My Selection Of Best Pictures Created By AI

Daily Challenge Images On NightCafe

6 min readMar 19, 2024

I recently read Stefano Roncoroni’s story, My Selection Of Best Pictures Created By AI. And my immediate thought was that it was a great idea. Also it was time that I should compile my AI artwork created on NightCafe. I have created several pieces under the name Alfred Stevens on NightCafe. But for this story I have only selected a few of my entries in NightCafe's Daily Challenge.

In my response I wrote, I will always be a fan of AI images. They are magnificent. And they are art at your fingertips. Something I’ll never be able to afford.
Nice read, I enjoyed it.

The images portrayed by AI include computer software. Or the application is referred to as generative models. These networks learn from all samples available of real images, and in this way they generate new images. Here are two generative model types, VAEs and GANs are often attention-grabbers.

For image generation, prompts can specify the content, style, or desired features of the image.

When you want the artificial intelligence to draw up any specific image, you input the “prompt” to it. A prompt means commanding the AI to look at what you want to be in the drawing. For instance, you may pronounce something to the effect of:

“Portrait of a charming divine girl, shiny face, curly black hair, blue eyes, red lips, white teeth, earrings, necklace, blue fashion dress, heavenly smile, sun, nature . steampunk engine heavenly sunshine beams divine bright soft focus holy in the clouds cyberpunk 2099 blade runner 2049 neon”

This is specifically true of language models that rely on AI. For example, this is demonstrated in the context of OpenAI’s GPT-3 where end-users are expected to instigate the model by supplying the prompts and which kind output is expected, whether it is generating text or code or even images.

AI is the software that it will use as the way point to get a picture of what it wants to see. This can also help you to some extent hold on to which of the content the AI will incorporate into its creation. Thus, crucial discussions for this are fairness and biases in the AI system. Therefore, it also focuses on the ethical part by trying to ensure that AI-generated images are being responsible and also fair.

Weight

The weight is a parameter learnable in some machine learning, which is neural networks included. An ‘Weight’ encloses a signal in the neural network. The signal is the strength of the signal transferred from one neuron to another. A weight is a factor to express how strongly the input of an output will be.

Hyper realistic, splash art, concept art, mid shot, intricately detailed, color depth, dramatic, 2/3 face angle, side light, colorful background”

Negative Prompts

Negative prompts for realistic images is a revolutionary trend in the field of generative art. They behave like a manual with exact details that put no AI at risk, unlike human beings. It is essential that this focus is on v2 versions of Stable Diffusion where a negative prompt eliminates the noise that is ostensibly in those photos to sharpen or change them into more realistic ones.

Ugly
Bad anatomy
Bad proportions
Bad quality
Blurry
Cropped
Deformed
Disconnected limbs

Here you can see that this image is off. The character in the image is using two writing tools to accomplish her task. This is one of the flaws one will encounter when using an AI tool. The final image is stunning nevertheless.

The character in the image is a baby. Here we can see that the feet of the infant is badly deformed, giving it a clubbed feet effect. The hand is also missing a finger because a negative prompt wasn't used.