!
A blog related to original AI artworks. I make AI artworks using coined words and emoji as prompts.
Most image-generating AI, including Stable Diffusion, currently represent images in 16 bits per pixel, which is often insufficient.
When generating images with AI, it often produces images with very low contrast. Increasing the contrast in such images essentially represents the image with much less information than 16 bits. As a result, artifacts may appear in areas where colors should smoothly transition. Here is an example:
In particular, the sky area is concerning.
To avoid this, it’s necessary to increase the amount of information per pixel, for example, by using 32 bits per pixel.