Best wishes for these Festive days.

Best wishes for the Christmas days and may 2026 be a year of enjoyment and good health.

(based on a gouache painting I made on location in 's Hertogen Bosch december 2024).
 

AI images (5): the burning towers

 

Another AI experiment. I started with the images of the last post (ai-images (4) the wing, september), the drawing and its AI-image (top left and middle). Then I made a new drawing, which actually turned out to be horrible (not shown here).

I lost interest and eventually played around with the blue AI-image and combined it with another AI image theme that I created. Some burning towers in a desolate landscape (top right). That resulted into a new drawing (bottom left). 

Then I generated a new AI image with reverse prompting. But that one became rather like a fake watercolour drawing. But I also experimented with turning a sketch into an image using AI. Rather interesting what the system came up with.

AI images (4). The wing.

 

Next installment of my AI game. I used the Nike of Samothrace drawing (with colours) to generate a prompt. As stated in the last post, it resulted in different images. I liked the interior one the best, but as you can see the statue was gone. I replaced it for a single wing (representing the Nike statue), and added some smoke. See above drawing, made with a pilot Parrellel pen with non-waterproof ink. 

Then I put it in the generator again, which led to horrible results. All very soft, girly like, fake watercolour images of pigeons flying in clouds of flowers, in pink, red, yellow and green. Yuck. 

So I had to tweak the prompts a bit  to get my wing and ruined buildings back again. And I grew tired of all the red, so I guided the generator towards grey and sepia. The result (through GPT-4o and GPT image) is shown below (bottom right). I will use that one to make a new drawing. It's time for the wing to leave the building.

AI images experiment (3)

 

As you can see I gave the ink drawing of the last AI experiment installment some shades. (done with a pentel Parrallel pen which has non-permanent red ink). Now I put this through the reversed prompting program, which translates the image into a text (which you can reuse again to generate new images). Interesting enough, different text machines deliver totally different images, of which the focus is different. Sometimes drawing style, sometimes colour, sometimes subject. The last one (Florence) even specified a red bird, which the picture didn't even show. Pictures all made by Stable Diffusion 3.0. It clearly shows how you need to work your prompting to get the results you want.

The images: top left: original drawing; right: Joycaption; second row left: GPT4o-mini; right: Florence.

AI images experiment (2)

I made above ink drawing (another step after the last AI image of the previous session, I added the Greek statue of Nike of Samothrace, it can be seen in the Louvre). I did reverse prompting of this image, and used that text to generate new images. I used different AI generation programs, even though the themes are similar, the outcome is very different.

From top left to top right: GPT image1, google imagen4, Stable Diffusion 3.0. Bottom: Classipeint XL, Flux dev, google imagen3.

 

AI images (1): the pool

 

An experiment with creative use of AI. 

I used an old drawing (La piscine des erreurs, from september 2016), applied reverse prompting and used the resulting texts (with some adaptations) to generate a few AI images. I used stable diffusion 3.0 through the Openart platform. Most of the results were not to my liking, I noticed that you will need some tweaking of the prompts to get nice results. Then I used one of the better images as inspiration to make a new drawing. See above, the one with the drapes. 

Then I converted the drawing through reverse prompting into an AI image again, see below. I will use that one to draw something new again. Interesting results.