Hi, nice to meet you.

My name is Ninniku-san (Garic) and I'm a beginner in AI image generation. I'm Japanese and I rely on translations for English in my articles, so sorry if my English is weird.

The basic idea is to create an anime-style image, so keep that in mind.

Now, onto the main issue. this time, I attempted to create a certain character.

For the moment, we've assumed these characteristics.

Female.

Green coloured hair

Long hair

Gentle impression

Dress shirt

Skirts

Hat

Now let's start with hair colour.

Green, deep green, yellow-green,.

The above will immediately come to everyone's mind.

deep green hair, long hair,

green hair, long hair,

yellow green hair, long hair,

They're all pretty and cool!

But they don't feel right to me. So I decided to look at the Colour Illustrated Book via my teacher Google.

It's a very useful illustrated book that lists the colours of Japan and the western world.

I sorry, I'm using,,,

base model

[CalicoMix] v7.5

LORA.

Anime 3D Converter LoRA:0.1

Small Body and Small Tits:0.6~0.7

(I'm the one who gave up on having Stable Diffusion on my PC. (I'm the one who gave up on having Stable Diffusion on my PC, probably because my PC was eating up a lot of space with music-related stuff and it wasn't working.)

viridian

The colours are different from those in the illustrated book, and it's not stable.

Maybe [CalicoMix] doesn't recognise it correctly.

The colour of the second hair tip is the closest to Viridian lol

opal green

The luster of the pearls is affecting the rest of the hair.

It's a waste of hair colour, which I feel is delicious. lol

It could be controlled by "negate".

But this time it is pass off.

Also, green-type hair seems to be more prone to elfification. haha.

Well, I tried a lot of things like that.

I can't put all of my challenges here. Because the results of many challenges are already out of this world.

This time I chose

Yellow-type "olive".

((olive hair:1.4), long hair:1.2),

I chose a greenish olive from the yellow-type because green produces good elf-ears, and it fits in quite well.

Oh, and by the way, the prompts on eyes are as follows.

(Bordeaux eyes, tareme, frameless glasses:1.2),

I looked for a red colour to match my hair colour,

"Bordeaux" was felt to be a good eye.

BREAK

Below is the prompt without the image quality instructions.

Her appearance is now somewhat solidified.

woman standing, cute Asian females,

(happy, smile, closed mouth:1.2), ,

(petite:1.2), (medium breasts, pointy breasts:1.3),

(Detailed face:1.3),

((olive hair:1.4), long hair:1.2), (Bordeaux eyes, tareme, frameless glasses:1.2),

.....(( ￣´Д￣) y-~~ (smoking...)

I blame it on my bad day. More than half of the images were unsound.

wearing Dress shirt, Skirt, Hat,

　　　 ↓

White shirts are transparent. Red skirts are good!

The buttons on the collar seem to be more difficult.

However, the collar is wider than the "dress shirt" and the shape is more stable. I decided on a tops.

"button-down shirt"

White is transparent, so I'll go for black. The red skirt will also look good, and it should be a safe coordinate.

wearing black button-down shirt, red Skirt, Hat,.

A significant proportion of shirts turn white.

This is probably due to the strong parameters of the text that precedes it, and therefore there are many elements that are ignored.

Adjust the parameters, but before that, decide on the type of hat and also indicate the skirt length.

wearing black button-down shirt, red long Skirt, boater,

Various broken down lol.

Adjusting parameters

She meets the requirements about 70% of the time, but the length of her skirt and the type of hat she wears are not stable.

Also, there is a probability of red encroaching on the hair colour there.

If a lot of colours are specified, to increase the probability of success ”BREAK" to split them up.

I don't know if this is correct, but I often separate physical characteristics from external factors such as clothing and possessions.

woman standing, cute Asian females,

(happy, smile, closed mouth:1.2), ,

(petite:1.2), (medium breasts, pointy breasts:1.3),

(Detailed face:1.3),

((olive hair:1.4), long hair:1.2), (Bordeaux eyes, tareme, frameless glasses:1.2),

BREAK

shoulder bag,

(wearing (black button-down shirt:1.3), (red long Skirt:1.3), (boater:1.3):1.2),

(wearing (black button-down shirt:1.3), (red long Skirt:1.2), (boater:1.2):1.2),

Mm, I think I've made 20 cards and about 3 ～ 6 of them are up to standard.

The type of hat is actually straw material, but if I get hung up on that, I'll be on an endless journey.

The shape and size of the hats are somewhat expressive of what I was aiming for. And I think this is fine, because the non-straw hat is better suited to the co-ordination.

You don't agree?　Then I expect you to do your best. Goodluck!

Now, let's specify the pose and where you are.

We'll probably have to fix some of the parameters and text, but for now, let's go to the amusement park!



hi! I know, It's Failed image.

Ouch,,

Finally, a passable product.

This is how the girl with olive hair was made.

Finally.

Even in pornographic images, I'm quite obsessed with hair colour and clothes. I often give up halfway through. lol

I'm even more obsessed with healthy images.

Solo shots are fine, but as soon as we become two people, I often lose control.

This time, I had a man appear at the end, but I couldn't create this situation of 'woman with glasses' and 'man without glasses'.

They both wear glasses. Or no glasses.

If you have a good way of doing this, I would like to know.

In terms of hair colour, hairstyle, clothing and accessories, and their colour schemes, I felt that the instructions were at the edge of the number of items that would stabilise this time around.

If the base model changes, there may be a possibility that they will listen to more detailed instructions, and if we can make more effective use of LORA, we may be able to open the way.

I'd like to coordinate them more carefully.

More recently, I gave up the co-ordination of a hooded traveller's cloak (dark green) over overalls.

The overalls outfit can easily give birth to naked overalls if there are no inner instructions, which is like an old man riding a combine harvester in the southern US. If it were a woman, a lascivious woman would be born.

(Naked overalls are a combination of being an old man in the US South, which I think makes for the best fashion statement.)

Per to avoid it,

The AI did not understand the innerwear and overalls and the hooded traveller's cloak on top of that.

Numerous times one of them would be missing or the cloak would transform into a sweatshirt and go inside the overalls. Perhaps the AI, or the base model, cannot understand the idea of a cloak over the overalls!

That's how I struggle every day.

I suppose the correct way to use it is to generate a certain amount of fluffy character images with a swinging range.

What are you people obsessed with?　You might think to me.

But there are people like me who are stupid and clumsy.

If there are some people who would appreciate it,

I'll think about putting it in an article when I make Ballroom Dance or, perhaps, Elysia, the Albino Girl.

Well then, enjoy posting your images!