Brandenburg Gate in Berlin

The Brandenburg Gate is one of the most known places in Berlin. The image on the left was created with Nightcafe Studio. I added "moonlit" in the prompt in order to create a bit of a dreamy atmosphere.

Other than that it looks quite realistic, a bit too realistic I thought. So I decided to play with the image a bit.

I used it as a starter image and added in the prompt:

"... in the style of Van Gogh". 

You can see the result of this addition on the right: The building itself was not changed much, but the foreground and the sky have a definite Van Gogh touch.

Using a gouache painting as starter image

Some days ago I did an abstract painting with gouache:

Now I wanted to have an AI version of it, telling the AI to create an image of an "abstract painting in cubist style".

Then I thought why not having the AI Generator create something very different from the starter image? So I told it to create an image with "mushrooms in a forest".

And it did! I must admit I had to play with the weight of the image a bit till I got the result above. 

Green Goblin - AI versions

In one of my last blogposts I published one of my latest paintings - a green goblin. Of course I wanted to see what the AI could do with it. So I opended Nightcafe Studio and uploaded my painting in order to use it as a starter image.

There are two ways to influence the outcome:

1. the prompt

2. the weight you place on the image

I used Nightcafe Studio. The first result I got is very different from the original painting. It shows a green goblin with some funny headgear and two blades growing out of his body. Not a bad design, but too far away from the origjnal piece.

So I reduced the weight of the prompt, and I got an illustration which resembles the original far more.

What did I learn?
1. Sometimes it makes sense to just let the AI decide what it comes up with. Taking the horns of the goblin and changing them into some sharp blades is quite interesting!

2. If you want an image which looks more like the original you have to try a few times. Just play and have fun!

AI versions of my Acrylic Painting

In my last post I presented an acrylic painting which I had done on the last weekend. Today I will show you two AI versions of it:

This was done with Playground AI, using the acrylic painting as starter image. There is quite a difference between the original painting and its AI version. However I like the way the orange / yellow / ochre colors have found their way into the digital image.

Here is the AI version I got from Limewire AI. It is much more similar to the acrylic painting and has a painterly, watercoloury look. 

Fluffy monsters

 

I did this monster quite some time ago with gouache. My aim was to use a limited palette and to get more familiar with my gouache paints.

I scanned the painting and added it to the folder "scanned documents" - and there it was until today when I was looking for an image to use as starter image.

As I mentioned before not every AI Art Generator which allows the use of starter images creates good images. Playground AI is one of the AI Art generators I go to when I use starter images.

The image on the right has the same color palette as the starter image, but otherwise the differences are quite big. I like the fluffy guy, but I wanted to create something more similar to the original painting.

Now there are two ways to do that: You can change the weight of the starter image, and you can change the seed number.

I did that, and look what I got now:

This image is way more similar to the original starter image. The style is that of a 2D illustration, and the only thing which differs from the starter image are the eyes in his face.

Goblins, created with AI and a starter image

I did not have a special character in mind when I did the illustration on the left quite some time ago. I was interested in the roundness of the shapes, and in giving the creature some kind of headgear. 

Now, with AI Generators allowing starter images I thought: "Why not create a goblin out of this creature?". Up to now I got pretty decent results with using Playground AI, so I used that Generator in this case as well.

I used the prompt: "portrait of a goblin, colored pencil illustration, intricate".

I am quite pleased with the first result: the shape of the head and the colors are similar, and the headgear looks quite nice.

The only thing which could be made better are the eyes. They are big, yes, but still looking quite human. 

So I changed the prompt to: "portrait of a goblin, big eyes, colored pencil illustration, intricate"
 

Playground created this image. Look at the big bulging eyes! It was just what I wanted to get.

Cubist cityscapes - a comparison

 Many AI art generators allow the use of an init or start image. Some do not. In my humble opinion the use of start images (which in my case are scans of my own paintings or illustrations) allows to create images which are special and unusual in a certain way. In this blogpost I will give an example.

The subject I chose is "Cubist cityscape". 

One of the AI Art generators that does allow using a start image is PicFinder AI. But in this case I did not want to use this way of creating images; I used the prompt "cubist cityscape". You can see one of the results on the left. Yes, the style is cubist, but the result is not very spectacular.

So I went to Nightcafe Studio and uploaded a painting which I did years ago with acrylic paints.

I remember that I had something of a cityscape in mind when I painted the image. But I do not do much planning when I am painting. Often I just paint, add one colour to the other, draw lines, paint lines and so on. 

I did not think much of the painting, and so the painting itself stayed in the folder, and the scan on my hard drive, until the moment when I thought it could be used as a starter image.

And here's what Nightcafe Studio did with the starter image. The style is cubism; the angle is taken from the starter image as well as some of the colors. They are muted, not as happy and vivid as in the image which I created with picFinder.

I think that using the starter image made it possible to create something away from the mainstream.

 

Combining my traditional art and AI Art

With some AI Generators you can upload your own images and use them as starter images (or Init images, as they are sometimes called. I'd already done this in the very beginning of my AI adventure, and then switched over to create images without any init images.

A few days ago I watched a video generated by Playground AI which described the process of creating images from other images (also called image-image process) in detail. Well, I thought, why not doing that again?

In 2018 I had focussed on painting / drawing masks with markers. Some of these paintings are quite okay, and some are not, like the one on the left. Now I wondered what would an AI Generator create out of this rather not so impressive drawing?

I chose one AI Art Generators, Playground AI, using the YouTube video as a guide. I have always fun learning new things, and this was a very good occasion to learn and practise.

So I uploaded the image on the left to Playground AI, and it created the following images:

As you can see, this image is somehow similar to the original painting, but only somehow. The round shape has changed to oval, and the color of the nose changed as well.

So I tried my luck a second time and raised the noise of the original image, thereby giving it more importance,

The image below has more of a handdrawn look to it, and the shape of the face is rounder. If you look at the leaves, you will notice that they look more like in the original drawing. 

I learnt two things:

1. How to change the result of the creation process by raising and lowering the noise slider. Sometimes a wee movement to the left or to the right on the slider can have a big impact.

2. To keep the scans of my handpainted or handdrawn images even when I think that they are not very good or even downright bad. I can always create something out of them with AI!