A blog related to original AI artworks. I make AI artworks using coined words and emoji as prompts.

Experimenting with Learning from Photographers to Create Art

So far, I have been using AI to learn from painters (specifying their names) and create artwork. However, this approach may raise copyright issues when learning from existing artists. Therefore, I thought about trying to learn from photographs and creating art. If I can produce paintings that don’t look too much like photos, it should avoid any copyright problems. Although I have only tried a few, the results are promising enough to continue this endeavor.

Since I don’t know many names of photographers, I searched the web and used them with Stable Diffusion 2.1. As usual, I provided the term “painting” as the positive and negative prompts to convert them into paintings. I created an equal number of paintings with the photographer’s name as the positive prompt and with both positive and negative prompts. However, most of the following paintings were generated using only the positive prompt because they turned out more impressive.

One of the experiments involved a Swedish photographer named Christian Aslund. You can find many of his photos, including those from Antarctica, on his Instagram. Here are some examples of the generated paintings, all of which had Aslund’s name as the negative prompt. Paintings generated with the positive prompt did not turn out well and resembled photos, so I haven’t included them.

The following paintings were generated by specifying “Erik Almas” only as the positive prompt. Although his Instagram account features many photos of people and nature, Stable Diffusion creates paintings of cityscapes.

Since specifying negative prompts did not result in particularly interesting artwork, I will not feature them. This applies to the following photographers as well. In other words, Aslund was the only one who could create interesting artwork with negative prompts.

Next is Florian Ritter (Florainritter). These paintings were generated by specifying “Florainritter” as the positive prompt.

Next is a photographer named Kait Robinson. These paintings were also generated with only the positive prompt. Although various pages can be found by searching on the web, it is unknown whether they belong to this photographer. Therefore, it is also unknown if the generated paintings are connected to the photographer.

Next is Andrea Gjestvang. These paintings were also generated with only the positive prompt.

Next is Jeremy Cowart. Although known as a photographer, he also creates art. These paintings were generated with only the positive prompt.

The dog painting (or photo?!) looks like a photo taken through glass with water droplets, so it might be a copy.

Next, let’s talk about Mareen Fischinger. For some reason, the majority of the paintings created are flowers. These paintings were also generated using only the positive prompt.

Next is Joe Mcnally. These paintings were also generated using only the positive prompt.

I have also generated paintings by other photographers, including Japanese photographers, but I haven’t compiled them yet.