Using ChatGPT for creating prompts for AI Art

 I had created a few monster AI images using my own art as start images. Well, there are some AI Art Generators which do not allow the use of starter images. I wanted to create monsters with them as well and was thinking about possible prompts. But I had no ideas regarding the prompts, so I went to ChatGPT and wrote:

I am an AI artist and I want to create images depicting cute monsters. Could you provide me with some prompts, please!

ChatGPT provided me with 10 prompts; most of them would never have come to my mind. I picked one of them:

A shy monster peeking out from behind a curtain with a bashful smile

and used it with Lexica AI.

Now we could discuss whether the monster in the image on the left has a "bashful smile" - but it has a smile.

Of course the prompt given by ChatGPT could be specified: We could ask the AI Art Generator to create a curtain in a certain color or texture: we could go for more details regarding the monster.

The main point is that I got 10 prompts full of ideas and contexts for the monster.

Why I participate in Challenges on Nightcafe Studio

I have been creating with Nightcafe Studio more than one year now. They have got daily challenges, and you can win additional credits by getting a certain amount of votes for your submitted image. There are a lot of discussions in the facebook group where people express their disappointment about not getting votes for their creations. I must admit that for some time this kept me from participating in challenges. People do not see that opinions on art, on images whether they have been created traditionally or digitally are subjective. I had to learn that as an artist myself - but I always created what I wanted to create and not what I thought people would like. Of course I am happy when people like what I create, but that is not the main reason for creating art for me. But I digress.

For some days I have rediscovered the challenges on Nightcafe Studio for a very special reason: It makes me try out a certain theme on various AI Art Generators. Here is an example, "Sharks" respectively "shark linocuts":

This is how Deep Dream Generator interpreted the prompt "beautiful colorful linocut of a shark swimming in the water". Yes, it looks like very simple linocut, and I like the pose of the shark, but I must say I otherwise I am not very impressed. In the beginning of my AI Art activities I used this Generator quite a lot, but nowadays it belongs to my least favorite AI tools. The results are often disappointing.

Same prompt but this image looks quite different, doesn't it? I used Lexica AI here, and since a month or so Lexica is one of my favorite AI Generators. I still have a free account. At the moment they offer you 100 images per week (one prompt generates four images).

And here's the result I got from Nightcafe Studio itself. I hadn't added "beautiful colorful" in the prompt, so I got this black and white image. I never did a linocut myself, but for me the creation really looks like a linocut.

So participating in the challenges on Nightcafe Studio does two things for me:

1. It gives me ideas for creations. 

2. It motivates me to try out various AI Art Generators. 

Lexica, one of my favorite AI Art Generators

 Last year Lexica underwent some changes, and that process obviously met with some obstacles. I have a free account there, with being able to create 100 images monthly, and suddenly I could not generate any images at all. So I decided to forget this AI Art Generator.

Then, suddenly at the end of 2023, things went smooth again, and I now can generate 100 images on a weekly basis. I am very pleased now, because I very much like the results I am getting with Lexica. Here are some examples:

Prompt: colored linocut of a shark swimming in the water

abstract oilpainting

Racoon reading the paper in the bus

Using my own paintings as starter images with AI Art Generators

Regarding my traditional art doing abstract paintings with acrylic ink, watercolor ink or alcohol based markers has always been one of my favorite art activities. Now, that I do know a bit more about the possibilities of AI Art generators I wanted to use more of my traditional paintings. So I selected this painting, done with acrylic inks:

I did this painting in 2020. It is not one which I consider very well done, but I like the colors and the movement of the painting.

Now I tried out several Art Generators. I tried Blue Willow - which allows Image to Image generation, and I got a description of how to do it - but I somehow I failed again and again. I tried Dreamlike AI, which allows image to image generation and makes it easy to do it - but was not convinced of the result. Then I tried Lexica, and here is what I got:

As Lexica always generates four variations of an image, I had to select one of four. I must say I am pretty impressed with the result.

I also tried Deep Dream Generator with the same original painting. 

In my opinion the Deep Dream Generator version is more playful than the Lexica one. However I like both results, and will continue with combining my traditional art with AI Art.

Rose woodcarving

Woodcarvings are something I highly appreciate. In Germany and Austria an old tradition of carving nativity figures exist which are then sold internationally. There is also a German artist called "Tilman Riemenschneider" who was famous for his wooden carved out altarpieces. 

I paint, I draw - but I could never carve. I haven't got the eye for it, not to speak of eye-hand-coordination. But with AI Art Generators I am able to carve respectively generated images with carvings on them.

I decided to go for roses, and here is the first one, a wooden carving of a rose, created with the use of Nightcafestudio. 

The prompt I submitted was:

wood carving of a rose, highdefinition, hyperrealistic, intricate

Nice, isn't it?

However I decided to go to Lexica AI and submitted this prompt:

Wood carving of a beautiful rose, hyperrealistic, highdefinition, intricate

As you may notice, there is one additional word in the second prompt: "beautiful". Now have a look at the result: In my opinion the word "beautiful" makes the image richer, more detailed,

I must say, I prefer this image to the first one.

However I wanted to find out how Dreamlike AI handels the prompt. This time I used the very same prompt:

Wood carving of a beautiful rose, hyperrealistic, highdefinition, intricate

Dreamlike generated something slightly different, as you can see in the image below.

The wooden carving is not a realistic rendering of a rose, but very detailed and decorative. It could be an ornament on a piece of furniture.

Creating images with Lexica ai

 I havve been using three ai art generators so far:

1. Nightcafestudio
2. Deep Dream Generator
3. Lexica Ai

All of them can be used for free. Whereas Nightcafestudio and Deep Dream Generator are based on a credit /energy points system, Lexica offers you 100 images per month for free. This may seem to be quite a lot, but you have to keep in mind that every time you generate an image, you get four images. 

Here is an example:

You can easily see that you really get four different images with the following prompt:

jungle with moss covering the dirt, colorful flowers, trees with thick roots, in the style of Ivan Bilibin, Ernst Haeckel, Daniel Merriam, watercolor and ink

You can choose the dimensions, either square, portrait or landscape. I downloaded one of the four images and it had 2560 px width and height. After postprocessing and downscaling it looks like this:

As you can see, the style is a bit dreamy and fantasy like - but that's okay with me.

So I intend to continue to have fun with Lexica - having to watch myself not to create too many images!

Using Lexica in order to create images with DALL-E2

 Lexica is a website where you can do two things:

• Generate images with Stable Diffusion
• Search for images and get the prompts which created them. 

On Lexica, you have to pay in order to be able to create a certain number of images (Mind you, Nightcafe is free, but you have to do something in order to get credits for your creations - log in daily, vote in challenges, put your images on instagram and twitter). The  most important difference for me, however, is that you can upscale your images in Nightcafe,- paying with credits, and you cannot in Lexica. So I will not use Lexica for creating images.

However: Browsing through the images in Lexica gives you hundreds of prompts. In many cases the prompt and the generated image differ. I tried out some prompts in DALL-E2 on Nightcafe - and the result was more similar to the prompt than the image on Lexica itself. 

Here's an example:

I copied the prompt from Lexica and changed it a bit. It now reads:

Photo of a golden statue of a lion with huge wings

The original prompt included "a collar", which I did not want in the image.