Craiyon, my preferred AI image generator, got a new major release out just in time to stop me from switching to Nightcafe. The new version includes a redefined selection of style models, modifiable aspect ratios, and most notably, base image support. In response, I pitted my some of original art pieces against the AI. While I was initially a little jarred at seeing my art both upscale and deep-fried, I found a new way to use it.

The way I look at it, base image generation works by showing the AI a blurred version of the base image and having it reconstruct it from the provided prompt. The process results in the AI mixing impurities with the provided framework, which can create useful and valuable results. In the roughly 15 minutes I've been using it for so far, I've run across two useful character designs. With active engineering and further development, I should be able to make something of both quantitative and qualitative value!