Sign In

A small town on a flower petal

0

May 25, 2026

musing
A small town on a flower petal

I have long tried, using the graphical capabilities of AI, to place a small town on a flower petal in such a way that it would look beautiful. Older models generated the town either in the center of the flower or on the leaves. Some advanced models of the previous generation could complete the task, but the images were not beautiful. And now, with the ChatGPT Images 2.0 model, I was able to get an image that I consider generally matching my idea.

2026-05-01T22.56.50_15.png

https://civitai.red/images/131752340

Right now, I believe that the ChatGPT Images 2.0 model is the optimal choice for a beginner in the field of AI generation. About a year ago, I wanted to bring an artistic technique into an image by describing it as follows. 'The image has a delicate structure and is built as if from colored threads, all shapes, all filamentous'. At that time, I did not know that the models available then could not understand from my prompt that I wanted to achieve. But ChatGPT Images 2.0 generated an image in which you can see exactly what I specified. The image is far from perfect, but this is the first attempt, and the generation result is encouraging.

artbreeder-prompter-2026-05-03T03_36_50.299Z.jpeg

https://civitai.red/images/131751920

The model also has drawbacks. For example, the model cannot place chess pieces in the starting position or draw a chessboard with the correct number of squares. And a competing model can do that. I also saw a hand with six fingers, although that is no longer a common occurrence. Some images are unpleasantly surprising with their ugliness. Fortunately, there aren't many of these. And yet the strengths of the model ChatGPT Images 2.0 model outweigh the drawbacks, providing a good product with a quite reasonable generation cost and a good level of artistic style. Moreover, it has the highest percentage ratio of successful to unsuccessful images for me. Now it can be said that beginners in the field of image generation will no longer have to struggle in fruitless attempts to realize most of their ideas, and if something doesn’t work out, they can turn to a competing but more expensive model. And, considering the editing capabilities, we have an excellent toolkit available. All that’s left is to come up with ideas.

0