Throughout the spring of 2023 I began to experiment with ChatGPT to interpret and draw well documented paintings. The first was one of my personal favorites, "The Potato Eaters" by Vincent Van Gogh. I knew from past experience that ChatGPT generally must be told what format to create a drawing in. In this case, I asked the following:
"I'm trying to make a drawing in SVG format that mimics a painting titled "The Potato Eaters" by Vincent Van Gogh. Use the same aspect ratio, shapes and color palette to the best of your ability."
ChatGPT Response:
"As an AI text-based model, I can't actually create visual images or SVG files. However, I can certainly help guide you through the process of creating an SVG drawing that mimics "The Potato Eaters" by Vincent Van Gogh.
First, let's understand the key elements of the painting:
Aspect ratio: The original painting's dimensions are 82 cm × 114 cm. Therefore, the aspect ratio is roughly 1:1.39.
Shapes: The painting depicts a group of peasants around a table in a dimly lit room. Key shapes you would need to represent include the figures of the people, their facial features, their clothes, the table, the dishes and potatoes on the table, and the room around them.
Color palette: Van Gogh uses a dark, earthy color palette in this painting, with lots of browns, blacks, greys, and yellows. When creating your SVG, you might choose hexadecimal color values like #654321 (dark brown), #000000 (black), #808080 (grey), and #FFFF00 (yellow) to start with."
ChatGPT also provided a simple SVG code example. The resulting drawing was much like you would expect from an insecure child who was trying to recreate a masterpiece. After several rounds of instructions, I decided was about the best ChatGPT can render the concept using simple vector graphics.
At no time did I modify the code provided by ChatGPT. Rather, this is the exact output of how the Ai visually communicates it's understanding of Van Gogh's Potato Eaters with basic human instruction. It is rumored that OpenAi may soon add DALL-E functionality to ChatGPT, which will dramatically improve the Ai's ability to return complex imagery based on text prompts. What you see here is a snapshot in time when ChatGPT was still a child.
Learn more about my work at www.RobertMatheson.com and @Robert_Matheson