Underdog shows off a 32MB speech model in new demo

Fecha de publicación: 6 de noviembre de 2026, 13:02

Despite its small size, Underdog's new model can generate both speech and music.