Underdog shows off a 32MB speech model in new demo

Posted Friday, November 6, 2026 8:02 am

Ivan Mehta (via TechCrunch)

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

AI, local AI, underdog