The Color You See is an interatice prototype that transforms words into a vivid palette of colors. Lying at the intersection between enginnering and visual design, it allows people to define the primary color (RGB) through their own perceptual lens, and then construct the entire color space based on this subjective foundation, thus the color of words is generated based on people’s personal connection with colors.
We primarily engage with algorithms through language, but the same word in different languages creates varied visual realities in Generative AI. Color is a prime example:
"Blue" As Prompt
"蓝色 (Blue in Chinese) " As Prompt
Generated from a vast dataset scraped from the internet, blue is seen by AI mostly as a female figure, decorated with visual cliches responding to the input language. These images of blue is accurate to the algorithm in the sense that it presents a certain type of visual that is defined by the naming and tagging of the word blue spread across the medium where the data is gathered.
And yet what is accurate to the algorithm might not be as accurate to each one of us. Color is a natural phenomenon, but it is also a complex cultural construct that resists the reduction to a single name of a color. Color is not something to be seen and named but to be felt and experienced. And this experiential quality of color varies among each individual.
Using the Word2Vec model, a technique in NLP for vector representations of words, I extended it to translate word vectors into three-dimensional RGB color vectors. Each word vector is mapped onto the RGB space, with dimensions reduced and correlated to specific colors. Participants can define their perceptions of red, green, and blue, creating a subjective relationship between words and colors.
Full Documentation Link