AI images, AI Fairies…
I recently decided to step away from the collection of vintage imagery on this site and explore the more contemporary imagery of AI (artificial intelligence). After sampling a number of the available online image generators such as Adobe Firefly, Dream Studio, Midjourney and others, I’ve been working with Dream Studio primarily to generate a variety of AI images which are shown in the gallery below. These are very elementary images, as I have been exploring the various techniques of generating text based images with filters such as anime, digital art, comic book, fantasy art and so forth, and also adding various techniques or filters on top – watercolor, pen and ink, oil, sharpening and similar.
My understanding of AI generated art is that one writes a description, such as ‘fairy with wings riding a dragon’ and the prompt kicks the image generator into action. The generator then matches those verbal prompts to existing imagery in its vast (basically the internet) database of available image data, and generates an image based on those descriptors. The image then can be enhanced further by expanded text prompts, and selecting areas of focus. The primary challenge is to select the right language to create what one wants, similar to searching google by asking the right questions.
A couple of things to consider have arisen during this short exploration. To begin with, ai generated content, including imagery, voice, music, news, etc… is a copyright no man’s land. As the law generally reads currently, ai generated art is not able to be copyrighted, primarily due to the fact that the engine is the creator and not a human, or in other words ‘For a product to be copyrighted, a human creator is needed. AI-generated content can’t be copyrighted because it isn’t considered to be the work of a human creator‘. On top of that is the consideration that any ai image may or may not contain parts and pieces, or layers of other copyrighted work, hence the lawsuits that several ai music generators are currently involved in for example. Another curious aspect of ai imagery is the sheer volume that can be created, which seems to add a layer of numerics to the overall copyright discussion, such as thousand variations of a thousand images.
The other interesting thing to arise is since the ai image generator is collecting imagery from the internet or other collections, it taps into the most recent interpretations of topics. For example, just typing in the word ‘fairy’ generates a young woman with or without wings in a flowing dress in one of the most popular filters,, and which represents the general interpretation of fairies today in modern culture. Notice the consistency of the fairy images in the examples below. Working away from these ‘defaults’ requires more refined text prompts, exploration and patience.
One other observation worth mentioning is that while the art below is pretty, it lacks any real artist expression, something seems to be missing. Which brings up my final thought, that ai is a tool, not much different than a brush, pen or paint. It certainly expands one’s creativities, but at the end of the day, the image created is based on my text prompts and working within the constraints of the program. To try to capture that emotional edge, that unique perspective, that unexpected expression, is the challenge as an artist. I will certainly continue to work with these ai image generators, and see where things go.
References –
