Some say that a single skill can provide an endless array of possibilities from which to draw and evoke in their work. A photographer may never touch a digital camera and still produce great images. Painters may only use acrylics their entire lives and still draw us into their imaginations. So why expand one’s skill set at all? Why add more tools to one’s tool box?
For several years I was into batik art, which entailed translating printed images from the computer to fabric using carbon paper. Then, I applied hot wax to areas where I wanted to resist the dyes. I repeated this process layer by layer and, after it was all done, I melted out the wax. This process was very different from oil painting. I enjoy both methods for different reasons.
In 2019 I went all in with Deep Dream Generator or DDG, which uses AI or, more specifically, image style transfer to generate images (see above). Some of the images require the use of Photoshop to layer or composite more than one DDG image. Back then I challenged myself to create at least one image per day that I liked and, in the process of doing this, I learned how to choose better base or source images and styles for the tool.
More recently, I started using Midjourney, which is an AI program commonly referred to as a natural language processor or NLP. Unlike DDG Midjourney uses text prompts to generate images such as “The Painted Lady” image above. For several weeks I enjoyed merging my love of creative writing (prompts) with digital imaging. Then, I discovered outpainting in DALL E 2 (another NLP tool).
Outpainting lets users extend an image to create larger-scale scenes based on the image’s original visual elements, such as shadows, reflections and textures. The use of outpainting in the image directly above was inspired by the art of French post-impressionist painter Henri Rousseau. I added his name to my prompt and DALL E 2 merged the artist’s style with the existing image. Then, I used Photoshop to add more elements to the composition. I simulated outpainting using a different tool for the following image.
“Shalimar Women” was ‘outpainted’ by combining two MJ images and manipulated by hand in Photoshop. Familiarity with or mastery of a variety of digital tools enables this. This week I returned to Deep Dream Generator and discovered “Text 2 Dream”, a new feature that combines image style transfer with natural language processing. I chose a Midjourney image to be used as the base/source image, then I entered a different text prompt. Here are the two images:
This result brings me full circle back to when I primarily used DDG to create AI-generated images, except this time there is a new tool in DDG to explore. Emerging technologies continue to expand artists’ toolkits, opening doors to many exciting possibilities. These processes keep artists (and the critics) on their toes, including regarding copyright. When I have more time I plan to dig out my oil paint and canvas to re-create one of the Midjourney images. As an artist I can do that or whatever comes to mind.