

that would tremendously lower the token cost

you can later merged in image editing software and compile your work

[ancient magnet]

How would I keep them consistent between emotions?
For reference this was the source image for the above one:

+Also how do I generate BG seperately? It adds backgrounds automatically without me prompting for it

[ancient magnet]

This was also with setting the reference scale to High

[wanton belfry]

this is example of a BG

this is a example of transparency induced character

[ancient magnet]

I think I found the transparency setting under the advanced settings

But it's resulting in stuff like this

[wanton belfry]

dont make prompt too big or it'll malfunction

[ancient magnet]

samurai ninja outfit, annoyed, hand on hip

That was the prompt

[wanton belfry]

then transparency is not enabled properly

[ancient magnet]

You mean it's more than this?

[wanton belfry]

it should appear like this

[wanton belfry]

yes

[ancient magnet]

What else is needed?

[wanton belfry]

probably its some issue of new Graphic interface, try in the legacy version

[ancient magnet]

Where do I find legacy? I'm not finding anything called that under Models

Oh no wait, found it

[wanton belfry]

extreme right

oh wait I think there was also a limitation of not being able to use Content reference etc with transparency

[ancient magnet]

That seems to be working a lot better 😊
But the quality is also lower now

[wanton belfry]

there is also option of generating and removing the BG

[ancient magnet]

It's on the same Leonardo Anime XL generator as pre-legacy

Honestly I don't need a background for these images

With the game engine I'm using the Background would be a seperate image file

[wanton belfry]

also fair warning, dont try to involve weapons, AI still is pretty messy in rendering those

[ancient magnet]

It's doing that automatically because of Samurai, but good to know xD

[wanton belfry]

swords still are fine, but hammers..axes...guns etc are pretty sad

terribly dissappointing sink of tokens for nothing

[ancient magnet]

Now I'm just trying to figure out outfit consistency across images

As well as seeing if I can boost the quality to be closer to the original images (Alchemy ain't doing it)

[wanton belfry]

thats more of model oriented thing

depends on how capable is model

[ancient magnet]

The quality or the outfit?

[wanton belfry]

well that can be too...yes

[ancient magnet]

Sorry, meant to type 'or' not 'of'

[wanton belfry]

man, I am gonna sleep now...its getting pretty hot out here and I am sleepy.

one last thing, you can still render a Character without transparency, dont prompt in a BG and then when it generates remove the BG

[ancient magnet]

That was my old method

It just seemed to be leaving some BG behind this time

That might've been cause it wasn't legacy though 🤔

Anyways, good night! Thanks for the help! o/

[ancient magnet]

I feel like there's something I'm missing here with the Image Guidance 🤔

Back when I used "Image to Image" I could recreate very similiar images all the time, but with a lack of much alteration to pose and such

Now with "Character Reference" it seems to be changing what the character looks like constantly, even when on High setting and using the fixed seed setting.

And the "Pose to Image" just doesn't seem to be applying, even at maximum strength.

It seems to work perfectly during video guides though, so there has to be some element I'm doing that I shouldn't, or an element I'm not doing that I should. And I can't figure out which it is.

[brittle glen]

Busg bonny gamer

[pure ivy]

¡Hola! Soy Verrocchio, un asistente de IA aquí para ayudarte a obtener acceso a nuestro servidor. Por favor, cuéntame más acerca de tu interés con Leonardo.Ai. ¿Estás interesado en la creación de imágenes o tienes alguna otra pregunta sobre la plataforma?

[real lynx]

Hola

[hardy sentinel]

Hi, I guess you have used character reference?

[hardy sentinel]

do you have any image that you like as a basis ?

then put that image into the canvas editor

switch the mode to inpainting

mask the area of the face

adjust your sizes and render density

then pick the same model that you have used to create the image

now, you can write the expression into the prompt

if you do not have many tokens, pls set your image generation into 1

it will take you longer

to select canvas editor pls click nr. 4

(marked deep blue)

this is the screenshot of canvas editor

Nr. 1 is to adjust the number of images

I would recommend to set it to one

Nr. 2 is to adjust the size of your work area

you can click 768 x 768

alternatively you can type in your "W" and "H"

Nr. 3 is to adjust between outpainting and inpainting

click once to go to inpainting mode

Nr. 4 is to pick your masking tool