I've been playing with AI generated images. One thing I've been doing is exploring "models." Models are trained on different image sets. I've been thinking of trying to build my own models using only public domain images as a way to address one of the ethical issues with models trained on images not licensed for that purpose. As part of this exploration I've been using the reference phrase "disapproving bunny" and the seed of 8675309 on different models to see how different the results can be. The results have been interesting.
Before settling on the shorter "disapproving bunny" as my reference phrase I generated a few with the phrase, "Reference image of a disapproving bunny created by different models." The results were a bit more varied.
No comments:
Post a Comment