A blog related to original AI artworks. I make AI artworks using coined words and emoji as prompts.
Negative prompts are used to diversify AI art images, but it has been found that in some cases, diversity may decrease or almost disappear when negative prompts are provided.
The following images were generated with “Kanabstracd” and the painter Hieronymus Bosch as positive prompts, and Salvador Dali as the negative prompt.
Various random numbers and parameters were changed, but the patterns are almost identical. Without negative prompts, diversity is not compromised to this extent.
In the following case, diversity is not compromised to the same extent, but when Kanabstracd and Giorgio de Chirico are given as positive prompts, and “gray” or “Chirico” as negative prompts, the following images were created.
When these are used as negative prompts, most images feature a road in the center with buildings on either side. Whether it is Bosch or Chirico, why does this happen?