AI Art

I’ve been exploring using various artificial intelligence tools such as Midjourney, Dall-E, DreamStudio and ChatGPT to create art and inspiration.

MidJourney Inspiration & Exploration

AI image generators can be used to create artistic output in their own right but I’ve also been using them to generate inspiration for my coded generative art. Exploring outputs by keyword engineering is a new way of creating which feels immediately natural.

amygoodchild_illustration_of_thousands_of_tiny_creatures_futuri_b48013ee-a60c-4238-a096-56b44c8679a2.png

amygoodchild_aesthetic_graph_8fa17f7a-0edf-4864-bc6a-df922e180865_3.png

amygoodchild_graph_aesthetic_hand_drawn_watercolour_35c4b665-b8bb-4355-9534-4789d6aadcfb_1.png

amygoodchild_aesthetic_graph_8fa17f7a-0edf-4864-bc6a-df922e180865_1.png

amygoodchild_hand_drawn_scientific_diagram_labelled_with_glyp_23631528-8bca-48d9-98f7-a8ceecc89497_1.png

amygoodchild_graph_aesthetic_hand_drawn_watercolour_35c4b665-b8bb-4355-9534-4789d6aadcfb_3.png

amygoodchild_theosophy_and_cosmology_theory_interconnectednes_7a265e02-3423-4eed-ab06-69395503bd0e_2.png

amygoodchild_mysterious_crt_monitor_in_a_forest_early_evening_fd44d252-d09f-4ec9-9066-eed4a1ec877a_0.png

ChatGPT x Sol LeWitt

ChatGPT produces text output including code, so it can be asked to produce p5js that draws images on an html canvas. I used this technique implement some of the Sol LeWitt wall drawings. My blog article about the whole process is here and it’s a fun read!

Wall Drawing #766 - ChatGPT Javascript output

Wall Drawing #118 - ChatGPT javascript output

Wall Drawing #415D - Human Drawn

Wall Drawing #415D - ChatGPT output

Midjourney x Sol LeWitt

The natural next step after my ChatGPTxSolLeWitt experiments was to use Midjourney to implement the Wall Drawing Instructions. With these I strayed further from the prompts and sought out aesthetically interesting outputs. You can read my blog article about the process here.

Generative Morphing

I made this video using Ganbreeder, a tool powered by BigGan, which has been trained on tens of thousands of categorised images. By editing the ‘genes’ of the output, I generated 200 images as keyframes.

I and then used Gantools to morph between the keyframes and create a smooth animated output.

Art to Art

DreamStudio is powered by Stable Diffusion and allows for a combination of keyword and image inputs. I’ve been using it to produce amended versions of my own artworks as new pieces.