I trained a 128×128 DCGAN on my Macbook M3 and deployed it on a Raspberry Pi 4 connected to a LILYGO TTGO T-Display ESP32. The whole thing runs headlessly as a systemd service and generates hallucinated face hybrids at the press of a button. It is a 6-block generator (latent → 4×4 → 8×8 → 16×16 → 32
