Spectrum Miami (December 3–7, 2025), booth #1112 — The FaCells are portraits of 15 facial attributes learned by a neural network trained on line drawings. More than 200,000 images were converted into sets of lines — sequences of (x, y) points — and used as input for an LSTM model. Each attribute corresponds to one neuron, or “cell,” in the network’s final layer.
| 0 – Arched_Eyebrows | 1 – Attractive | 2 – Bangs |
| 3 – Blond_Hair | 4 – Eyeglasses | 5 – Heavy_Makeup |
| 6 – High_Cheekbones | 7 – Male | 8 – Mouth_Slightly_Open |
| 9 – No_Beard | 10 – Smiling | 11 – Wavy_Hair |
| 12 – Wearing_Hat | 13 – Wearing_Lipstick | 14 – Young |
By hacking the model, it can be made to generate attributes instead of simply classifying them. The plotter connected to the system acts as a medium between the digital and the physical worlds, drawing each FaCell on paper as the network and the plotter express what they have learned — revealing their own distinctive style and aesthetic.
Close your eyes and imagine the idea of eyeglasses. Think about how that image exists within your mind — how it is stored and recalled. The FaCells illustrate, through accessible mathematics and visible process, how machines can learn to represent and retrieve visual ideas too, using the language of lines.
The FaCells is one of the earliest art installations where AI stands as the main element — and they’re not for sale. If you want one, you need to attend Spectrum and make a post on social media. Come visit!